Raffelt - Stars As Laboratories For Fundamental Physics

Georg G.
Raffelt
Stars as Laboratories
for Fundamental Physics
The Astrophysics of Neutrinos, Axions, and Other
Weakly Interacting Particles
Final Manuscript, October 1995
Published by the University of Chicago Press, 1996

Georg G. Raffelt is a staff researcher at the Max-Planck-Institut für Physik (MPP)
in Munich, Germany, and a member of the Sonderforschungsbereich Astroteilchen-
physik (Special Research Center Astroparticle Physics) at the Technische Univer-
sität München. He is co-editor of the journal Astroparticle Physics. He studied
physics at the University of Munich and the University of California at Berkeley,
received a Master of Arts from the latter, and a Diplom and in 1986 a doctorate
from the former. Before taking up his position at the MPP in 1990 he has completed
several years of postdoctoral research at the Astronomy Department in Berkeley,
the Institute for Geophysics and Planetary Physics in Livermore, and the Institute
for Advanced Study in Princeton.
To be published by the University of Chicago Press, Chicago.

c University of Chicago Press (1996)
!
Table of Contents
Preface (xiv)
Acknowledgments (xxii)
1. The Energy-Loss Argument
1.1 Introduction (1)
1.2 Equations of Stellar Structure (5)
1. Hydrostatic Equilibrium (5) ! 2. Generic Cases of Stellar Structure (7) !
3. Energy Conservation (10) ! 4. Energy Transfer (11) ! 5. Gravitational
Settling (13)
1.3 Impact of Novel Particles (14)
1. Energy Loss (14) ! 2. Application to the Sun (16) ! 3. Radiative Energy
Transfer (17) ! 4. Opacity Contribution of Arbitrary Bosons (18) ! 5. Solar
Bound on Massive Pseudoscalars (20)
1.4 General Lesson (21)
2. Anomalous Stellar Energy Losses Bounded

by Observations
2.1 Stages of Stellar Evolution (23)
1. The Main Sequence (23) ! 2. Becoming a Red Giant (28) ! 3. Helium
Ignition (33) ! 4. The Horizontal Branch (34) ! 5. From Asymptotic Giants
to White Dwarfs (35) ! 6. Type I Supernovae (36) ! 7. Intermediate-Mass
Stars (37) ! 8. Massive Stars and Type II Supernovae (37) ! 9. Variable
Stars (39)
2.2 White-Dwarf Cooling (42)
1. Theoretical and Observed White-Dwarfs Properties (42) ! 2. Cooling
Theory (45) ! 3. Neutrino Cooling (47) ! 4. Cooling by Boson Emission (50)
! 5. Period Decrease of Variable White Dwarfs (52)
2.3 Neutron Stars (54)
1. Late-Time Cooling (54) ! 2. X-Ray Observations (56) ! 3. Nonstandard
Cooling and Heating Effects (58) ! 4. Cooling by Particle Emission (59)
v
vi Table of Contents
2.4 Globular-Cluster Stars (60)

1. Observables in the Color-Magnitude Diagram (60) ! 2. Theoretical Re-
lations (65) ! 3. Observational Results (70) ! 4. Interpretation of the
Observational Results (74) ! 5. An Alternate Analysis (76) ! 6. Systematic
Uncertainties (78)
2.5 Particle Bounds from Globular-Cluster Stars (79)
1. Helium-Burning Lifetime (79) ! 2. Helium Ignition (83)
2.6 Summary (87)
3. Particles Interacting with Electrons

and Baryons
3.2 Compton Process (91)
1. Vector Bosons (91) ! 2. Scalars (93) ! 3. Pseudoscalars (94) ! 4. Neutrino
Pairs (95) ! 5. Energy-Loss Rates (96) ! 6. Applying the Energy-Loss
Argument (98)
3.3 Pair Annihilation (99)
3.4 Free-Bound and Bound-Free Transitions (100)
3.5 Bremsstrahlung (101)
1. Nondegenerate, Nonrelativistic Medium (101) ! 2. High Degeneracy:
Pseudoscalars (103) ! 3. High Degeneracy: Neutrino Pairs (105) ! 4. Neutron-
Star Crust (106) ! 5. Applying the Energy-Loss Argument (107)
3.6 Astrophysical Bounds on Yukawa and Gauge Couplings (109)
1. Pseudoscalars (Axions) (109) ! 2. Energy Loss by Scalar and Vector
Bosons (110) ! 3. Long-Range Forces (112) ! 4. Leptonic and Baryonic
Gauge Interactions (114)
3.7 Graviton Emission from Stars (116)
4. Processes in a Nuclear Medium

4.2 Axionic Bremsstrahlung Process (119)
1. Matrix Element for N N → N N a (119) ! 2. Energy-Loss Rate (120) !
3. Nondegenerate Limit (121) ! 4. Degenerate Limit (123) ! 5. Bremsstrah-
lung Emission of Scalars (124) ! 6. Mixture of Protons and Neutrons (125)
4.3 Neutrino Pair Emission (126)
1. Structure Function (126) ! 2. Bremsstrahlung Emission of Neutrino
Pairs (130)
4.4 Axion Opacity (131)
Table of Contents vii
4.5 Neutrino Opacity (132)

1. Elastic Scattering (132) ! 2. Pair Absorption (132) ! 3. Inelastic Scattering
(133)
4.6 Structure Functions (136)
1. Formal Definition (136) ! 2. Nonrelativistic Limit (138) ! 3. The f-Sum
Rule (139) ! 4. Long-Wavelength Properties (141) ! 5. Axion Emission
in the Classical Limit (143) ! 6. Classical vs. Quantum Result (146) !
7. High-Density Behavior (147)
4.7 Effective Nucleon Mass and Coupling (151)
4.8 The URCA Processes (152)
4.9 Novel Phases of Nuclear Matter (155)
1. Pion Condensate (155) ! 2. Quark Matter (157) ! 3. Bubble Phase (159)
4.10 Emission of Right-Handed Dirac Neutrinos (160).
5. Two-Photon Coupling of Low-Mass Bosons

5.1 Electromagnetic Coupling of Pseudoscalars (165)
5.2 Primakoff Process in Stars (168)
1. Screened Cross Section and Emission Rate (168) ! 2. Plasmon Decay
and Coalescence (170) ! 3. Axion Emission from Electromagnetic Plasma
Fluctuations (172) ! 4. Solar Axion Spectrum (175) ! 5. Globular-Cluster
Bound on gaγ (176)
5.3 Search for Cosmic Axions (176)
5.4 Axion-Photon Oscillations (179)
1. Mixing Equations (179) ! 2. Solar Axions (181) ! 3. Shining Light through
Walls (182) ! 4. Vacuum Birefringence (183).
5.5 Astrophysical Magnetic Fields (185)
1. Transitions in Magnetic Fields of Stars (185) ! 2. Birefringence in a
Pulsar Magnetosphere (186) ! 3. Conversion of Stellar Arions in the Galactic
Field (188) ! 4. Polarimetry of Distant Radio Sources (190) ! 5. Temperature
Fluctuations in the Cosmic Microwave Background (191).
5.6 Summary of Constraints on gaγ (191)
6. Particle Dispersion and Decays in Media

6.2 Particle Dispersion in Media (196)
1. Refractive Index and Forward Scattering (196) ! 2. Particle Momentum
and Velocity (198) ! 3. Wave-Function Renormalization (200)
viii Table of Contents
6.3 Photon Dispersion (202)

1. Maxwell’s Equations (202) ! 2. Linear Response of the Medium (204) !
3. Isotropic Polarization Tensor in the Lorentz Gauge (206) ! 4. Lowest-
Order QED Calculation of Π (209) ! 5. Dispersion Relations (212) !
6. Renormalization Constants ZT,L (217)
6.4 Screening Effects (219)
1. Debye Screening (219) ! 2. Correlations and Static Structure Factor (222)
! 3. Strongly Coupled Plasma (223) ! 4. Screened Coulomb Scattering (224)
6.5 Plasmon Decay in Neutrinos (227)
1. Millicharged Neutrinos (227) ! 2. Neutrino Dipole Moments (228) !
3. Standard-Model Couplings (229) ! 4. Summary of Decay Rates (231) !
5. Energy-Loss Rates (232) ! 6. Astrophysical Bounds on Neutrino Electro-
magnetic Properties (234)
6.6 Neutrino Form Factors in Media (237)
6.7 Neutrino Refraction (241)
1. Neutrino Refractive Index (241) ! 2. Higher-Order Effects (245) ! 3. The
Sun a Neutrino Lens? (247)
6.8 Majoron Decay (248)
7. Nonstandard Neutrinos
7.1 Neutrino Masses (251)
1. The Fermion Mass Problem (251) ! 2. Dirac and Majorana Masses (253) !
3. Kinematical Mass Bounds (254) ! 4. Neutrinoless Double-Beta Decay (257)
! 5. Cosmological Mass Bounds (258).
7.2 Neutrino Mixing and Decay (260)
1. Flavor Mixing (260) ! 2. Standard-Model Decays of Mixed Neutrinos (263)
7.3 Neutrino Electromagnetic Form Factors (267)
1. Overview (267) ! 2. Single-Photon Coupling (268) ! 3. Two-Photon
Coupling (271)
7.4 Electromagnetic Processes (272)
7.5 Limits on Neutrino Dipole Moments (275)
1. Scattering Experiments (275) ! 2. Spin-Flip Scattering in Supernovae (277)
! 3. Spin-Flip Scattering in the Early Universe (277) ! 4. Search for Radiative
Neutrino Decays (278) ! 5. Plasmon Decay in Stars (279).
Table of Contents ix
8. Neutrino Oscillations
8.2 Vacuum Oscillations (282)
1. Equation of Motion for Mixed Neutrinos (282) ! 2. Two-Flavor Oscilla-
tions (284) ! 3. Distribution of Sources and Energies (287) ! 4. Experimental
Oscillation Searches (289) ! 5. Atmospheric Neutrinos (290)
8.3 Oscillations in Media (293)
1. Dispersion Relation for Mixed Neutrinos (293) ! 2. Oscillations in Homo-
geneous Media (296) ! 3. Inhomogeneous Medium: Adiabatic Limit (297)
! 4. Inhomogeneous Medium: Analytic Results (299) ! 5. The Triangle and
the Bathtub (301) ! 6. Neutrino Oscillations without Vacuum Mixing (303)
8.4 Spin and Spin-Flavor Oscillations (304)
1. Vacuum Spin Precession (304) ! 2. Spin Precession in a Medium (306) !
3. Spin-Flavor Precession (306) ! 4. Twisting Magnetic Fields (308)
9. Oscillations of Trapped Neutrinos

9.2 Kinetic Equation for Oscillations and Collisions (313)
1. Stodolsky’s Formula (313) ! 2. Matrix of Densities (315) ! 3. Free
Evolution: Flavor Oscillations (316) ! 4. Interaction with a Background
Medium (317)
9.3 Neutral-Current Interactions (320)
1. Hamiltonian (320) ! 2. Neutrino Refraction (321) ! 3. Kinetic Terms (322)
! 4. Recovering Stodolsky’s Formula (324) ! 5. Weak-Damping Limit (325)
! 6. Small Mixing Angle (328) ! 7. Flavor Conversion by Neutral Cur-
rents? (328)
9.4 Charged-Current Interactions (329)
1. Hamiltonian (329) ! 2. Kinetic Terms (329) ! 3. Weak-Damping
Limit (330)
9.5 Flavor Conversion in a SN Core (332)
1. Rate Equation (332) ! 2. Neutrino Interaction Rates (334) ! 3. Time
Scale for Flavor Conversion (335)
9.6 Sterile Neutrinos and SN 1987A (338)
10. Solar Neutrinos

10.2 Calculated Neutrino Spectrum (347)
1. Individual Sources (347) ! 2. Standard Solar Models (351) ! 3. Uncer-
tainties of Standard Neutrino Predictions (353)
x Table of Contents
10.3 Observations (357)

1. Absorption Reactions for Radiochemical Experiments (357) ! 2. Chlo-
rine Detector (Homestake) (360) ! 3. Gallium Detectors (SAGE and
GALLEX) (362) ! 4. Water Cherenkov Detector (Kamiokande) (365) !
5. Summary (372)
10.4 Time Variations (372)
1. Day-Night Effect (372) ! 2. Seasonal Variation (372) ! 3. Correlation
with Solar Cycle at Homestake (373) ! 4. Summary (376)
10.5 Neutrino Flux Deficits (377)
1. Boron Flux (377) ! 2. Beryllium Flux (378) ! 3. An Astrophysical
Solution? (379)
10.6 Neutrino Oscillations (380)
1. Which Data to Use? (380) ! 2. Vacuum Oscillations (381) ! 3. Resonant
Oscillations (MSW Effect) (384)
10.7 Spin and Spin-Flavor Oscillations (387)
10.8 Neutrino Decay (389)
10.9 Future Experiments (390)
1. Superkamiokande (390) ! 2. Sudbury Neutrino Observatory (SNO) (392)
! 3. BOREXINO (393) ! 4. Homestake Iodine Detector (394) ! 5. Sum-
mary (394)
11. Supernova Neutrinos

11.1 Stellar Collapse and Supernova Explosions (395)
1. Stellar Collapse (395) ! 2. Deleptonization and Cooling (399) ! 3. Su-
pernova Explosions (401) ! 4. Nucleosynthesis (405)
11.2 Predicted Neutrino Signal (407)
1. Overall Features (407) ! 2. Energies and Spectra (408) ! 3. Time
Evolution of the Neutrino Signal (411)
11.3 SN 1987A Neutrino Observations (414)
1. Supernova 1987A (414) ! 2. Neutrino Observations (415) ! 3. Analysis of
the Pulse (423) ! 4. Neutrino Mass and Pulse Duration (426) ! 5. Anomalies
in the Signal? (427)
11.4 Neutrino Oscillations (430)
1. Overview (430) ! 2. Prompt νe Burst (432) ! 3. Cooling-Phase ν e ’s
(434) ! 4. Shock Revival (436) ! 5. R-Process Nucleosynthesis (437) !
6. A Caveat (442)
11.5 Neutrino Propulsion of Neutron Stars (443)
11.6 Future Supernovae (445)
Table of Contents xi
12. Radiative Particle Decays from Distant

Sources
12.1 Preliminaries (449)
12.2 Laboratory Experiments (451)
1. Spectrum of Decay Photons (451) ! 2. Electron Neutrinos from Reac-
tors (453) ! 3. Heavy Neutrinos from Reactors (455) ! 4. Neutrinos from
a Beam Stop (456)
12.3 Particles from the Sun (458)
1. Electron Neutrinos (458) ! 2. Heavy Neutrino Admixtures (461) ! 3. New
Particles (462)
12.4 Supernova 1987A (462)
1. Decay Photons from Low-Mass Neutrinos (462) ! 2. SMM Observa-
tions (465) ! 3. Radiative Decay Limit: Low-Mass Neutrinos (467) !
4. Decay Photons from High-Mass Neutrinos (469) ! 5. Radiative Decay
Limits: High-Mass Neutrinos (472) ! 6. Summary of ν → ν ! γ Limits (477)
! 7. Limit on ντ → νe e+ e− (480) ! 8. Heavy, Sterile Neutrinos (482) !
9. Axions (483) ! 10. Supernova Energetics (483)
12.5 Galactic Supernovae and ντ → νe e+ e− (484)
1. Bounds on the Positron Flux (484) ! 2. Can the Tau Neutrino Be
Heavy? (485)
12.6 Neutrinos from All Stars (486)
12.7 Cosmological Bounds (488)
1. Neutrinos (488) ! 2. Axions (491)
13. What Have We Learned from SN 1987A?

13.2 Basic Characteristics of the Neutrino Burst (494)
1. Fluence (494) ! 2. Energy Distribution (495) ! 3. Prompt νe Burst (496)
! 4. Nonobservation of a γ-Ray Burst (497)
13.3 Dispersion Effects (497)
1. Photons vs. Antineutrinos (497) ! 2. Neutrinos vs. Antineutrinos (498)
! 3. Intrinsic Dispersion of the ν e -Pulse (499)
13.4 Duration of Neutrino Emission (501)
1. A General Argument (501) ! 2. Analytic Criterion in the Free-Streaming
Limit (504) ! 3. Trapping Limit (506)
13.5 Axions (508)
1. Numerical Studies (508) ! 2. Impact of Multiple-Scattering Effects (511)
13.6 How Many Neutrino Flavors? (513)
xii Table of Contents
13.7 Neutrino Opacity (514)

13.8 Right-Handed Neutrinos (516)
1. Dirac Mass (516) ! 2. Right-Handed Currents (519) ! 3. Magnetic
Dipole Moments (521) ! 4. Millicharges (522) ! 5. Charge Radius (523)
14. Axions
14.1 The Strong CP-Problem (524)
14.2 The Peccei-Quinn Mechanism (521)
1. Generic Features (526) ! 2. Axions as Nambu-Goldstone Bosons (528)
! 3. Pseudoscalar vs. Derivative Interaction (531) ! 4. The Onslaught of
Quantum Gravity (532)
14.3 Fine Points of Axion Properties (534)
1. The Most Common Axion Models (534) ! 2. Axion Mass and Coupling
to Photons (535) ! 3. Model-Dependent Axion-Fermion Coupling (536)
14.4 Astrophysical Axion Bounds (538)
14.5 Cosmological Limits (540)
15. Miscellaneous Exotica

15.1 Constancy of Fermi’s Constant (545)
15.2 Constancy of Newton’s Constant (546)
1. Present-Day Constraints from Celestial Mechanics (546) ! 2. Big-Bang
Nucleosynthesis (547) ! 3. Properties of the Sun (549) ! 4. White
Dwarfs (551) ! 5. Globular Clusters (551)
15.3 Test of the Equivalence Principle (554)
15.4 Photon Mass and Charge (555)
15.5 Free Quarks (556)
15.6 Supersymmetric Particles (557)
15.7 Majorons (558)
1. Particle-Physics and Cosmological Motivations (558) ! 2. Majorons and
Stars (562)
15.8 Millicharged Particles (564)
16. Neutrinos: The Bottom Line

16.1 Standard Neutrinos (568)
16.2 Minimally Extended Standard Model (570)
1. Cosmological Mass Limit for All Flavors (570) ! 2. Oscillations of Solar
Neutrinos (571) ! 3. Oscillation of Supernova Neutrinos (572) ! 4. Elec-
tromagnetic Properties (573)
Table of Contents xiii
16.3 New Interactions (574)

1. Majorana Masses (574) ! 2. “Heavy” Neutrinos and Fast Decays (574)
! 3. Electromagnetic Properties (576) ! 4. Summary (579)
Appendices
A. Units and Dimensions (580)
B. Neutrino Coupling Constants (583)
C. Numerical Neutrino Energy-Loss Rates (585)
1. Plasma Process (585) ! 2. Photoneutrino and Pair-Annihilation Pro-
cess (586) ! 3. Bremsstrahlung (588) ! 4. Total Emission Rate (589)
D. Characteristics of Stellar Plasmas (591)

D.1 Normal Matter (591)
1. Temperatures and Densities (591) ! 2. Relativistic Conditions for Elec-
trons (593) ! 3. Electron Degeneracy (593) ! 4. Plasma Frequency (596) !
5. Screening Scale (596) ! 6. Summary (598)
D.2 Nuclear Matter (600)
1. The Ideal p n e νe Gas (600) ! 2. Kinetic and Chemical Equilibrium (600)
! 3. Cold Nuclear Matter (601) ! 4. Hot Nuclear Matter (602)
References (606)
Acronyms (642)
Symbols (644)
Subject Index (649)
Preface
Ever since Newton proposed that the moon on its orbit follows the same
laws of motion as an apple falling from a tree, the heavens have been
a favorite laboratory to test the fundamental laws of physics, notably
classical mechanics and Newton’s and Einstein’s theories of gravity.
This tradition carries on—the 1993 physics Nobel prize was awarded
to R. A. Hulse and J. H. Taylor for their 1974 discovery of the binary
pulsar PSR 1913+16 whose measured orbital decay they later used to
identify gravitational wave emission. However, the scope of physical
laws necessary to understand the phenomena observed in the super-
lunar sphere has expanded far beyond these traditional fields. Today,
astrophysics has become a vast playing ground for applications of the
laws of microscopic physics, in particular the properties of elementary
particles and their interactions.
This book is about how stars can be used as laboratories to probe
fundamental interactions. Apart from a few arguments relating to grav-
itational physics and the nature of space and time (Is Newton’s constant
constant? Do all relativistic particles move with the same limiting ve-
locity? Are there novel long-range interactions?), most of the discussion
focusses on the properties and nongravitational interactions of elemen-
tary particles.
There are three predominant methods for the use of stars as particle-
physics laboratories. First, stars are natural sources for photons and
neutrinos which can be detected on Earth. Neutrinos are now routinely
measured from the Sun, and have been measured once from a collapsing
star (SN 1987A). Because these particles literally travel over astronom-
ical distances before reaching the detector one can study modifications
of the measured signal which can be attributed to propagation and dis-
persion effects, including neutrino flavor oscillations or axion-photon
oscillations in intervening magnetic fields. It is well known that the
discrepancy between the calculated and measured solar neutrino spec-
tra is the most robust, yet preliminary current indication for neutrino
oscillations and thus for nonvanishing neutrino masses.
xiv
Preface xv
Second, particles from distant sources may decay, and there may
be photons or even measurable neutrinos among the decay products.
The absence of solar x- and γ-rays yields a limit on neutrino radia-
tive decays which is as “safe” as a laboratory limit, yet nine orders
of magnitude more restrictive. An even more restrictive limit obtains
from the absence of γ-rays in conjunction with the SN 1987A neutri-
nos which allows one to conclude, for example, that even ντ must obey
the cosmological limit of mν < ∼ 30 eV unless one invents new invisible
decay channels.
Third, the emission of weakly interacting particles causes a direct
energy-loss channel from the interior of stars. For neutrinos, this effect
has been routinely included in stellar evolution calculations. If new low-
mass elementary particles were to exist such as axions or other Nambu-
Goldstone bosons, or if neutrinos had novel interactions with the stellar
medium such as one mediated by a putative neutrino magnetic dipole
moment, then stars might lose energy too fast. A comparison with the
observed stellar properties allows one to derive restrictive limits on the
operation of a new energy-loss or energy-transfer mechanism and thus
to constrain the proposed novel particle interactions.
While these and related arguments as well their application and re-
sults are extensively covered here, I have not written on several topics
that might be expected to be represented in a book on the connection
between particle physics and stars. Neutron stars have been speculated
to consist of quark matter so that in principle they are a laboratory to
study a quark-gluon plasma. As I am not familiar enough with the liter-
ature on this interesting topic I refer the reader to the review by Alcock
and Olinto (1988) as well as to the more recent proceedings of two top-
ical conferences (Madsen and Haensel 1992; Vassiliadis et al. 1995).
I have also dodged some important issues in the three-way relation-
ship between cosmology, stars, and particle physics. If axions are not
the dark matter of the universe, it is likely filled with a “background
sea” of hypothetical weakly interacting massive particles (WIMPs) such
as the lightest supersymmetric particles. Moreover, there may be ex-
otic particles left over from the hot early universe such as magnetic
monopoles which are predicted to exist in the framework of typical
grand unified theories (GUTs). Some of the monopoles or WIMPs
would be captured and accumulate in the interior of stars. GUT
monopoles are predicted to catalyze nucleon decay (Rubakov-Callan-
effect), providing stars with a novel energy source. This possibility can
be constrained by analogous methods to those presented here which
limit anomalous energy losses. The resulting constraints on the pres-
xvi Preface
ence of GUT monopoles in the universe have been reviewed, for exam-
ple, in the cosmology book of Kolb and Turner (1990). Because nothing
of substance has changed, a new review did not seem warranted.
WIMPs trapped in stars would contribute to the heat transfer be-
cause their mean free path can be so large that they may be orbit-
ing almost freely in the star’s gravitational potential well, with only
occasional collisions with the background medium. Originally it was
thought that this effect could reduce the central solar temperature
enough to solve the solar neutrino problem, and to better an alleged
discrepancy between observed and predicted solar p-mode frequencies.
With the new solar neutrino data it has become clear, however, that a
reduction of the central temperature alone cannot solve the problem.
Worse, solving the “old solar neutrino problem” by the WIMP mecha-
nism now seems to cause a discrepancy with the observed solar p-mode
frequencies (Christensen-Dalsgaard 1992). In addition, a significant ef-
fect requires relatively large scattering cross sections and thus rather
contrived particle-physics models. Very restrictive direct laboratory
constraints exist for the presence of these “cosmions” in the galaxy.
Given this status I was not motivated to review the topic in detail.
The annihilation of dark-matter WIMPs captured in the Sun or
Earth produces high-energy neutrinos which are measurable in terres-
trial detectors such as Kamiokande or the future Superkamiokande,
NESTOR, DUMAND, and AMANDA Cherenkov detectors. This in-
direct approach to search for dark matter may well turn into a serious
competitor for the new generation of direct laboratory search experi-
ments that are currently being mounted. This material is extensively
covered in a forthcoming review Supersymmetric Dark Matter by Jung-
man, Kamionkowski, and Griest (1995); there is no need for me to
duplicate the effort of these experts.
The topics covered in my book revolve around the impact of low-
mass or massless particles on stars or the direct detection of this ra-
diation. The highest energies encountered are a few 100 MeV (in the
interior of a SN core) which is extremely small on the high-energy scales
of typical particle accelerator experiments. Therefore, stars as labora-
tories for fundamental physics help to push the low-energy frontier of
particle physics and as such complement the efforts of nonaccelerator
particle experiments. Their main thrust is directed at the search for
nonstandard neutrino properties, but there are other fascinating top-
ics which include the measurement of parity-violating phenomena in
atoms, the search for neutron or electron electric dipole moments which
would violate CP, the search for neutron-antineutron oscillations, the
Preface xvii
search for proton decay, or the search for particle dark matter in the
galaxy by direct and indirect detection experiments (Rich, Lloyd Owen,
and Spiro 1987). It is fascinating that the IMB and Kamiokande water
Cherenkov detectors which had been built to search for proton decay
ended up seeing supernova (SN) neutrinos instead. The Fréjus detec-
tor, instead of seeing proton decay, has set important limits on the
oscillation of atmospheric neutrinos. Kamiokande has turned into a
major solar neutrino observatory and dark-matter search experiment.
The forthcoming Superkamiokande and SNO detectors will continue
and expand these missions, may detect a future galactic SN, and may
still find proton decay.
"
All currently known phenomena of elementary particle physics are

either perfectly well accounted for by its standard model, or are not
explained at all. The former category relates to the electroweak and
strong gauge interactions which have been spectacularly successful at
describing microscopic processes up to the energies currently available
at accelerators. On the other side are, for example, the mass spectrum
of the fundamental fermions (quarks and leptons), the source for CP
violation, or the relationship between the three families (why three?).
There must be “physics beyond the standard model!”
In the standard model, neutrinos have been assigned the most min-
imal properties compatible with experimental data: zero mass, zero
charge, zero dipole moments, zero decay rate, zero almost everything.
Any deviation from this simple picture is a sensitive probe for physics
beyond the standard model—thus the enthusiasm to search for neu-
trino masses and mixings, notably in oscillation experiments, but also
for neutrino electromagnetic properties, decays, and other effects. In
astrophysics, even “minimal neutrinos” play a major role for the en-
ergy loss of stars as they can escape unscathed from the interior once
produced. Moreover, in spite of their weak interaction there are two
astrophysical sites where they actually reach thermal equilibrium: the
early universe up to about the nucleosynthesis epoch and in a SN core
for a few seconds after collapse. Neutrinos thus play a dominant role
in the cosmic and SN dynamical and thermal evolution—little wonder
that these environments are important neutrino laboratories.
Nonstandard neutrino properties such as small Majorana masses
or magnetic dipole moments would be low-energy manifestations of
novel physics at short distances. Another spectacular interloper of high-
energy physics in the low-energy world would be a Nambu-Goldstone
xviii Preface
boson of a new symmetry broken at some large energy scale. Such a

particle would be massless or nearly massless. The most widely dis-
cussed example is the “invisible axion” which has been postulated as
an explanation for the observed absence of a neutron electric dipole
moment which in QCD ought to be about as large as its magnetic one.
The axion also doubles as a particle candidate for the dark matter of
the universe; two beautiful experiments to search for galactic axions are
about to go on line in Livermore (California) and Kyoto (Japan). Be-
cause of their weak interactions, the role of axions and similar particles
in stellar evolution is closely related to that of neutrinos.
"
The complete or near masslessness of the particles studied here (neu-

trinos, photons, axions, etc.) opens up a rich phenomenology in its own
right. One intriguing issue is the production and propagation of these
objects in a hot and dense medium which modifies their dispersion rela-
tions in subtle but significant ways. One of the most important neutrino
production processes in stars is the “photon decay” γ → νν which is
enabled by the modified photon dispersion relation in a medium, and
by an effective neutrino-photon coupling mediated by the ambient elec-
trons. Another example is the process of resonant neutrino oscillations
(MSW effect) which is instrumental at explaining the measured solar
neutrino spectrum, and may imply vast modifications of SN physics,
notably the occurrence of r-process nucleosynthesis. The MSW effect
depends on a flavor-birefringent term of the neutrino dispersion rela-
tion in a medium. In a SN, neutrinos themselves make an important
contribution to the “background medium” so that their oscillations be-
come a nonlinear phenomenon. In the deep interior of a SN core where
neutrinos are trapped, a kinetic treatment of the interplay between os-
cillations and collisions requires a fascinating “nonabelian Boltzmann
collision equation.”
One may think that low-energy particle physics in a dense and hot
medium requires the tools of field theory at finite temperature and
density (FTD) which has taken a stormy development over the past
fifteen years. In practice, its contribution to particle astrophysics has
been minor. The derivation of dispersion relations in a medium can be
understood in kinetic theory from forward scattering on the medium
constituents—dispersion is a lowest-order phenomenon. The imaginary
part of a particle self-energy in the medium is physically related to its
emission and absorption rate which in FTD field theory is given by
Preface xix
cuts of higher-order graphs. These methods have been applied only

in a few cases of direct interest to particle astrophysics where the re-
sults seem to agree with those from kinetic theory in the limits which
have been relevant in practice. A systematic formulation of astrophys-
ically relevant particle dispersion, emission, and absorption processes
in the framework of FTD field theory is a project for another author.
Tanguy Altherr had begun to take up this challenge with his collabo-
rators in a series of papers. However, his premature death in a tragic
climbing accident on 14 July 1994 has put an abrupt end to this line
of research.
My presentation of dispersion, emission, and absorption processes
is based entirely on the old-fashioned tools of kinetic theory where, say,
an axion emission rate is given by an integral of a squared matrix ele-
ment over the thermally occupied phase space of the reaction partners.
Usually this approach is not problematic, but it does require some tin-
kering when it comes to the problem of electromagnetic screening or
other collective effects. I would not be surprised if subtle but important
collective effects had been overlooked in some cases.
A simple kinetic approach is not adequate in the hot nuclear medium
characteristic for a young SN core. However, in practice quantities like
the neutrino opacities and axion emissivities have been calculated as
if the medium constituents were freely propagating particles. At least
for the dominant axial-vector current interactions this approach is not
consistent as one needs to assume that the spin-fluctuation rate is small
compared with typical thermal energies while a naive calculation yields
a result much larger than T . Realistically, it probably saturates at
O(T ), independently of details of the assumed interaction potential.
This conjecture appears to be supported by our recent “calibration” of
SN opacities from the SN 1987A neutrino signal. However, a calculation
of either neutrino opacities or axion emissivities on the basis of first
principles is not available at the present time because FTD effects as
well as nuclear-physics complications dominate the problem.
"
It has been challenging to hammer the multifarious and intertwined

aspects of my topic into the linear shape required by the nature of
a book. Chapters 1−6 are mainly devoted to the stellar energy-loss
argument. It is introduced in Chapter 1 where its general aspects are
developed on the basis of the stellar structure equations. Chapter 2
establishes the observational limits on those stellar evolutionary time
xx Preface
scales that have been used for the purposes of particle astrophysics.
For each case the salient applications are summarized. Chapters 3−6
deal with the interaction of “radiation” (neutrinos, axions, other low-
mass bosons) with the main constituents of stellar plasmas (photons,
electrons, nucleons). In these chapters all information is pulled together
that pertains to the given interaction channel, even if it is not directly
related to the energy-loss argument. For example, in Chapter 5 the
limits on the electromagnetic coupling of pseudoscalars with photons
are summarized; the stellar energy-loss ones are the most restrictive
which justifies this arrangement.
Chapter 6 develops the topic of particle dispersion in media. The
medium-induced photon dispersion relation allows for the plasma pro-
cess γ → νν which yields the best limit on neutrino dipole moments
by virtue of the energy-loss argument applied to globular-cluster stars.
The neutrino dispersion relation is needed for the following discussion
of neutrino oscillations, establishing a link between the energy-loss ar-
gument and the dispersion arguments of the following chapters.
Dispersion and propagation effects are particularly important for
massive neutrinos with flavor mixing. To this end the phenomenology
of massive, mixed neutrinos is introduced in Chapter 7. Vacuum and
matter-induced flavor oscillations as well as magnetically induced spin
oscillations are taken up in Chapter 8. If neutrinos are in thermal equi-
librium as in a young SN core or the early universe, neutrino oscillations
require a different theoretical treatment (Chapter 9).
Chapters 10−13 are devoted to astrophysical sources where neu-
trinos have been measured, i.e. the Sun and supernovae (for the lat-
ter only the SN 1987A signal exists). Neutrino oscillations, notably
of the matter-induced variety, play a prominent role in Chapter 10
(solar neutrinos) and Chapter 11 (SN neutrinos). Radiative particle
decays, especially of neutrinos, are studied in Chapter 12 where the
Sun and SN 1987A figure prominently as sources. In Chapter 13 the
particle-physics results from SN 1987A are summarized, including the
ones related to the energy-loss and other arguments.
Chapters 14−16 give particle-specific summaries. While axions play
a big role throughout this text, only the structure of the interaction
Hamiltonian with photons, electrons, and nucleons is needed. Thus
everything said about axions applies to any pseudoscalar low-mass bo-
son for which they serve as a generic example. In Chapter 14 these
results are interpreted in terms of axion-specific models which relate
their properties to those of the neutral pion and thus establish a nearly
unique relationship between their mass and interaction strength. Often-
Preface xxi
quoted “astrophysical bounds on the axion mass” are really transformed

bounds on their interaction strength. This chapter also summarizes
recent developments of the putative cosmological role of axions. Chap-
ter 15 takes up a variety of hypotheses which can be tested by the
methods developed in this book. Finally, Chapter 16 is an attempt at
a bottom line of what we have learned about neutrino properties in the
astrophysical laboratory.
"
Parts of my presentation are devoted to theoretical and calcula-

tional fine points of particle dispersion and emission effects in media. I
find some of these issues quite intriguing in their own right. Still, the
main goal has been to provide an up-to-date overview of what we know
about elementary particles and their interactions on the basis of estab-
lished stellar properties and on the basis of measured or experimentally
constrained stellar particle fluxes.
All those whose lives are spent searching for truth are well
aware that the glimpses they catch of it are necessarily fleet-
ing, glittering for an instant only to make way for new and
still more dazzling insights. The scholar’s work, in marked
contrast to that of the artist, is inevitably provisional. He
knows this and rejoices in it, for the rapid obsolescence of
his books is the very proof of the progress of scholarship.
(Henri Pirenne, 1862−1935)
In spite of this bittersweet insight, and in spite of some inevitable

errors of omission and commission, I hope that my book will be of some
use to researchers, scholars, and students interested in the connection
between fundamental physics and stars.
Munich, May 1995.

Acknowledgments
While writing this book I have benefitted from the help and encourage-
ment of many friends and colleagues. In particular, I need to mention
Hans-Thomas Janka, Lothar Luh, Günter Sigl, Pierre Sikivie, Thomas
Strobel, and Achim Weiss who read various parts of the manuscript,
and the referees Josh Frieman and J. Craig Wheeler who read all or
most of it. Their comments helped in no small measure to improve the
manuscript and to eliminate some errors. Hans-Thomas, in particular,
has spared no effort at educating me on the latest developments in the
area of supernova physics. Several chapters were written during a visit
at the Center for Particle Astrophysics in Berkeley—I gratefully ac-
knowledge the fine hospitality of Bernard Sadoulet and his staff. Stay-
ing for that period as a guest in Edward Janelli’s house made a huge
difference. During the entire writing process Greg Castillo provided
encouragement and a large supply of Cuban and Puerto Rican music
CDs which have helped to keep my spirits up. At the University of
Chicago Press, I am indebted to Vicki Jennings, Penelope Kaiserlian,
Stacia Kozlowski, and Eleanore Law for their expert handling of all
editorial and practical matters that had to be taken care of to trans-
form my manuscript into a book. David Schramm as the series editor
originally solicited this opus (“just expand your Physics Report a little
bit”). It hasn’t quite worked that way, but now that I’m finished I am
grateful that David persuaded me to take up this project.
xxii
Chapter 1
The Energy-Loss Argument
Weakly interacting, low-mass particles such as neutrinos or axions con-

tribute to the energy loss or energy transfer in stars. The impact of
an anomalous energy-loss mechanism is discussed qualitatively and in
terms of homology relations between standard and perturbed stellar
models. The example of massive pseudoscalar particles is used to il-
lustrate the impact of a new energy-loss and a new radiative-transfer
mechanism on the Sun.
1.1 Introduction
More than half a century ago, Gamow and Schoenberg (1940, 1941)
ushered in the advent of particle astrophysics when they speculated that
neutrinos may play an important role in stellar evolution, particularly
in the collapse of evolved stars. Such a hypothesis was quite bold for
the time because neutrinos, which had been proposed by Pauli in 1930,
were not directly detected until 1954. That their existence was far from
being an established belief when Gamow and Schoenberg wrote their
papers is illustrated by Bethe’s (1939) complete silence about them in
his seminal paper on the solar nuclear fusion chains.
Even after the existence of neutrinos had been established they
seemed to interact only by β reactions of the sort e− + (A, Z) →
(A, Z−1) + νe or (A, Z−1) → (A, Z) + e− + ν e , the so-called URCA
reactions which Gamow and Schoenberg had in mind, or by fusion pro-
cesses like pp → de+ νe . The URCA reactions and related processes
become important only at very high temperatures or densities because
of their energy threshold. While the Sun emits two neutrinos for every
helium nucleus fused from hydrogen, the energy loss in neutrinos is only
a few percent of the total luminosity and thus plays a minor role.
1
2 Chapter 1
Still, neutrinos can be important in normal stars. This became

clear when in 1958 Feynman and Gell-Mann as well as Sudarshan and
Marshak proposed the universal V −A interaction law which implied
a direct neutrino-electron interaction with the strength of the Fermi
constant. Pontecorvo (1959) realized almost immediately that this in-
teraction would allow for the bremsstrahlung radiation of neutrino pairs
by electrons, and that the absence of a threshold renders this process
an important energy loss mechanism for stars. Of course, their typical
energies will correspond to the temperature of the plasma (about 1 keV
in the Sun) while neutrinos from nuclear reactions have MeV energies.
On the basis of the bremsstrahlung process, Gandel’man and Pinaev
(1959) calculated the approximate conditions for which neutrino losses
would “outshine” the photon luminosity of stars. For the Sun, thermal
neutrino emission is found to be irrelevant.
Subsequently, the neutrino emissivity was calculated by many au-
thors. It was quickly realized that the dominant emission processes from
a normal stellar plasma are the photoneutrino process γe− → e− νν, the
bremsstrahlung process e− +(A, Z) → (A, Z)+e− +νν, and the plasma
process (“photon decay”) γ → νν which is possible because photons
have an effective mass in the medium. These and related reactions will
be discussed at length in Chapters 3−6.
In the late 1960s Stothers and his collaborators1 established that
the observed paucity of red supergiants is best explained by the fast
rate with which these carbon-burning stars spend their nuclear fuel
due to neutrino emission. Therefore, the existence and approximate
magnitude of the direct electron-neutrino coupling was at least ten-
tatively established by astrophysical methods several years before its
experimental measurement in 1976.
While standard neutrino physics today is an integral part of stellar
evolution and supernova theory, they could have novel couplings to the
plasma, for example by a magnetic dipole moment. Then they could
be emitted more efficiently than is possible with standard interactions.
Moreover, new concepts of particle physics have emerged that could be
equally important despite the relatively low energies available in stellar
interiors. In various extensions of the standard model, the spontaneous
breakdown of a symmetry of the Lagrangian of the fundamental inter-
actions by some large vacuum expectation value of a new field leads
to the prediction of massless or nearly massless particles, the Nambu-
Goldstone bosons of the broken symmetry. The most widely discussed
1
For a summary see Stothers (1970, 1972).
The Energy-Loss Argument 3
example is the axion (Chapter 14) which arises as the Nambu-Goldstone

boson of the Peccei-Quinn symmetry which explains the puzzling ab-
sence of a neutron electric dipole moment, i.e. it explains CP conserva-
tion in strong interactions. The production of axions in stars, like that
of neutrinos, is not impeded by threshold effects.
Clearly, the emission of novel weakly interacting particles or the
emission of neutrinos with novel properties would have a strong impact
on the evolution and properties of stars. Sato and Sato (1975) were
the first to use this “energy-loss argument” to derive bounds on the
coupling strength of a putative low-mass Higgs particle. Following this
lead, the argument has been applied to a great variety of particle-
physics hypotheses, and to a great variety of stars.
In the remainder of this chapter the impact of a novel energy-loss
mechanism on stars will be discussed in simple terms. The main mes-
sage will be that the emission of weakly interacting particles usually
leads to a modification of evolutionary time scales. By losing energy in
a new channel, the star effectively burns or cools faster and thus shines
for a shorter time. In the case of low-mass red giants, however, particle
emission leads to a delay of helium ignition and thus to an extension
of the red-giant phase. Either way, what needs to be observationally
established is the duration of those phases of stellar evolution which
are most sensitive to a novel energy-loss mechanism. In Chapter 2 such
evolutionary phases will be identified, and the observational evidence
for their duration will be discussed. In the end one will be able to
state simple criteria for the allowed rate of energy loss from plasmas at
certain temperatures and densities.
One may be tempted to think that one should consider the hottest
and densest possible stars because no doubt the emission of weakly in-
teracting particles is most efficient there. However, this emission com-
petes with standard neutrinos whose production is also more efficient
in hotter and denser objects. Because neutrinos are thermally emitted
in pairs their emission rates involve favorable phase-space factors which
lead to a temperature dependence which is steeper than that for the
emission of, say, axions. Thus, for a given axion coupling strength the
relative importance of axion emission is greater for lower temperatures.
Of course, the temperatures must not be so low that neither neutrino
nor axion emission is important at all relative to the photon luminos-
ity. Consequently, the best objects to use are those where neutrinos just
begin to have an observational impact on stellar observables. Exam-
ples are low-mass red giants, horizontal-branch stars, white dwarfs, and
old neutron stars. They all have masses of around 1 M" (solar mass).
4 Chapter 1
For low-mass stars plenty of detailed observational data exist, allowing

one to establish significant limits on possible deviations from standard
evolutionary time scales.
From the astrophysical perspective, the energy-loss argument re-
quires establishing evolutionary time scales and other observables that
are sensitive to a novel “energy sink.” This problem will be taken up
in Chapter 2. From the particle-physics perspective, one needs to cal-
culate the energy-loss rate of a plasma at a given temperature, density,
and chemical composition into a proposed channel. For this purpose
it is not necessary to know any of the underlying physics that leads to
the hypothesis of nonstandard neutrino properties or the prediction of
new particles such as axions. All one needs is the structure of the inter-
action Hamiltonian with the constituents of stellar plasmas (electrons,
nucleons, nuclei, and photons). For example, it is enough to know that
an axion is a low-mass particle with a pseudoscalar interaction with
electrons to calculate the axion emission rate. As such “axion” stands
for any particle with a similar interaction structure. The motivation
for “real axions” and their detailed model-dependent properties are not
discussed until Chapter 14. In Chapters 3−5 and partly in Chapter 6
various interaction structures will be explored and constraints on the
overall coupling strengths will be derived or summarized. Also, they
will be put into the context of evidence from sources other than the
stellar energy-loss argument.
Putative novel particles almost inevitably must be very weakly in-
teracting or else they would have been seen in laboratory experiments.
Still, one is sometimes motivated to speculate about particles which
may be so strongly interacting that they cannot freely escape from
stars; their mean free path may be shorter than the stellar radius. It
is often incorrectly stated that such particles would be astrophysically
allowed. Nothing could be further from the truth. Particles which
are “trapped,” i.e. which cannot freely stream out once produced, con-
tribute to the radiative transfer of energy in competition with photons.
Radiative energy transfer occurs because particles produced in a hot
region get absorbed in a neighboring somewhat cooler region. This
mechanism is more efficient if the mean free path is larger because then
ever more distant regions with larger temperature differences are ther-
mally coupled. Therefore, particles which interact more weakly than
photons would dominate the radiative energy transfer and thus have a
tremendous impact on the structure of stars.
Very roughly speaking, then, novel particles must be either more
strongly interacting than photons, or more weakly interacting than neu-
trinos, to be harmless in stars. Of course, depending on whether the

particles are bosons or fermions, and depending on the details of their
interaction structure, this statement must be refined. Still, the impact
of novel particles on stellar structure and evolution is maximized when
their mean free path is of order the geometric dimension of the system.
The energy-transfer argument is equally powerful as the energy-loss ar-
gument. The only reason why it has not been elaborated much in the
literature is because there is usually little motivation for considering
“strongly” interacting novel particles.
When it comes to the evolution of a supernova (SN) core after col-
lapse even neutrinos are trapped. Such a newborn neutron star is so
hot (T of order 30 MeV) and dense (ρ exceeding nuclear density of
3×1014 g cm−3 ) that neutrinos take several seconds to diffuse to the
surface. Particles like axions can then compete in spite of the extreme
conditions because they freely stream out if their coupling is weak
enough. The energy-loss argument can be applied because the neu-
trino cooling time scale has been established by the SN 1987A neutrino
observations. While the energy-loss argument in this case is fundamen-
tally no different from, say, white-dwarf cooling, the detailed reasoning
is closely intertwined with the issue of neutrino physics in supernovae,
and with the details of the SN 1987A neutrino observations. There-
fore, it is taken up only in Chapter 13. The groundwork concerning
the interactions of neutrinos and axions with nucleons, however, is laid
in Chapter 4 within the series of chapters devoted to various modes of
particle interactions with the constituents of stellar plasmas.
1.2 Equations of Stellar Structure

1.2.1 Hydrostatic Equilibrium
To understand the impact of a novel energy-loss mechanism on the
evolution of stars one must understand the basic physical principles that
govern stellar structure. While a number of simplifying assumptions
need to be made, the theory of stellar structure and evolution has been
extremely successful at modelling stars with a vast range of properties.
For a more detailed account than is possible here the reader is referred
to the textbook literature, e.g. Kippenhahn and Weigert (1990).
One usually assumes spherical symmetry and thus excludes the ef-
fects of rotation, magnetic fields, tidal effects from a binary companion,
and large-scale convective currents. While any of those effects can be
important in special cases, none of them appears to have a noticeable
6 Chapter 1
impact on the overall picture of stellar structure and evolution, with

the possible exception of supernova physics where large-scale convective
overturns may be crucial for the explosion mechanism (Chapter 11).
Second, one usually assumes hydrostatic equilibrium, i.e. one ignores
the macroscopic kinetic energy of the stellar medium. This approxima-
tion is inadequate for a study of stellar pulsation where the inertia of
the material is obviously important, and also inadequate for “hydrody-
namic events” such as a supernova explosion (Sect. 2.1.8), and perhaps
the helium flash (Sect. 2.1.3). For most purposes, however, the changes
of the stellar structure are so slow that neglecting the kinetic energy is
an excellent approximation.
Therefore, as a first equation one uses the condition of hydrostatic
equilibrium that a spherical shell of the stellar material is held in place
by the opposing forces of gravity and pressure,
dp GN M r ρ
=− . (1.1)
dr r2
Here, GN is Newton’s constant, p and ρ are the pressure and mass den-
! r # #2
sity at the radial position r, and Mr = 4π 0 dr ρ r is the integrated
mass up to the radius r.
With apologies to astrophysicists I will usually employ natural units
where h̄ = c = kB = 1. In Appendix A conversion factors are given
between various units of mass, energy, inverse length and time, tem-
perature, and so forth. Newton’s constant is then GN = m−2 Pl with the
19 −5
Planck mass mPl = 1.221×10 GeV = 2.177×10 g. Stellar masses are
always denoted with the letter M to avoid confusion with an absolute
bolometric brightness which is traditionally denoted by M .
In general, the pressure is given in terms of the density, temperature,
and chemical composition by virtue of an equation of state. For a
classical monatomic gas p = 23 u with u the density of internal energy.
One may multiply Eq. (1.1) on both sides with 4πr3 and integrate
from the center (r = 0) to the surface (r = R). The! r.h.s. gives the
total gravitational energy while the l.h.s. yields −12π 0R dr p r2 after a
partial integration with the boundary condition p = 0 at the surface.
With p = 23 u this is −2U with U the total internal energy of the star.
Because for a monatomic gas U is the sum of the kinetic energies of the
atoms one finds that on average for every atom
$Ekin % = − 12 $Egrav %. (1.2)
This is the virial theorem which is the most important tool to under-
stand the behavior of self-gravitating systems.
As a simple example for the beauty and power of the virial theo-
rem one may estimate the solar central temperature from its mass and
radius. The material is dominated by protons which have a gravita-
tional potential energy of order −GN M" mp /R" = −2.14 keV where
M" = 1.99×1033 g is the solar mass, R" = 6.96×1010 cm the solar
radius, and mp the proton mass. The average kinetic energy of a
proton is equal to 32 T (remember, kB has been set equal to unity),
yielding an approximate value for the solar internal temperature of
T = 13 2.14 keV = 0.8×107 K. This is to be compared with 1.56×107 K
found for the central temperature of a typical solar model. This exam-
ple illustrates that the basic properties of stars can be understood from
simple physical principles.
1.2.2 Generic Cases of Stellar Structure
a) Normal Stars
There are two main sources of pressure relevant in stars, thermal pres-
sure and degeneracy pressure. The third possibility, radiation pressure,
never dominates except perhaps in the most massive stars. The pres-
sure provided by a species of particles is proportional to their density,
to their momentum which is reflected on an imagined piston and thus
exerts a force, and to their velocity which tells us the number of hits on
the piston per unit time. In a nondegenerate nonrelativistic medium
a typical particle velocity and momentum is proportional to T 1/2 so
that p ∝ (ρ/µ) T with ρ the mass density and µ the mean molecu-
lar weight of the medium constituents. For nonrelativistic degenerate
electrons the density is ne = p3F /3π 2 (Fermi momentum pF ), a typical
momentum is pF , and the velocity is pF /me , yielding a pressure which
is proportional to p5F or to n5/3
e and thus to ρ5/3 .
The two main pressure sources determine two generic forms of be-
havior of overall stellar models, namely normal stars such as our Sun
which is dominated by thermal pressure, and degenerate stars such as
white dwarfs which are dominated by degeneracy pressure. These two
cases follow a very different logic.
A normal star is understood most easily if one imagines how it ini-
tially forms from a dispersed but gravitationally bound gas cloud. It
continuously loses energy because photons are produced in collisions
between, say, electrons and protons. The radiation carries away energy
which must go at the expense of the total energy of the system. If it is
roughly in an equilibrium configuration, the virial theorem Eq. (1.2) in-
8 Chapter 1
forms us that a decrease of $Ekin +Egrav % causes the gravitational energy

to become more negative, corresponding to a more tightly bound and
thus more compact system. At the same time the average kinetic energy
goes up which corresponds to an increased temperature if the system
can be considered to be locally in thermal equilibrium. Therefore, as
the system loses energy it contracts and heats up. Self-gravitating sys-
tems have a negative specific heat!
As a protostar contracts it becomes more opaque and soon the en-
ergy loss is limited by the speed of energy transfer from the inner parts
to the surface, i.e. essentially by the photon diffusion speed. Before an
internal energy source for stars was known it was thought that gravita-
tional energy provided for their luminosity. Stellar lifetimes seemed to
be given by the “Kelvin-Helmholtz time scale” for thermal relaxation
which is fixed by the speed of energy transfer. The total reservoir of
gravitational energy of the Sun is estimated by GN M2" /2R" and its
luminosity is L" = 3.85×1026 W = 3.85×1033 erg/s so that the thermal
relaxation scale is about τKH ≈ 12 GN M2" R" −1 −1
L" = 1.6×107 yr. Pressed
by thermodynamic theory and the authority of Lord Kelvin, geologists
tried to adjust the age of the Earth to this short time scale against
sound evidence to the contrary. At the beginning of our century the
discovery of radioactivity and thus of nuclear processes revealed that
stars had another source of energy and consequently could live much
longer than indicated by τKH .
A further contraction is thus intercepted by the onset of hydrogen
burning which commences when the temperature is high enough for
protons to penetrate each other’s electrostatic repulsive potential. The
hydrogen-burning reactions are more fully discussed in Chapter 10; the
bottom line is that four protons and two electrons combine to form
a helium nucleus (α particle), releasing 26.73 MeV of energy. A few
percent are immediately lost in the form of two neutrinos which must
emerge to balance the electron lepton number, but most of the en-
ergy is available as heat. Because the nuclear reaction rates have a
steep temperature dependence, a further contraction and heating of
the star leads to much more nuclear energy generation, which quickly
increases the average Ekin of the nuclei and thus leads to an expansion
and cooling by the same virial-theorem logic that led to contraction and
heating when energy was lost. Thus a stable configuration of “thermal
equilibrium” is reached where the energy lost is exactly balanced by
that produced from nuclear reactions. Stars as fusion reactors are per-
fectly regulated by the “negative specific heat” of a self-gravitating
system!
The most salient feature of a normal stellar configuration is the in-

terplay of its negative specific heat and nuclear energy generation. Con-
versely, if the pressure were dominated by electron degeneracy it would
be nearly independent of the temperature. Then this self-regulation
would not function because heating would not lead to expansion. Thus,
stable nuclear burning and the dominance of thermal pressure go insep-
arably hand in hand. Another salient feature of such a configuration is
the inevitability of its final demise because it lives on a finite supply of
nuclear fuel.
b) Degenerate Stars
Everything is different for a configuration dominated by degeneracy
pressure. Above all, it has a positive heat capacity so that a loss
of energy no longer implies contraction and heating. The star actu-
ally cools. This is what happens to a brown dwarf which is a star
so small (M < 0.08 M" ) that it did not reach the critical condi-
tions to ignite hydrogen: it becomes a degenerate gas ball which slowly
“browns out.”
The relationship between radius and mass is inverted. A normal
star is geometrically larger if it has a larger mass; very crudely R ∝
M. When mass is added to a degenerate configuration it becomes
geometrically smaller as the reduced size squeezes the electron Fermi
sea into higher momentum states, providing for increased pressure to
balance the increased gravitational force. The l.h.s. of Eq. (1.1) can be
approximated as p/R where p ∝ ρ5/3 is some average pressure. Because
ρ ≈ M/R3 one finds p/R ∝ M5/3 /R6 while the r.h.s. of Eq. (1.1) is
proportional to Mρ/R2 and thus to M2 /R5 . Therefore, a degenerate
configuration is characterized by R ∝ M−1/3 .
Increasing the mass beyond a certain limit causes the radius to
shrink so much that the electrons become relativistic. Then they move
with a velocity fixed at c (or 1 in natural units), causing the pressure
to vary only as p4F or ρ4/3 . In this case adding mass no longer leads to a
sufficient pressure increase to balance for the extra weight. Beyond this
“Chandrasekhar limit,” which is about 1.4 M" for a chemical composi-
tion with Ye = 12 (number of electrons per baryon), no stable degenerate
configuration exists.
In summary, the salient features of a degenerate configuration are
the inverse mass-radius relationship R ∝ M−1/3 , the Chandrasekhar
limit, the absence of nuclear burning, and the positive specific heat
which allows the configuration to cool when it loses energy.
10 Chapter 1
c) Giant Stars
A real star can be a hybrid configuration with a degenerate core and
a nondegenerate envelope with nuclear burning at the bottom of the
envelope, i.e. the surface of the core. The core then follows the logic
of a degenerate configuration. The envelope follows the self-regulating
logic of a normal star except that it is no longer dominated by its self-
gravity but rather by the gravitational force exerted by the compact
core. Amazingly, the envelopes of such stars tend to expand to huge
dimensions. It does not seem possible to explain in a straightforward
way why stars become giants2 except that “the equations say so.” Low-
mass red giants will play a major role in Chapter 2—a further discussion
of their fascinating story is deferred until then.
1.2.3 Energy Conservation

Returning to the basic principles that govern stellar structure, energy
conservation yields another of the stellar structure equations. If the lo-
cal sources and sinks of energy balance against the energy flow through
the surface of a spherical mass shell one finds
dLr /dr = 4πr2 (ρ, (1.3)
where Lr is the net flux of energy through a spherical shell of radius r

while ( is the effective rate of local energy production (units erg g−1 s−1 ).
It is a sum
( = (nuc + (grav − (ν − (x , (1.4)
where (nuc is the rate by which nuclear energy is liberated, (ν is the

energy-loss rate by standard neutrino production, and (x is for novel
particles or nonstandard neutrinos with, say, large magnetic dipole mo-
ments. Further,
(grav = cp T (∇ad ṗ/p − Ṫ /T ) (1.5)
is the local energy gain when T and p change because of expansion or

contraction of the star. This term is the dominant heat source in a
red-giant core which contracts because of the mass gained by hydrogen
shell burning. Conversely, it is a sink which absorbs energy when helium
ignites in such a star: the would-be explosion is dissipated by expansion
2
For recent attempts see Faulkner and Swenson (1988), Eggleton and Cannon
(1991), and Renzini et al. (1992).
against the force of gravity. In Eq. (1.5) cp is the heat capacity at

constant pressure. Further, the quantity ∇ad ≡ (∂ ln T /∂ ln p)s , taken
at constant entropy density s, is the “adiabatic temperature gradient.”
It is not really a gradient. It is a thermodynamic quantity characteristic
of the medium at the local conditions of ρ, p, T , and the chemical
composition.
The calculation of (x for a number of hypotheses concerning the
existence of novel particles or novel properties of neutrinos will be a
major aspect of this book. Even for a given interaction law between
the particles and the medium constituents this is not always a straight-
forward exercise because the presence of the ambient medium can have
a significant impact on the microscopic reactions.
This is equally true for the nuclear energy generation rates. Nuclear
reactions are slow in stars because the low temperature allows only few
nuclei to penetrate each other’s Coulomb barriers. Therefore, screening
effects are important, and the nuclear cross sections need to be known at
energies so low that they cannot be measured directly in the laboratory.
Much of the debate concerning the solar neutrino problem revolves
around the proper extrapolation of certain nuclear cross sections to
solar thermal energies.
1.2.4 Energy Transfer

The transfer of energy is driven by the radial temperature gradient. In
the absence of convection heat is carried by photons and electrons mov-
ing between regions of different temperature, i.e. by radiative transfer
and by conduction. In this case the relationship between the energy
flux and the temperature gradient is
4πr2 d(aT 4 )
Lr = − , (1.6)
3κρ dr
where aT 4 is the energy stored in the radiation field (a = π 2 /15 in

natural units) and κ is the opacity (units cm2 /g). It is given by a sum
κ−1 = κ−1 −1 −1
γ + κc + κx , (1.7)
where κγ is the radiative opacity, κc the contribution from conduction

by electrons, and κx was included for a possible contribution from novel
particles. The quantity (κγ ρ)−1 = $λγ %R is the “Rosseland average”
of the photon mean free path—its precise definition will be given in
Sect. 1.3.4.
12 Chapter 1
One of the main difficulties at calculating the opacity is that heavier

elements, notably iron, are only partially ionized for typical conditions.
Resonant transitions of electrons between different bound states are
very important, an effect which causes stellar models to be rather sen-
sitive to the amount of “metals” (elements heavier than helium). The
construction of an opacity table is a major effort as it requires includ-
ing huge numbers of electronic energy levels. Widely used were the Los
Alamos and the Livermore Laboratory opacity tables.
Recently, the Livermore tables were systematically overhauled (Igle-
sias, Rogers, and Wilson 1990; Iglesias and Rogers 1991a,b), resulting
in the new OPAL tables which since have become the standard in stel-
lar evolution calculations. The main differences to the previous tables
are at moderate temperatures so that no substantial changes in the
deep interior of stellar structures have occurred. However, envelope
phenomena are affected, notably convection near the surface and stel-
lar pulsations. A number of previous discrepancies between theory and
observations in this area have now disappeared.
If the equation of state, the energy generation rate, and the opacity
are known one can construct a stellar model for an assumed compo-
sition profile by solving the stellar structure equations with suitable
boundary conditions. (It is not entirely trivial to define surface bound-
ary conditions because the star, strictly speaking, extends to infinity.
A crude approach is to take T = 0 and p = 0 at the photosphere.)
It may turn out, however, that in some locations this procedure yields
a temperature gradient which is so steep that the material becomes
unstable to convection—it “boils.”
An adequate treatment of convection is one of the main problems
of stellar evolution theory. A simplification occurs because convection
is extremely effective at transporting energy and so the temperature
gradient will adjust itself to a value very close to the “adiabatic gra-
dient” which marks the onset of the instability. At this almost fixed
temperature gradient a nearly arbitrary energy flux can be carried by
the medium. This approximation tends to be justified for regions in the
deep interior of stars while the “superadiabatic convection” found near
the surface requires a substantial refinement. One usually applies the
“mixing length theory” which contains one free parameter, the ratio be-
tween the convective mixing length and the pressure scale height. This
parameter is empirically fixed by adjusting the radius of a calculated
solar model to the observed value.
Main-sequence stars like our Sun with M < ∼ M" have a radiative
interior with a convective surface which penetrates deeper with decreas-
ing M; stars with M < ∼ 0.25 M" are fully convective. For M ∼ M"
>
the outer regions are radiative while the core is convective out to an ever
increasing mass fraction of the star with increasing M. A star with M
near 1 M" is very special in that it is radiative almost throughout; the
Sun is thought to have only a relatively minor convective surface layer.
Besides transporting energy, convection also moves matter and thus
affects the composition profile of a star. This is seen, for example, in
the upper panels of Fig. 2.4 where the hydrogen depletion of a solar
model (which is radiative) is a function of the local nuclear burning
rates while for the convective helium core of a horizontal-branch (HB)
star the helium depletion reaches to much larger radii than nuclear
burning. The long lifetimes of HB stars cannot be understood without
the convective supply of fuel to the nuclear furnace at the center. The
Sun, on the other hand, will complete its main-sequence evolution when
hydrogen is depleted at the center, corresponding to about a 10% global
depletion only.
The extent of convective regions can change during the course of
stellar evolution. They can leave behind composition discontinuities
which are a memory of a previous configuration. For example, on the
lower red-giant branch (RGB) the convective envelope reaches so deep
that it penetrates into the region of variable hydrogen content caused
by nuclear burning. Later, the convective envelope retreats from the ad-
vancing hydrogen-burning shell which encounters a discontinuity in the
hydrogen profile. This causes a brief “hesitation” on the RGB ascent
and thus a “bump” in the distribution of stars in the color-magnitude
diagram of globular clusters on the lower RGB. This bump has been
identified in several clusters (Figs. 2.18 and 2.19); its location is in good
agreement with theoretical expectations (Fusi Pecci et al. 1990).
1.2.5 Gravitational Settling

The composition profile of a star can also change by diffusion, and no-
tably by gravitational settling of the heavier elements. This effect was
ignored in most evolution calculations because the time scales are very
large. Still, the settling of helium will displace hydrogen from the cen-
ter of a hydrogen-burning star and thus accelerate the depletion and
main-sequence turnoff. The gravitational settling of metals will lead
to an opacity increase in the central regions. Helium settling reduces
the inferred globular cluster ages by 1 Gyr or more which is about a
10% effect (Proffitt and Michaud 1991; Chaboyer et al. 1992). Be-
cause the inferred globular cluster ages are larger than the expansion
14 Chapter 1
age of the universe, this “cosmic age problem” is slightly alleviated by

gravitational settling.
Helium and metal settling was recently included in solar models
(Bahcall and Pinsonneault 1992, 1995; Kovetz and Shaviv 1994; Proffitt
1994). It increased the predicted neutrino fluxes on the 10−30% level,
depending on the specific treatment of gravitational settling and on the
neutrino source reaction. These modifications are not huge, but surely
not entirely negligible. They go in the direction of aggravating the solar
neutrino problem.
1.3 Impact of Novel Particles

1.3.1 Energy Loss
One of the most interesting possibilities to use stars as particle-physics
laboratories is to study the backreaction of the novel energy-loss rate
(x implied by the existence of new low-mass particles such as axions, or
by nonstandard neutrino properties such as magnetic dipole moments.
The impact on degenerate stars such as white dwarfs is rather obvi-
ous: the new energy-loss rate accelerates the cooling. Therefore, the
observationally established cooling speed allows one to constrain this
process or to detect evidence for it.
The impact of a novel energy-loss mechanism on a nondegenerate
star like the Sun is less obvious. According to the virial theorem one
expects that the loss of energy leads to heating and contraction up to
a point where the temperature has risen enough that increased nuclear
burning provides for the extra energy loss. Because of the steep temper-
ature dependence of (nuc one expects that the overall stellar structure
changes very little in response to (x —the main impact is to accelerate
the consumption of nuclear fuel and thus the completion of hydrogen
burning.
This argument was cast into a quantitative form by Frieman, Di-
mopoulos, and Turner (1987). They asked how a given equilibrium
structure of a star would change in response to turning on a new energy-
loss rate (x . The main simplifying assumption is that the perturbed
configuration is obtained by a homology transformation, so that “the
distance between any two points is altered in the same way as the ra-
dius of the configuration.” Thus, if the new radius of the star is given
by R# = yR with a dimensionless scaling factor y, then every point
in the star is mapped to a new position r# = yr. The mass interior
to the new radius is identical with that interior to the old location,
M# (r# ) = M(r), and the chemical composition at r# is the same as

that at r. The density is transformed by ρ# (r# ) = y −3 ρ(r), and from
Eq. (1.1) one finds that the pressure scales as p# (r# ) = y −4 p(r). The
equation of state for a nondegenerate, low-mass star is approximately
given by the ideal-gas law where p ∝ ρT /µ, where µ is the average
molecular weight of the electrons and nuclei. Since µ# (r# ) = µ(r) by
assumption, the temperature is found to scale as T # (r# ) = y −1 T (r), and
the temperature gradient as dT # (r# )/dr# = y −2 dT (r)/dr.
The assumption that the star reacts to new particle emission by a
homologous contraction imposes restrictions on the constitutive rela-
tions for the effective energy generation rate and the opacity. In par-
ticular, for a chemically homogeneous star one needs to assume that
( ∝ ρn T ν and κ ∝ ρs T p . (1.8)
For the opacity, Frieman et al. took the Kramers law with s = 1
and p = −3.5 which is found to be a reasonable interpolation for-
mula throughout most lower main-sequence interiors. Hence, the local
energy flux scales as
L# (r# ) = y −1/2 L(r) . (1.9)
The hydrogen-burning rate (nuc also has the required form with n = 1,
and for the pp chain ν = 4−6; it dominates in the Sun and in stars
with lower mass. It is assumed that the new energy-loss rate (x follows
the same proportionality; the standard neutrino losses (ν are ignored
because they are small on the lower main sequence. If the star is not
in a phase of major structural readjustment one may also ignore (grav
in Eq. (1.4) so that in Eq. (1.3)
( = (1 − δx ) (nuc , (1.10)
where δx < 1 is a number which depends on the interaction strength of
the new particles. From Eq. (1.3) one concludes that
L# (r# ) = y −(3+ν) (1 − δx ) L(r), (1.11)
leading to
y = (1 − δx )2/(2ν+5) . (1.12)
Assuming δx + 1, Frieman et al. then found for the fractional changes
of the stellar radius, luminosity, and interior temperature,
δR −2δx δL δx δT 2δx
= , = , = . (1.13)
R 2ν + 5 L 2ν + 5 T 2ν + 5
Therefore, the star contracts, becomes hotter, and the surface photon
luminosity increases—it overcompensates for the new losses. Moreover,
16 Chapter 1
even if the luminosity Lx in “exotics” is as large as the photon lumi-

nosity (δx = 12 ) the overall changes in the stellar structure remain mod-
erate. The predominant effect is an increased consumption of nuclear
fuel at an almost unchanged stellar structure, leading to a decreased
duration of the hydrogen-burning phase of
δτ /τ ≈ −δx . (1.14)
The standard Sun is halfway through its main-sequence evolution so
that a conservative constraint is δx < 12 .
In general, the exotic losses do not have the same temperature and
density dependence as the nuclear burning rate, implying a breakdown
of the homology condition. However, to lowest order these results will
remain valid if one interprets δx as a suitable average over the en-
tire star,
δx = Lx /(Lx + Lγ ), (1.15)
with the photon luminosity Lγ and that in exotics Lx . To lowest order
Lx can be computed from an unperturbed stellar model.
For a convective structure (M < ∼ 0.25 M" main-sequence stars)
Frieman et al. found by a similar treatment
δR −2δx δL −5δx δT 2δx
= , = , = . (1.16)
R 2ν + 11 L 2ν + 11 T 2ν + 11
These stars also contract, and the internal temperature increases, but
the surface luminosity decreases.
1.3.2 Application to the Sun

For the Sun, the radius and luminosity are very well measured and
so one may think that small deviations δR and δL from a standard
model were detectable. This is not so, however, because a solar model
is defined to produce the observed radius and luminosity at an age
of 4.5 Gyr. The unknown presolar helium abundance Yinitial is chosen
to reproduce the present-day luminosity, and the one free parameter
of the mixing-length theory relevant for superadiabatic convection is
calibrated by the solar radius.
In a numerical study Raffelt and Dearborn (1987) implemented ax-
ion losses by the Primakoff process in a 1 M" stellar model, metal-
licity Z = 0.02, which was evolved to 4.5 Gyr with different amounts
of initial helium and different axion coupling strengths. Details of the
emission rate as a function of temperature and density are studied in
Table 1.1. Initial helium abundance for solar models with axion losses.
g10 Yinitial δx Xc
0 0.274 0.00 0.362
10 0.266 0.16 0.307
15 0.256 0.32 0.292
20 0.241 0.51 0.245
25 0.224 0.65 0.151
Sect. 5.2. For the present discussion the axion losses represent some
generic energy-loss mechanism with a rate proportional to the square
of the axion-photon coupling strength gaγ .
Without exotic losses a presolar helium abundance of Yinitial = 0.274
was needed to reproduce the present-day Sun. For several values of
g10 ≡ gaγ /10−10 GeV−1 Raffelt and Dearborn found the initial helium
values given in Tab. 1.1 necessary to produce the present-day luminos-
ity. The values for δx in Tab. 1.1 are defined as in Eq. (1.15) with Lx
the axion luminosity of the (perturbed) present-day solar model which
has Lγ = L" . Also, the central hydrogen abundance Xc of the present-
day model is given. For g10 = 30, corresponding to δx ≈ 0.75, no
present-day Sun could be constructed for any value of Yinitial .
The primordial helium abundance is thought to be about 23%, and
the presolar abundance is certainly larger. Still, a value of δx less than
about 0.5 is hard to exclude on the basis of this calculation. Therefore,
the approximate solar constraint remains δx < 1 <
∼ 2 or Lx ∼ L" as found
from the analytic treatment in the previous section.
One may be able to obtain an interesting limit by considering the
oscillation frequencies of the solar pressure modes. Because of the excel-
lent agreement between standard solar models and the observed p-mode
frequencies there is little leeway for a modified solar structure and com-
position. This method has been used to constrain a hypothetical time
variation of Newton’s constant (Sect. 15.2.3).
1.3.3 Radiative Energy Transfer

If novel particles are so weakly interacting that they escape freely from
the star once produced their role is that of a local energy sink. Neu-
trinos are of that nature, except in supernova cores where they are
“trapped” for several seconds. One could imagine new particles with
18 Chapter 1
such large interactions that they are even trapped, say, in the Sun.
Because their mean free path (mfp) is now less than the geometric di-
mension of the star, they remove energy from one region and deposit it
at an approximate distance of one mfp. This is precisely the mechanism
of radiative energy transfer: the particles now contribute to the opac-
ity. In a transition region where the mfp is on the order of the stellar
radius this mode of energy transfer couples distant regions and thus
cannot be described in the form of the differential equation (1.6). In
this case the difference between an energy-loss and an energy-transfer
mechanism is blurred.
Equation (1.6) is justified when the mfp is less than the temper-
ature scale height (d ln T /dr)−1 . For radiative transfer (κρ)−1 is an
average mfp. Therefore, Eq. (1.6) informs us that the energy flux car-
ried through a sphere of radius r is a product of a numerical factor, the
area, the photon mfp, the number density of photons, and the temper-
ature gradient. Radiative transfer is more efficient for a larger mfp, i.e.
for a more weakly interacting particle!
If a second photon existed with a coupling strength α# instead of
1
α = 137 , it would contribute more to the energy transfer for α# < α.
The observed properties of the Sun and other stars confirm that the
standard opacities certainly cannot be wrong by more than a factor
of a few. Therefore, the new photon must interact about as strongly
as the standard one to be in agreement with the properties of stars,
or it must interact so weakly that it freely escapes and the integrated
volume emission Lγ ! is less than the photon luminosity Lγ .
1.3.4 Opacity Contribution of Arbitrary Bosons

In order to implement the energy-transfer argument one needs a prop-
erly defined expression for the Rosseland opacity contribution of arbi-
trary bosons. Following the textbook derivation of the photon radiative
opacity such a definition was provided by Carlson and Salati (1989) as
well as Raffelt and Starkman (1989).
For a sufficiently short mean free path - the radiation field of the
new bosons is taken to be locally isotropic. The local energy flux is then
found to be Fω = − 13 βω -ω ∇Bω where the index ω indicates that a quan-
tity refers to the boson energy ω. Moreover, Bω and Fω are understood
to be “specific,” i.e. an energy density and flux per unit energy interval.
The velocity is βω = [1 − (m/ω)2 ]1/2 for a boson with mass m. (Recall
that natural units with h̄ = c = kB = 1 are always used.) In local
thermal equilibrium the energy density for massive bosons correspond-
ing to the phase-space element d3 p is g ω (eω/T − 1)−1 d3 p/(2π)3 where

g is the number of polarization degrees of freedom and (eω/T − 1)−1
is the thermal boson occupation number. With an angular integration
d3 p becomes 4πp2 dp where p = |p|. Using p = (ω 2 − m2 )1/2 one finds
p dp = ω dω so that
g ω 2 (ω 2 − m2 )1/2
Bω = . (1.17)
2π 2 eω/T − 1
The total energy flux is found by integrating over all frequencies,
" ∞
F = − 13 ∇T dω βω -ω ∂T Bω , (1.18)
m
where ∇Bω = ∂T Bω ∇T was used with ∂T = ∂/∂T .

For photons, one usually writes F = −(3κγ ρ)−1 ∇aT 4 where aT 4
is the total energy density in photons (a = π 2 /15). Together with
Eq. (1.18) this defines the photon Rosseland mean opacity κγ . For
other bosons one defines a corresponding quantity,
1 1 "∞
≡ dω -ω βω ∂T Bω . (1.19)
κx ρ 4aT 3 m
The “exotic” opacity thus defined appears in the stellar structure equa-
tion in the way indicated by Eq. (1.7).
The production and absorption of bosons involves a Bose stimu-
lation factor. This effect is taken account of by including a factor
(1 − e−ω/T ) under the integral in Eq. (1.19). The “absorptive” opacity
thus derived is usually referred to as the reduced opacity κ∗ which is
the quantity relevant for energy transfer. In practice, it is not very
different from κ as ω is typically 3T for massless bosons.
In the large-mass limit (m , T ) the reduction factor may be ignored
entirely and one finds to lowest order,
# $3 " ∞
1 15 m
=g 4 e−m/T dy -ω (y) y e−y , (1.20)
κx ρ 4π T 0
where y ≡ βω2 m/2T was used so that the energy of a nonrelativistic

boson is given by ω = m + yT .
20 Chapter 1
1.3.5 Solar Bound on Massive Pseudoscalars

In order to illustrate the energy-loss and energy-transfer argument for
a boson with arbitrary mass I use the example of pseudoscalars χ which
couple to electrons by a “fine-structure constant” αx . This is the only
case in the literature where both arguments have been applied without
assuming that the particle mass mx is small relative to T .
The new bosons can be produced and absorbed by a variety of
reactions. For purposes of illustration I focus on the Compton-type
process γe ↔ eχ. For mx + T the energy-loss rate per unit mass will
be given in Eq. (3.23) in a more general context. For arbitrary masses
it was derived by Raffelt and Starkman (1989). The result is
160ααx Ye T 6
(x = F (mx /T ), (1.21)
π mu m4e
√
where F (z) = 1 for mx + T and F (z) = (20 2)−1 z 9/2 e−z for T +
mx + me . Further, Ye is the number of electrons per baryon and mu is
the atomic mass unit. (It was used that approximately ne /ρ = Ye /mu
with the electron density ne .)
For T + mx + me the Compton absorption rate is found to be
Γx = 4πααx m2x m−4
e ne , leading to an opacity contribution of the pseudo-
Fig. 1.1. Effects of massive pseudoscalar particles on the Sun (interior tem-
perature about 1 keV). Above the dashed line they contribute to the radia-
tive energy transfer, below they escape freely and drain the Sun of energy.
The shaded area is excluded by this simple argument. (Adapted from Raffelt
and Starkman 1989.)
scalars of
2(2π)9/2 ααx Ye T 5/2 emx /T
κx = 1/2
45 mx m4e
= 4.4×10−3 cm2 g−1 × αx Ye m−0.5 2.5 mx /T
keV TkeV e , (1.22)
where mkeV = mx /keV and TkeV = T /keV.
The observed properties of the Sun then allow one to exclude a large
range of parameters in the mx -αx -plane. The energy-loss rate integrated
over the entire Sun must not exceed L" . Moreover, κ−1
x must not exceed
−1 2
the standard photon contribution κγ ≈ 1 g/cm , apart from perhaps
a factor of order unity. Taking a typical solar interior temperature
of 1 keV, these requirements exclude the shaded region in Fig. 1.1.
The dashed line marks the parameters where the mfp is of order the
solar radius.
1.4 General Lesson

What have we learned? Weakly interacting particles, if they are not
trapped in stars, carry away energy. For a degenerate object this en-
ergy loss leads to additional cooling, for a burning star it leads to an
accelerated consumption of nuclear fuel. New particles would cause sig-
nificant effects only if they could compete with neutrinos which already
carry away energy directly from the interior.
If new particles are trapped because of a short mfp, they contribute
to the energy transfer. They dominate unless their mfp is shorter than
that of photons. Such particles probably do not exist or else they
would have been found in laboratory experiments. This justifies the
usual focus on the energy-loss argument or “free streaming” limit.
Still, the impact of new low-mass particles is maximized for an mfp
of order the stellar radius, a fact which is often not appreciated in the
literature.
If the particles were so heavy that they could not be produced by
thermal processes in a stellar plasma they would be allowed for any cou-
pling strength. How heavy is heavy? The average energy of blackbody
photons is about 3T . Therefore, if mx < ∼ 3T the particle production
will not be significantly suppressed. However, the solar example of
Fig. 1.1 illustrates that for a sufficiently large coupling strength even
a particle with a mass of 30 T could have a significant impact. While
it can be produced only by plasma constituents high up in the tails
of the thermal distributions, the Boltzmann suppression factor e−mx /T
22 Chapter 1
can be compensated by a strong coupling. For particles with a different

interaction structure the equivalent of Fig. 1.1 would look qualitatively
similar, albeit different in detail.
Particles are not harmless in stars just because they are trapped,
or just because their mass exceeds a typical temperature. However, in
practice new particles are usually thought to be very weakly interacting
and either essentially massless on the scales of stellar temperatures, or
else very massive on those scales. In this book I will focus on low-mass
particles with very weak interactions so that the energy-loss argument
will be in the foreground of the discussion.
The opposite case of weakly interacting massive particles (WIMPs)
is of great interest in the framework of a hypothesis which holds that
such objects are the dark matter of the universe. (Another dark matter
candidate are axions which fall into the low-mass category covered in
this book.) WIMP masses would be in the 10 GeV range and above so
that they cannot be produced in stars, but they can be captured from
the galactic background. They would contribute to the energy transfer
by conduction. Hence their role in stars is similar to that of electrons,
except for their large mfp which allows them to contribute significantly
even if their concentration is low. I do not treat this subject because
it requires entirely different formal tools from those relevant for low-
mass particles. The fascinating story of WIMPs in stars will be told
comprehensively within a forthcoming review paper Supersymmetric
Dark Matter by Jungman, Kamionkowski, and Griest (1995).
Chapter 2
Anomalous
Stellar Energy Losses
Bounded by Observations
After a description of the main phases of stellar evolution, a review is

given of the observations that have been used to constrain novel stellar
energy losses. The main arguments involve the cooling speed of white
dwarfs and old neutron stars, the delay of helium ignition in low-mass
red giants, and the helium-burning lifetime of horizontal-branch stars.
The latter two arguments, which are based on observations of globular-
cluster stars, are cast into the form of an easy-to-use analytic criterion
that allows for a straightforward application in many different cases.
The cooling speed of nascent neutron stars is discussed in Chapter 11
in the context of supernova neutrinos.
2.1 Stages of Stellar Evolution

2.1.1 The Main Sequence
To identify those observables of stellar structure and evolution that can
be used to constrain or perhaps discover a novel energy-loss mechanism
it is necessary to understand how stars live and die. In Chapter 1 the
basics of the theory of stellar structure have already been discussed.
It is now time to give an overview of the stages of stellar evolution,
and how they manifest themselves in the observable properties of stars.
Because stellar evolution is an old subject, the literature is vast. I
will give reference only to a few recent original papers that I need to
support specific points. Otherwise, detailed quantitative accounts of
23
24 Chapter 2
stellar structure and evolution and their observational consequences

are found in the textbook and review literature.3
How do stars form in the first place? While a detailed understanding
of this process is still elusive, for the present discussion it is enough to
know that gravitationally bound clouds of gas ultimately fragment and
condense because of their continuous energy loss by electromagnetic
radiation. As discussed in Sect. 1.2.2, the negative specific heat of a
self-gravitating system enforces its contraction and heating. For the
fragmentation process a variety of parameters may be important such
as overall pressure, angular momentum, or magnetic fields.
Because the formation process of stars is poorly understood the
initial mass function (IMF) is not accounted for theoretically. Typically,
the number of stars per mass interval may be given by Salpeter’s IMF,
dN/dM ∝ (M/M! )−1.35 , which means that there are a lot more small
than large stars. However, the overall range of stellar masses is quite
limited. The largest stars have masses of up to 100 M! ; a value beyond
this limit does not seem to allow for a stable configuration. At the
small mass end, stars with M < 0.08 M! never become hot enough to
ignite hydrogen; such “brown dwarfs” have never been unambiguously
detected. It has been speculated that they could make up the dark
matter of spiral galaxies. A search in our galaxy by the gravitational
microlensing technique has yielded first candidates (Alcock et al. 1993,
1995; Aubourg et al. 1993, 1995).
The first stars consist of a mixture of X ≈ 0.75 (mass fraction of
hydrogen) and Y ≈ 0.25 (mass fraction of helium), the material left over
from the big bang of the universe. Subsequent generations also contain
a small amount of “metals” which in astronomers’ language is anything
heavier than helium; our Sun has a metallicity Z ≈ 0.02 (mass fraction
of metals). These heavier elements were bred by nuclear fusion in earlier
generations of stars which returned some of their mass to the interstellar
3
An excellent starting point for the nonexpert is Shu (1982). The classics are
Chandrasekhar (1939), Schwarzschild (1958), Clayton (1968), and Cox and Giuli
(1968). A recent general textbook is Kippenhahn and Weigert (1990). Recent mono-
graphs specializing on the Sun in general are Stix (1989), and on solar neutrinos
Bahcall (1989). The theory of stellar pulsation is covered in Cox (1980). A classic on
the physics of compact stars is Shapiro and Teukolsky (1983). A recent monograph
on neutron stars is Lipunov (1992) and on pulsars Mészáros (1992). Recent collec-
tions of papers or conference proceedings are available on the physics of red giants
(Iben and Renzini 1981), white dwarfs (Barstow 1993), pulsating stars (Schmidt
1989), neutron stars (Pines, Tamagaki, and Tsuruta 1992), and supernovae (Brown
1988; Petschek 1990). Many excellent reviews are found in the Annual Review of
Astronomy and Astrophysics.
Anomalous Stellar Energy Losses Bounded by Observations 25
medium. Mass loss in advanced stages of stellar evolution can take

on the benign form of a “stellar wind,” or it can occur in gigantic
“supernova explosions.” The end state of stellar evolution must be
complete disruption, or a compact object (white dwarf, neutron star, or
black hole), because a normal star, supported by thermal pressure, can
exist only as long as its nuclear fuel lasts (Sect. 1.2.2). The masses of
degenerate stars are constrained by their Chandrasekhar limit of about
1.4 M! for white dwarfs, and a similar value for neutron stars. Most
massive stars seem to return enough mass to the interstellar medium
that their typical end states are degenerate stars, not black holes. The
primordial material evolves chemically because it is cycled through one
generation of stars after another.
Today, the continuing birth and death of stars takes place mostly in
the disks of spiral galaxies. Such galaxies also have a population of old
halo stars and of globular clusters, about 150 in our Milky Way, each of
which is a gravitationally bound system of about 106 stars (Fig. 2.1).
Most of the globular clusters are in the galactic halo, far away from the
disk. The gravitational escape velocity for stars or gas from a cluster
is rather small, about 10 km s−1 . A supernova explosion, on the other
hand, ejects its material with typical velocities of several 103 km s−1 .
A single supernova is enough to sweep a globular cluster clean of all
Fig. 2.1. Globular cluster M3. (Image courtesy of Palomar/Caltech.)

26 Chapter 2
Fig. 2.2. Zero-age main sequence in the Hertzsprung-Russell diagram, with a

composition of 68.5% hydrogen, 29.4% helium, and 2.1% metals. (Adapted
from Kippenhahn and Weigert 1990.)
gas; then no further star formation is possible.4 Therefore, these sys-

tems are very clean laboratories for studies of stellar evolution as they
contain an early generation of stars (typical metallicities 10−3 −10−4 )
which in a given cluster are all coeval and have almost equal chemical
compositions. To a first approximation the stars in a globular cluster
differ only in one parameter—their initial mass.
The virial theorem explains (Sect. 1.2.2) that the negative specific
heat of a protostellar cloud leads to contraction and heating until it has
reached a thermal equilibrium configuration where the nuclear burning
rate exactly balances its overall luminosity. At this point the configu-
ration is determined entirely by its mass, apart from a small influence
of the metal content. Therefore, after the initial contraction hydrogen-
burning stars of different mass form the “zero-age main sequence” in
a Hertzsprung-Russell diagram where the effective surface temperature
is plotted on the horizontal axis and the stellar luminosity on the verti-
cal axis (Fig. 2.2). As hydrogen is consumed small adjustments of the
configuration occur. However, essentially a star remains at its initial
main-sequence location for most of its life.
The escape velocity from our galaxy is about 500 km s−1 . However, in a galactic
4
spiral arm the interstellar medium is dense enough to dissipate a SN explosion which
is only able to blow a hole into the interstellar material.
A 1 M! star lives about 10 Gyr (1 Gyr = 109 yr) on the main

sequence; our Sun is thought to have completed about half of this
episode. Heavier stars burn brighter (crudely L ∝ M3 ) and thus
live shorter lives. Because the universe is 10−20 Gyr old, stars with
M < ∼ 0.7−0.9 M! have not yet completed their hydrogen-burning
phase, even if they formed shortly after the big bang. Because glob-
ular clusters formed very early one expects a main-sequence turnoff
Fig. 2.3. Color magnitude diagram for the globular cluster M3 according
to Buonanno et al. (1986), based on the photometric data of 10,637 stars.
Following Renzini and Fusi Pecci (1988) the following classification has been
adopted for the evolutionary phases. MS (main sequence): core hydrogen
burning. BS (blue stragglers). TO (main-sequence turnoff): central hydro-
gen is exhausted. SGB (subgiant branch): hydrogen burning in a thick shell.
RGB (red-giant branch): hydrogen burning in a thin shell with a growing
core until helium ignites. HB (horizontal branch): helium burning in the
core and hydrogen burning in a shell. AGB (asymptotic giant branch): he-
lium and hydrogen shell burning. P-AGB (post-asymptotic giant branch):
final evolution from the AGB to the white-dwarf stage. (Original of the
figure courtesy of A. Renzini.)
28 Chapter 2
near this mass, depending somewhat on their metallicity. Indeed, the

color-magnitude diagram5 of the cluster M3 (Fig. 2.3) clearly shows this
effect. The location of the turnoff (TO) point in such a diagram can be
converted to the mass and thus to the age of a star which is just about
to complete hydrogen burning. Globular clusters, therefore, allow one
to establish a significant lower limit for the age of the universe.
2.1.2 Becoming a Red Giant
A main-sequence star naturally is hottest and densest at the center

where hydrogen is depleted fastest. (See Fig. 2.4 for the overall struc-
ture of a typical solar model which represents a low-mass star about
halfway through its main-sequence evolution.) After all of the hy-
drogen is consumed at the center, the further evolution depends on
the total mass—I begin with the fascinating story of low-mass stars
(M < ∼ 2 M! ).
After central hydrogen exhaustion a new configuration establishes
itself where the “helium ashes” which have accumulated at the cen-
ter form an ever denser core surrounded by a hydrogen burning shell
(Fig. 2.5). At the same time the material above the core begins to ex-
pand which implies that the stellar surface becomes larger and thus the
surface temperature lower according to the Stefan-Boltzmann law. Put
another way, the star becomes redder, and ultimately a “red giant.”
In the color-magnitude diagram of the globular cluster M3 (Fig. 2.3),
the collection of stars which begins to form a contracting core and an
expanding envelope is marked as the subgiant branch (SGB).
5
The brightness of a star is photometrically measured on the basis of the lumi-
nosity in a wavelength band defined by a filter with a well-defined spectral response.
There exist many different color systems, corresponding to many different filters. In
Fig. 2.3 the visual brightness is defined by V = −2.5 log LV + const. with LV the lu-
minosity measured with the “visual” filter, centered in the yellow-green waveband,
and a similar definition applies to the brightness B in the blue. The “color” B − V
is a measure of the surface temperature with lower temperatures (redder color) to
the right. For technical definitions see Allen (1963). Note that the downward turn
of the horizontal branch in the blue is determined by the V filter; bolometrically
the HB is truly horizontal. An absolute bolometric brightness is given by
Mbol = 4.74 − 2.5 log(L/L" ).
It is a dimensionless number, but often the unit mag (magnitude) is appended for
clarity. The visual brightness in Fig. 2.3 is given in apparent magnitudes which de-
pend on the distance of the cluster. However, because of the logarithmic brightness
definition a different distance leads only to a vertical shift of the entire picture; the
color and relative brightnesses remain unchanged.
Fig. 2.4. Left panels: A typical solar model (Bahcall 1989) as an ex-
ample for a low-mass star which is halfway through its hydrogen-burning
(main-squence) phase. Right panels: Horizontal-branch star (central helium
burning, shell hydrogen burning) with a metallicity Z = 0.004 after about
2.5×107 yr which is about a quarter of the HB lifetime. (Model from the
calculations of Dearborn et al. 1990.)
30 Chapter 2
Fig. 2.5. Main evolutionary phases of low-mass stars. The envelope and core
dimensions depend on the location on the RGB, HB, or AGB, respectively.
The given radii are only meant to give a crude orientation.
As hydrogen burning continues it dumps more and more helium on

the core which at first supports itself by thermal pressure. Soon it
becomes so dense, however, that the electrons become degenerate, in-
verting the mass-radius relationship to R ∝ M−1/3 (Sect. 1.2.2). Thus,
as the helium core mass Mc grows, the core radius Rc shrinks. The
gravitational potential Φc at the edge of the core is determined entirely
by the core because the envelope contributes little due to its large exten-
sion so that Φc ≈ −GN Mc /Rc ∝ M4/3 c . Because the hydrogen-burning
shell above the core still supports itself by thermal pressure, the tem-
perature near the core edge is determined by Φc ∝ M4/3 c . The growing
core mass causes the hydrogen burning shell to become ever hotter!
Because of the steep T dependence of the hydrogen burning rates, the
growing core causes the star to become ever brighter (Fig. 2.6). In this
Fig. 2.6. Evolutionary track of a 0.8 M! star (Z = 0.004) from zero age
to the asymptotic giant branch. The evolutionary phases are as in Fig. 2.3.
(Calculated with Dearborn’s evolution code.)
32 Chapter 2
shell-burning phase the luminosity is determined almost entirely by the

core mass. On the MS it was determined by the total mass, a parameter
which now hardly matters.
The red-giant branch (RGB) ascension is a fast process compared
with the main-sequence (MS) evolution (Fig. 2.6). Therefore, the red
giants in the observational color-magnitude diagram of Fig. 2.3 differ
only by very small amounts of their initial mass. The stars beyond
the MS have nearly identical properties while those on the MS differ
by their total mass. Therefore, the evolved stars in Fig. 2.3 essentially
trace out the evolutionary path of a single star with an approximate
mass corresponding to the MS turnoff (TO). For example, the stars on
the RGB in Fig. 2.3 essentially constitute snapshots of the single-star
evolutionary track shown in the upper panel of Fig. 2.6.
A red giant is a star with a compact energy source at the center and
a large convective envelope. In the present case we have a degenerate
helium core with shell hydrogen burning, but other configurations are
possible. The core is a star unto itself with near to negligible feed-
back from the envelope. The envelope, on the other side, is strongly
influenced by the nontrivial central boundary conditions provided by
the core. Convective configurations with a fixed mass and a prescribed
luminosity from a central point source occupy the “Hayashi line” in
the Hertzsprung-Russell diagram. A low-mass red giant ascends the
Hayashi line corresponding to its envelope mass.
The inflation of a star to red-giant dimensions is a remarkable phe-
nomenon which defies an intuitive explanation, except that the stellar
structure equations allow for such solutions. Another conceivable struc-
ture is a much smaller envelope with radiative rather than convective
energy transfer—upper MS stars are of that nature. Prescribing a core
with a given luminosity one can imagine these two extreme configura-
tions. Indeed, bright stars (L > 2
∼ 10 L! ) are usually found either near
the MS or near the RGB which together form a V-shaped pattern in
the Hertzsprung-Russell diagram (Fig. 2.9). The empty space in the
V is known as the “Hertzsprung gap”—the few stars found there are
thought to be on the move from one arm of the V to the other, i.e.
from a radiative to a convective envelope structure. Massive stars af-
ter completing hydrogen burning move almost horizontally across the
Hertzsprung gap and inflate to red-giant dimensions. Remarkably, they
can deflate and move horizontally back, executing a “blue loop.”
2.1.3 Helium Ignition
The core of a red giant reaches its limiting mass when it has become
so hot and dense that helium ignites. Because the nucleus 8 Be which
consist of two α particles (He nuclei) is not stable, He burning proceeds
directly to carbon, 3α → 12 C (“triple-α reaction”), via an intermediate
8
Be state. Because it is essentially a three-body reaction its rate de-
pends sensitively on ρ and T . In a red-giant core helium ignites when
Mc ≈ 0.5 M! with central conditions of ρ ≈ 106 g cm−3 and T ≈ 108 K
where the triple-α energy generation rate per unit mass, #3α , varies ap-
proximately as ρ2 T 40 . This steep temperature dependence allows one
to speak of a sharp ignition point even though there is some helium
burning at any temperature and density.
The helium core of a red giant is like a powder keg waiting for a
spark. When the critical temperature is reached where #3α exceeds
the neutrino losses a nuclear runaway occurs. Because the pressure
is mainly due to degenerate electrons the energy production at first
does not lead to structural changes. Therefore, the rise in temperature
is unchecked and feeds positively on the energy generation rate. As
this process continues the core expands nearly explosively to a point
where it becomes nondegenerate and the familiar self-regulation by the
gravitational negative specific heat kicks in (Chapter 1). The “explosion
energy” is absorbed by the work necessary to expand the core from
about 106 g cm−3 to about 104 g cm−3 . The core temperature of the final
configuration remains at about 108 K because of the steep temperature
dependence of #3α which allows only for a narrow range of stationary
burning conditions.
The final configuration with a helium-burning core and a hydrogen-
burning shell (Fig. 2.5) is known as a horizontal-branch (HB) star, a
term which is justified by the location of these objects in the color-
magnitude diagram Fig. 2.3. Note that the overall luminosity has de-
creased by the process of helium ignition (Fig. 2.6) because of the core
expansion which lowers the gravitational potential at the core edge and
thus the temperature in the hydrogen-burning shell which continues to
be regulated by the core mass and radius. The total luminosity of an
HB star is given to about 1/3 by He burning and 2/3 by hydrogen shell
burning (Fig. 2.4, right panels).
Because helium ignition is an almost explosive process on dynamical
time scales it is known as the “helium flash.” For the same reason, a
realistic numerical treatment does not seem to exist (for a review, see
Iben and Renzini 1984). There are two main problems. First, normal
34 Chapter 2
stellar-evolution codes use hydrostatic stellar structure equations which

ignore kinetic energies which are small in stationary phases or phases
of slow expansion or contraction (Sect. 1.2.1). However, in a phase of
fast expansion kinetic energies can be large. It is not known whether
the helium flash is a hydrostatic or a hydrodynamic event. Secondly,
convection undoubtedly plays a large role at transferring energy from
the ignition point. Note that helium ignites off-center because neutrino
losses cause a temperature dip at the center. A fundamental theory
of convection does not exist; it is likely that one would have to per-
form a three-dimensional hydrodynamic calculation to develop a full
understanding of the helium flash.
2.1.4 The Horizontal Branch
After helium ignition the overall luminosity has dropped as explained

above, and the surface has shrunk considerably, leading to a substan-
tially hotter (bluer) configuration. Precisely how blue an HB star be-
comes depends on the size of its envelope and thus on its total mass.
A substantial amount of mass loss occurs on the RGB where the pho-
tosphere of the star is so inflated that it is only weakly gravitationally
bound; a red giant may typically lose about 0.2 M! of its envelope
before helium ignites. Presumably, the amount of mass loss is not ex-
actly the same from star to star; a small spread of order 0.03 M! is
enough to explain the wide range of surface radii and thus colors found
for these objects which form a horizontal branch in the color-magnitude
diagram.6 The downward turn of the HB in Fig. 2.3 is an artifact of the
filter which determines the visual brightness—the bolometric brightness
is the same. It is fixed by the core properties.
In practice, the HB morphology found in different clusters is com-
plicated. Some clusters have widely spread HBs such as that shown in
Fig. 2.3 while others have stars only near the red or blue end, and gaps
and blue tails occur. One important parameter is the metallicity of the
envelope which determines the opacity to which the envelope structure
is very sensitive. However, one finds clusters with the same chemical
composition but different HB morphologies, a phenomenon which has
prompted a search for the “second parameter.” Many suggestions have
been made by some and refuted by others. One candidate second pa-
rameter which appears to remain viable is the cluster age, or rather,
the MS mass of the stars observed today on the HB. The envelope mass
6
For recent synthetic HBs see Lee, Demarque, and Zinn (1990, 1994).
remaining for an HB star after mass loss on the RGB will still depend
on its initial value, rendering this a rather natural possibility. How-
ever, it implies that the globular clusters of our galaxy did not form
at the same time—an age spread of several Gyr is required with the
older clusters predominantly at smaller galactocentric distances (Lee,
Demarque, and Zinn 1994 and references therein). Apart from some
anomalous cases such an age spread appears to be enough to explain
the second parameter phenomenon.
Returning to the inner structure of an HB star, it evolves quietly
at an almost fixed total luminosity (Fig. 2.6). The core constitutes
essentially a “helium main-sequence star.” Because its inner core is
convective it dredges helium into the nuclear furnace at the very center,
leaving a sharp composition discontinuity at the edge of the convective
region (Fig. 2.4). After some 12 C has been built up, 16 O also forms. At
the end of the HB phase, the helium core has developed an inner core
consisting of carbon and oxygen.
2.1.5 From Asymptotic Giants to White Dwarfs

After the exhaustion of helium at the center a degenerate carbon-
oxygen (CO) core forms with helium shell burning and continuing hy-
drogen shell burning (Fig. 2.5). Again, the star grows progressively
brighter and inflates. Put another way, it becomes very similar to a
star which first ascended the RGB: it ascends its Hayashi line for a
second time. The track in the Hertzsprung-Russell diagram asymp-
totically approaches that from the first ascent (“asymptotic giants”).
The upper RGB can be observationally difficult to distinguish from the
asymptotic giant branch (AGB) even though they are reasonably well
separated in Fig. 2.3.
In low-mass stars carbon and oxygen never ignite. The shell sources
extinguish when most of the helium and hydrogen has been consumed
so that the star has lost its entire envelope either by hydrogen burning
or by further mass loss on the AGB. The remaining degenerate CO
star continues to radiate the heat stored in its interior. At first these
stars are rather hot, but geometrically very small with a typical ra-
dius of 104 km. Their small surface area restricts their luminosity in
spite of the high temperature and so they are referred to as “white
dwarfs.” Because they are supported by electron degeneracy pressure,
their remaining evolution is cooling by neutrino emission from the in-
terior and by photon emission from the surface until they disappear
from visibility.
36 Chapter 2
Fig. 2.7. Ring Nebula in Lyra (M57), a planetary nebula. (Image courtesy
of Palomar/Caltech.)
For stars which start out sufficiently large the mass loss on the
AGB can be so dramatic that one may speak of the ejection of the
entire envelope. It forms a large shell of gas which is illuminated by
its central star, the newborn white dwarf. Such systems are known as
“planetary nebulae” (Fig. 2.7).
2.1.6 Type I Supernovae
A white dwarf can make a spectacular reappearance if it is in a binary

system. If the other member expands because it is in an earlier evolu-
tionary phase it can transfer mass, allowing for renewed nuclear burning
on the white-dwarf surface. This adds mass to the CO configuration
which ultimately becomes so hot and dense that these elements ignite,
leading to a (subsonic) deflagration or (supersonic) detonation front
which sweeps through the star and disrupts it entirely. This course of
events is the standard scenario for type I supernova explosions which
are among the most energetic events known in the universe. They must
be carefully distinguished from type II supernovae which are related to
the collapse of evolved, massive stars.
2.1.7 Intermediate-Mass Stars
The evolutionary scenario described so far applies to stars with a mass

below 2−3 M! . For larger masses the core conditions evolve contin-
uously to the ignition of helium which is then a quiet process. For a
given chemical composition the transition between the two scenarios is
a sharp function of the total mass. Because red giants are very bright
they dominate the total luminosity of an old stellar population so that
this transition affects its integrated brightness in a discontinuous way.
Some authors have used the concept of an “RGB phase transition” to
describe this phenomenon (Sweigart, Greggio, and Renzini 1990, and
references therein; Bica et al. 1991).
Stars with masses of up to 6−8 M! are expected to end up as CO
white dwarfs. Their evolution on the AGB can involve many interesting
phenomena such as “thermal pulses”—for a key to the literature see
Iben and Renzini (1984). Intermediate-mass stars have thus far played
little role for the purposes of particle astrophysics, and so their evolution
does not warrant further elaboration in the present context.
2.1.8 Massive Stars and Type II Supernovae
The course of evolution for massive stars (M > 6−8 M! ) is qualita-

tively different because they ignite carbon and oxygen in their core,
allowing them to evolve further. This is possible because even after
mass loss they are left with enough mass that their CO core grows
toward the Chandrasekhar limit. Near that point the density is high
enough to ignite carbon which causes heating and thus temporarily re-
lieves the electron degeneracy. Next, the ashes of carbon burning (Ne,
Mg, O, Si) form a degenerate core which ultimately ignites Ne burning,
and so forth. Ultimately, the star has produced a degenerate iron core,
surrounded by half a dozen “onion rings” of different burning shells.
The game is over when the iron core reaches its Chandrasekhar limit
because no more nuclear energy can be released by fusion. The tem-
perature is at 0.8×1010 K = 0.7 MeV, the density at 3×109 g cm−3 , and
there are about Ye = 0.42 electrons per baryon. Further contraction
leads to a negative feedback on the pressure as photons begin to dis-
sociate iron, a process which consumes energy. Electrons are absorbed
and converted to neutrinos which escape, lowering the electron Fermi
momentum and thus the pressure. Therefore, the core becomes un-
stable and collapses, a process which is intercepted only when nuclear
density (3×1014 g cm−3 ) is reached where the equation of state stiffens.
38 Chapter 2
Fig. 2.8. The Crab nebula, remnant of the supernova of A.D. 1054. (Image
courtesy of the European Southern Observatory.)
At this point a shock wave forms at the edge of the core and moves
outward. The implosion can be said to be reflected and thus turned into
an explosion. In practice, it is difficult to account for the subsequent
evolution as the shock wave tends to dissipate its energy by dissociating
iron. Currently it is thought that a revival of the stalled shock is needed
and occurs by neutrinos depositing their energy in the “hot bubble”
below the shock. This region has a low density yet high temperature,
and thus a high entropy per baryon by common astrophysical standards.
Within about 0.3 s after collapse the shock has moved outward and
ejects the entire overburden of the mantle and envelope. This course
of events is the scenario of a type II supernova (SN) explosion.
What remains is an expanding nebula such as the Crab (Fig. 2.8)
which is the remnant of the SN of A.D. 1054, and a central neutron
star (radius about 10 km, mass about 1 M! ) which often appears in
the form of a pulsar, a pulsating source of radiation in some or all
electromagnetic wave bands. The pulsed emission is explained by a
complicated interplay between the fast rotation and strong magnetic
fields (up to 1012 −1013 G) of these objects.
Returning to the moment after collapse of the iron core, it is so dense
(nuclear density and above) and hot (temperature of several 10 MeV),
that even neutrinos are trapped. Therefore, energy and lepton num-
ber are lost approximately on a neutrino diffusion time scale of several
seconds. The neutrinos from stellar collapse were observed for the first
and only time when the star Sanduleak −69 202 in the Large Magel-
lanic Cloud (a small satellite galaxy of the Milky Way) collapsed. The
subsequent explosion was the legendary SN 1987A.
After the exhaustion of hydrogen, massive stars move almost hori-
zontally across the Hertzsprung-Russell diagram until they reach their
Hayashi line, i.e. until they have become red supergiants. However, sub-
sequently they can loop horizontally back into the blue; the progenitor
of SN 1987A was such a blue supergiant.
The SN rate in a spiral galaxy like our own is thought to be about
one in a few decades, pessimistically one in a century. Because many
of the ones occurring far away in our galaxy will be obscured by the
dust and gas in the disk, one has to be extremely lucky to witness
such an event in one’s lifetime. The visible galactic SNe previous to
1987A were Tycho’s and Kepler’s in close succession about 400 years
ago. Both of them may have been of type I—no pulsar has been found
in their remnants.7 Of course, in the future it may become possible to
detect optically invisible galactic SNe by means of neutrino detectors
like the ones which registered the neutrinos from SN 1987A.
2.1.9 Variable Stars
Stars are held in equilibrium by the pull of gravity which is opposed

by the pressure of the stellar matter; its inertia allows the system to
oscillate around this equilibrium position. If the adiabatic relationship
between pressure and density variations is written in the form δP/P =
γ δρ/ρ, the fundamental oscillation period P of a self-gravitating ho-
mogeneous sphere (density ρ) is found to be8 P −1 = [(γ − 43 )GN ρ/π]1/2 .
This yields the period-mean density relationship P (ρ)1/2 ≈ const.
which is often written in the form P (ρ/ρ! )1/2 = Q with the average
solar density ρ! = 1.41 g cm−3 and the pulsation “constant” Q. It is
in the range 0.5−3 h, depending on the adiabiatic coefficient γ and the
7
Observationally, type I and II SNe are distinguished by the absence of hydrogen
spectral lines in the former which is explained by their progenitor being an accret-
ing white dwarf which explodes after carbon ignition. Therefore, it is difficult to
establish the type of a historical SN unless a pulsar is detected as in the Crab.
8
For γ < 4/3 the star is dynamically unstable. We have already encountered
this magic number for the relationship between pressure and density of a degenerate
relativistic electron gas where it led to Chandrasekhar’s limiting white-dwarf mass.
40 Chapter 2
detailed structure of the star.

When these oscillations are damped, stars are found in their equi-
librium configurations with constant color and brightness. However, it
is possible that a small deviation from equilibrium is amplified, leading
to a growing oscillation amplitude. Throughout a star there can be re-
gions which try to excite oscillations, and others which damp them. If
the driving mechanism is strong enough the star is found in a continuing
oscillation which manifests itself in an oscillating lightcurve.
Of particular interest are the Cepheid-type variables. Their oscilla-
tions are excited by the “κ mechanism” where the driving force is the
heat valve provided by the opacity of the stellar matter near the sur-
face. There, hydrogen and helium are only partially ionized so that a
temperature increase leads to increased ionization which increases the
opacity because the Thomson cross section on free electrons is much
larger than the Rayleigh one on neutral atoms. Therefore, the temper-
ature increase caused by compression makes it more difficult for energy
to escape: the valve shuts. In most regions of the star the opposite hap-
pens. Increased temperature makes it easier for energy to leak out from
the compressed region, the valve opens, and oscillations are damped.
Because the operation of the κ mechanism is determined by condi-
tions near the stellar surface which is characterized by its temperature
and luminosity, Cepheid-type variables are found on a certain locus of
color and brightness. This “instability strip” extends throughout the
Hertzsprung-Russell diagram (hatched band in Fig. 2.9); it applies to
stars of radically different internal structure. Variable stars are found
wherever the instability strip overlaps with an actual stellar population.
At the bright end of the strip which is close and nearly parallel to the
red-giant Hayashi line one finds the classical Cepheids (δ Cepheids) with
luminosities 300−30,000 L! and periods 1−50 days. The brightness
variation can be up to 1 mag (visual), or a luminosity which changes
by up to a factor of 3. Classical Cepheids are massive stars which cross
the otherwise unpopulated Hertzsprung gap from the main-sequence
to the red-giant region, and red giants and supergiants which execute
blue loops and so temporarily move to the blue of the Hayashi line.
The linear relationship between period and brightness of Cepheids as
well as their large intrinsic luminosities are the key for their prominent
role as standard candles and thus as astronomical distance indicators.
The crossing of the instability strip with the HB is the domain of
the RR Lyrae stars. Their luminosities are around 50−100 L! , their
periods 1.5−24 h. Their color coincides with the MS turnoff in globular
clusters (Fig. 2.3). RR Lyrae stars play an important role in the ongoing
Fig. 2.9. Schematic Hertzsprung-Russell diagram of the main types of stars.

The hatched band is the instability strip, the locus of Cepheid-type variables.
The “Hertzsprung gap” between the main sequence and the red-giant region
is populated with stars crossing between these branches, and by red giants
and supergiants which execute blue loops.
debate about the age of globular clusters.

Toward fainter luminosities the instability strip crosses the MS, a
region where the δ Scuti stars and dwarf Cepheids (periods of 1−3 h)
are located.
Fainter still, the only remaining population of stars are white dwarfs.
At the crossover of their locus with the instability strip the ZZ Ceti
variables are found; they have periods of a few minutes. Because the
oscillation period depends on the luminosity, an observed secular change
of the period of a ZZ Ceti star can serve as a sensitive diagnostic for a
decrease of its luminosity and thus for its cooling speed.
42 Chapter 2
2.2 White-Dwarf Cooling

2.2.1 Theoretical and Observed White-Dwarf Properties
After this general survey of how stars evolve it is time to study individ-
ual aspects in more detail, and notably, how stellar evolution is affected
by the emission of weakly interacting particles. I begin with the con-
ceptually most transparent case of stars for which the loss of energy
simply accelerates their cooling, i.e. white dwarfs and neutron stars.
The former represent the final state of the evolution of stars with
initial masses of up to several M! , perhaps up to 8 M! (Sect. 2.1).
For reviews of the theory and observed properties see Hubbard (1978),
Liebert (1980), Shapiro and Teukolsky (1983), Weidemann (1990), and
D’Antona and Mazzitelli (1990). Because white dwarfs (WDs) are sup-
ported by electron degeneracy pressure the hydrostatic and thermal
properties are largely decoupled. In Sect. 1.2.2 we had encountered
their inverted mass-radius relationship; in a polytropic approximation
of the WD structure one finds quantitatively (Shapiro and Teukol-
sky 1983)
R = 10,500 km (0.6M! /M)1/3 (2/µe )5/3 . (2.1)
Here, M is the stellar mass and µe = ρ m−1 u ne

−1
= Ye−1 the “mean
molecular weight of the electrons” with ρ the mass density, mu the
atomic mass unit, ne the electron density, and Ye the number of elec-
trons per baryon. WDs do not contain any hydrogen in their interior—it
would immediately ignite—so that µe = 2. Typically they consist of
carbon and oxygen, the end products of helium burning in the core of
the progenitor star. The central density of a polytropic model is
ρc = 1.46×106 g cm−3 (M/0.6 M! )2 (µe /2)5 , (2.2)
assuming nonrelativistic electrons.

If the mass is so large (the radius so small) that the electrons become
relativistic there exists no stable configuration, i.e. the masses of WDs
must lie below the Chandrasekhar limit (Shapiro and Teukolsky 1983)
MCh = 1.457 M! (2/µe )2 . (2.3)
Observationally the WD mass distribution is strongly peaked near M =

0.6 M! (Weidemann and Koester 1984) so that a nonrelativistic treat-
ment of the electrons is justified. The low mass of observed WDs is
understood by the large rate of mass loss during the red giant and
asymptotic giant evolution which can amount to an ejection of the en-

tire envelope and thus to the formation of a planetary nebula (Fig. 2.7).
A theoretical evolutionary track for a 3 M! star from the MS to the
WD stage was calculated, e.g. by Mazzitelli and D’Antona (1986). The
central stars of planetary nebulae are identified with nascent WDs. The
rate of WD formation inferred from the luminosity function discussed
below agrees within a factor of about 2 with the observed formation
rate of planetary nebulae, which means that both quantities agree to
within their statistical and systematic uncertainties.
The hottest and brightest WDs have luminosities of L ≈ 0.5 L!
while the faintest ones are observed at L ≈ 0.5×10−4 L! . Thus, be-
cause of their small surface area WDs are intrinsically faint (see the
Hertzsprung-Russell diagram Fig. 2.9). This implies that they can be
observed only in the solar neighborhood, for bright WDs out to about
100 pc (300 lyr). Because their vertical scale height in the galactic disk
is about 250 pc (Fleming, Liebert, and Green 1986) the observed WDs
homogeneously fill a spherical volume around the Sun. One may then
express the observed number of WDs in terms of a volume density; it
is of order 10−2 pc−3 .
The observed luminosity function (the space density of WDs per
brightness interval) is shown in Fig. 2.10 and listed in Tab. 2.1 according
to Fleming, Liebert, and Green (1986) and Liebert, Dahn, and Monet
(1988). The operation of a novel cooling mechanism can be constrained
by three important features which characterize the luminosity function:
its slope, which signifies the form of the cooling law, its amplitude,
which characterizes the cooling time and WD birthrate, and its sudden
break at log(L/L! ) ≈ −4.7, which characterizes the beginning of WD
formation. Even the oldest WDs have not had time to cool to lower
luminosities. From this break one can infer an age for the galactic
disk of 8−10.5 Gyr (Winget et al. 1987; Liebert, Dahn, and Monet
1988; Iben and Laughlin 1989; Wood 1992) while Hernanz et al. (1994)
find 9.5−12 Gyr on the basis of their cooling calculations which include
crystallization effects.
Because of the WD mass-radius relationship the surface tempera-
ture and luminosity are uniquely related for a given WD mass. There-
fore, instead of the luminosity function one may consider the temper-
ature distribution which, in principle, is independent of uncertain WD
distance determinations. Fleming, Liebert, and Green (1986) gave the
distribution of their sample of hot WDs in several temperature bins
(Tab. 2.2). Numerical cooling calculations of Blinnikov and Dunina-
Barkovskaya (1994) found good agreement with this distribution, as-
44 Chapter 2
Table 2.1. Observed WD luminosity function according to Fleming, Liebert,

and Green (1986) and Liebert, Dahn, and Monet (1988). For the hot and
bright degenerates (upper part of the table) a large fraction of their spectrum
lies in the ultraviolet, causing a large discrepancy between the absolute visual
magnitude MV and the absolute bolometric magnitude Mbol . For the hot
dwarfs the bins orginally had been chosen on the MV scale with a width
of 0.5 mag, centered on the half-magnitudes; the listed Mbol is the mean in
these intervals.
Mean Mean dN/dMbol

MV Mbol log(L/L! ) [pc−3 mag−1 ] log(dN/dMbol )
9.5 5.50 −0.31 1.22×10−6 −5.91 (+0.18,−0.31)
10.0 6.88 −0.86 1.01×10−5 −5.00 (+0.14,−0.21)
10.5 7.84 −1.25 2.16×10−5 −4.67 (+0.13,−0.18)
11.0 8.92 −1.68 9.56×10−5 −4.02 (+0.12,−0.16)
11.5 10.12 −2.16 1.21×10−4 −3.92 (+0.11,−0.15)
12.0 11.24 −2.61 1.51×10−4 −3.82 (+0.11,−0.16)
12.5 11.98 −2.90 2.92×10−4 −3.54 (+0.11,−0.16)
13.0 12.55 −3.13 6.07×10−4 −3.22 (+0.20,−0.39)
13.50 −3.51 0.89×10−3 −3.05 (+0.14,−0.21)
14.50 −3.91 1.34×10−3 −2.87 (+0.14,−0.20)
15.50 −4.31 0.24×10−3 −3.62 (+0.18,−0.31)
Table 2.2. Temperature distribution of hot WDs according to Fleming,

Liebert, and Green (1986).
Teff [103 K] Fraction of WDs

40−80 (7.06 ± 1.27) × 10−3
20−40 0.235 ± 0.026
12−20 0.759 ± 0.083
suming that the WD mass function was peaked around 0.7 M! which
is somewhat larger than the canonical value of 0.6 M! . For the present
purpose, however, the smallness of this difference is taken as a confir-
mation of the standard WD cooling theory.
Fig. 2.10. Observed WD luminosity function as in Tab. 2.1. The dot-

ted line represents Mestel’s cooling law with a constant WD birthrate of
B = 10−3 pc−3 Gyr−1 . The dashed line is from the numerical cooling curve
of a 0.6 M! WD (Koester and Schönberner 1986), including neutrino losses
and assuming the same constant birthrate.
2.2.2 Cooling Theory

A WD has no nuclear energy sources and so it shines on its residual ther-
mal energy: the evolution of a WD must be viewed as a cooling process
(Mestel 1952). Because electron conduction is an efficient mechanism
of energy transfer the interior can be viewed, to a first approximation,
as an isothermal heat bath with a total amount of thermal energy U .
Because the nondegenerate surface layers have a large “thermal resis-
tance,” they insulate the hot interior from the cold surrounding space,
throttling the energy loss Lγ by photon radiation. Of course, WDs can
also lose energy by neutrino volume emission Lν , and by novel particle
emission Lx . Hence, WD cooling is governed by the equation
dU/dt = −(Lγ + Lν + Lx ). (2.4)
This simple picture ignores the possibility of residual hydrogen burn-

ing near the surface, a possibly important luminosity source for young
WDs (e.g. Castellani, Degl’Innocenti, and Romaniello 1994; Iben and
Tutukov 1984). I will get back to this problem below.
46 Chapter 2
In order to translate this equation into the observable luminosity

function I assume a constant WD birthrate B so that the total number
density of degenerates is N = B tgal (age of the galactic disk tgal ).
Taking the above values N ≈ 10−2 pc−3 and tgal ≈ 9 Gyr one finds
B ≈ 10−3 pc−3 Gyr−1 . Because the number density of WDs in a given
magnitude interval dMbol is proportional to the time interval dt it takes
to cool through this magnitude range one readily obtains
dN dt dU/dMbol
=B = −B . (2.5)
dMbol dMbol Lγ + Lν + Lx
The photon luminosity is Lγ = 78.7 L! 10−2Mbol /5 in terms of the bolo-
metric magnitude, equivalent to log(Lγ /L! ) = (4.74 − Mbol )/2.5. Lγ is
related to the internal temperature T by the thermal conductance of
the surface layers so that one may derive a function T (Lγ ). The quanti-
ties U , Lν , and Lx are given in terms of T so that they can be expressed
in terms of Lγ and hence of Mbol .
In hot WDs the thermal energy is largely stored in the nuclei which
form a nearly classical Boltzmann gas. At low T the ideal-gas law
breaks down and eventually the nuclei arrange themselves in a lattice.
The internal energy is then a more complicated function of temperature.
The heat capacity per nucleon, which is 32 for an ideal gas, rises to
3 near the Debye temperature ΘD and then drops approximately as
(16π 4 /5) (T /ΘD )3 to zero (Shapiro and Teukolsky 1983). However, the
observed WDs have a relatively small ρ because of their small mass
around 0.6 M! so that even the oldest WDs have not yet crystallized.
Therefore, as a reasonable first approximation the internal energy is
U = C T with the ideal-gas heat capacity for the entire star of
3 M ! Xj L! Gyr M ! Xj
C= = 3.95×10−2 , (2.6)
2 mu j Aj 107 K M! j Aj
where Xj is the mass fraction of the element j, atomic mass Aj , and
mu = 1.661×10−24 g is the atomic mass unit.
The thermal conductance of the surface layers is calculated by solv-
ing the stellar structure equations. Using a Kramers opacity law,
κ = κ0 ρ T −7/2 , one finds (van Horn 1971; Shapiro and Teukolsky 1983)
" #7/2
−3 M T
L = 1.7×10 L! ≡ K T 7/2 , (2.7)
M! 107 K
where T is the internal temperature. The observed WD luminosities of
Tab. 2.1 vary between 0.5×10−4 and 0.5 L! , corresponding to a range
0.4−6 × 107 K of internal temperatures.
With these results the luminosity function is found to be
dN 4 ln(10) C (Lγ /K)2/7

= B . (2.8)
dMbol 35 Lγ + Lν + Lx
With B3 ≡ B/10−3 pc−3 Gyr−1 this is numerically

dN
= B3 2.2×10−4 pc−3 mag−1
dMbol
" #5/7 !
10−4Mbol /35 L! M Xj
× . (2.9)
78.7L! 10−2Mbol /5 + Lν + Lx M! j Aj
If one ignores Lν and Lx this is

$ %5/7
dN M ! Xj
= B3 2.9×10−6 pc−3 mag−1 102Mbol /7 .
dMbol M! j Aj
(2.10)
12 16
Taking M = 0.6 M! and an equal mixture of C and O one finds
2
log(dN/dMbol ) = 7
Mbol − 6.84 + log(B3 ) , (2.11)
a behavior known as Mestel’s cooling law. For B3 = 1 this function

is shown as a dotted line in Fig. 2.10. Detailed cooling curves and
luminosity functions have been calculated, for example, by Lamb and
van Horn (1975), Shaviv and Kovetz (1976), Iben and Tutukov (1984),
Koester and Schönberner (1986), Winget et al. (1987), Iben and Laugh-
lin (1989), Segretain et al. (1994), and Hernanz et al. (1994).
From Fig. 2.10 it is evident, however, that Mestel’s cooling law pro-
vides a surprisingly good representation for intermediate luminosities
where it is most appropriate. At the bright end, the luminosity func-
tion is slightly depressed, providing evidence for neutrino cooling. It
rapidly falls off at the faint end, presumably indicating the beginning
of WD formation as discussed above.
2.2.3 Neutrino Cooling

For the hottest WDs volume neutrino emission is more important than
surface photon cooling. The photon luminosity of Eq. (2.7) can be
expressed as an effective energy-loss rate per unit mass of the star, #γ =
Lγ /M = 3.3×10−3 erg g−1 s−1 T73.5 with T7 = T /107 K. For the upper
relevant temperature range, neutrinos are emitted mostly by the plasma
48 Chapter 2
process γ → νe ν e which is studied in detail in Chapter 6; numerical

emission rates are discussed in Appendix C. The neutrino energy-loss
rate as a function of temperature is shown in Fig. C.6; for a WD the
short-dashed curve (2ρ/µe = 106 g cm−3 ) is most appropriate.
Neutrino cooling causes a depression of the WD luminosity func-
tion at the bright end. The dashed line in Fig. 2.10 represents a nu-
merical cooling calculation for a 0.6 M! WD which included neutrinos
(Koester and Schönberner 1986); the “neutrino dip” is clearly visible.
The photon decay γ → νν is a neutrino process made possible by
the medium-induced photon dispersion relation and by the medium-
induced effective photon-neutrino coupling (Chapter 6). As such this
process is not observable in the laboratory, although there is no doubt
about its reality. Still, it is encouraging to see it so plainly in the WD
luminosity function.
If neutrino emission were much stronger than standard, the neu-
trino dip would be correspondingly deeper. Stothers (1970) used the
observation of several bright WDs in the Hyades cluster to constrain the
efficiency of neutrino cooling; at that time the existence and magnitude
of a direct neutrino-electron interaction had not yet been experimen-
tally established. Stothers found that an emission rate 300 times larger
than standard could be conservatively excluded.
Recently, Blinnikov and Dunina-Barkovskaya (1994) have studied
this subject in detail. Their motivation was the possible existence of
a neutrino magnetic dipole moment which would enhance the effec-
tive neutrino-photon coupling and thus the efficiency of the plasma
process. For the present general discussion it is enough to think of
their study as an arbitrary variation of the neutrino emissivity, even
though the magnetic-moment induced plasma emission rate has a dif-
ferent density dependence—see Eq. (6.94) for details. For the perti-
nent conditions the energy-loss rate induced by the assumed dipole
moment µν is roughly 0.06 µ212 times the standard one where µ12 ≡
µν /10−12 µB (Bohr magneton µB ). Blinnikov and Dunina-Barkovskaya
(1994) calculated the luminosity function for 0.6 M! WDs for the stan-
dard case (µ12 = 0, dashed line in Fig. 2.11) and for a roughly 25-fold
increased rate of neutrino cooling (µ12 = 20, dotted line in Fig. 2.11).
The birthrate of WDs was assumed constant and adjusted to opti-
mize the agreement with the observations. For µ12 = 0 they needed
B = 0.62×10−3 pc−3 Gyr−1 while for µ12 = 20 the best fit was achieved
for 0.67 in these units.
Blinnikov and Dunina-Barkovskaya (1994) pointed out that a par-
ticularly sensitive observable for the neutrino dip is the temperature dis-
Fig. 2.11. Luminosity function for 0.6 M! WDs for two values of µ12 (Blin-
nikov and Dunina-Barkovskaya 1994) compared with the observations quoted
in the upper part of Tab. 2.1 (Fleming, Liebert, and Green 1986).
Fig. 2.12. Relative number of 0.6 and 0.8 M! WDs in the two hot tempera-
ture bins of Tab. 2.2 as a function of the anomalous neutrino cooling implied
by a magnetic dipole moment µν (Blinnikov and Dunina-Barkovskaya 1994).
The number of WDs in the temperature range 12,000−40,000 K is normal-
ized to unity. The shaded bands correspond to the observations (Tab. 2.2).
50 Chapter 2
tribution of hot WDs derived by Fleming, Liebert, and Green (1986)—

see Tab. 2.2. They calculated cooling sequences for M = 0.6 M! and
0.8 M! with varying amounts of nonstandard cooling. The relative
number of WDs in the two hot temperature bins of Tab. 2.2 are shown
as a function of µν in Fig. 2.12.
These results seem to indicate that a significantly enhanced rate of
neutrino emission can be conservatively excluded. However, this view
may be challenged if one includes the possibility of residual hydro-
gen burning near the WD surface which could mask neutrino cooling
because it would fill in some of the “neutrino dip” in the luminosity
function. Because it is not known how much hydrogen is retained by
a WD after the planetary nebula phase one has an adjustable parame-
ter to provide a desired amount of heating (Castellani, Degl’Innocenti,
and Romaniello 1994). However, preliminary investigations seem to in-
dicate that even when residual hydrogen burning is included the impact
of µν is masked only in one of the temperature bins used by Blinnikov
and Dunina-Barkovskaya (1994) so that a significant deformation of the
luminosity function appears to remain (Blinnikov and Degl’Innocenti
1995, private communication).
2.2.4 Cooling by Boson Emission

Standard or exotic neutrino emission from WDs (or neutron stars) has
the important property that it switches off quickly as the star cools
because of the steep temperature dependence of the emission rates.
Therefore, neutrinos cause a dip at the hot end of the luminosity func-
tion while older WDs are left unaffected, even for significantly enhanced
neutrino cooling (Fig. 2.11). One may construct other cases, however,
where this is different. One example is when a putative low-mass bo-
son is emitted in place of a neutrino pair, say, in the bremsstrahlung
process e + (Z, A) → (Z, A) + e + νν. The reduced final-state phase
space then reduces the steepness of the temperature dependence of the
energy-loss rate. The possible existence of such particles is motivated
by theories involving spontaneously broken global symmetries. The
most widely discussed example is the axion which will be studied in
some detail in Chapter 14. For the present discussion all that matters
is the temperature variation of an assumed energy-loss rate.
The bremsstrahlung rate for pseudoscalar bosons will be calculated
in Chapter 3. For the highly degenerate limit the result is given in
Eq. (3.33) where α# = g 2 /4π is the relevant “fine-structure constant.”
Because this rate depends on the density only weakly through a factor
F which includes Coulomb screening by ion correlations and electron

relativistic corrections one may easily “integrate” over the entire star.
For an assumed equal mixture of carbon and oxygen one finds for the
luminosity in pseudoscalar “exotica”
Lx = α26 2.0×10−3 L! (M/M! ) (F ) T74 , (2.12)
where α26 = α# /10−26 and T7 = T /107 K (internal temperature T ).

Further, (F ) ≈ 1.0 within a few 10% (Sect. 3.5.2).
This energy-loss rate varies with internal temperature almost as
the surface photon luminosity of Eq. (2.7). If Lx dominates in Mes-
tel’s cooling law Eq. (2.11) the slope 10 35
Mbol is replaced with the al-
most identical value 1235
M bol . Thus, for a given WD birthrate the main
impact of pseudoscalars is to reduce the amplitude of the luminosity
function. Conversely, the inferred birthrate is larger by about a factor
1 + Lx /Lγ ≈ 1 + α26 . Because the formation rate of planetary nebulae,
the progenitors of WDs, agrees with the standard inferred birthrate to
within a factor of about 2, one finds α26 < ∼ 1.
9
The observed break of the luminosity function at the faint end

(Fig. 2.10) has been interpreted as the beginning of WD formation.
If boson cooling were dominant the faintest luminosities would have
been reached in a shorter amount of time, reducing the inferred age of
the galactic disk; standard cooling implies tgal = 8−12 Gyr. Even the
solar system is 4.5 Gyr old and so a reduction of tgal by more than a
factor of 2 is ruled out. This implies Lx < Lγ so that α26 < 1.0 as
before.
The galactic age constraint appears to be much more reliable than
the one based on the formation rate of planetary nebulae. The former
could be avoided if the observations reported by Liebert, Dahn, and
Monet (1988) were crudely incomplete at the faint end of the luminos-
ity function, i.e. if many faint WDs had been overlooked so that the
break in the luminosity function of Fig. 2.10 were a major observational
selection effect. Barring this remote possibility the limit on α26 is con-
servative. Even for a much smaller value of α26 , the inferred value for
tgal is reduced by an approximate factor (1 + α26 )−1 which may still be
significant.
Wang (1992) calculated numerical 1 M! WD sequences with vary-
ing amounts of pseudoscalar cooling while Blinnikov and Dunina-Bar-
kovskaya (1994) performed a more detailed study for 0.6 M! WDs.
9
In the original derivation (Raffelt 1986b) ion correlations were ignored, leading
to (F ) ≈ 3 and to the limit α26 <
∼ 0.3.
52 Chapter 2
They calculated the distribution for the temperature bins of Tab. 2.2
in analogy to the discussion of neutrino dipole moments in the previous
section. Their results are shown in Fig. 2.13 as a function of α26 where
I have corrected from (F ) = 3, which they used in order to compare
with Raffelt (1986b), to the more appropriate value (F ) = 1.
Fig. 2.13. Relative number of 0.6 M! WDs in the hot and intermediate
temperature bins of Tab. 2.2 as a function of the “fine-structure constant”
α26 = α# /10−26 of pseudoscalar bosons (Blinnikov and Dunina-Barkovskaya
1994). I have corrected from (F ) = 3 to the more appropriate value 1. The
number of WDs in the temperature range 12,000−40,000 K is normalized to
unity. The shaded bands correspond to the observations (Tab. 2.2).
This temperature method is now relatively insensitive because axion

cooling leaves the shape of the luminosity function nearly unchanged,
except that the neutrino dip is washed out for strong cooling. In detail,
however, important changes occur. It is noteworthy that boson cool-
ing has a significant impact on the shape of the luminosity function
even for α26 < 1. It appears that the possibility of residual hydrogen
burning in young WDs would make the present argument more con-
servative because it has the same effect as boson emission, namely to
reduce or wash out the neutrino dip in the luminosity function.
2.2.5 Period Decrease of Variable White Dwarfs

The luminosity function allows one to determine the WD cooling speed
because one is looking at a large WD ensemble of different age. Recently
it has become possible for the first time to measure WD cooling directly
for a single object by virtue of the period change of a ZZ Ceti star.
White dwarfs have residual atmospheres which may be hydrogen

rich (DA white dwarfs) or helium rich (DB); the DA stars are about
four times more frequent. The surface layers of DA white dwarfs are
not fundamentally different from those of main-sequence or giant stars
and so for appropriate conditions one expects Cepheid-type pulsations.
Indeed, where the faint continuation of the Cepheid instability strip
intersects with the locus of WDs in the Hertzsprung-Russell diagram
(Fig. 2.9) one finds the DA variables (DAV), also known as ZZ Ceti
stars after their prototype example. They have pulsation periods of a
few minutes.
The oscillation period depends on the temperature of the layer
which exhibits the κ instability (Sect. 2.1.9) and thus excites the pul-
sations, and also on the radius of the star. Therefore, the slowing
down of the period P is a direct measure of the temperature decrease
and thus of the WD cooling speed. Standard pulsation theory yields
Ṗ /P ≈ −a Ṫ /T + b Ṙ/R where the dimensionless constants a and b
are of order unity (Winget, Hansen, and van Horn 1983). Because
a WD has an almost fixed radius, the second term may be ignored.
The time scale of cooling and of the period change are then related by
T /Ṫ = −a P/Ṗ . ZZ Ceti stars have surface temperatures in the neigh-
borhood of 13,000 K where the cooling time scale is of order 1 Gyr.
For a period of a few minutes one is talking of a period decrease
Ṗ = O(10−14 s s−1 ), not an easy quantity to measure.
After upper limits on Ṗ had been established over the years for a
number of cases, Kepler et al. (1991) succeeded at a measurement for
the DAV star G117–B15A (Tab. 2.3) using the Whole Earth Telescope
which allows for nearly 24 h a day coverage of a given object.
A variety of model calculations give Ṗ = 2−5 × 10−15 s s−1 , some-
what smaller than the measured value, i.e. the star appears to cool
Table 2.3. Properties of the DAV star G117–B15A.
Surface temperature Teff 13,200 K

Luminosity log(L/L! ) −2.3
Bolometric brightness Mbol 10.49 mag
Mass M (0.49 ± 0.03) M!
Pulsation period P (215.197,387 ± 0.000,001) s
Period change Ṗ (12 ± 4) × 10−15 s s−1
P/Ṗ (0.57 ± 0.17) Gyr
54 Chapter 2
faster than predicted by these models (Kepler et al. 1991). It should

be noted, however, that part of the observed period change can be at-
tributed to a Doppler shift if G117–B15A is a physical binary with its
proper-motion companion G117–B15B.
Isern, Hernanz, and Garcı́a-Berro (1992) speculated that an addi-
tional cooling agent may be operating and notably that this star is
cooled by axion emission. For a fiducial WD model (M = 0.5 M! ,
internal temperature T = 1.8×107 K) the cooling time scale is found
to be T /Ṫ = 1.0 Gyr while P/Ṗ ≈ 1.4 Gyr. Therefore, Isern et al.
found that Lx = 0.5−2.6 Lγ was needed to account for the observed Ṗ ,
although other models needed little or no axion cooling. For their
fiducial model the required axion cooling yielded a coupling constant
α26 = 0.2−0.8. This interpretation is speculative, of course, but appar-
ently not in conflict with any other constraints on the electron coupling
of pseudoscalars (Sect. 3.6.1).
The most conservative interpretation of the Ṗ measurement of the
star G117–B15A is that it agrees with theoretical calculations within
the observational and model uncertainties. Therefore, it provides in-
dependent evidence that the WD cooling speed is known to within a
factor of O(1) so that any novel cooling agent is constrained to be less
efficient than O(Lγ ).
2.3 Neutron Stars

2.3.1 Late-Time Cooling
Neutron stars are born when the degenerate iron core of an evolved
massive star becomes unstable and collapses to nuclear densities, an
implosion which is partly reflected at the core bounce and leads to a
type II supernova explosion (Sect. 2.1.8). These events and the first few
seconds of neutron star cooling are discussed more fully in Chapter 11.
For a few seconds the star emits most of its binding energy in the form
of MeV neutrinos which were observed from SN 1987A. Afterward, the
temperature at the neutrino sphere (the analogue of the photosphere in
ordinary stars) has dropped so much that the detectors are no longer
sensitive to the neutrino flux although the star continues to cool by
surface neutrino emission.
After 10−100 yr the internal temperature has dropped to about
9
10 K ≈ 100 keV where the neutron star becomes entirely transpar-
ent to neutrinos and continues to cool by neutrino volume emission.
After about 105 yr it reaches an inner temperature of about 2×108 K,
a point at which photon emission from the surface becomes the dom-
inant form of cooling. In Fig. 2.14 the central temperature, surface
temperature, neutrino luminosity, and photon luminosity are shown as
functions of age according to a numerical calculation of Nomoto and
Tsuruta (1987).
Fig. 2.14. Cooling of a neutron star with baryon mass 1.4 M! (gravitational
mass 1.3 M! ) according to Nomoto and Tsuruta (1987). The solid line is
for an equation of state of intermediate stiffness (model FP), the dotted
line for a stiff model (PS). All “nonstandard” effects were ignored such as
nucleon superfluidity, a meson condensate, magnetic fields, and so forth.
Temperatures and luminosities are local, ignoring the gravitational redshift.
56 Chapter 2
2.3.2 X-Ray Observations
Because the surface temperature of neutron stars is in the keV regime

they can be observed only by x-ray satellites such as the Einstein Obser-
vatory which was launched in 1979, and more recently by EXOSAT and
Table 2.4. Selected x-ray observations of SN remnants and pulsars.a
Age [yr] Dist.

∞
Remnant (in 1995) [kpc] Pulsarb Teff [106 K] c d
3C58 (814) (2.6) — 2.2 ± 0.2 E 1

Crab 941 1.7–2 0531+21 < 1.6 R 2
(r, o, x, γ)
RCW 103 1500 ± 500 2 — 2.15 ± 0.15 E 1
< 1.2 R 3
MSH15–52 1850 ± 250 4.2 (r, x) detected E 1
Vela X ∼ 12,000 0.5 0833–45 0.95 ± 0.15 E 1
(r, o, x, γ) 1.6 ± 0.2 R 4
Monogem ∼ 110,000 (0.5) 0656+14 0.90 ± 0.04 R 5
Ring (?) (r, x) (0.30 ± 0.05)
— ∼ 340,000 0.15–0.4 Geminga 0.52 ± 0.10 R 6
(x, γ)
— ∼ 540,000 0.5–1.5 1055–52 0.65 ± 0.15 X 7
(r, x, γ) 0.75 ± 0.06 R 8
a Adapted from Tsuruta (1986) and updated.
b Pulsed radiation: r = radio, o = optical, x = x-rays, γ = γ-rays.
c E = Einstein, R = ROSAT, X = EXOSAT observation.
d References:
1. See the review by Tsuruta (1986) for references to the original literature.
2. Becker and Aschenbach (1995).
3. Becker et al. (1992).
4. Ögelman, Finley, and Zimmermann (1993).
5. Thompson et al. (1991). Finley, Ögelman, and Kiziloğlu (1992).
6. Halpern and Ruderman (1993).
7. Brinkmann and Ögelman (1987).
8. Ögelman and Finley (1993). Anderson et al. (1993).
ROSAT. X-rays have been observed from a number of compact sources

in supernova remnants and from several isolated pulsars. However,
with the limited spectral resolution of these instruments it is difficult
to extract the actual thermal surface emission because there can be
significant nonthermal x-ray emission from the magnetosphere.
There are three candidates in the 103 yr age category (3C58, the
Crab pulsar, RCW 103) from which x-rays have been observed by the
Einstein Observatory (Tab. 2.4). However, ROSAT did not detect a
compact source in RCW 103; only an upper flux limit has been re-
ported (Tab. 2.4). The Crab pulsar is x-ray bright mostly from non-
thermal emission so that ROSAT could only establish an upper limit on
its surface temperature. The remaining Einstein source 3C58 probably
should also be interpreted as an upper limit on thermal surface emis-
sion. These upper limits are in agreement with the standard cooling
curves of Nomoto and Tsuruta (1987) shown in Fig. 2.15.
The Einstein data point for the Vela pulsar at an age of about
4
10 yr is somewhat low, a fact which has given rise to the speculation
that “exotic” cooling effects may be operating such as neutrino emission
by virtue of a meson condensate. However, a blackbody spectral fit to
the ROSAT observations yields a much higher temperature (Tab. 2.4,
Fig. 2.15. Surface temperature (observed at infinity) of a neutron star with

baryon mass 1.4 M! (gravitational mass 1.3 M! ) according to Nomoto and
Tsuruta (1987). The solid line is for an equation of state of intermediate
stiffness (model FP), the dotted line for a stiff (PS), and the dashed line for
a soft model (BPS). All “nonstandard” effects were ignored. The measure-
ments of Tab. 2.4 are also shown.
58 Chapter 2
open circle in Fig. 2.15). Such a high temperature is not compatible

with the total x-ray luminosity unless the radius of the neutron star is
very small (3−4 km). Either way, it may be premature to reach definite
conclusions regarding neutron star cooling on the basis of Vela.
At still larger ages, ROSAT measurements of the surface temper-
ature of the pulsars PSR 0656+14, PSR 1055–52, and Geminga have
been reported which lie close to the theoretical standard cooling curves.
However, the inferred temperature values depend sensitively on the as-
sumed circumstellar atmospheric models which can modify the spec-
trum and thus lead to an erroneous temperature assignment. Still, the
old isolated pulsars give one a first realistic observational handle at the
issue of neutron star cooling.
2.3.3 Nonstandard Cooling and Heating Effects

The so-called “standard” neutron star cooling scenario should be called
a “reference” or “minimal” scenario because there are many effects that
will alter the cooling history; no doubt at least some of them will be in
operation in some or all neutron stars. For reviews see Tsuruta (1986,
1992), for recent numerical cooling curves including various nonstan-
dard effects see Umeda, Tsuruta, and Nomoto (1994).
The occurrence of nucleon superfluidity slows the neutrino emission
by the URCA process (Sect. 4.8). However, the cooling curves including
superfluidity (Nomoto and Tsuruta 1987) do not seem to differ signif-
icantly from the reference curves at ages above a few hundred years
unless extreme assumptions are made.
When nucleon superfluidity is important, νν bremsstrahlung emis-
sion by electrons in the crust dominates. However, electron band-
structure effects may suppress this process, and the crust mass may be
smaller than previously thought (Pethick and Thorsson 1994). There-
fore, the cooling may be slowed even further. Slowed cooling may also
occur by a number of heating effects (accretion, polar cap heating, vor-
tex creep, and others), although such effects become important only
for relatively old neutron stars (t > 4
∼ 10 yr). Of course, heating effects
related to accretion will not be important in isolated pulsars which thus
are preferred laboratories to study neutron star cooling.
The cooling is accelerated if the equation of state provides enough
protons to allow for the direct URCA process (Sect. 4.8). In this case
the surface temperature drops catastrophically at an age of about 100 yr
(Page and Applegate 1992; Lattimer et al. 1994) until superfluidity sets
in which essentially stops neutrino cooling from the core. The temper-
ature then stays almost constant for a long time until photon cooling
from the surface begins to dominate. In this scenario the cooling curve
depends sensitively on the “on switch” set by the occurrence of the di-
rect URCA process anywhere in the star, and by the “off switch” from
superfluidity. As the occurrence of these effects depends on fine points
of the equation of state as well as on the density and thus the stellar
mass there may not be a universal cooling curve for all neutron stars.
Another effect which would accelerate cooling is the occurrence of
a meson condensate (Sect. 4.9.1) because of the increased efficiency of
neutrino emission. Again, this effect depends sensitively on the equa-
tion of state and thus on the density and the stellar mass. The most
recent numerical study of neutron-star cooling with a pion condensate
was performed by Umeda, Nomoto, and Tsuruta (1994).
2.3.4 Cooling by Particle Emission

The emission of novel particles would also accelerate the cooling of
neutron stars. Iwamoto (1984) considered axion emission e + (Z, A) →
(Z, A) + e + a in the crust and found unacceptably fast cooling unless
α# <
∼ 10
−25
where α# is the axionic fine-structure constant (Chapter 3).
However, this result is quite uncertain, notably in view of the above
Pethick and Thorsson (1994) band-structure suppression of the brems-
strahlung rate. Moreover, in view of the white-dwarf and globular-
cluster bounds of α# < ∼ 10
−26
it appears that crust cooling by axions is
not important in neutron stars.
In the interior of neutron stars, axions can be emitted by the neu-
tron bremsstrahlung process nn → nna (Sect. 4.2). With a numerical
implementation of Iwamoto’s (1984) bremsstrahlung rate Tsuruta and
Nomoto (1987) found a limit gan < ∼ 10
−10
for the axion-neutron Yukawa
coupling, based on a comparison with the 103 yr old sources. In their
calculation the effect of superfluidity apparently was not included which
would diminish the bound. Conversely, including protons in a regime
where neither protons nor neutrons are superfluid would increase the
emission rate (Sect. 4.2.6). Of course, the more recent ROSAT results
suggest that thermal surface emission has not been observed for any of
the 103 yr old sources anyway.
Most recently, Iwamoto et al. (1995) considered the plasma decay
process γ → νν in the crust under the assumption of a large neutrino
magnetic dipole moment. They found that for µν of order 10−10 µB a
significant effect would obtain, but that a value as large as 5×10−7 µB
would be consistent with current data. In view of the globular-cluster
60 Chapter 2
limit of µB < −12

∼ 3×10 µB (Sect. 6.5.6), I interpret these results to mean
that a neutrino magnetic dipole moment leaves neutron-star cooling
unaffected—one less nonstandard effect to worry about.
The rough agreement between the reference cooling curves and the
data points in Fig. 2.15, notably for the old pulsars, suggests that “non-
standard” effects cannot be much more efficient than standard neutrino
cooling unless all of the old isolated pulsar surface temperatures have
been incorrectly assigned—not a likely scenario. Therefore, it is clear
that these and future observations of cooling neutron stars will be piv-
otal as laboratories to study novel phenomena such as the occurrence
of nonstandard phases of nuclear matter.
However, at the present time it is not clear if the emission of weakly
interacting particles such as axions could still have an interesting impact
on neutron star cooling. At the present time it appears easier to make
the reverse statement that such cooling effects likely are not important
in view of the restrictive limits on the interaction strength of axions or
nonstandard neutrinos set by other astrophysical objects. Therefore,
at present it appears that for the more narrowly defined purposes of
particle physics the role of old neutron stars as laboratories is less useful
than had been thought in the early works discussed above. It also
appears that novel weakly interacting particles usually would have a
more dramatic impact on the first few seconds of Kelvin-Helmholtz
cooling of a protoneutron star (Chapter 11) than they do on the cooling
of old pulsars.
2.4 Globular-Cluster Stars
2.4.1 Observables in the Color-Magnitude Diagram
Globular clusters are gravitationally bound associations of typically

106 stars (Fig. 2.1); the clusters themselves (about 150 in our galaxy)
form an approximately spherical galactic halo. The metallicity is in the
range Z = 10−4 −10−2 ; it is usually the same for all stars in one cluster.
The low metallicity is one indicator for their great age—like isolated
halo stars they belong to the Population II which formed early from
a relatively “uncontaminated” hydrogen-helium mixture left over from
the big bang of the universe. Stars found in the galactic disk belong
to the later Population I which continue to form even today from the
interstellar gas. Clusters of disk stars are usually less populous and less
tightly bound—the so-called “open clusters.”
Fig. 2.16. Observables in the color-magnitude diagram of a typical globular

cluster (here M15 after Sandage 1986). Depending on the metallicity, the
red-giant “bump” can also appear below the HB brightness.
In the color-magnitude diagram of a globular cluster (Fig. 2.3) the

stars arrange themselves in a characteristic pattern which is schemati-
cally shown in Fig. 2.16. Each branch corresponds to a certain evo-
lutionary phase as discussed in Sect. 2.1. Typically, all stars in a
given cluster are coeval and have a fixed chemical composition; they
differ only in their mass. After formation, they all began their evo-
lution on the zero-age main-sequence as in Fig. 2.2. The more mas-
sive stars evolve faster, become red giants and explode as supernovae
(M > <
∼ 8 M! ), or become white dwarfs (M ∼ 8 M! ). Recently, many
pulsars, the remnants of type II supernovae and thus of massive stars,
have indeed been found in globular clusters. The upper main sequence
(MS) is depleted of stars down to a limiting value below which they
have not had time to complete hydrogen burning. Therefore, the stel-
lar mass corresponding to the MS turnoff (TO) is a precise measure
of the cluster age. A typical value is 0.7−0.9 M! , depending on the
precise age and metallicity.
The color-magnitude diagram of a globular cluster represents an
“isochrone” of a stellar population. It shows the locus of coeval stars
with different initial masses. It is to be distinguished from the evolu-
tionary track of a single star (Fig. 2.6) which shows a star of a fixed
mass at different ages. However, the evolution beyond the MS is very
62 Chapter 2
fast. Therefore, the stars along the red-giant branch (RGB), horizon-
tal branch (HB), and asymptotic giant branch (AGB) in a globular
cluster have almost identical initial masses whence for these phases a
single-star track is practically identical with an isochrone. On the other
hand, the TO region requires the construction of detailed theoretical
isochrones to compare theory and observations and thus to determine
the ages of globular clusters.
In order to associate a certain stellar mass with the TO in a cluster
one needs to know the absolute brightness of the stars at the TO, i.e.
one needs to know the precise distance. All else being equal, the inferred
age varies with the TO luminosity as ∂ log(age)/∂ log LTO = −0.85 or
∂ log(age)/∂VTO = 0.34 (Iben and Renzini 1984). Therefore, a 0.1 mag
error in VTO leads to an 8% uncertainty in the inferred cluster age.
Put another way, because L ∝ (distance)2 a 10% uncertainty in cluster
distances leads to an 18% age uncertainty. This is the main problem
with the age determination of globular clusters.
A particularly useful method to measure the distance is to use
RR Lyrae stars as standard candles. As discussed in Sect. 2.1, their
luminosity is determined almost entirely by their core mass (apart from
a dependence on chemical composition), which in turn is fixed by he-
lium ignition on the RGB which, again, depends only on the chemical
composition and not on the red-giant envelope mass. Therefore, the
brightness of the HB is nearly independent of stellar mass. Conse-
TO
quently, the brightness difference ∆VHB between the HB and the TO
is a distance-independent measure of the TO mass and thus of the clus-
ter age. Moreover, because the color of RR Lyrae stars coincides with
that of the TO region it is not necessary to convert from the measured
brightness with a certain filter (e.g. visual brightness V ) to a bolomet-
ric brightness, i.e. there is no need for a bolometric correction (BC).
Also, RR Lyrae stars are bright and easily identified because of their
TO
pulsations. Therefore, ∆VHB is one of the most important observables
in the color-magnitude diagram of globular clusters (Iben and Renzini
1984; Sandage 1986).
TO
As an example the recent ∆VHB determinations of Buonanno, Corsi,
and Fusi Pecci (1989) in 19 globular cluster are shown in Fig. 2.17 as
a function of metallicity; the logarithmic metallicity measure [Fe/H] is
defined in Eq. (2.15). The best linear fit is
TO
∆VHB = (3.54 ± 0.13) − (0.008 ± 0.078) [Fe/H], (2.13)
so that the HB brightness varies with metallicity almost exactly as the
TO brightness. The measured points are in agreement with a Gaussian
Fig. 2.17. Brightness difference between main-sequence turnoff (TO) and

horizontal branch (HB) in 19 galactic globular clusters according to Buon-
anno, Corsi, and Fusi Pecci (1989). The shaded band indicates the 1σ sta-
tistical error of the best fit Eq. (2.13).
distribution about the mean. This result illustrates the level of precision
TO
that presently can be achieved at determining ∆VHB .
In principle, the color of the TO also specifies the location on the
MS and thus the TO mass and cluster age. In practice, the color of
a star is theoretically less well determined than its luminosity because
it depends on the treatment of the photosphere and on the surface
area and thus on the radius which is partly fixed by the treatment
of convection. Still, the TO color is a useful measure for the relative
ages between clusters of identical metallicities. Notably, the distance-
independent color difference ∆(B − V ) between the TO and the base
of the RGB (Fig. 2.16) has been used to establish an age difference of
about (3 ± 1) Gyr between the clusters NGC 288 and 362 (VandenBerg,
Bolte, and Stetson 1990; Sarajedini and Demarque 1990).
In order to constrain the operation of a novel energy-loss mechanism
the brightness of the RGB tip is particularly useful because particle
emission (neutrinos, axions) from a red-giant core delays helium igni-
tion. This delay allows the stars to develop a more massive core and
thus to turn brighter before they become HB stars. Again, the inferred
luminosity of the brightest star on the RGB depends on the distance
whence the most useful observable is the distance-independent bright-
ness difference between the RGB tip and the HB. However, because
the color of RR Lyrae stars and red giants is very different one must
64 Chapter 2
tip
convert to an absolute bolometric brightness difference ∆MHB rather
tip
than using the visual brightness difference ∆VHB .
Fig. 2.18. Evolutionary speed on the RGB for a model with M = 0.8 M! ,
metallicity Z = 10−4 , and initial helium abundance Y = 0.240 (Raffelt and
Weiss 1992). The dashed line is the tangent near the bright end.
Fig. 2.19. Luminosity function of the RGB of the globular cluster NGC 2808
which has [Fe/H] = −1.37 (Fusi Pecci et al. 1990).
A very important measure of the speed of evolution of a single star

along the RGB, HB, and AGB is the relative number of stars found on
these branches in a given cluster. Because for these advanced evolution-
ary phases a single-star track is essentially an isochrone, these number
ratios R give us directly the relative amounts of time spent on these
branches (“R-method”). For example, a novel energy-loss mechanism
that operates mostly in the nondegenerate core of an HB star would
shorten the helium-burning lifetime. This possibility is constrained by
the relative number of stars found on the HB and RGB where the RGB
in this context refers to that part which is brighter than the HB.
When a star ascends the RGB, its hydrogen-burning shell encoun-
ters at some point a discontinuity in the composition profile left behind
by a previous deep penetration of the envelope convection into the
region of varying hydrogen content caused by nuclear burning. The as-
cent is briefly interrupted and the star stays at a fixed luminosity for a
brief period of time. Afterward, the hydrogen shell works itself through
a constant composition profile which was prepared by the convective
envelope. Therefore, at a brightness near the HB one expects to find
a “bump” in the number of stars on the RGB, i.e. in the distribution
∂N/∂Mbol which essentially corresponds to dt/dMbol for a single-star
evolutionary track (Fig. 2.18). The bump was recently identified in a
number of clusters; for a particularly beautiful example see Fig. 2.19.
It has been suggested to use it as a standard candle to calibrate the
RR Lyrae brightness-metallicity relation (Fusi Pecci et al. 1990).
An important observable is the absolute brightness of RR Lyrae
stars which are not members of globular clusters. No reason is known
why these field stars for a given metallicity should be any different from
those found in a cluster and so brightness determinations of nearby
RR Lyrae stars provide important information about their luminosity
calibration.
2.4.2 Theoretical Relations
In order to test the standard stellar-evolution picture against the ob-

servables introduced in Sect. 2.4.1 they need to be related to stellar
properties such as mass and chemical composition. In the past, exten-
sive grids of stellar evolutionary sequences have been calculated and
have been used to derive analytic approximations for the connection
between various stellar parameters; as a canonical standard I use the
evolutionary HB and RG sequences of Sweigart and Gross (1976, 1978).
66 Chapter 2
a) Composition Parameters
The chemical composition is characterized by the helium content and

the metallicity. The convective envelope on the RGB at some point
reaches down into the region of a variable composition profile caused
by nuclear burning, causing a certain amount of processed material
(helium) to be dredged up. Therefore, the helium mass fraction Yenv
of the envelope is slightly larger than the value Y at formation. The
amount of dredge-up is approximately given by (Sweigart, Renzini, and
Tornambé 1987)
∆Ys ≡ Yenv − Y ≈ 0.0136 + 0.0055 Z13 , (2.14)
where the metallicity parameter Z13 is defined in Eq. (2.16) below.

However, because the exact amount of convective dredge-up is some-
what model dependent, and because measurements of 12 C/13 C ratios
in metal-poor field red giants seem to indicate that the dredge-up of
processed material is more efficient than predicted by standard calcula-
tions (Sneden, Pilachewski, and VandenBerg 1986) it is best to use Yenv
rather than Y as an independent parameter to characterize red giants
and HB stars.10
Gravitational settling of helium throughout the MS evolution has
the opposite effect of reducing the envelope abundance relative to the
initial homogeneous value. In recent evolutionary sequences which were
calculated to estimate the effect of helium diffusion on the inferred
globular-cluster ages a decrease between ∆Ys = −0.009 and −0.015 was
found (Proffitt and Michaud 1991; Chaboyer et al. 1992). Therefore,
the effect of MS gravitational settling and that of RG convective dredge-
up appear to cancel each other more or less so that Yenv appears to be
much closer to the initial value Y than had been thought previously.
The metallicity is usually characterized by the mass fraction Z of
elements heavier than helium. Because iron is most important for the
opacities one often uses the abundance of iron relative to hydrogen
as a metallicity measure. It is characterized by the quantity [Fe/H]
which is the logarithmic abundance of iron over hydrogen relative to
the solar value. If the solar metallicity is taken to be Z! = 0.02 so that
10
In their calculation of RG sequences, Sweigart and Gross (1978) used the symbol
Y to denote the MS helium abundance; the envelope abundance near the helium
flash can be inferred from their tabulation of ∆Ys values for each sequence. In
their 1976 study of HB sequences, they used the symbol Y to denote the envelope
abundance which is here consistently called Yenv .
log Z! = −1.7 one finds

log Z = [Fe/H] + log Z! = [Fe/H] − 1.7. (2.15)
Because globular-cluster metallicities cover a range of Z = 10−4 to 10−2
a typical average value is log Z = −3 or [Fe/H] = −1.3. The primordial
helium abundance is thought to be about 23%. Therefore, it is useful
to employ the “reduced” composition parameters
Y23 ≡ Yenv − 0.23,
Z13 ≡ log Z + 3 = [Fe/H] + 1.3, (2.16)
which are zero for a typical globular cluster.
b) Core Mass at Helium Ignition

One of the most important quantities to be affected by a novel energy-
loss mechanism is the core mass at the helium flash because helium
ignition is an extremely sensitive function of the temperature. Based
on the Sweigart and Gross (1978) models Raffelt (1990b) has derived
the analytic approximation
Mc = 0.500 − 0.22 Y23 − 0.011 Z13 − 0.021 M7 + δMc , (2.17)
where all stellar masses are understood in units of the solar mass M! .
Here, Mc is the core mass at helium ignition and M7 ≡ M − 0.7 is
the “reduced total mass.” I have increased Mc by 0.004 relative to
the original calculation to account for the corrected plasma neutrino
emission rate (Haft, Raffelt, and Weiss 1994). Within about 0.003 M!
Raffelt and Weiss (1992) found the same expression (when corrected for
the plasma rates) except for a slightly shallower metallicity dependence.
Recently, Sweigart (1994) has reviewed the core-mass calculations
at the helium flash. All workers seem to agree within a few 10−3 M!
except for Mazzitelli (1989) who found core masses larger by some
0.020 M! . Sweigart (1994) claims that this disagreement cannot be
attributed to the algorithm adopted to accomplish the “shell shifting”
of the numerical grid which represents the star on a computer.
Because of substantial mass loss on the RGB the meaning of the
total mass M in this equation is not entirely obvious. If there were
enough time to relax to equilibrium one would think that it is the in-
stantaneous mass at the helium flash. Indeed, Raffelt and Weiss (1992)
found in an evolutionary sequence with mass loss that the end mass de-
termined Mc . However, in this calculation the mass loss was stopped
68 Chapter 2
sometime before helium ignition and so this finding is not surprising.

Castellani and Castellani (1993) studied RG sequences with mass loss
in more detail and found the surprising result that the core mass at
helium ignition was determined by the initial stellar mass while the en-
velope structure followed the instantaneous envelope mass. Apparently,
in these calculations the core retained memory of a previous configu-
ration. The total amount of mass loss on the RGB may be of order
0.2 M! , causing a maximum discrepancy between the two scenarios of
about 0.005 M! in the expected Mc .
In Eq. (2.17) a deviation δMc was explicitly included which rep-
resents nonstandard changes of Mc . The core-mass increase δMc is
the main quantitity to be constrained by observations. It should be
thought of as a function of the parameters which govern the physics
which causes the delay of helium ignition such as coupling constants of
particles which contribute to the energy loss, or a more benign param-
eter such as the angular frequency of core rotation.
c) Brightness at Helium Ignition

Next, the brightness at helium ignition is needed, identical with the
brightness at the tip of the RGB. In the Sweigart and Gross (1978)
calculations, both the luminosity and the core mass at the helium flash
are functions of Z, Yenv , and M. For the present purposes, however, the
core mass at helium ignition must be viewed as another free parameter
which is controlled, for example, by the amount of energy loss by novel
particle emission. In order to determine how the luminosity at helium
ignition varies with Mc if all other parameters are held fixed Raffelt
(1990b) considered ∂ log L/∂Mc for a grid of Sweigart and Gross tracks
near the flash. An interpolation yields
log Ltip = 3.328 + 0.68 Y23 + 0.129 Z13 + 0.007 M7 + 4.7 Mc ,
(2.18)
with Ltip the luminosity at the RGB tip in units of L! . Ignoring the
dependence on the total mass here and in Eq. (2.17) one finds for the
absolute bolometric brightness of the RGB tip11
Mtip = −3.58 + 0.89 Y23 − 0.19 Z13 − 11.8 δMc , (2.19)
slightly different from the results of Raffelt (1990b) who used the mass
of RR Lyrae stars for the total mass in Eq. (2.18).
11
Recall that the absolute bolometric brightness is given by M = 4.74 − 2.5 log L
for L in units of L" , M in magnitudes.
d) Brightness of RR Lyrae Stars

Theoretically, details of the evolution of stars on the HB are difficult to
account for. Notably, they may move in and out of the RR Lyrae insta-
bility strip so that the stars found there cannot be trivially associated
with a specific age after the beginning of helium burning. Therefore,
it is easiest to use the absolute bolometric brightness of zero-age HB
stars with a mass chosen such that they fall into the RR Lyrae strip
(Buzzoni et al. 1983; Raffelt 1990b)
MRR = 0.66 − 3.5 Y23 + 0.16 Z13 − ∆RR − 7.3 δMc , (2.20)
where ∆RR is an unknown amount of deviation between real RR Lyrae

stars and the zero-age HB models of Sweigart and Gross (1976) that
served to derive this relation.
It is expected that ∆RR is a positive number of order 0.1 mag, i.e.
on average RR Lyrae stars are thought to be somewhat brighter than
zero-age HB star models. This conclusion is supported by Sandage’s
(1990a) investigation of the vertical height of the HB by means of the
pulsational properties of RR Lyrae stars. Sandage found an intrinsic
width between 0.2 mag for the most metal-poor and about 0.4 mag for
the most metal-rich clusters, i.e. an average deviation between 0.1 and
0.2 mag between zero-age and average HB stars.
Lee, Demarque, and Zinn (1990) have constructed synthetic HBs
for a range of metallicities and helium content on the basis of new
evolutionary sequences. For a MS helium content of 0.20, which in their
calculation amounts to Yenv = 0.22, they found (see also Lee 1990)
MRR = 0.70 + 0.22 Z13 . (2.21)
Comparing this with Eq. (2.20) at Y23 = −0.01 and δMc = 0 one
finds ∆RR ≈ 0. This is not in contradiction with the brightening of
RR Lyrae stars relative to zero age, it only means that there is a slight
offset relative to the analytic representation Eq. (2.20) derived from the
Sweigart and Gross (1976) calculations. ∆RR shall always refer to the
brightness difference of real RR Lyrae stars relative to Eq. (2.20), it
does not refer to an offset relative to real zero-age HB stars.
e) Brightness Difference between HB and RGB Tip

The main observable to constrain a deviation from the standard core
mass at the helium flash is the brightness difference between the HB
70 Chapter 2
(at the RR Lyrae strip) and the RGB tip for which one finds with
Eqs. (2.19) and (2.20)
tip
∆MHB ≡ MRR − Mtip
= 4.24 − 4.4 Y23 + 0.35 Z13 − ∆RR + 4.5 δMc (2.22)
which is defined such that it is a positive number.
f ) Ratio of HB/RGB Stars

The relative duration of the HB vs. RGB phase is given by the number
ratio of the stars on these branches where the RGB is defined as that
part which is brighter than the HB. The HB lifetime is found to be
(Buzzoni et al. 1983; Raffelt 1990b)
log(tHB /yr) = 8.01 + 0.37 Y23 + 0.06 Z13 − 2.9 δMc . (2.23)
The RGB lifetime cannot be expressed easily in terms of a simple linear

formula because of the RGB bump discussed earlier. A simple approx-
imation to the lifetime ratio R = tHB /tRGB is (Buzzoni et al. 1983;
Raffelt 1990b)
log R = 0.105 + 2.29 Y23 + 0.029 Z13 + 0.33 ∆RR − 0.70 δMc .
(2.24)
This quantity is particularly sensitive to the helium content of the stars

and almost independent of the core mass at helium ignition.
2.4.3 Observational Results

a) Brightness at the RGB Tip
The brightness at the tip of the RGB can be estimated by the brightest
red giant in a given globular cluster. A homogeneous set of observations
of the brightest RGs in 33 globular clusters are those of Cohen, Frogel,
and Persson (1978), Da Costa, Frogel, and Cohen (1981), Cohen and
Frogel (1982) and Frogel, Persson, and Cohen (1981, 1983). According
to Frogel, Cohen, and Persson (1983) only in 26 of the 33 clusters
the brightest giant was likely observed; for those cases the bolometric
brightness difference between the brightest RGs and the HB are shown
as a function of metallicity in Fig. 2.20. Also shown is a linear fit
tip
∆MHB = 4.06 + 0.38 Z13 . The observational errors are thought to be
less than about 0.05 mag.
Fig. 2.20. Bolometric brightness difference between the horizontal branch

and the brightest red giant in 26 globular clusters from the observations
referenced in the text.
However, because there are relatively few stars near the tip on the
RGB (on average about 10 mag−1 in the observed clusters), the bright-
est RG is on average about 0.1 mag below the actual RGB tip. Raffelt
(1990b) estimated the richness of the RGB near the tip for each cluster
on the basis of the first few brightest stars provided by the observa-
tions and thus estimated the expected brightness difference between
the brightest RG and the tip. This yields a linear regression
tip
∆MHB = (4.19 ± 0.03) + (0.41 ± 0.06) Z13 , (2.25)
about 0.13 mag brighter than the fit shown in Fig. 2.20. The slope of
Eq. (2.25) agrees very well with the theoretical expectation Eq. (2.22),
provided that ∆RR does not introduce a large modification.
More recent observations are those of Da Costa and Armandroff
(1990) who also found excellent agreement between the theoretical slope
of the brightness of the RGB tip luminosity as a function of metallicity.
Because the coefficient of the metallicity dependence agrees well
with the predicted value one may restrict a further comparison be-
tip
tween theory and observation to ∆MHB at a given metallicity for which
it is best to use the average value [Fe/H] = −1.48 or Z13 = −0.18
of the globular clusters used in Fig. 2.20. Inserting these values into
Eqs. (2.22) and (2.25) and adding the errors quadratically one finds
4.4 Y23 + ∆RR − 4.5 δMc = 0.06 ± 0.03. (2.26)
If ∆RR = 0.2 mag this result implies that the envelope helium abun-
72 Chapter 2
dance, which presumably is very close to the primordial value, has to

be 0.20 to achieve perfect agreement, or else the core mass at helium
ignition has to be 0.030 M! larger than given by the standard the-
ory. Alternatively, the zero-age HB models may represent the bright-
ness of RR Lyrae stars better than anticipated so that ∆RR ≈ 0
in which case there is perfect agreement with the standard values
Yenv ≈ Yprimordial ≈ 0.23 and δMc = 0.
b) Ratio of HB/RGB Stars

Buzzoni et al. (1983) have determined the number of stars on the HB
(including RR Lyrae stars),12 NHB , and on the RGB brighter than the
HB, NRGB , in 15 globular clusters. The resulting values for the number
ratio R = NHB /NRGB is shown in Fig. 2.21 as a function of metallicity.
The individual errors are found by assuming standard deviations of
N 1/2 for the number counts and adding the errors of NHB and NRGB
quadratically (Raffelt 1990b). The data are fit by a linear regression
log R = (0.162 ± 0.016) + (0.065 ± 0.032) Z13 , (2.27)
also shown in Fig. 2.21.
Fig. 2.21. Number ratio of HB/RGB stars for 15 globular clusters according
to Buzzoni et al. (1983).
Comparing this result with the theoretical prediction of Eq. (2.24)

one finds that the metallicity dependence essentially agrees within the
12
Buzzoni et al. (1983) call this quantity NHB+RR .
stated uncertainty; it is very shallow anyway. Therefore, one compares

at the average metallicity of the 15 clusters, [Fe/H] = −1.54 or Z13 =
−0.24, adds the errors quadratically, and finds
Y23 + 0.14 ∆RR − 0.31 δMc = 0.021 ± 0.008. (2.28)
With ∆RR ≈ 0.2 mag and δMc ≈ 0 this confirms a primordial helium
abundance of around 23%.
c) RR Lyrae Absolute Brightness

The RR Lyrae absolute brightness as well as the precise variation of
MRR with metallicity is the single most discussed issue about globular-
cluster color-magnitude diagrams because the distance and thus the
age determination depends critically on this quantity. For example,
TO
when using ∆VHB measurements like the ones shown in Fig. 2.17, the
inferred relative ages of globular clusters depend crucially on the slope
a of MRR = a [Fe/H] + b.
One possibility to determine a is the use of the pulsation frequencies
of these variable stars. The result is about 0.35, almost twice as large
as that obtained by theoretical zero-age HB models or by synthetic
HBs, an issue known as the Sandage period shift effect (Sandage 1990
and references therein; Iben and Renzini 1984; Renzini and Fusi Pecci
1988). While this issue is crucial for a relative age determination of
globular clusters, it is of relatively minor importance for the present
discussion where the zero point b for an intermediate metallicity is the
most crucial quantity.
Probably the most direct determination of MRR is to use nearby
field RR Lyrae stars for which, in principle, a distance determination by
parallax measurements is possible. Barnes and Hawley (1986) applied
the method of statistical parallaxes to a sample of 142 stars and found
a mean absolute visual brightness of (0.68 ± 0.14) mag. Assuming a
bolometric correction for RR Lyrae stars of −0.06 mag this leads to
(MRR ) = (0.62 ± 0.14) mag. (2.29)
The average metallicity of this sample is probably [Fe/H] ≈ −1.4.

The Baade-Wesselink method applied to a total of 25 field RR Lyrae
stars, and using a bolometric correction of −0.06, leads to
MRR = 0.72 + 0.19 Z13 (2.30)
(Sandage and Cacciari 1990, and references therein).

74 Chapter 2
Most recently, the brightness of RR Lyrae stars in the Large Mag-

ellanic Cloud was measured; it has a distance which is thought to be
well determined by other methods. Walker (1992) found
MRR = 0.48 + 0.15 Z13 , (2.31)
if the same bolometric correction −0.06 is assumed.

In summary, a reasonably conservative estimate of the absolute
RR Lyrae bolometric brightness is
MRR = (0.60 ± 0.15) + 0.17 Z13 . (2.32)
A comparison with Eq. (2.20) yields
3.5 Y23 + ∆RR + 7.3 δMc = 0.06 ± 0.15. (2.33)
This is in good agreement with the standard values Yenv = 0.23, δMc =
0, and ∆RR ≈ 0.1 mag.
2.4.4 Interpretation of the Observational Results

In order to interpret the observational results it is first assumed that
there is no anomalous core-mass increase. However, considering the
uncertainties entering the calculation of Mc such as the precise value of
the relevant total stellar mass, uncertainties in the electron conductive
opacities, etc., it appears that a plausible range of uncertainty is δMc =
±0.010 M! even in the absence of any novel phenomena. Adopting this
uncertainty and adding it quadratically to the previous uncertainties,
the three observables from Eq. (2.26), (2.28), and (2.33) yield
tip
Yenv = (0.244 ± 0.012) − 0.23 ∆RR from ∆MHB ,
Yenv = (0.251 ± 0.009) − 0.14 ∆RR from R,
Yenv = (0.247 ± 0.048) − 0.29 ∆RR from MRR . (2.34)
The primordial helium abundance likely is in the range 22−24%; the

envelope abundance in globular clusters is probably slightly larger, de-
pending on details of gravitational settling on the MS and convective
dredge-up on the RGB. Therefore, with ∆RR between 0 and 0.2 mag
these results are perfectly consistent.
In order to constrain Yenv and δMc simultaneously it is assumed
that ∆RR = 0.1 ± 0.1. Adding this error quadratically in Eq. (2.26),
(2.28), and (2.33) one finds

tip
Yenv − 1.0 δMc = 0.239 ± 0.024 from ∆MHB ,
Yenv − 0.3 δMc = 0.237 ± 0.016 from R,
Yenv + 2.1 δMc = 0.247 ± 0.043 from MRR . (2.35)
Bands of allowed values for Yenv and δMc are shown in Fig. 2.22.
From Fig. 2.22 one reads off that approximately
Yenv = 0.24 ± 0.02 and |δMc | < 0.025 M! . (2.36)
Therefore, the core mass at the helium flash is tightly constrained to
lie within 5% of its standard value.
This result allows one to constrain the operation of a novel energy-
loss mechanism in the core of a red giant, i.e. at densities of around
106 g cm−3 and at a temperature of about 108 K. In some cases, how-
ever, an energy-loss mechanism is suppressed by degeneracy effects in
a RG core, while it may be fairly efficient in the helium-burning core
of an HB star (ρ ≈ 104 g cm−3 , T ≈ 108 K) which is essentially nonde-
generate. In this case δMc ≈ 0 while the energy loss from the HB core
Fig. 2.22. Allowed values for the envelope helium abundance of evolved
globular-cluster stars and of an anomalous core-mass excess at helium igni-
tip
tion. The limits were derived from the observed brightness difference ∆MHB
between the HB and the RGB tip, from the “R-method” (number counts
on the HB vs. RGB), and from the brightness determination of nearby field
RR Lyrae stars (MRR ) by statistical parallaxes and the Baade-Wesselink
method as well as the brightness of RR Lyrae stars in the LMC.
76 Chapter 2
may be substantial and thus would shorten the helium-burning lifetime

as the nuclear fuel would be consumed faster.
The HB/RGB number ratio R indicates that the acceleration of the
HB evolution must not be too extreme. From Eq. (2.24) together with
the observational result Eq. (2.27) one finds
δ log R = 0.01 ± 0.06, (2.37)
if one adopts Yenv = 0.24 ± 0.02, ∆RR = 0.1 ± 0.1, and δMc =
±0.010 M! . Therefore, R cannot be smaller by much more than about
10% of its standard value. Put another way, the helium-burning life-
time of low-mass stars is determined within about 10% from the ratio
of HB/RGB stars in globular clusters.
With lesser statistical significance this result is corroborated by
number counts of clump giants in open clusters which consist of more
recently formed stars in the galactic disk (Population I). Clump giants
in open clusters correspond to HB stars in globular clusters; instead
of forming a horizontal branch they are concentrated in a clump near
the base of the RGB. Cannon (1970) compared the number of clump
giants in the open cluster M67 with the number of stars per luminos-
ity interval near the MS turnoff and found tHe ≈ 1.5×108 yr, with a
large statistical uncertainty, however, because there were only 5 clump
giants. (Open clusters tend to be much less populous than globular
ones.) Tinsley and Gunn (1976) derived tHe = (1.27 ± 0.29) × 108 yr
from low-mass giants of the old galactic disk population. These results
are in full agreement with Eq. (2.23).
2.4.5 An Alternate Analysis

The above discussion of the globular-cluster limits is a somewhat up-
dated version of my own previous work (Raffelt 1990b). Very recently,
Catelan, de Freitas Pacheco, and Horvath (1995) have provided an in-
dependent new and extended analysis. While they closely follow the
line of reasoning of Raffelt (1990b) they have changed numerous de-
tip
tails. Of the 26 globular clusters which enter the ∆MHB argument,
and of the 15 clusters which enter the R-method, they have discarded
several with an extreme HB morphology. For the absolute brightness
of RR Lyrae stars they have employed Walker’s (1992) values which
depend on a precise knowledge of the LMC distance. For the bright-
ness difference between zero-age HB stars and RR Lyrae stars they use
∆RR = 0.31 + 0.10 [Fe/H] = 0.18 + 0.10 Z13 .
These authors have introduced a fourth observable, the so-called

mass to light ratio A of RR Lyrae stars. This observable has also been
used by Castellani and Degl’Innocenti (1993) to constrain a possible
core-mass excess. It amounts to a determination of the core mass of an
RR Lyrae star on the basis of its luminosity and pulsation period.
Going through similar steps as in the previous sections, Catelan,
de Freitas Pacheco, and Horvath (1995) then found the results
tip
Yenv − 0.9 δMc = 0.207 ± 0.039 from ∆MHB ,
Yenv − 0.3 δMc = 0.218 ± 0.018 from R,
Yenv + 2.1 δMc = 0.250 ± 0.049 from MRR ,
Yenv + 1.5 δMc = 0.238 ± 0.008 from A. (2.38)
Bands of allowed values for Yenv and δMc are shown in Fig. 2.23 which
is analogous to Fig. 2.22.
From this analysis one infers a best-fit value for the envelope he-
lium abundance which is somewhat low. The primordial abundance
probably is not lower than 22%. The only possibility to reduce the
envelope abundance from this level is by gravitational settling which
is counteracted by convective dredge-up, and perhaps by other effects
that might eject helium into the envelope from the core. Therefore,
Fig. 2.23. Allowed values for the envelope helium abundance of evolved
globular-cluster stars and of an anomalous core-mass excess at helium igni-
tion according to the analysis of Catelan, de Freitas Pacheco, and Horvath
(1995). For comparison see Fig. 2.22.
78 Chapter 2
Yenv > 0.22 is probably a conservative estimate. This takes us back

to the conclusion that δMc ≈ 0 within at least 0.025 M! . Again, the
core mass at helium ignition is found to agree with the standard result
to better than 5%.
2.4.6 Systematic Uncertainties

The globular-cluster argument yields some of the most restrictive lim-
its on novel modes of energy loss. Therefore, it is important to under-
stand some of the systematic uncertainties that enter the nominal limit
δMc < ∼ 0.025 M! .
Core rotation has often been quoted as an effect to change Mc .
However, as it would actually delay helium ignition it cannot be invoked
to compensate for anomalous cooling effects. In addition, if fast core
rotation were an important effect one would expect it to vary from star
tip
to star, causing a random broadening of the distribution of ∆MHB ,
tip
an effect not indicated by the observations. The scatter of ∆MHB is
completely within observational errors and within the scatter caused
by the effect that the brightest red giant is not exactly at the tip of the
RGB in a given cluster. Further discussions of the rotational impact
on Mc are found in Catelan, de Freitas Pacheco, and Horvath (1995).
The uncertainty of the conductive opacities are relatively large as
stressed by Catelan, de Freitas Pacheco, and Horvath (1995). However,
conceivable modifications of Mc do not seem to exceed the 0.010 M!
level even with extreme assumptions.
One of the main theoretical weaknesses of the helium-ignition ar-
gument is that the helium flash has never been properly calculated.
Because helium ignites off-center one expects that convection plays a
major role in the process of heating the entire core and its expansion.
One may worry that in the process of the flash, parts of the core are
ejected into the stellar envelope, reducing its post-flash size. However,
if significant amounts of helium were ejected, the inferred Yenv would be
changed dramatically. Even the ejection of 0.010 M! of helium would
increase Yenv by 0.03 if one assumes an envelope mass of 0.3 M! and
thus would brighten RR Lyrae stars by 0.12 mag. Within the stated
limit of δMc < 0.025 M! , mass ejection from the core is a dramatic
effect that would be hard to hide in the data.
One significant systematic uncertainty arises from the relative abun-
dance of metals among each other which is usually fixed by the solar
mixture (Ross and Aller 1976). The assumption of a Ross-Aller mix-
ture for metal-poor systems like globular-cluster stars has been called
into question in recent years (e.g. Wheeler, Sneden, and Truran 1989).
It is thought that in these systems the “α elements” (mostly oxygen)
are enhanced relative to iron. RG sequences calculated by Raffelt and
Weiss (1992) with somewhat extreme α enhancements indicated that
the red-giant core mass and luminosity at the helium flash are only
moderately changed. Most of the change that did occur was due to
the reduction of Fe at constant Z because of the α enhancement. Put
another way, if Z13 = [Fe/H] + 1.3 is used as the defining equation for
the above “reduced metallicity” the effect of α enhancements appear
to be rather minimal.
Catelan, de Freitas Pacheco, and Horvath (1995) have taken the
point of view that for enhanced α elements one should rescale the
metallicity according to a recipe given by Chieffi, Straniero, and Salaris
(1991). Put another way, the metallicity parameter of Eq. (2.16) that
was used in the previous sections should be redefined as
Z13 ≡ [Fe/H] + 1.3 + log(0.579 f + 0.421). (2.39)
Here, f is the enhancement factor of the abundance of α elements
relative to the solar value. For f = 3 one finds that Z13 must be
offset by +0.33. Catelan, de Freitas Pacheco, and Horvath (1995) went
through their analysis with this modification, causing the allowed bands
of Fig. 2.23 to be slightly shifted relative to each other. However, the
overall change is small, well within the stated upper limit on δMc .
In summary, no effect has been discussed in the literature that would
cause the predicted core mass at helium ignition to deviate from its
standard value beyond the adopted limit of ±0.025 M! . Therefore, any
new energy-loss mechanism that would cause a significantly larger core-
mass excess would have to be compensated by a hitherto unidentified
other novel effect.
2.5 Particle Bounds from Globular-Cluster Stars

2.5.1 Helium-Burning Lifetime
A particularly simple argument to constrain the properties of novel
particles arises from the observed duration of helium burning of low-
mass stars, i.e. form the lifetime of stars on the horizontal branch (HB).
In Sect. 2.4.4 it was argued that the number ratio R of stars on the HB
vs. RGB in globular clusters agreed with standard predictions to within
10%. Therefore, the helium-burning lifetime tHe agrees with standard
predictions to within this limit. A less significant confirmation arises
80 Chapter 2
from the number of clump giants in open clusters and from the old
galactic disk population.
In Sect. 1.3.1 it was shown that the main impact of a nonstandard
energy-loss rate on a star is an acceleration of the nuclear fuel con-
sumption while the overall stellar structure remains nearly unchanged.
The temperature dependence of the helium-burning (triple-α reaction)
energy generation rate #3α ∝ T 40 is much steeper than the case of hy-
drogen burning discussed in Sect. 1.3.1 and so the adjustment of the
stellar structure is even more negligible. With L3α the standard helium-
burning luminosity of the core of an HB star and Lx the nonstandard
energy-loss rate integrated over the core, tHe will be reduced by an ap-
proximate factor L3α /(Lx + L3α ). Demanding a reduction by less than
10% translates into a requirement Lx < ∼ 0.1 L3α . Because this con-
straint is relatively tight one may compute both Lx and L3α from an
unperturbed model. If the same novel cooling mechanism has delayed
the helium flash and has thus led to an increased core mass only helps
to accelerate the HB evolution. Therefore, it is conservative to ignore
a possible core-mass increase.
The standard value for L3α is around 20 L! ; see Fig. 2.4 for the
properties of a typical HB star. Because the core mass is about 0.5 M!
the core-averaged energy generation rate is (#3α ) ≈ 80 erg g−1 s−1 . Then
a nonstandard energy-loss rate is constrained by
(#x ) < −1 −1
∼ 10 erg g s . (2.40)
Previously, this limit had been stated as 100 erg g−1 s−1 , overly con-
servative because it was not based on the observed HB/RGB number
ratios in globular clusters. However, in practice Eq. (2.40) does not
improve the constraints on a novel energy-loss rate by a factor of 10
because the appropriate average density and temperature are somewhat
below the canonical values of ρ = 104 g cm−3 and T = 108 K.
For a simple estimate the energy-loss rate may be calculated for
average conditions of the core. Typically, #x will depend on some small
power of the density ρ, and a somewhat larger power of the temperature
T . For the HB star model of Fig. 2.4 the core-averaged values (ρn ) and
(T n ) are shown in Fig. 2.24 as a function of n. The dependence on n
is relatively mild so that the final result is not sensitive to fine points
of the averaging procedure.
In order to test the analytic criterion Eq. (2.40) in a concrete exam-
ple consider axion losses by the Primakoff process. The energy-loss rate
will be derived in Sect. 5.2.1. It is found to be proportional to T 7 /ρ
and to a coupling constant g10 = gaγ /10−10 GeV−1 . For a typical HB
Fig. 2.24. Average values of ρn and T n for the HB star model of Fig. 2.4
where ρ4 = ρ/104 g cm−3 and T8 = T /108 K.
core I find (T87 /ρ4 ) ≈ 0.3 which leads to # ≈ g10

2
30 erg g−1 s−1 . Thus,
for g10 = 1 one concludes that the helium-burning lifetime should be
reduced by a factor 80/(80 + 30) = 0.7.
This result may be compared with numerical evolution sequences
for 1.3 M! stars with an initial helium abundance of 25% and a metal-
licity of Z = 0.02 (Raffelt and Dearborn 1987). The helium-burning
lifetime was found to be 1.2×108 yr which was modified to 0.7×108 yr
for g10 = 1. This means that the axion losses on the HB led to a tHe
reduction by a factor 0.6, in good agreement with the analytic estimate.
This comparison corroborates that the analytic criterion represents the
claimed impact of a novel energy loss on the helium-burning lifetime
with a reasonable precision.
Equation (2.40) may now be applied to many different cases. An
overview over the most salient results is given in Tab. 2.5 where the
original references are given, and the sections of this book are indi-
cated where a more detailed discussion can be found. Apart from the
listed examples, the argument has also been applied to low-mass super-
symmetric particles (Fukugita and Sakai 1982; Bouquet and Vayonakis
1982; Ellis and Olive 1983). However, it is now thought that super-
symmetric particles, if they exist, are probably not light enough to be
produced in stars whence the interest in this case has waned.
82 Chapter 2
Table 2.5. Constraints on low-mass particles from the observed duration

of helium burning in globular-cluster stars. Details are discussed in the
indicated sections of this book.
Particle property Dominant process Constrainta [Ref.]

Sect.
Yukawa coupling of Bremsstrahlung gS < 1.3×10−14 [1]
scalar (vector) boson e + He → He + e + φ gV < 0.9×10−14
4 4
3.5.5
φ to electrons
Yukawa coupling Compton gS < 4.3×10−11 [1]

to baryons γ + 4 He → 4 He + φ gV < 3.0×10−11 3.6.2
Photoproduction of Photoproduction σ< −50

∼ 3×10 cm
2
[2]
X ◦ boson γ + 4 He → 4 He + X ◦
Yukawa coupling of Compton g < 4.5×10−13 [3]

pseudoscalar boson γ+e→e+a 3.2.6
a to electrons
Yukawa coupling of Compton g < 3.2×10−13 [4]

“paraphoton” γ # γ + e → e + γ# 3.2.6
γγ coupling of Primakoff gaγ < 0.6×10−10 [5]

pseudoscalar boson γ + 4 He → 4 He + a GeV−1 5.2.5
Neutrino dipole Plasmon decay µeff < 1×10−11 µB [6]

moment γ → νν 6.5.6
a Asderived in this book; may differ from the quoted references.
References:
1. Grifols and Massó (1986); Grifols, Massó, and Peris (1989).
2. van der Velde (1989); Raffelt (1988b).
3. Dicus et al. (1978, 1980); Georgi, Glashow, and Nussinov (1981);
Barroso and Branco (1982); Fukugita, Watamura, and Yoshimura (1982a,b);
Pantziris and Kang (1986); Raffelt (1986a).
4. Hoffmann (1987).
5. Raffelt (1986a); Raffelt and Dearborn (1987).
6. Sutherland et al. (1976); Fukugita and Yazaki (1987);
Raffelt and Dearborn (1988); Raffelt, Dearborn, and Silk (1989).
2.5.2 Helium Ignition

Another powerful constraint arises from the agreement between the
predicted and observationally inferred core mass at the helium flash
(Sect. 2.4.4). An energy-loss mechanism which is efficient in a degen-
erate medium (ρ ≈ 106 g cm−3 , T ≈ 108 K) can delay helium ignition.
To establish the core-mass increase as a function of nonstandard par-
ticle parameters one needs to evolve red giants numerically to the he-
lium flash.
However, for a simple analytic estimate one observes that the core
mass of a red giant grows by hydrogen shell burning. Because it is
a degenerate configuration its radius shrinks and so the core releases
a large amount of gravitational binding energy which amounts to an
average energy source (#grav ). If a novel energy-loss rate (#x ) is of the
same order then helium ignition will be delayed.
In order to estimate (#grav ) one may treat the red-giant core as
a low-mass white dwarf. Its total energy, i.e. gravitational potential
energy plus kinetic energy of the degenerate electrons, is found to be
(Chandrasekhar 1939)
3 GN M 2
E=− . (2.41)
7 R
The radius of a low-mass white dwarf (nonrelativistic electrons!) may
be expressed as R = R∗ (M! /M)1/3 with R∗ = 8800 km so that
$ %1/3
Ė G N M! M Ṁ
(#grav ) = − = . (2.42)
M R∗ M! M!
From the numerical sequences of Sweigart and Gross (1978) one finds
that near the helium flash M ≈ 0.5 M! and Ṁ ≈ 0.8×10−15 M! s−1
so that (#grav ) ≈ 100 erg g−1 s−1 . Therefore, one must require (#x ) *
100 erg g−1 s−1 in order to prevent the helium flash from being delayed.
In order to sharpen this criterion one may use results from Sweigart
and Gross (1978) and Raffelt and Weiss (1992) who studied numerically
the delay of the helium flash by varying the standard neutrino losses
with a numerical factor Fν where Fν = 1 represents the standard case.
The results are shown in Fig. 2.25. Note that for Fν < 1 the standard
neutrino losses are decreased so that helium ignites earlier, causing
δMc < 0. It is also interesting that for Fν = 0 helium naturally
ignites at the center of the core while for Fν > 1 the ignition point
moves further and further toward the edge (Fig. 2.26). This behavior is
84 Chapter 2
Fig. 2.25. Change of the red-giant core mass at helium ignition, Mc , as a

function of a factor Fν which multiplies the standard neutrino energy-loss
rate. Triangles: Metallicity Z = 10−3 (Sweigart and Gross 1978). Squares:
Z = 10−4 (Raffelt and Weiss 1992).
Fig. 2.26. Variation of the red-giant core mass at helium ignition, Mc , as a

function of Fν as in Fig. 2.25 for Z = 10−4 , Y = 0.22, and M = 0.8 M! .
Also shown is the mass coordinate Mig where helium ignites (Raffelt and
Weiss 1992).
understood from the relatively steep density dependence of the plasma

neutrino emission rate which is most efficient at the center.
In order to prevent the core mass from exceeding its standard value
by more than 5% (δMc < 0.025 M! ) one must require Fν < ∼ 3 or
(#x ) <
∼ 2(# ν ). According to the Sweigart and Gross (1978) red-giant se-
quences the neutrino luminosity of the core at helium ignition is approx-
imately 1 L! so that (#ν ) ≈ 4 erg g−1 s−1 . Therefore, an approximate
analytic criterion to constrain a nonstandard energy loss is
#x < −1 −1
∼ 10 erg g s , (2.43)
where #x is to be evaluated in a helium plasma at the average density of
the core of about 2×105 g cm−3 (the central density is about 106 g cm−3 )
and at the almost constant temperature of T = 108 K. The standard
neutrino plasma emission rate evaluated at 2×105 g cm−3 and 108 K
is 3 erg g−1 s−1 , in good agreement with the above average neutrino
luminosity of the Sweigart and Gross models.
Table 2.6. Increase of the core mass at helium ignition because of the emis-
sion of pseudoscalars (Raffelt and Weiss 1995).
α# [10−26 ] δMc [M! ]

0.0 0.000
0.5 0.022
1.0 0.036
2.0 0.056
A simple application of Eq. (2.43) is the case of bremsstrahlung

emission of pseudoscalars e + 4 He → 4 He + e + a. For a degenerate
medium the emission rate is given in Eq. (3.33); for T = 108 K it
is approximately α# 2×1027 erg g−1 s−1 with the “axion fine-structure
constant” α# . Then, Eq. (2.43) yields
α# < −26
∼ 0.5×10 . (2.44)
The same case was treated numerically by Raffelt and Weiss (1995)
who implemented the energy-loss rate Eq. (3.33) with varying values
of α# in several red-giant evolutionary sequences.13 They found the
13
In a previous numerical treatment by Dearborn, Schramm, and Steigman (1986)
the correct emission rate had not yet been available and so they overestimated the
energy-loss rate by as much as a factor of 10 at the center of a red-giant core.
86 Chapter 2
Fig. 2.27. Increase of the core mass of a red giant at helium ignition due
to the emission of pseudoscalars according to Tab. 2.6 (Raffelt and Weiss
1995).
core-mass increases given in Tab. 2.6 and shown in Fig. 2.27. The re-
quirement that the core not exceed its standard value by more than 5%
reproduces the analytic bound. This example nicely corroborates the
surprising precision of the simple criterion Eq. (2.43).
Another important case where a detailed numerical study is avail-
able is the emission of neutrinos by the plasma process γ → νν when
they have nonstandard magnetic dipole moments µν . The emission
rates are derived in Sect. 6.5.5 and the simple criterion Eq. (2.43) is
applied in Sect. 6.5.6. It yields a limit µ12 < −12
∼ 2 where µ12 = µν /10 µB
with the Bohr magneton µB = e/2me . The numerical variation of the
core mass with µν is shown in Fig. 2.28 according to Raffelt and Weiss
(1992). An analytic approximation is
& '
3/2
δMc = 0.025 M! (µ212 + 1)1/2 − 1 − 0.17 µ12 . (2.45)
The requirement δMc <

∼ 0.025 M! then translates into
µν < −12
∼ 3×10 µB , (2.46)
a result which, again, is almost identical with the analytic treatment,

supporting the power of the simple criterion stated in Eq. (2.43).
Fig. 2.28. Increase of the core mass of a red giant at helium ignition as a
function of an assumed neutrino dipole moment according to Raffelt and
Weiss (1992) with a total stellar mass 0.8 M! and an initial helium abun-
dance of Y = 0.22 or 0.24 (the core-mass increase is found to be the same).
The triangles refer to the metallicity Z = 10−3 , the squares to 10−4 . The
open circles are the corresponding results of Castellani and Degl’Innocenti
(1992) with the same stellar mass, Y = 0.23, and Z = 2×10−4 . The solid
line is the analytic fit Eq. (2.45).
2.6 Summary
What is the bottom line after studying the impact of an anomalous
energy-loss rate on a variety of stellar-evolution phases? While the Sun
remains an interesting object from a pedagogical point of view, its main
use is that of a distant particle source for terrestrial experimentation,
notably to study neutrino properties (Chapters 10 and 12). Supernovae
and their collapsed cores (newborn neutron stars) remain very impor-
tant; they will be discussed in Chapters 11−13. At the present time
the cooling of old neutron stars appears to be a useful laboratory to
study conventional phenomena (Does the direct URCA process occur?
Are there meson condensates or other exotic phases? What is the role
of magnetic fields?). The emission of weakly interacting particles other
than standard neutrinos appears to play a lesser role in view of other
astrophysical limits on their interaction strength.
Then, apart from the supernova arguments to be discussed later,
the most useful and reliable observables to constrain the operation of a
88 Chapter 2
nonstandard energy-loss mechanism in stars are the white-dwarf lumi-

nosity function, the helium-burning lifetime of horizontal-branch stars,
and the nondelay of helium ignition in low-mass red giants as observed
by the brightness of the tip of the red-giant branch in globular clus-
ters. It was possible to condense the latter two arguments into two
exceedingly simple criteria, namely that an anomalous energy-loss rate
in the cores of HB stars as well as in red-giant cores before helium
ignition must not exceed about 10 erg g−1 s−1 . The emission rate is to
be calculated at the pertinent plasma conditions, i.e. at an approx-
imate temperature of 108 K = 8.6 keV, an electron concentration of
Ye = 0.5, and an average density of about 0.6×104 g cm−3 (HB stars)
or 2×105 g cm−3 (red giants). The former case corresponds to roughly
nondegenerate conditions, the latter case to degenerate ones so that it
depends on the density dependence of the emission rates which of these
cases will yield a more restrictive limit.
A red-giant core is essentially a 0.5 M! helium white dwarf. The
observed “real” white dwarfs have typical masses of about 0.6 M! ;
they are thought to consist mostly of carbon and oxygen. Both the
helium-ignition argument and the white-dwarf luminosity function al-
low one to constrain a novel energy-loss mechanism roughly on the level
of standard neutrino emission. Therefore, it is no surprise that bounds
derived from both arguments tend to be very similar.
There may be other objects or phenomena in the universe that mea-
sure novel particle-physics hypotheses even more sensitively than the
cases discussed here. They still need to make their way into the particle
astrophysics literature.
Chapter 3
Particles Interacting with

Electrons and Baryons
The stellar energy-loss argument is applied to weakly interacting par-

ticles which couple to electrons and baryons. The emission from a nor-
mal stellar plasma can proceed by a variety of reactions, for example
the Compton process γe− → e− χ where χ stands for a single particle
(axion, paraphoton, etc.) or a neutrino pair νν. Other examples of
practical interest are electron bremsstrahlung e− (Z, A) → (Z, A)e− χ
or e− e− → e− e− χ, electron free-bound transitions, and pair annihila-
tion e− e+ → γχ or e− e+ → νν. The corresponding energy-loss rates
of stellar plasmas are studied for temperatures and densities which are
of interest for stellar-evolution calculations. For particles coupled to
baryons some of the same processes apply in a normal plasma if one sub-
stitutes a proton or a helium nucleus for the electron. Reactions specific
to a nuclear medium are deferred to Chapter 4. In addition to draining
stars of energy, scalar or vector bosons would mediate long-range forces.
Leptonic long-range forces would be screened by the cosmic neutrino
background. Thermal graviton emission from stars is mentioned.
3.1 Introduction
If weakly interacting particles couple directly to electrons, they can
be produced thermally in stellar plasmas without nuclear processes.
For neutrinos, a direct electron coupling was first contemplated after
the universal V −A theory for their interactions had been proposed in
1958. It was realized immediately that such a coupling would allow
for thermal pair production by bremsstrahlung e− (Z, A) → (Z, A)e− νν
89
90 Chapter 3
(Pontecorvo 1959; Gandel’man and Pinaev 1959) or by photoproduc-

tion γe− → e− νν (Ritus 1961; Chiu and Stabler 1961). The set of
processes important for normal stars was completed by Adams, Rud-
erman, and Woo (1963) who discovered the plasma process γ → νν.
Neutrino emission by these reactions is now a standard aspect of stellar-
evolution theory.
If other weakly interacting particles were to exist which couple di-
rectly to electrons they could essentially play the same role and thus
add to the energy loss of stars. The main speculation to be followed
up in this chapter is the possible existence of weakly interacting bosons
that would couple to electrons. Among standard particles the only low-
mass bosons are photons and probably gravitons. The former dominate
the radiative energy transfer in stars, the latter are so weakly interact-
ing that their thermal emission is negligible (Sect. 3.7). Why worry
about others?
Such a motivation arises from several sources. Low-mass bosons
could mediate long-range forces between electrically neutral bodies for
which gravity is the only standard interaction. It is an interesting
end in itself to set the best possible bounds on possible other forces
which might arise from the exchange of novel scalar or vector bosons.
Their existence seemed indicated for some time in the context of the
“fifth-force” episode alluded to in Sect. 3.6.3 below. It is also possible
that baryon or lepton number play the role of physical charges similar
to the electric charge, and that a new gauge interaction is associated
with them. The baryonic or leptonic photons arising from this hypoth-
esis are intriguing candidates for weakly interacting low-mass bosons
(Sect. 3.6.4). It will turn out, however, that typically massless bosons
which mediate long-range forces are best constrained by experiments
which test the equivalence principle of general relativity. Put another
way, to a high degree of accuracy gravity is found to be the only long-
range interaction between neutral bodies.
Long-range leptonic forces can be screened by the cosmic neutrino
background. In this case the stellar energy-loss argument remains of
importance to limit their possible strength (Sect. 3.6.4).
The remaining category of interesting new bosons are those which
couple to the spin of fermions and thus do not mediate a long-range
force between unpolarized bodies. In the simplest case their CP-con-
serving coupling would be of a pseudoscalar nature. Such particles arise
naturally as Nambu-Goldstone bosons in scenarios where a global chiral
U(1) symmetry is spontaneously broken at some large energy scale. The
most widely discussed example is the Peccei-Quinn symmetry that was
Particles Interacting with Electrons and Baryons 91
proposed as an explanation of CP conservation in strong interactions.

It leads to the prediction of axions which will be discussed in some detail
in Chapter 14. The most restrictive limits on their coupling strength
arise from the stellar energy-loss argument, and it cannot be excluded
that in fact they play an important role in the evolution of some stars
(Sect. 2.2.5). No wonder that axions have played a primary role in
studies concerning the impact of new weakly interacting particles on
stellar evolution.
The main focus of this chapter is an application of the stellar energy-
loss argument to weakly interacting bosons which couple to electrons by
a variety of interaction structures. The relevant processes are entirely
analogous to those which emit neutrino pairs except for the plasma pro-
cess which requires a two-body final state. Therefore, I will presently
study photo and bremsstrahlung production of weakly interacting par-
ticles, including standard neutrinos.
One may consider the same processes with protons substituted for
electrons. For neutrino emission this variation is of no interest because
the rate is much smaller. It is significant for low-mass bosons which
couple only to baryons.
The energy-loss argument will be systematically applied, yielding
restrictive limits on the possible Yukawa and gauge couplings of novel
bosons to electrons and baryons. For very low-mass scalar or vector
bosons these limits are discussed in the context of those arising from
the absence of novel long-range interactions.
3.2 Compton Process
3.2.1 Vector Bosons
The simplest process for the emission of weakly interacting particles

from the hot and dense interior of a star is the Compton process where
a photon from the heat bath interacts with an electron and is thus con-
verted into a neutrino pair, an axion, or some other boson (Fig. 3.1).
These processes are analogous to the usual Compton scattering of pho-
tons. Therefore, I begin with this well-known case which is based on the
standard electron-photon interaction Lint = ie ψ e γµ ψe Aµ with the elec-
tron charge e, the electron Dirac field ψe , and the photon field A. With
the fine structure constant α = e2 /4π ≈ 1/137, the electron mass me ,
and σ0 ≡ πα2 /m2e the total Compton cross section is (e.g. Itzykson and
92 Chapter 3
Fig. 3.1. Compton processes for photon scattering as well as for axion and
neutrino pair production (“photoneutrino process”). In each case there is
another amplitude with the vertices interchanged.
Zuber 1980)
! "
16 ŝ + 1 2 (ŝ2 − 6 ŝ − 3)
σ = σ0 + + ln(ŝ) . (3.1)
(ŝ − 1)2 ŝ2 (ŝ − 1)3
√
Here, ŝ ≡ s/m2e with s the CM (center of mass) energy.14 This cross
section is shown in Fig. 3.2 as a function of the CM photon energy ω.
For ω & me the CM frame is the electron rest frame, ŝ → 1, and one
recovers the Thomson cross section σ = 83 σ0 . For ω ' me one has
ŝ ' 1 and so σ = (σ0 /ŝ) [2 ln(ŝ) + 1] = (πα2 /ω 2 ) [ln(2ω) + 14 ] because
in this limit s = (2ω)2 .
The standard Compton cross section can also be used to study the
photoproduction of novel low-mass vector particles which couple to
electrons in the same way as photons except that the fine-structure
constant must be replaced by the new coupling α" . Then
σ0 ≡ παα" /m2e with α" ≡ g 2 /4π, (3.2)
a definition that pertains to all bosons which couple to electrons with
a dimensionless Yukawa or gauge coupling g.
If novel vector bosons such as “paraphotons” exist (Holdom 1986)
they likely couple to electrons by virtue of an induced magnetic moment
rather than by a tree-level gauge coupling (Hoffmann 1987),
Lint = (g/4me ) ψ e σµν ψe F µν , (3.3)
where g is a dimensionless effective coupling constant and F the para-
14
The square of the CM energy is s = (P + K)2 with P = (E, p) and K = (ω, k)
the four-vectors of the initial-state electron and photon, respectively. The CM frame
is defined by p = −k so that s = (E + ω)2 = [(m2e + ω 2 )1/2 + ω]2 with ω the initial-
state photon energy in the CM frame. In the frame where the target electron is at
rest (p = 0) one finds s = 2ωme + m2e where now ω is the photon energy in the
electron frame. Thus, with ω the photon energy in the respective frames,
# √
1
ω 2 (ŝ − 1)/ ŝ in the CM frame,
= 1
me (ŝ − 1) in the electron rest frame.
2
photon field tensor. In the nonrelativistic limit this yields the same to-
tal cross section as the interaction with pseudoscalars Eq. (3.6) below.
This is seen if one compares the matrix elements between two electron
states i and f for the two cases. For paraphotons (momentum k, polar-
ization vector !) it is g (f |eik·r (k × !) · σ|i* while for pseudoscalars it is
g (f |eik·r k · σ|i*. After an angular average the two expressions are the
same. Of course, one must account for the two paraphoton polarization
states by an extra factor of 2.
Fig. 3.2. Total cross section for the Compton process with a final-state
vector, scalar, or pseudoscalar boson according to Eqs. (3.1), (3.5), and
(3.9), respectively, with σ0 defined in Eq. (3.2) and ω the CM initial photon
energy.
3.2.2 Scalars
Grifols and Massó (1986) studied the stellar emission of scalars φ which
couple according to
Lint = g ψ e ψe φ. (3.4)
Integrating their differential cross section I find
! "
−16 1 − 3ŝ (ŝ + 3)2
σ = σ0 + + ln(ŝ) (3.5)
(ŝ − 1)2 2ŝ2 (ŝ − 1)3
shown in Fig. 3.2 (dashed line). For small and large photon energies
this is half the cross section for massless vector bosons which have
two polarization degrees of freedom. For intermediate energies the two
results are not related by a simple factor.
94 Chapter 3
3.2.3 Pseudoscalars
Next, turn to the photoproduction of low-mass pseudoscalars φ which
couple to electrons by the interaction
Lint = (1/2f ) ψ e γµ γ5 ψe ∂ µ φ or Lint = −ig ψ e γ5 ψe φ, (3.6)
where f is an energy scale and g a dimensionless coupling constant.
Both interaction laws yield the same Compton cross section with the
identification g = me /f (see the discussion in Sect. 14.2.3).
If a pseudoscalar (frequency ω, wavevector k) is emitted in a transi-
tion between the nonrelativistic electron states |i* and |f *, the matrix
element is
g 1 $ %% ik·r % &
%
Mpseudoscalar = √ f %e σ · k %i . (3.7)
2me 2ω
This is to be compared with the corresponding matrix element for pho-
ton transitions,
−e 1 $ %% ik·r % &
Mphoton = √ f %e [2! · p + σ · (k × !)]%%i , (3.8)
2me 2ω
with the photon polarization vector ! and the electron momentum oper-
ator p. Therefore, transitions involving pseudoscalars closely compare
with photonic M1 transitions, a fact that was used to scale nuclear or
atomic photon transition rates to those involving axions (Donelly et al.
1978; Dimopoulos, Starkman, and Lynn 1986a,b).
The relativistic Compton cross section for massive pseudoscalars
was first worked out by Mikaelian (1978). The most general discussion
of the matrix element was provided by Brodsky et al. (1986) and by
Chanda, Nieves, and Pal (1988) who also included an effective photon
mass relevant for a stellar plasma. For the present purpose it is enough
to consider massless photons and pseudoscalars. With σ0 as defined in
Eq. (3.2) one finds
' (
ln(ŝ) 3ŝ − 1
σ = σ0 − , (3.9)
ŝ − 1 2ŝ2
shown in Fig. 3.2 (dotted line).
For ω ' me one finds σ = (παα" /2ω 2 ) [ln(2ω) − 34 ], similar to the
scalar case. For ω & me , however,
σ = 43 σ0 (ω/me )2 , (3.10)
so that the cross section is suppressed at low energies. This reduction
is related to the M1 nature of the transition.
3.2.4 Neutrino Pairs

The photoneutrino process was first studied by Ritus (1961) and by
Chiu and Stabler (1961). The effective neutral-current Hamiltonian is
GF
Hint = √ ψ e γµ (CV − CA γ5 )ψe ψ ν γ µ (1 − γ5 )ψν , (3.11)
2
where GF is the Fermi constant, and the dimensionless couplings CV
and CA are given in Appendix B. Of course, in the early sixties neutral
currents were not known—one used Fierz-transformed charged currents
which gave CV = CA = 1. For general CV ’s and CA ’s the cross section
was first calculated by Dicus (1972) who found15 (Fig. 3.3)
) *
σ = σ0 (CV2 + CA2 ) σ̂+ − (CV2 − CA2 ) σ̂− ,
49 5 (13 ŝ − 7) 15 − 117 ŝ − 55 ŝ3

σ̂+ = + +
12 (ŝ − 1)2 12 ŝ2
25 − 28 ŝ − 27 ŝ2 − 2 ŝ3 + 2 ŝ4
+ ln(ŝ),
(ŝ − 1)3
120 ŝ 8 ŝ − 1 12 (2 + 2 ŝ + 5 ŝ2 + ŝ3 )
σ̂− = −39 + − + ln(ŝ),
(ŝ − 1)2 ŝ2 (ŝ − 1)3
α G2F m2e
σ0 = . (3.12)
9 (4π)2
In the nonrelativistic (NR) limit this is (CM photon energy ω)
σNR = σ0 6
35
(CV2 + 5CA2 ) (ŝ − 1)4
= σ0 96
35
(CV2 + 5CA2 ) (ω/me )4 . (3.13)
In the extreme relativistic (ER) limit it is
σER = σ0 (CV2 + CA2 ) 2ŝ [ ln(ŝ) − 55

24
]
= σ0 (CV2 + CA2 ) 16 (ω/me )2 [ ln(2ω/me ) − 55
48
], (3.14)
where σ− does not contribute.

15
With CV2 = CA 2
= 1 this result agrees with that of Ritus (1961) while Chiu and
Stabler (1961) appear to have an extra factor 2ŝ/(ŝ + 1). In the nonrelativistic limit
with ŝ → 1 this deviation makes no difference while in the extreme relativistic limit
their result is a factor of 2 larger than that of Ritus (1961) and Dicus (1972).
96 Chapter 3
Fig. 3.3. Dimensionless total cross sections σ̂+ and σ̂− for the photoneutrino
process γe− → e− νν according to Eq. (3.12) with ω the initial photon energy
in the CM frame.
3.2.5 Energy-Loss Rates

The Compton-type processes are typically important when the elec-
trons are nondegenerate (otherwise bremsstrahlung dominates) and
nonrelativistic (otherwise e+ e− annihilation dominates). In these lim-
its one may use the cross sections without Pauli blocking corrections.
Because the recoil of the target electron is neglected, the energy ω of
a photon impinging on an electron is identical with the energy carried
away by the new boson or neutrino pair. Therefore, the energy-loss
rate per unit volume is a simple integral over the initial-state photon
phase space, weighted with their Bose-Einstein occupation numbers,
+
2 d3 k σ ω
Q = ne , (3.15)
(2π)3 eω/T − 1
where ne is the number density of electrons, T the temperature, and

the factor 2 is for two photon polarization states.
In this expression the photon “plasma mass” ωP has been neglected.
If ωP >
∼ 3T , corresponding to a typical thermal photon energy, the pho-
ton dispersion relation would have to be included properly in both the
phase-space integration and in the cross section calculation. However,
these are insignificant fine points for the cases to be studied below.
In the nonrelativistic limit it is easy to estimate a suppression fac-
tor Fdeg by electron degeneracy. If recoil effects can be neglected, the
initial- and final-state electrons have the same momentum, reducing the
calculation to an average of the Pauli blocking factor over all electrons
, -
1 + 2 d3 p 1 1
Fdeg = 1 − (E−µ)/T , (3.16)
ne (2π)3 e(E−µ)/T + 1 e +1
where µ is the electron chemical potential and E 2 = m2e + p2 . Then,
1 +∞ ex
Fdeg = p E dE , (3.17)
n e π 2 me (ex + 1)2
where x ≡ (E − µ)/T . For degenerate conditions the integrand is
strongly peaked near x = 0 so that one may replace p and E with the
values pF and EF at the Fermi surface (x = 0), and one may extend the
lower limit of integration to −∞. The integral then yields T so that
Fdeg = 3EF T /p2F , (3.18)
where ne = p3F /3π 2 was used.
Returning to the nonrelativistic, nondegenerate limit note that the
cross sections are of the form σ = σ∗ (ω/me )p so that
σ∗ ne T p+4 + ∞ xp+3 (p + 3)! ζp+4 σ∗ ne T p+4
Q= dx = . (3.19)
π 2 mpe 0 ex − 1 π2 mpe
Here, ζn = ζ(n) is the Riemann zeta function which shall be set equal
to unity.16 In a medium of mass density ρ the electron density is ne =
Ye ρ/mu where Ye is the electron number fraction per baryon and mu
the atomic mass unit. Therefore, the energy-loss rate per unit mass is
(p + 3)! Ye σ∗ T p+4
-= . (3.20)
π2 mu mpe
The average energy of the photons which are converted into weakly
interacting particles is
(ω* = (p + 3) T. (3.21)
For p = 0 one recovers (ω* = 3T for the average energy of blackbody
photons.17
16
ζn → 1 rather quickly with increasing n; for example ζ4 = π 4 /90 ≈ 1.082.
Therefore, the error is small if one takes ζn = 1 which corresponds to using a
Maxwell-Boltzmann rather than a Bose-Einstein distribution. They differ at small
ω where an effective photon mass in the plasma should be taken into account anyway.
Therefore, at the crude level of accuracy where one uses massless photons nothing
is gained by using the Bose-Einstein distribution.
17
With a Bose-Einstein distribution it is (ω*/T = 3 ζ4 /ζ3 = π 4 /30ζ3 ≈ 2.70.
98 Chapter 3
Beginning with the case of scalars, the low-energy cross section is

constant at σ∗ = 43 παα" /m2e , leading to
8αα" Ye T 4
-scalar = = α" 5.7×1029 erg g−1 s−1 Ye T84 , (3.22)
π mu m2e
where T8 = T /108 K. Vector bosons carry an extra factor of 2 for their

polarization states.
Turning to pseudoscalars, the low-energy cross section was found to
be σ∗ = 43 (παα" /m2e ) (ω/me )2 , i.e. p = 2, so that
160 αα" Ye T 6
-pseudo = = α" 3.3×1027 erg g−1 s−1 Ye T86 . (3.23)
π mu m4e
The average energy is (ω* ≈ 5T or (ω*/me ≈ 0.08 T8 .
Finally, turn to neutrino pair production for which the NR cross
section was given in Eq. (3.13). The energy-loss rate is
, -8
96 α G2F m6e T
-νν = (CV2 + 5CA2 ) 4
Ye
π mu me
= (CV2 + 5CA2 ) 0.166 erg g−1 s−1 Ye T88 . (3.24)
Because p = 4 for this process, (ω* ≈ 7T . Relativistic corrections be-

come important at rather low temperatures. For example, at T = 108 K
the true emission rate is about 25% smaller than given by Eq. (3.24).
3.2.6 Applying the Energy-Loss Argument

After the derivation of the energy-loss rates it is now a simple matter
to apply the energy-loss argument. In Chapter 2 it was shown that the
most restrictive limits obtain from the properties of globular-cluster
stars; two simple criteria were derived in Sect. 2.5 which amount to the
requirement that a novel energy-loss rate must not exceed 10 erg g−1 s−1
for the typical conditions encountered in the core of a horizontal-branch
star, and in the core of a red giant just before helium ignition which
both have T ≈ 108 K.
The Compton process is suppressed by degeneracy effects in a dense
plasma as discussed above—see Eq. (3.18). Therefore, at a fixed tem-
perature the emissivity per unit mass decreases with increasing density.
Because a red-giant core is nearly two orders of magnitude denser than
the core of an HB star it is enough to apply the argument to the latter
case.
The energy-loss rates for scalars and pseudoscalars in Eqs. (3.22)

and (3.23) are independent of density, and proportional to T 4 and T 6 ,
respectively. The averages over the core of a typical HB star are (T84 * =
0.40 and (T86 * = 0.37 (Fig. 2.24). With Ye = 0.5 appropriate for helium,
carbon, and oxygen the 10 erg g−1 s−1 limit yields
.
0.9×10−28 scalar,
α <
"
∼ (3.25)
1.6×10−26 pseudoscalar.
These bounds apply to bosons with a mass below a few times the tem-
perature, m < ∼ 20−30 keV. For larger masses the limits are significantly
degraded because only the high-energy tail of the blackbody photons
can produce the particles. For massive pseudoscalars this effect was
explicitly studied in Sect. 1.3.5 in the context of solar limits.
For vector bosons which interact by means of a Yukawa coupling,
the same limits to α" apply except that they are more restrictive by a
factor of 2 because of the two polarization states which increases the
emission rate.
For vector bosons which couple by means of a “magnetic moment”
as the “paraphotons” in Eq. (3.3), the bound on g is the same as for
pseudoscalars apart from an extra factor of 2 in the emission rate from
the two polarization states.
3.3 Pair Annihilation

Electron-positron pair annihilation can produce new bosons by the
“crossed” version of the Compton amplitude while the conversion into
neutrino pairs does not require the participation of a photon (Fig. 3.4).
For pseudoscalars the cross section for e+ e− → γa is (Mikaelian 1978)
! , / -"
2παα" s
σ= 2
ln 1+ 1 − 4m2e /s . (3.26)
s − 4me 4m2e
For e+ e− → νν it is (’t Hooft 1971; Dicus 1972)
G2F (CV2 + CA2 ) (s − m2e ) + 3 (CV2 − CA2 ) m2e

σ= / . (3.27)
12π 1 − 4m2e /s
Because pair annihilation requires the presence of positrons it is impor-

tant only for relativistic plasmas.
In this limit σ ∝ s−1 ln(s) for the production of bosons (pseu-
doscalar, scalar, vector) and σ ∝ s for neutrino pairs. Therefore, the
100 Chapter 3
Fig. 3.4. Pair annihilation processes for the production of neutrino pairs or
new bosons where a second amplitude with the vertices interchanged is not
shown.
importance of stellar energy-loss rates into new scalars relative to neu-

trino pairs decreases with increasing temperature and density. The
impact of new bosons on stellar evolution relative to neutrinos is then
expected to be most pronounced for low-mass stars where other pro-
cesses such as photoproduction dominate. Consequently, the pair pro-
cess has not played any significant role at constraining the interactions
of new bosons.
3.4 Free-Bound and Bound-Free Transitions

Photons, new bosons, or neutrino pairs can be emitted in transitions
where a free electron is captured by an ion to form a bound state.
For the case of axions this effect was dubbed “axio-recombination”
(Dimopoulos et al. 1986). In the Sun, it contributes about 4% of the
total axion flux which is mostly from bremsstrahlung. The energy-loss
rate scales as T 3/2 , bremsstrahlung as T 5/2 , and Compton emission as
T 6 . Thus, axio-recombination is of importance in low-mass stars which
have low internal temperatures; for main-sequence stars with M < ∼
0.2 M% it would be the dominant axion emission process. However,
given the limits on the coupling of pseudoscalars to electrons from other
arguments, no observable effects can be expected.
Of some practical interest is the inverse process where an axion un-
binds an atomic electron, the “axio-electric effect” (Dimopoulos, Stark-
man, and Lynn 1986a,b). It serves to constrain the solar flux of axions
or other pseudoscalars which could produce keV electrons in a Ge spec-
trometer designed to search for double-β decay (Avignone et al. 1987).
Unfortunately, the resulting bound of α" < ∼ 10
−21
is not very restrictive.
Pseudoscalars saturating this limit would be a major energy drain of
the Sun and thus not compatible with its observed properties.
For neutrinos, free-bound transitions were first discussed by Pinaev
(1963). Recently, Kohyama et al. (1993) studied the corresponding
stellar energy-loss rate in detail and found that it dominates the other
neutrino emission processes only in such regions of temperature and
density where the overall neutrino luminosity is very small. Therefore,
free-bound transitions do not seem to be of practical importance as a
stellar energy-loss mechanism.
3.5 Bremsstrahlung
3.5.1 Nondegenerate, Nonrelativistic Medium
The last emission process to be discussed is bremsstrahlung (Fig. 3.5)
where an electron emits a boson or a neutrino pair when scattering
off the Coulomb field of a nucleus. Conceptually, bremsstrahlung is
closely related to the Compton process (Fig. 3.1) because in both cases
the electron interacts with electromagnetic field fluctuations of the am-
bient medium which have nonvanishing power for all wavenumbers and
frequencies. The Compton process corresponds to wavevectors which
satisfy the photon dispersion relation so that a real (on-shell) excita-
tion is absorbed. However, in a typical stellar plasma there is more
power in the “off-shell” electromagnetic field fluctuations associated
with the charged particles.18 Moreover, degeneracy effects do not sup-
press bremsstrahlung at high densities, in contrast with the Compton
process.
Fig. 3.5. Bremsstrahlung emission of bosons or neutrino pairs by an electron

which scatters from the Coulomb field of a nucleus of charge Ze. In each
case a second amplitude with the vertices interchanged is not shown.
For neutrino pairs bremsstrahlung dominates over other processes

only in the highly degenerate regime (Appendix C). For pseudoscalars,
18
In a typical stellar plasma there are many more charged particles than black-
body photons. In the solar center, for example, the electron density is 6×1025 cm−3
while for photons at a temperature of 1.3 keV it is 2ζ(3) T 3 /π 2 = 6×1022 cm−3 .
102 Chapter 3
however, it is important even in environments which are approximately

nondegenerate and so I begin with this simple case.
The calculation amounts to a straightforward evaluation of the ma-
trix element corresponding to the amplitude of Fig. 3.5 and an inte-
gration over the Maxwell-Boltzmann distributions of the electrons. In
order to account for screening effects the Coulomb propagator is mod-
ified according to Eq. (6.72), |q|−4 → [q2 (q2 + kS2 )]−1 where kS is a
screening wave number (Sect. 6.4). For the emission of pseudoscalars
one then finds for the energy-loss rate per unit volume to lowest order
in kS2 (Krauss, Moody, and Wilczek 1984; Raffelt 1986a)
, -5/2
128 α2 α" T
Q= √ ne
45 π me me
! ' ( ' ("
0 √ 5 kS2 5 kS2
× nj Zj2 2 1− + Zj 1− . (3.28)
j 8 me T 4 me T
Here, the sum is extended over all nuclear species with charges Zj e and
1
number densities nj ; note that ne = j Zj nj . The term quadratic in Zj
corresponds to electron-nucleus collisions while the linear term is from
electron-electron scattering which yields a nonnegligible contribution
under nondegenerate conditions.19
Raffelt (1986a) incorrectly used Eq. (6.61) as a modification of the
Coulomb propagator, a procedure which enhances the terms propor-
tional to kS2 by a factor of 2. Either way, screening is never an impor-
tant effect. The screening scale is kS2 = kD
2
+ ki2 because both electrons
and nuclei (ions) contribute. Then
, -
kS2 4πα 0
2
= ne + n j Zj , (3.29)
me T me T 2 j
which is about 0.12 at the center of the Sun and 0.17 in the cores of
horizontal-branch (HB) stars.
Ignoring screening effects, the energy-loss rate per unit mass is
0 2 √ 3
- = α" 5.9×1022 erg g−1 s−1 T82.5 Ye ρ Yj Zj2 + Zj / 2 , (3.30)
j
8 −3
where T8 = T /10 K, ρ is in g cm , and Yj = Xj /Aj is the number
fraction of nuclear species j relative to baryons while Xj is the mass
fraction, Aj the mass number.
19
Note that e− e− → e− e− γ vanishes to lowest order because two particles of
equal mass moving under the influence of their Coulomb interaction do not produce
a time-varying electric dipole moment because their center of mass and “center of
charge” coincide. However, the emission of pseudoscalars corresponds to M1 rather
than E1 transitions and so it is not suppressed.
Grifols, Massó, and Peris (1989) have worked out the bremsstrah-
lung rate for a scalar boson; in this case e− e− collisions can be ignored
relative to electron-nucleus scattering. They found in the nonrelativis-
tic and nondegenerate limit, ignoring screening effects which are small,
2α2 α" ne T 1/2 0 Xj Zj2
-= a
π 3/2 mu m3/2
e j Aj
0 Xj Zj2
= α" 2.8×1026 erg g−1 s−1 T80.5 Ye ρ , (3.31)
j Aj
where a is an angular integral which numerically is found to be 8.36,

mu is the atomic mass unit, and ρ is in g cm−3 . For low-mass vector
bosons the same result pertains with an extra factor of 2 for the two
polarization states.
3.5.2 High Degeneracy: Pseudoscalars

Bremsstrahlung is a particularly important effect under conditions of
degeneracy. In this case one neglects e− e− collisions entirely which are
suppressed by degeneracy relative to the the electron-nucleus process.
If the target nuclei are taken to be infinitely heavy, Raffelt (1990) found
for the volume emissivity of pseudoscalars (“axions”)
+ ∞ + E1 + +
4α2 α" 0 2 dΩ2 dΩa
Q= Zj n j dE1 f1 dE2 (1 − f2 )
π2 j me me 4π 4π
4 5
|p1 | |p2 | ω 2 2
2 P1 P2 − me + (P2 − P1 )K P1 K P2 K
× 2 2 2
2ω +2− − ,
q (q + κ ) (P1 K)(P2 K) P2 K P1 K
(3.32)
where P1 is the four-vector of the incoming, P2 of the outgoing electron,
f1,2 are the electron Fermi-Dirac occupation numbers at energies E1,2 ,
temperature T , and chemical potential µ. Further, K is the four-vector
of the outgoing pseudoscalar (energy ω = E1 − E2 , direction Ωa , a for
axion), q = p1 − p2 − k is the momentum transfer to the nucleus, and
kS a screening scale. The Coulomb propagator was modified according
to Eq. (6.72). As long as the plasma is not strongly coupled one may
use kS = ki while the electrons do not contribute because kTF & ki in
a strongly degenerate plasma.
If the electrons are very degenerate, the energy integrals can be
done analytically. Moreover, all electron momenta are close to the
Fermi surface, |p1 | ≈ |p2 | ≈ pF with pF the Fermi momentum defined
104 Chapter 3
by ne = p3F /3π 2 . This also implies that |q|2 ≈ |p1 − p2 |2 ≈ 2p2F (1 − c12 )
where c12 is the cosine of the angle between p1 and p2 . With these
approximations and the velocity at the Fermi surface βF ≡ pF /EF =
pF /(m2e + p2F )1/2 one finds
, -
π 2 α2 α" T 4 0 2
Q= n j j F,
Z (3.33)
15 m2e j
where
+ +
dΩ2 dΩa (1 − βF2 ) [2 (1 − c12 ) − (c1a − c2a )2 ]
F =
4π 4π (1 − c1a βF ) (1 − c2a βF ) (1 − c12 )(1 − c12 + κ2 )
(3.34)
with κ2 ≡ kS2 /2p2F .
For a single species of nuclei with charge Ze and atomic weight A
the energy-loss rate per unit mass is
π 2 α2 α" Z 2 T 4
-= F
15 A mu m2e
Z2 4
= α" 1.08×1027 erg g−1 s−1 T F, (3.35)
A 8
where again T8 = T /108 K. Because F is of order unity for all condi-
tions, the bremsstrahlung rate mostly depends on the temperature and
chemical composition, and - is not suppressed at high density.
Expanding Eq. (3.34) in powers of βF for nonrelativistic or partially
relativistic electrons one finds
' ( ! ' ( "
2 2 + κ2 2 + 5κ2 2 + κ2 2 2
F = ln + ln − β + O(βF4 ).
3 κ2 15 κ2 3 F
(3.36)
Therefore, in contrast to the nondegenerate calculation this expression
would diverge in the absence of screening.
Another approximation can be made if one observes that Coulomb
scattering is mostly forward, i.e. the main contribution to the integral
is from c12 ≈ 1 which implies c1a ≈ c2a . With c2a = c1a only in the
denominator one obtains
' ( ! ' ( "
2 2 + κ2 2 + 3κ2 2 + κ2
F = ln + ln − 1 f (βF ) (3.37)
3 κ2 6 κ2
with
' (
3 − 2βF2 3 (1 − βF2 ) 1 + βF
f (βF ) = − ln . (3.38)
βF2 2βF3 1 − βF
The function f (βF ) is 0 at βF = 0 and rises monotonically to 1 for
βF → 1. Hence in the relativistic limit

' (
2 + κ2 2 + κ2
F = ln − 1, (3.39)
2 κ2
somewhat different from what Iwamoto (1984) found who used the
Thomas-Fermi wave number as a screening scale.
For a strongly coupled, degenerate plasma typical for white dwarfs
the factor F was calculated numerically by Nakagawa, Kohyama, and
Itoh (1987) and Nakagawa et al. (1988) who also gave analytic approxi-
mation formulae for the axion emission rate, applicable to nonrelativis-
tic and relativistic conditions. For a 12 C plasma with densities in the
range 104 −106 g cm−3 and temperatures of 106 −107 K it is found that
F = 1.0 within a few tens of percent. Therefore, for simple estimates
this value is a satisfactory approximation.
Altherr, Petitgirard, and del Rı́o Gaztelurrutia (1994) have calcu-
lated the bremsstrahlung process with the methods of finite tempera-
ture and density (FTD) field theory. The main point is that one com-
putes directly the interaction of the electrons with the electromagnetic
field fluctuations which are induced by the ambient charged particles.
The result for the emission rate is similar to the one derived above.
3.5.3 High Degeneracy: Neutrino Pairs

Neutrino pair bremsstrahlung (Fig. 3.5) was the first nonnuclear neu-
trino emission process ever proposed (Pontecorvo 1959; Gandel’man
and Pinaev 1959). A detailed calculation of the energy-loss rate in
a degenerate medium, relativistic and nonrelativistic, was performed
by Festa and Ruderman (1969) while conditions of partial degeneracy
were studied by Cazzola, de Zotti, and Saggion (1971). The Festa and
Ruderman calculation was extended by Dicus et al. (1976) to include
neutral-current interactions. After a calculation very similar to the one
presented above for pseudoscalars the emission rate for neutrino pairs is
, -) *
2πα2 2 6 0
Q= GF T nj Zj 12 (CV2 + CA2 )F+ + 12 (CV2 − CA2 )F− .
2
189 j
(3.40)
With a single species of nuclei (charge Z, atomic mass A) the energy-
loss rate per unit mass is
- = 0.144 erg g−1 s−1 (Z 2 /A) T86 [ . . . ], (3.41)
8
where T8 = T /10 K and the square bracket is from Eq. (3.40). The
temperature dependence is steeper than for axions by two powers.
106 Chapter 3
The factors F+ and F− are of order unity. Dicus et al. (1976) de-
rived analytic expressions in terms of a screening scale and the Fermi
velocity of the electrons. In fact, because F− is always much smaller
than F+ and CV2 − CA2 is much smaller than CV2 + CA2 , the “minus”
term may be neglected entirely. Therefore, the Dicus et al. (1976) re-
sult is identical with that of Festa and Ruderman (1969). Either one
is correct only within a factor of order unity because Eq. (6.61) was
used as a screening prescription with the Thomas-Fermi wave number
as a screening scale. However, in a degenerate medium electrons never
dominate screening. The most important effect is from the ion corre-
lations which, in a weakly coupled plasma (Γ < ∼ 1), can be included by
Eq. (6.72) with the Debye scale ki of the ions as a screening scale. While
it is easy to replace kTF with ki in these results, the modification of the
Coulomb propagator according to Eq. (6.72) cannot be implemented
without redoing the entire calculation.
A systematic approach to include ion correlations (i.e. screening
effects) was pioneered by Flowers (1973, 1974) who showed clearly how
to separate the ion correlation effects in the form of a dynamic structure
factor from the matrix element of the electrons and neutrinos. This
approach also allows one to include lattice vibrations when the ions
form a crystal in a strongly coupled plasma. In a series of papers Itoh
and Kohyama (1983), Itoh et al. (1984a,b), and Munakata, Kohyama,
and Itoh (1987) followed this approach and calculated the emission rate
for all conditions and chemical compositions.
As an estimate, good to within a factor of order unity, one may use
F+ = 1 and F− = 0. Moreover, inspired by the axion results one can
guess a simple expression which can be tested against the numerical
rates of Itoh and Kohyama (1983). I find that F− = 0 and
' (
2 + κ2 κ2
F+ ≈ ln + (3.42)
κ2 2 + κ2
is a reasonable fit even for strongly coupled conditions (Appendix C).
3.5.4 Neutron-Star Crust

The degenerate bremsstrahlung emission of νν pairs is relevant in such
diverse environments as the cores of low-mass red giants, white dwarfs,
and in neutron-star crusts. Pethick and Thorsson (1994) noted that
in the latter case the medium is so dense that band-structure effects
of the electrons become important. The band separations can be up
to O(1 MeV), suppressing electron scattering and bremsstrahlung pro-

cesses for temperatures below this scale (T < 9
∼ 5×10 K). Bremsstrah-
lung of νν pairs from the crust was thought to dominate neutron-star
cooling for some conditions while Pethick and Thorsson (1994) now
find that it may never be important. These findings also diminish
Iwamoto’s (1984) axion bound based on the bremsstrahlung emission
from neutron-star crusts.
3.5.5 Applying the Energy-Loss Argument

One may now easily derive astrophysical limits on the Yukawa couplings
of scalars and pseudoscalars, in full analogy to Sect. 3.2.6 where the
Compton emission rates were used. I begin with the same case that was
considered there, namely the restriction - < −1 −1
∼ 10 erg g s in the cores
of horizontal-branch stars. For nondegenerate conditions the emission
rates are Eq. (3.30) and Eq. (3.31), respectively. They are proportional
to ρ T 0.5 (pseudoscalar) and ρ T 2.5 (scalar). With (ρ4 * = 0.64, (T80.5 * =
0.82, (T82.5 * = 0.48, and a chemical composition of pure helium one finds
.
1.4×10−29 scalar,
α <
"
∼ (3.43)
1.6×10−25 pseudoscalar.
Comparing these limits with those from the Compton process Eq. (3.25)
reveals that for scalars the present bremsstrahlung limit is more restric-
tive, for pseudoscalars the Compton one.
Because bremsstrahlung is not suppressed in a degenerate plasma,
one can also apply the helium-ignition argument of Sect. 2.5.2 which
again requires - < −1 −1 8
∼ 10 erg g s at the same T ≈ 10 K, however at a
6 −3
density of around 10 g cm . For these conditions the plasma is degen-
erate but weakly coupled (Appendix D) so that Debye screening should
be an appropriate procedure. Therefore, for the emission of pseu-
doscalars the emission rate Eq. (3.35) with F from Eq. (3.36) should
be a reasonable approximation. The screening scale is dominated by
the ions, kS = ki = 222 keV, the Fermi momentum is pF = 409 keV so
that κ2 = 0.15 while βF = 0.77, yielding F ≈ 1.8.
In Fig. 3.6 the energy-loss rate of a helium plasma at T = 108 K
is plotted as a function of density, including the Compton process,
and the degenerate (D) and nondegenerate (ND) bremsstrahlung rates.
This figure clarifies that bremsstrahlung, of course, is suppressed by
degeneracy effects relative to the ND rates, but it is not a significantly
decreasing function of density. In this figure a simple interpolation
(solid line) between the regimes of high and low degeneracy is shown
108 Chapter 3
Fig. 3.6. Energy-loss rate of a helium plasma at T = 108 K as a function

of density for pseudoscalars with the coupling α" = 10−26 . The Compton
rate is given in Eq. (3.23), suppressed with the factor Fdeg of Eq. (3.18) at
high density. The nondegenerate (ND) and degenerate (D) bremsstrahlung
rates are from Eqs. (3.28) and (3.35), respectively. The solid line is the
interpolation formula of Raffelt and Weiss (1995).
that was used in a numerical study of axion emission from red giants
by Raffelt and Weiss (1995); details of how it was constructed can be
found there.
At T = 108 K the compound rate (solid line) coincidentally is almost
independent of density. The requirement - < −1 −1
∼ 10 erg g s then yields
a constraint α" <
∼ 10
−26
for any density whence the helium-burning life-
time of HB stars as well as the core mass at helium ignition yield an
almost identical constraint. In detail one needs to apply the helium-
ignition argument at an average core density of 2×105 g/cm3 (the cen-
tral density is about 106 g/cm3 ) and at the almost constant tempera-
ture of T = 108 K. The degenerate emission rate is then approximately
α" 2×1027 erg g−1 s−1 so that
α" < −26

∼ 0.5×10 . (3.44)
This is the most restrictive available bound on the Yukawa coupling of

pseudoscalars to electrons.
The same case was treated numerically by Raffelt and Weiss (1995)
who implemented the compound energy-loss rate of Fig. 3.6 with vary-
ing values of α" in several red-giant evolutionary sequences. The re-
sults of this work have been discussed in Sect. 2.5.2 where it was used
as one justification for the simple 10 erg g−1 s−1 energy-loss constraint
that was derived there. This detailed numerical study yielded the same
limit Eq. (3.44) on a pseudoscalar Yukawa coupling to electrons.
3.6 Astrophysical Bounds on Yukawa and Gauge

Couplings
3.6.1 Pseudoscalars (Axions)
The main concern of this chapter was novel bosons which interact with
electrons by a dimensionless Yukawa coupling. It will become clear
shortly that these results can be easily translated into limits on bary-
onic couplings as well. It may be useful to pull together the main
results found so far, and discuss them in the context of other sources
of information on the same quantities.
The best studied case of boson couplings to electrons is that of
pseudoscalars because the existence of such particles is motivated by
their role as Nambu-Goldstone bosons of a spontaneously broken chiral
symmetry of the fundamental interactions. Within this class, axions
(Chapter 14) have been most widely discussed; they usually serve as a
generic example for low-mass pseudoscalars. For vector bosons which
couple by means of a “magnetic moment” (Eq. 3.3) such as “para-
photons” the same limits apply apart from an extra factor of 2 in the
emission rate from the two polarization states of these particles.
The simplest constraint on the Yukawa coupling ga of pseudoscalars
(axions) to electrons (αa = ga2 /4π) arises from the argument that the
age of the Sun precludes any novel energy-loss mechanism to be more
efficient than the surface photon luminosity (Sect. 1.3.2). The relevant
emission processes are the Compton reaction with the energy-loss rate
given in Eq. (3.23), and bremsstrahlung with electrons scattering on
electrons, protons, and helium nuclei; the emission rate was given in
Eq. (3.28). An integration over a typical solar model yields an axion
luminosity (Raffelt 1986a)
La = αa 6.0×1021 L% , (3.45)
with about 25% from the Compton process, 25% from ee bremsstrah-
lung, and 50% from bremsstrahlung by electrons scattering on nuclei.
110 Chapter 3
The requirement La < <

∼ L% yields a constraint αa ∼ 1.7×10 .
−22
A much more restrictive limit arises from the white-dwarf luminos-

ity function as axion emission would accelerate white-dwarf cooling.
This argument was studied in detail in Sect. 2.2.4 as a generic case
for the use of the white-dwarf luminosity function; the resulting con-
straint is given in Tab. 3.1. The cooling speed of white dwarfs was
also established from a measurement of the period decrease of the DA
variable (ZZ Ceti) star G117–B15A which thus yields a similar limit.
However, the period decrease of this star may be slightly faster than
can be attributed to standard cooling processes; it has been speculated
that axions with a coupling strength of about αa ≈ 0.5×10−26 could be
responsible (Sect. 2.2.5). The limit Eq. (3.44) that was derived in the
previous section from the helium-ignition argument in globular clusters
is of a similar magnitude, but not restrictive enough to exclude this
hypothesis.
All of these constraints apply to low-mass bosons. The most restric-
tive one is based on the helium ignition argument with T ≈ 108 K =
8.6 keV. Therefore, these constraints apply if m < ∼ 10 keV. However,
it would be incorrect to think that for larger masses there was no con-
straint. There is one, but it is degraded because threshold effects limit
the particle production to the high-energy tails of the thermal distri-
butions of the plasma constituents. For massive pseudoscalars, this
question has been studied in Sect. 1.3.5 in the context of general solar
particle constraints. For the more restrictive globular-cluster limits,
such a detailed investigation does not exist in the literature.
3.6.2 Energy Loss by Scalar and Vector Bosons

The couplings of low-mass scalar or vector particles φ are easy to con-
strain by the same methods. Because the energy-loss rates have been
calculated only for nondegenerate conditions, only the arguments in-
volving the solar age and the helium-burning lifetime can be employed.
The latter yields a constraint on the φ-e coupling of
αφe < −29

∼ 1.4×10 . (3.46)
It is based on the bremsstrahlung process e + 4 He → 4 He + e + φ as

discussed above in Sect. 3.5.5. For vector bosons which couple by a
current-current structure analogous to photons the same results apply
except for a factor of two in the emission rate which improves the limit
by a factor of 2.
Tab. 3.1. Astrophysical bounds on the Yukawa coupling ga of pseudoscalars (axions) to electrons.
Upper limit Astrophysical Dominant emission Detailed

αa = ga2 /4π observable process discussion
1.7×10−22 Solar age Bremsstrahlung (75%), Compton (25%) here
e + X → X + e + a (X = e, p, 4 He)
γ+e→e+a
1.0×10−26 Galactic age inferred from break Bremsstrahlung (degenerate) Sect. 2.2.4
in white-dwarf luminosity function e + 12 C (16 O) → 12 C (16 O) + e + a
similar Period decrease in DAV star G117–B15A same Sect. 2.2.5
1.6×10−26 Helium-burning lifetime of Compton Sect. 3.2.6

Particles Interacting with Electrons and Baryons
horizontal-branch stars γ+e→e+a
0.5×10−26 Core mass at helium ignition in Bremsstrahlung (degenerate) Sect. 2.5.2

low-mass red giants e+A→A+e+a Sect. 3.5.5
(A = 4 He, 12 C, 16 O)
111
112 Chapter 3
One may also consider the Yukawa coupling of scalar (vector) bosons
to baryons (Grifols and Massó 1986; Grifols, Massó, and Peris 1989b).
In this case one may use the Compton process γ + 4 He → 4 He + φ on
a helium nucleus for which the emission rate is given mutatis mutandis
by the same formula as for nonrelativistic electrons. Assuming the
same coupling to protons and neutrons the emission rate is coherently
enhanced by a factor 42 . Moreover, in Eq. (3.22) one must replace
Ye with YHe (number of 4 He nuclei per baryon) which is 14 for pure
helium, me is to be replaced by mHe ≈ 4mu , and α → 4α to account
for the coherent photon coupling. Pulling these factors together one
finds - = αφN 0.7×1023 erg g−1 s−1 for pure helium. Because the helium-
burning lifetime argument limits this energy-loss rate to 10 erg g−1 s−1
one finds
αφN <
∼ 1.5×10
−22
(3.47)
as a limit on the coupling of a scalar boson to a nucleon N . Again, for
vector bosons this limit is a factor of 2 more restrictive.
3.6.3 Long-Range Forces

Scalar or vector particles mediate long-range forces between macro-
scopic bodies. For pseudoscalars this is not the case because their CP-
conserving coupling to fermions has a pseudoscalar structure, i.e. in the
nonrelativistic limit they couple to the fermion spin. Therefore, even
if the mass of the new particles is very small or exactly zero, they do
not mediate a long-range force between unpolarized bodies. The resid-
ual force caused by the simultaneous exchange of two pseudoscalars is
found to be extremely small (e.g. Grifols and Tortosa 1994).
Consider a scalar of mass m = λ−1 (Compton wave length λ) which
couples to nucleons with a Yukawa strength g. It mediates an at-
tractive force between two nucleons given in terms of the potential
−(g 2 /4π) r−1 e−r/λ . Two macroscopic test bodies of geometric dimen-
sion much below λ are then attracted by virtue of the total potential
m1 m2
V (r) = −GN (1 + βe−r/λ ), (3.48)
r
where GN = m−2 19
Pl is Newton’s constant, mPl = 1.22×10 GeV is the
Planck mass, m1 and m2 are the masses of the bodies, and
g 2 m2Pl
β≡ . (3.49)
4π u1 u2
Here, u1,2 = m1,2 /N1,2 with N1,2 the total number of nucleons in each
body. Apart from small binding-energy effects which are different for
different materials u1 ≈ u2 ≈ mu is the atomic mass unit, approxi-

mately equal to a nucleon mass mN . For this force not to compete
with gravity one needs g < O(10−19 ) or α = g 2 /4π < O(10−37 ) so that
low-mass scalars, if they exist, need to have extremely feeble couplings
to matter. Their smallness implies that such particles would not have
any impact whatsoever on the energy loss of stars.
Therefore, for boson masses so small that the Compton wave length
λ is macroscopic, the most restrictive limits on g obtain from analyzing
the forces between macroscopic bodies, not from the energy-loss argu-
ment. The existence of a composition-dependent “fifth force” in nature
with a strength of about 1% of gravity and a range λ of a few hun-
dred meters seemed indicated by a reanalysis of Eötvös’s original data
(Fischbach et al. 1986). Subsequently many experiments were carried
out to search for this effect, with no believable positive outcome (for
a review see Fischbach and Talmadge 1992). However, these investi-
gations produced extremely restrictive limits on β, depending on the
assumed range λ of the new force. The limits also depend on the pre-
sumed coupling; if the new force couples to baryon number one finds
that β < −3 < −9
∼ 10 for λ in the cm range, or β ∼ 10 for λ of order the
Earth-Sun distance and above. Thus, novel long-range forces must be
much weaker than gravity. No bounds on β seem to exist for λ below
the cm range, i.e. for boson masses of order 10−3 eV and above, apart
from the stellar energy-loss argument.
The effect of a novel force with intermediate range on the equa-
tions of stellar structure was discussed by Glass and Szamosi (1987,
1989). Solar models including such a force and the impact on the
solar oscillation frequencies were discussed by Gilliland and Däppen
(1987) and Kuhn (1988). For a force so weak or weaker than indi-
cated by the laboratory limits, no observable consequences for stel-
lar structure and evolution seem to obtain. Also, the impact of the
new field on the value of fundamental coupling constants even at com-
pact objects such as neutron stars is far below any observable limit
(Ellis et al. 1989).
A significant bound obtains from the orbital decay of the Hulse-
Taylor binary pulsar. In order for the energy loss in the new scalars to
remain below 1% of the gravitational wave emission the Yukawa cou-
pling to baryons must satisfy g <∼ 3×10
−19
(Mohanty and Panda 1994).
<
This translates into β ∼ 1 for scalar boson masses below the orbital pul-
sar frequency of 2π/P = 2.251×10−4 s−1 , i.e. for λ−1 < −19
∼ 1.5×10 eV
or λ > 14
∼ 1.3×10 cm. This “fifth-force limit,” however, is weaker than
those derived by terrestrial laboratory methods.
114 Chapter 3
3.6.4 Leptonic and Baryonic Gauge Interactions

The physical motivation for considering long-range interactions medi-
ated by vector bosons, besides the fifth-force episode, is the hypothesis
that baryon number or lepton number could play the role of physical
charges similar to the electric one (Lee and Yang 1955; Okun 1969).
Their association with a gauge symmetry would provide one explana-
tion for the strict conservation of baryon and lepton number which so
far has been observed in nature. In the framework of this hypothesis
one predicts the existence of baryonic or leptonic photons which couple
to baryons or leptons by a charge eB or eL , respectively. The novel
gauge bosons would be massless like the ordinary photon.
Therefore, the limits established in the previous sections on the di-
mensionless couplings of vector bosons can be readily restated as limits
on the values of putative baryonic or leptonic charges. The energy-loss
argument applied to helium-burning stars yields
eL <∼ 1×10 ,
−14
eB <∼ 3×10 ,
−11
(3.50)
according to Eqs. (3.43) and (3.47), respectively. Tests of the equiv-
alence principle (i.e. of a composition-dependent fifth force) on solar-
system scales yield β < −9
∼ 10 (Sect. 3.6.3) so that
eB <∼ 1×10 .
−23
(3.51)
Apparently this is the most restrictive limit on eB that is currently
available.
One may be tempted to apply the limits from the equivalence prin-
ciple also to a leptonic charge eL . However, in this case one has to worry
about the fact that even neutrinos would carry leptonic charges. The
universe is probably filled with a background neutrino sea in the same
way as it is filled with a background of microwave photons. This neu-
trino medium would constitute a leptonic plasma which screens sources
of the leptonic force just as an electronic plasma screens electric charges
(Zisman 1971; Goldman, Zisman, and Shaulov 1972; Çiftçi, Sultansoi,
and Türköz 1994; Dolgov and Raffelt 1995).
Debye screening will be studied in Sect. 6.4.1. For the screening
wave number one finds the expression
+ ∞
kS2 = 2 (eL /π)2 dp fp p (v + v −1 ), (3.52)
0
where p = |p| is the momentum (isotropy was assumed), v = p/E

the velocity of the charged particles, and fp the occupation number of
mode p. The overall factor of 2 relative to Eq. (6.55) arises because

neutrinos and antineutrinos contribute equally.
The smallest kS (the largest radius over which leptonic fields remain
unscreened) arises if the background neutrinos have such small masses
that they are still relativistic
6
today (v = 1).
6∞
The neutrino number
density is given by nν+ν = 2 fp d p/(2π) = 0 fp p2 dp/π 2 . Therefore,
3 3
in the relativistic limit one finds roughly

−1/3
kS−1 ≈ e−1 −1
L nν+ν ≈ eL 0.2 cm, (3.53)
independently of details of the neutrino momentum distribution. In

this estimate the predicted number density of background neutrinos in
each flavor of nν+ν ≈ 100 cm−3 was used.
The largest conceivable kS would obtain if neutrinos were nonrel-
ativistic today, and if some of them were bound to the galaxy. The
escape velocity from the galaxy is vesc ≈ 500 km s−1 so that the maxi-
mum momentum of a gravitationally bound neutrino is pmax = mν vesc .
Because neutrinos obey Fermi statistics, the largest conceivable galactic
neutrino density is nmax ≈ p3max ≈ m3ν vesc
3
. A typical neutrino velocity
is of order the galactic velocity dispersion, i.e. of order vesc . Therefore,
−1
from Eq. (3.52) one estimates kS2 ≈ e2L n2/3 2 2
max vesc ≈ eL mν vesc . Because the
largest cosmologically allowed neutrino mass is about 30 eV one finds
that neutrino screening cannot operate on scales below e−1 L 10
−5
cm.
Therefore, the screening scale is reduced by no more than six orders
of magnitude by the fact that cosmic neutrinos could be nonrelativis-
tic today.
The stellar energy-loss result Eq. (3.50) informs us that for rela-
tivistic neutrinos kS > 13
∼ 10 cm ≈ 1 AU where 1 AU = 1.5×10 cm
13
(astronomical unit) is the distance to the Sun. If neutrinos were non-

relativistic, leptonic forces could be screened over distances six orders
of magnitude smaller, i.e. over about 100 km. However, it is proba-
bly safe to assume that leptonic forces with a strength comparable to
gravity would have been noticed in terrestrial experiments searching
for a composition-dependent fifth force. Therefore, even with neutrino
screening it appears inconceivable that eL could exceed about 10−19 .
In that case the screening scale would always exceed about 0.1 AU so
that terrestrial limits would easily apply. Thereby one could gain a
few orders of magnitude in the limit, and so one could even use so-
lar system constraints, taking one back to a result of order the bary-
onic one Eq. (3.51). A similar conclusion was reached by Blinnikov
et al. (1995).
116 Chapter 3
Finally, the matter of the galaxy would exert a leptonic force on

neutrinos propagating, say, from a distant supernova to us. This effect
would cause an energy-dependent dispersion of the measurable neutrino
burst (Sect. 13.3.3). However, when eL > −20
∼ 10 , which is necessary to
cause an interesting effect on supernova neutrinos, then the galactic
leptonic charge is completely screened over the relevant length scales,
even if the cosmic background neutrinos are relativistic.
3.7 Graviton Emission from Stars

The one nonelectromagnetic long-range force that is actually known
to exist is gravity. Whatever the ultimate quantum theory of gravity,
there is little doubt that there will be quantized wave excitations, the
gravitons. They would be massless spin-2 particles. In fact, gravity is
the only possible force that can be mediated by a massless spin-2 boson
because as a source it needs a conserved rank-2 tensor. The energy-
momentum tensor, which acts as a source for the gravitational field, is
the only example.
Classical gravitational waves are an inevitable consequence of Ein-
stein’s theory of general relativity. The orbital decay of the binary
pulsar PSR 1913+16 (Hulse and Taylor 1975) yields firm evidence for
their emission (Taylor and Weisberg 1989; Damour and Taylor 1991).
Gravitons can be produced in hot plasmas in analogy to axions or
neutrino pairs; typical processes are bremsstrahlung e + p → e + p + g
(graviton g) and the Primakoff effect (gravitons have a two-photon
coupling). Because of their weak interaction gravitons can freely es-
cape once produced in the interior of stars. Early calculations of the
emission rates were summarized by Papini and Valluri (1977). More
recent discussions include Schäfer and Dehnen (1983), Gould (1985),
and del Campo and Ford (1988). As the graviton coupling involves the
inverse of the Planck mass (1.2×1019 GeV) the graviton luminosity of
stars is inevitably small. For the Sun it is about 1015 erg s−1 ≈ 10−19 L% ,
much too small to be of any observational relevance. The same conclu-
sion holds for other stars.
Therefore, gravity itself illustrates that the mediation of long-range
forces is a far more important effect of low-mass bosons than their
thermal emission from stellar plasmas. Unless, of course, they only
couple to the fermion spins rather than to a “charge.” Pseudoscalars
such as axions are in that category.
Chapter 4
Processes in a Nuclear
Medium
The interaction rates of neutrinos and axions with nucleons in a nuclear

medium are studied with a focus on neutral-current processes such as
bremsstahlung emission of axions N N → N N a and of neutrino pairs
N N → N N νν as well as neutrino scattering. A severe problem with
the perturbative rate calculations at high density is discussed which has
a strong impact on axion emissivities and the dominant axial-vector
contribution to the neutrino opacities.
4.1 Introduction
New particles which couple to nucleons are emitted from ordinary stars
by analogous processes to those discussed in Chapter 3 for electrons.
For example, axions can be produced by the Compton process γp → pa;
the previous results can be easily adapted to such reactions. Presently
I will focus on processes involving neutrinos or axions that are specific
to a nuclear medium, i.e. to supernova (SN) cores or neutron stars.
The main focus of the literature which deals with microscopic pro-
cesses in a nuclear medium was inspired by the problem of late-time
neutron-star cooling. The recent progress of x-ray astronomy has led
to reasonably safe ROSAT identifications of thermal surface emission
from a number of old pulsars (Sect. 2.3). Together with the spin-down
age of these objects one can begin to test neutron-star cooling sce-
narios, notably those that involve novel phases of nuclear matter such
as superfluidity, meson condensates, quark matter, etc. (Shapiro and
Teukolsky 1983; Tsuruta 1992).
117
118 Chapter 4
From the perspective of particle physics, however, a more interesting

nuclear environment is a young neutron star for the first few seconds
after the progenitor collapsed. The nuclear medium here is so hot that
it is essentially nondegenerate, and neutrinos are trapped. Therefore,
the production of even more weakly interacting particles such as axions
or right-handed neutrinos can compete with neutrino energy transfer
which is essentially a diffusion process. For a quantitative understand-
ing of the emissivities of the new particles, but also for the conventional
transport of energy and lepton number by neutrinos, a knowledge of
the microscopic interaction rates is needed.
The neutrino opacities that went into standard SN collapse and ex-
plosion calculations as well as the particle emissivities that went into
the derivation of, say, axion bounds from SN 1987A were based on the
assumption that the hot nuclear medium can be treated as an ideal
Boltzmann gas of free particles, except for degeneracy effects that are
easy to include. It turns out, however, that the approximations made
are internally inconsistent, notably for the dominant processes which
involve couplings to the nucleon spin (axial-vector current interactions).
In order for the “naive” neutral-current neutrino opacities to be correct
one needs to assume that the nucleon spins do not fluctuate too fast on
a time scale set by the temperature. A naive perturbative calculation,
however, yields a spin-fluctuation rate which is much larger than this
limit. This large rate went into the axion emissivities. Therefore, the
existing studies of SN axion bounds are based on microscopic interac-
tion rates which simultaneously make use of the opposite limits of a
spin-fluctuation rate very small and very large compared with T .
Beyond the ideal-gas approximation there is virtually no literature
on the microscopic interaction rates in a hot nuclear medium, presum-
ably because of the historical focus on old neutron stars, and presum-
ably because degenerate nuclear matter is more reminiscent of actual
nuclei. Thus there is a dearth of reliable microscopic input physics for
conventional studies of SN evolution, and for variations involving novel
particle-physics ideas.
Most of this chapter focusses on the dominant axial-vector current
interactions of neutrinos and axions with nucleons in a dense and hot
medium. Most of the material is based on a series of papers which
I have co-authored (Raffelt and Seckel 1991, 1995; Keil, Janka, and
Raffelt 1995; Keil et al. 1995) and as such does not represent a com-
munity consensus. On the other hand, I am not aware of a significant
controversy. Rather, it appears that very little serious interest has
been taken in the difficult question of weakly interacting particles in-
Processes in a Nuclear Medium 119
teracting with a hot nuclear medium, even though these issues are of
paramount importance for a proper quantitative understanding of SN
physics where the interaction of neutrinos with the medium dominates
the thermal and dynamical evolution. My discussion can only be a
starting point for future work that may actually yield some answers to
the questions raised.
4.2 Axionic Bremsstrahlung Process

4.2.1 Matrix Element for N N → N N a
The simplest neutral-current process of the kind to be discussed in this
chapter is bremsstrahlung emission of axions or other pseudoscalars
(Fig. 4.1) because the single-particle axion phase space is particularly
simple. It will turn out that the result thus derived can be applied
to neutrino processes almost without modification. The interaction
Hamiltonian with nucleons is of the form
CN
Hint = − ψ γµ γ5 ψN ∂ µ φ, (4.1)
2fa N
where fa is an energy scale (the Peccei-Quinn scale for axions), CN with
N = n or p is a dimensionless, model-dependent coupling constant of
order unity, the ψN are the proton and nucleon Dirac fields, and φ is
the axion field or any other pseudoscalar Nambu-Goldstone boson.
Consider a single species of nonrelativistic nucleons interacting by a
one-pion exchange (OPE) potential. The spin-summed squared matrix
element is (Brinkmann and Turner 1988; Raffelt and Seckel 1995)
$ %2 $ %2
! 16 (4π)3 απ2 αa  k2 l2
|M|2 = + 2
spins 3m2N k2 + m2π l + m2π
&
k2 l2 − 3 (k · l)2
+ 2 . (4.2)
(k + m2π )(l2 + m2π )
Here, αa ≡ (CN mN /fa )2 /4π and απ ≡ (f 2mN /mπ )2 /4π ≈ 15 with f ≈
1 are the axion-nucleon and pion-nucleon “fine-structure constants,”
respectively. Further, k = p2 −p4 and l = p2 −p3 with pi the momenta
of the nucleons Ni as in Fig. 4.1.
In a thermal medium k2 ≈ 3mN T so that [k2 /(k2 + m2π )]2 varies
between 0.86 for T = 80 MeV and 0.37 for T = 10 MeV. Therefore,
neglecting the pion mass causes only a moderate error in a SN core
(Brinkmann and Turner 1988; Burrows, Ressell, and Turner 1990).
120 Chapter 4
Fig. 4.1. Feynman graph for nucleon-nucleon axion bremsstrahlung. There

is a total of eight amplitudes, four with the axion attached to each nucleon
line, and an exchange graph each with N3 ↔ N4 .
Raffelt and Seckel (1995) showed that including mπ causes less than
a 30% reduction of typical neutrino or axion rates for T > 20 MeV. Be-
cause this will be a minor error relative to the dominant uncertainties
the term in square brackets is approximated as [3 − (k̂ · l̂)2 ].
The remaining (k̂ · l̂)2 term is inconvenient without yielding any
significant insights. In a degenerate medium it averages to zero in
expressions such as the axion emission rate while in a nondegenerate
medium it can be as large as about 1.31 (Raffelt and Seckel 1995),
leading to an almost 50% reduction of the emissivity. Still, for the
present discussion I will neglect this term and use
!
|M|2 = 16 (4π)3 απ2 αa m−2
N . (4.3)
spins
While this may seem somewhat arbitrary, it must be stressed that us-
ing an OPE potential to model the nucleon interactions in a nuclear
medium is in itself an approximation of uncertain precision. For the
present discussion a factor of order unity will not change any of the
conclusions.
4.2.2 Energy-Loss Rate

The axionic volume energy-loss rate of a medium is the usual phase-
space integral,
' ' (4
d3 k a d3 pi
Qa = ωa f f (1 − f3 )(1 − f4 )
3 1 2
2ωa (2π)3 i=1 2Ei (2π)
!
× (2π)4 δ 4 (P1 + P2 − P3 − P4 − Ka ) 14 |M|2 , (4.4)
spins
where P1,2 are the four-momenta of the initial-state nucleons, P3,4 are
for the final states, and Ka is for the axion. The factor 14 is a statistics
factor to compensate for double counting of identical fermions in the

initial and final state. The occupation numbers f1,2 and the Pauli
blocking factors (1 − f3,4 ) are for the nucleons while the axions are
assumed to escape freely so that a Bose stimulation factor as well as
backreactions (axion absorption) can be neglected.
In the nonrelativistic limit the nucleon mass mN is much larger than
all other energy scales such as the temperature or Fermi energies. The
nucleon momenta are then much larger than the momentum carried by
the radiation. A typical nonrelativistic nucleon kinetic energy is Ekin =
p2 /2mN so that a typical nucleon momentum is p = (2mN Ekin )1/2 .
In a bremsstrahlung process, the radiation typically takes the energy
Ekin with it, less in a degenerate medium, so that a typical radiation
momentum is ka = Ekin ' p. Therefore, one may ignore the radiation
in the law of momentum conservation so that Eq. (4.4) is simplified
according to
δ 4 (P1 + P2 − P3 − P4 − Ka ) →
→ δ(E1 + E2 − E3 − E4 − ωa ) δ 3 (p1 + p2 − p3 − p4 ). (4.5)
In this case, the second integral expression in Eq. (4.4), which “knows”
about axions only by virtue of the energy-momentum transfer Ka in
the δ function, is only a function of the axion energy ωa .
Therefore, in terms of a dimensionless function s(x) of the dimen-
sionless axion energy x = ωa /T one may write the energy-loss rate in
the form (baryon density nB )
$ %2 '
CN d 3 ka
Qa = n B Γσ ωa s(ωa /T ) e−ωa /T
2fa 2ωa (2π)3
α a nB Γ σ T 3 ' ∞
= dx x2 s(x) e−x . (4.6)
4π m2N 0
Γσ will turn out to represent the approximate rate of change of a nucleon

spin under the influence of collisions with other nucleons.
4.2.3 Nondegenerate Limit

To find Γσ and s(x) turn first to an evaluation of Eq. (4.4) in the
nondegenerate limit. The initial-state nucleon occupation numbers f1,2
are given by the nonrelativistic Maxwell-Boltzmann
)
distribution fp =
3/2 −p2 /2mN T
(nB /2) (2π/mN T ) e so that 2fp d p/(2π)3 = nB gives the
3
nucleon (baryon) density where the factor 2 is for two spin states. Pauli
blocking factors are omitted: (1 − f3,4 ) → 1.
122 Chapter 4
Because the matrix element has been assumed to be a constant

it can be pulled out of the integral which reduces to a phase-space
volume. Nonrelativistically, d3 pi /[2Ei (2π)3 ] = d3 pi /[2mN (2π)3 ] while
Ei = p2i /2mN in the energy δ function. The axion momentum is ignored
according to Eq. (4.5). One uses CM momenta p1,2 = p0 ±p and p3,4 =
p0 ± q where p0 = 12 (p1 + p2 ) = 12 (p3 + p4 ) and defines u2 ≡ p2 /mN T
and y ≡ v 2 ≡ q2 /mN T . Then one finds explicitly
√ −5/2
Γσ = 4 π απ2 nB T 1/2 mN ,
'
2
s(x) = 4 du dv u2 v 2 e|x|−u δ(u2 − v 2 − |x|)
' ∞ * +1/2 ,
= dy e−y |x|y + y 2 ≈ 1 + |x| π/4 . (4.7)
0
The analytic form is accurate to better than 2.2% everywhere; it has

the correct asymptotic behavior s(0) = 1 and s(|x|)1) = (|x|π/4)1/2 .
We will see in Sect. 4.6.3 that for |x| ) 1 the nondegenerate s(x)
must actually decrease, in conflict with this explicit calculation. This
problem reveals a pathology of the OPE potential which is too singular
at short distances. For the present discussion this is of no concern so
that I stick to the explicit OPE result in order to facilitate comparison
with the existing literature.
To determine the total
)
emission rate one uses the first representation
of s(x) and performs dx first to remove the δ function;
)∞ n
the remaining
integrals are easily done. Explicit results for sn ≡ 0 x s(x) e−x dx are
given in Tab. 4.1. The normalized axion energy spectrum dNa /dx =
x s(x)e−x /s1 is shown in Fig. 4.2 (solid line). The average axion energy
is *ωa +/T = s2 /s1 = 16/7. The total nondegenerate energy-loss rate is
)∞
Table 4.1. sn = 0 xn s(x) e−x dx for nondegenerate (ND) and degenerate
(D) conditions.
n sn (ND) sn (D)
1 8/5 2ζ3 + 6ζ5 /π 2
2 128/35 31π 4 /315
3 256/21 24ζ5 + (180/π 2 ) ζ7
4 4096/77 82π 6 /315
Fig. 4.2. Normalized axion spectrum x s(x)e−x /s1 from nucleon-nucleon

bremsstrahlung emission. The nondegenerate and degenerate functions s(x)
are given in Eqs. (4.7) and (4.9), respectively, while s1 (ND) and s1 (D) are
found in Tab. 4.1.
explicitly
128 αa απ2 n2B T 7/2

QND
a = √ 9/2
,
35 π mN
+ND
a = αa 1.69×1035 erg g−1 s−1 ρ15 TMeV
3.5
, (4.8)
where ρ15 = ρ/1015 g cm−3 , TMeV = T /MeV, and +ND

a = QND
a /ρ is the
energy-loss rate per unit mass.
4.2.4 Degenerate Limit

Details of the nucleon phase space in the degenerate limit can be found
in Friman and Maxwell’s (1979) calculation of νν emission. The axion
energy-loss rate is expressed as in Eq. (4.6). One finds
4απ2 pF T 3 (x2 + 4π 2 ) |x|

Γσ = and s(x) = , (4.9)
3π nB 4π 2 (1 − e−|x| )
where pF is the nucleon Fermi momentum. As in the nondegenerate

)∞ n
case
3 2 −x
s(0) = 1 while s(|x|)1) = |x| /4π . Values for sn = 0 x s(x) e dx
are given in Tab. 4.1. The normalized axion spectrum is shown in
Fig. 4.2 (dotted line), the average energy is *ωa +/T = s2 /s1 ≈ 3.16.
124 Chapter 4
The total energy-loss rate is
31π 2 pF T 6
QD 2
a = α a απ ,
945 m2N
−1 −1 −2/3
+D 31
a = αa 1.74×10 erg g s ρ15 6
TMeV , (4.10)
(Iwamoto 1984; Brinkmann and Turner 1988).

The degenerate and nondegenerate rates are best compared in terms
of a parameter ξ ≡ p2F /(2π mN T ) which approaches η/π in the degen-
erate limit (degeneracy parameter η),
QD 31π 4 −5/2
a
= √ ξ ≈ 1.39 ξ −5/2 . (4.11)
QND
a 1536 2
Therefore, they are equal for ξ ≈ 1, or a degeneracy parameter of

η ≈ 3.5. This defines the dividing line between the regimes where these
approximations can be reasonably used.
In the degenerate limit the nucleon phase-space integrals can be
done analytically with the inclusion of a nonzero mπ . The mπ = 0 rates
must be supplemented with a factor20 (Ishizuka and Yoshimura 1990)
- .
5u 2 u2
G(u) = 1 − arctan + +
6 u 3(u2 + 4)
$ √ %
u2 2 2u2 + 4
+ √ 2 arctan , (4.12)
6 2u + 4 u2
where u = mπ /pF . For only one species of nucleons (as approximately

1/3 −1/3
in a neutron star) pF = 515 MeV ρ15 so that u = 0.26 ρ15 with ρ15
the mass density in 1015 g/cm3 . For this case G is shown as a function
of ρ in Fig. 4.3 (solid line).
4.2.5 Bremsstrahlung Emission of Scalars

The previous results equally apply to pseudoscalars with a coupling
2
igaN ψ N γ5 ψN φ with αa ≡ gaN /4π if a derivative pion-nucleon interac-
tion is used (Sect. 14.2.3). However, for scalars which couple according
20
Eq. (4.12) differs from the corresponding result of Friman and Maxwell (1979)
which is identical with that of Iwamoto (1984) who apparently did not take the
third term of the matrix element Eq. (4.2) properly into account.
Fig. 4.3. Correction of the degenerate bremsstrahlung rate for a nonzero

pion mass. For pseudoscalars, G(mπ /pF ) is given explicitly in Eq. (4.12). It
is assumed that only one species of nucleons is present.
to g ψ N ψN φ the results are different. The degenerate bremsstrahlung

energy-loss rate was worked out by Ishizuka and Yoshimura (1990),
- .4
44 T
Qscalar = α# απ2 p5F Gscalar (mπ /pF ), (4.13)
153 mN
with α# ≡ g 2 /4π. The function Gscalar (u) is similar to Eq. (4.12); it is
plotted in Fig. 4.3 (dashed line).
4.2.6 Mixture of Protons and Neutrons

For a mixed medium of protons and neutrons one needs to consider the
individual Yukawa couplings gan = Cn mN /fa and gap = Cp mN /fa as
well as the isoscalar and isovector combinations g0,1 = 12 (gan ± gap ) and
the “fine-structure constants” αj = gj2 /4π with j = n, p, 0, 1. For equal
couplings αn = αp = α0 while α1 = 0.
In the nondegenerate limit, the main difference is that np scattering
benefits from
√ the exchange of charged pions which couple more strongly
by a factor 2. Depending on the chemical composition of the medium,
the emission rate will be increased by up to a factor of 2. On the other
hand, some reduction factors have been ignored such as the (k̂ · l̂)2 term
and the pion mass. Therefore, ignoring this enhancement essentially
compensates for the previously introduced errors.
In the degenerate limit the changes are more dramatic. The role
of the Fermi momentum is played by pF → (3π 2 nB )1/3 . It sets the
126 Chapter 4
scale for the proton and neutron Fermi momenta which are pn,p F =
(3π 2 nB )1/3 Yn,p
1/3
. According to a result of Brinkmann and Turner (1988)
the effective coupling is then
αa → αn Yn1/3 + αp Yp1/3 + ( 28 α +
3 0
20
α ) (Yn2/3
3 1
+ Yp2/3 )1/2
$ %
1 |Yn2/3 − Yp2/3 |
× √ 2− 2/3 2/3
. (4.14)
2 2 Yn + Yp
The third term is from np collisions; it has the remarkable feature that it
does not vanish as Yp → 0. For equal couplings (Cn = Cp ) the variation
of the emission rate is shown in Fig. 4.4 as a function of Yp = 1 − Yn .
For all proton concentrations the np contribution dominates.
Fig. 4.4. Variation of the degenerate nucleon-nucleon bremsstrahlung rate

with Yp (proton number fraction) according to Eq. (4.14) with equal cou-
plings so that αn = αp = α0 and α1 = 0.
4.3 Neutrino Pair Emission

4.3.1 Structure Function
Neutrino pair emission in nucleon-nucleon collisions (Fig. 4.5) is anal-
ogous to that of axions. Therefore, instead of embarking on a new
calculation it is worth understanding their common features—the neu-
trino rates can be obtained for free on the basis of the axion ones.
This approach amounts to defining the dynamical structure functions
Fig. 4.5. Bremsstrahlung emission of neutrino pairs in nucleon-nucleon colli-

sions. There is a total of eight amplitudes, four with the neutrinos attached
to each nucleon line, and an exchange graph each with N3 ↔ N4 .
of the medium, quantities which are of utmost importance to under-

stand the general properties of the emission, absorption, and scattering
rates independently of phase-space details of the neutrinos or axions.
Contrary to the νν bremsstrahlung emission in electron-nucleus col-
lisions (Sect. 3.5.3), nonrelativistically only the nucleon axial-vector
coupling contributes in nucleon-nucleon collision (Friman and Maxwell
1979). This difference originates from the interaction potential of the
colliding particles which involves a spin-dependent force between nucle-
ons so that the spin fluctuations caused by collisions are more dramatic
than those of the velocity—see Sect. 4.6.5 below. Hence, the part of the
interaction Hamiltonian relevant for neutrino pair bremsstrahlung is
CAN GF
Hint = √ ψ N γµ γ5 ψN ψ ν γ µ (1 − γ5 )ψν , (4.15)
2
where CAp ≈ −CAn ≈ 12 ; see Appendix B for a discussion of the appro-
priate values in a nuclear medium.
The interaction Hamiltonian Eq. (4.15) has the same structure as
that for axions Eq. (4.1). The squared matrix elements are then of the
general form
 √
!  (C N GF / 2)2 Mµν N µν for neutrinos,
A
|M|2 = (4.16)
 (C /2f )2 M K µ K ν for axions.
spins N a µν a a
Here, Ka is the axion four-momentum while

* +
N µν = 8 K1µ K2ν + K2µ K1ν − K1 · K2 g µν − i+αβµν K1α K2β (4.17)
with the neutrino and antineutrino four momenta K1 and K2 (Gaemers,

Gandhi, and Lattimer 1989). N µν and Kaµ Kaν are the squared matrix
elements of the neutrino and axion current, respectively.
128 Chapter 4
The matrix M µν is the nuclear part of the squared matrix element.

It is exactly the same for axion or neutrino interactions because in
Eq. (4.16) the global coupling constants have been explicitly pulled
out. Therefore, one may go one step further and perform the entire
nucleon phase-space integration for both cases directly on M µν ,
1 ' (
4
d3 pi
S µν ≡ f1 f2 (1 − f3 )(1 − f4 )
nB i=1 2Ei (2π)3
× (2π)4 δ 4 (P1 + P2 − P3 − P4 + K) M µν . (4.18)
Here, the Pi are the nucleon four momenta, K = −Ka for axion emis-
sion, and K = −(K1 + K2 ) for neutrino pairs. Thus, K is the energy-
momentum transfer from the radiation (axions or neutrino pairs) to
the nucleons. Because S µν knows about the radiation only through
the energy-momentum δ function it is only a function of K = (ω, k),
apart from the temperature and chemical potentials of the medium.
The slightly awkward definition of the sign of K follows the common
definition of the structure function where a positive energy transfer ω
refers to energy given to the medium.
The energy-loss rates by νν or axion emission are then the phase-
space integrals
$ %2 '
CAN GF d3 k 1 d3 k 2
Qνν = √ nB Sµν N µν (ω1 + ω2 ),
2 2ω1 (2π)3 2ω2 (2π)3
$ %2 '
CN d 3 ka
Qa = nB Sµν Kaµ Kaν ωa . (4.19)
2fa 2ωa (2π)3
Here, it was assumed that both axions and neutrinos can escape freely
from the medium so that final-state Pauli blocking or Bose stimulation
factors can be ignored.
In the nonrelativistic limit the nucleon current in Eqs. (4.1) and
(4.15) reduces to χ† τi χ where χ is a nucleon two-spinor and τi (i =
1, 2, 3) are Pauli matrices representing the nucleon spin operator. Put
another way, in the nonrelativistic limit the axial-vector current rep-
resents the nucleon spin density. Therefore, it has only spatial com-
ponents so that S µν → S ij (i, j = 1, 2, 3). In order to construct the
most general tensorial structure for S ij in an isotropic medium only δij
is available. Recall that in the nonrelativistic limit S µν does not know
about the momentum transfer k because of Eq. (4.5). There is then no
vector available from which a spatial tensor can be constructed. Thus,

the structure function has the most general form
Sij (ω) = Sσ (ω) δij . (4.20)
For nonrelativistic nucleons all axion or axial-vector neutrino processes

involve only one scalar function Sσ (ω) of the energy transfer.
The contraction of δij with Kaµ Kaν for axion emission yields k2a =
2
ωa . (The spatial Kronecker δ should be viewed as a Lorentz tensor
with a zero in the 00 position.) Therefore, Sµν Kaµ Kaν → ωa2 Sσ (−ωa )
in Eq. (4.19). For neutrino emission, K1 · K2 = ω1 ω2 − k1 · k2 and
the contraction of δij with gµν is −3. Therefore, the contraction of
δij with N µν yields 8ω1 ω2 (3 − cos θ) where θ is the angle between the
ν and ν momenta. The neutrino phase-space integration will always
average cos θ to zero so that it may be dropped. Therefore, Sµν N µν →
24 ω1 ω2 Sσ (−ω1 −ω2 ) in Eq. (4.19). One may then immediately perform
the integration over one of the neutrino energies and is left with an
integration over the energy transfer.
Thus, in the nonrelativistic limit the energy-loss rates Eq. (4.19) are
of the form
$ %2
CAN GF nB ' ∞
Qνν = √ dω ω 6 Sσ (−ω),
2 20π 4 0
$ %2
CN nB ' ∞
Qa = dω ω 4 Sσ (−ω). (4.21)
2fa 4π 2 0
The medium properties are embodied in a common function which may
be expressed in the form
2
Γσ 1 for ω > 0,
Sσ (ω) = 2 s(ω/T ) × (4.22)
ω eω/T for ω < 0.
Because of the “detailed-balance relationship” between positive and
negative energy transfers to be discussed more fully below, s(x) must
be an even function. In the nondegenerate and degenerate limits Γσ
and s(x) have been determined above. They allow one to calculate the
νν emission rate without any effort.
130 Chapter 4
4.3.2 Bremsstrahlung Emission of Neutrino Pairs

In order to calculate the neutrino pair emission rate explicitly turn
first to the nondegenerate limit for which one uses Γσ and s(x) given
in Eq. (4.7) and the integrals of Tab. 4.1. The average energy of a
neutrino pair is (s4 /s3 ) T ≈ 4.36 T and the total energy-loss rate is
2 11/2
2048 2 2 2 nB T
QND
νν = C G α ,
385 π 7/2 A F π m5/2 N
+ND
νν = 2.4×1017 erg g−1 s−1 ρ15 TMeV
5.5
, (4.23)
with ρ15 = ρ/1015 g cm−3 , TMeV = T /MeV, and +ND ND

νν = Qνν /ρ is the
energy-loss rate per unit mass.
For the degenerate rate one uses Eqs. (4.9). The average energy of
a neutrino pair is (s4 /s3 ) T ≈ 5.78 T and the total energy-loss rate is21
41π 2 2 2
QD
νν = C G α pF T 8 ,
4725 A F π
−1 −1 −2/3
+D 13
νν = 4.4×10 erg g s ρ15 8
TMeV . (4.24)
One can correct for a nonvanishing value of the pion mass by virtue of
Eq. (4.12).
For a mixture of protons and neutrons the same remarks as in
Sect. 4.2.6 apply. Apart from a small correction the neutrino cou-
pling is isovector (CAp ≈ −CAn ) so that α0 ≈ 0 while the other α’s are
approximately equal (Appendix B). For degenerate conditions, with a
small modification the dependence on the proton concentration is the
same as that shown in Fig. 4.4. Again, the absolutely dominating con-
tribution is from np collisions unless protons are so rare that they are
nondegenerate.22
21
Friman and Maxwell’s (1979) total energy-loss rate is 2/3 of the one found here.
Apparently they did not include the crossterm in the squared matrix element, i.e.
the third term in Eq. (4.2).
22
This conclusion, based on the work of Brinkmann and Turner (1988), is in
conflict with the results of Friman and Maxwell (1979). They found that Qνν was
proportional to the proton Fermi momentum which is relatively small in neutron-
star matter. On the other hand, for small proton concentrations Brinkmann and
Turner’s pF in Eq. (4.24) approaches the neutron Fermi momentum. I am in no
position to decide between these conflicting results.
4.4 Axion Opacity

When axions or other pseudoscalar bosons interact so “strongly” that
they are trapped in a young SN core, they still contribute to the ra-
diative transfer of energy and can thus have an impact on the cooling
speed. In order to include axions in a numerical evolution calculation
one needs to determine the opacity of the medium to axions (Burrows,
Ressell, and Turner 1990).
The “reduced” Rosseland mean opacity relevant for radiative trans-
port by relativistic bosons (Sect. 1.3.4) is defined by
1 15 ' ∞
= 2 3 dω 3ω (1 − e−ω/T )−1 ∂T Bω (T ), (4.25)
ρκa 4π T 0
where ρ is the mass density of the medium, 3ω the boson mean free path
against absorption, and Bω (T ) = (2π 2 )−1 ω 3 (eω/T − 1)−1 is the boson
spectral density for one spin degree of freedom. After ∂T = ∂/∂T has
been taken one finds
1 15 ' ∞ x4 e2x
= 4 dx 3x x , (4.26)
ρκa 8π 0 (e − 1)3
where x ≡ ω/T , and (1 − e−x )−1 = ex (ex − 1)−1 was used.
The axion opacity thus defined is to be added to the photon opacity
by κ−1 −1 −1
tot = κγ + κa . The energy flux is then given by the usual ex-
pression F = −(3κtot ρ)−1 ∇aγ T 4 where aγ = π 2 /15 gives the radiation
density of photons. Burrows, Ressell, and Turner (1990) have defined
an axion opacity which is to be used in conjunction with the axion radi-
ation density which is half as large because there is only one spin degree
of freedom, i.e. aa = π 2 /30. Their κ−1a is twice that in Eq. (4.26) so
that the energy flux is the same. I prefer the present definition because
it allows for the usual addition of all opacity contributions.
It is easy to determine the axion absorption rate from the discussion
in Sect. 4.3.1. Starting
)
from Qa in Eq. (4.21) one removes the phase-
space integral 0∞ dω 4πω 2 /(2π)3 and a factor ω because Qa was an
energy-loss rate. One includes a factor eω/T to account for the detailed-
balance relationship (see Eq. 4.43) so that altogether
$ %2
CN nB Γσ s(x)
3−1
ω = , (4.27)
2fa T 2x
where x = ω/T . Therefore, the reduced Rosseland mean opacity is
$ %2
CN Γσ
κa = κ̂, (4.28)
2fa T mN
132 Chapter 4
where ρ/nB ≈ mN was used. The dimensionless opacity is

'
−1 15 ∞ x4 e2x 2x
κ̂ ≡ 4 dx x . (4.29)
8π 0 (e − 1)3 s(x)
For the nondegenerate case s(x) was given in Eq. (4.7), for the degen-
erate one in Eq. (4.9). Then one finds numerically κ̂ND = 0.46 and
κ̂D = 1.53. The respective Γσ ’s were given in Eqs. (4.7) and (4.9).
4.5 Neutrino Opacity

4.5.1 Elastic Scattering
The trapping of neutrinos in a hot SN core allows them to escape only
by diffusion. This implies that they transport energy in a fashion similar
to radiative transfer. Apart from proper spectral weights as in the
Rosseland mean opacity, the main figure of merit that determines the
transport efficiency is the neutrino mean free path (mfp). While the
charged-current absorption of νe ’s is very important for practical SN
core-cooling calculations it has no direct bearing on the main issues of
interest here.
However, all neutrino flavors interact by neutral-current interactions
which allow for the scattering on nucleons. The inverse mean free path
is then simply the cross section times the nucleon density. If there is
only one species of nondegenerate nucleons one easily finds
λ−1 = (CV2 + 3CA2 ) G2F nB ω12 /π, (4.30)
where ω1 is the energy of the incident neutrino, assumed to be much

smaller than mN so that recoil effects can be neglected. For a mixture
of protons and neutrons one has to take a proper average with the
coupling constants CV and CA for protons and neutrons (Appendix B).
For νe ’s which have a large chemical potential, a Pauli blocking factor
must be included, and the same for nucleons if they are degenerate.
4.5.2 Pair Absorption

Because the neutrinos are assumed to be trapped there is an ambient
bath of ν’s and ν’s which allows for neutrino absorption by the inverse
bremsstrahlung process νν N N → N N , i.e. Fig. 4.5 read from right to
left. The rate for this process is closely related to that for pair emission:
it is based on the same matrix element with the neutrinos “crossed”

into the initial state. Then one finds in analogy to Sect. 4.3.1
$ %2
−1 C A GF nB ' d 3 k2
λ = √ 3
f2 Sµν N µν
2 2ω1 2ω2 (2π)
3CA2 G2F nB ' ∞
= dω2 ω22 f2 Sσ (ω1 + ω2 ), (4.31)
2π 2 0
where 1 refers to the ν for which the mfp is being determined while
2 refers to a ν from the thermal environment (occupation number
f2 ). With the detailed-balance relationship Sσ (ω) = Sσ (−ω) eω/T (see
Eq. 4.43) and writing Sσ (|ω|) = (Γσ /ω 2 ) s(ω/T ) as before one finds
'
−1
∞ x22 s(x1 + x2 )
λ = 3CA2 G2F n B Γσ T dx2 f2 , (4.32)
0 2π 2 (x1 + x2 )2
where xi = ωi /T . Then one may use the previously determined Γσ and
s(x) to find the mfp for given conditions (degenerate or nondegenerate).
The ratio between the inverse mfp’s from elastic scattering and pair
absorption is, ignoring the contribution from CV2 ,
λ−1
pair Γσ ' ∞ x22 s(x1 + x2 )
= dx 2 f 2 . (4.33)
λ−1
scat T 0 2π x21 (x1 + x2 )2
An average with regard to a thermal x1 distribution yields about 0.02
for the dimensionless integral where Fermi-Dirac distributions with
chemical potentials µν = 0 were used. The main figure of merit, how-
ever, is Γσ /T , the ratio of a typical spin-fluctuation rate and the ambi-
ent temperature. In the nondegenerate limit one finds with Eq. (4.7)
- .1/2
Γσ nB ρ 30 MeV
γσ ≡ = 4π 1/2 απ2 5/2 ≈ 16 , (4.34)
T mN T 1/2 ρ0 T
with the nuclear density ρ0 = 3×1014 g/cm3 .
4.5.3 Inelastic Scattering

It appears that for typical conditions of a young SN core, pair absorp-
tion is almost as important as elastic scattering. However, even though
the quantity γσ is larger than unity, the pair-absorption rate has an
unfavorable phase-space factor from the initial-state ν so that the di-
mensionless integral in Eq. (4.33) is a small number. This would not
be the case for the inelastic scattering process νN N → N N ν shown in
134 Chapter 4
Fig. 4.6. Inelastic neutrino-nucleon scattering. There is a total of eight

amplitudes, four with the neutrinos attached to each nucleon line, and an
exchange graph each with N3 ↔ N4 .
Fig. 4.6. In this process, neutrinos can give or take energy even though
recoil effects for heavy nucleons are small in the elastic scattering pro-
cess νN → N ν.
This reaction is identical with pair absorption with the antineutrino
line crossed into the final state. The corresponding mfp is given by the
same expression as for pair absorption, except that the index 2 now
refers to the final-state ν, the initial-state occupation number f2 is to
be replaced with a final-state Pauli blocking factor (1 − f2 ), and the
energy transfer is ω = ω2 − ω1 ,
3CA2 G2F nB ' ∞

λ−1 = dω2 ω22 (1 − f2 ) Sσ (ω1 − ω2 ). (4.35)
2π 2 0
However, because the energy transfer can be zero, and because Sσ (ω) ∝
ω −2 , this expression diverges. The ω −2 behavior seemed harmless before
because it was moderated by powers of ω from the phase space of axions
or neutrinos.
The occurrence of this divergence could have been predicted without
a calculation by inspecting Fig. 4.6. If one cuts the intermediate-state
nucleon line, this graph falls into two sub-processes (nucleon-nucleon
scattering and nucleon-neutrino scattering) which are each permitted
by energy-momentum conservation, allowing the intermediate nucleon
in the compound process to go “on-shell.” Therefore, the pole of the
propagator which corresponds to real particles causes a divergence of
the cross section. Physically, the divergence reflects a long-range inter-
action which occurs because the intermediate nucleon can travel arbi-
trarily far when it is on its mass shell.
Still, the inelastic scattering process is an inevitable physical possi-
bility. For nonzero energy transfers its differential rate is given by the
unintegrated version of Eq. (4.35). For a vanishing energy transfer it is
elastic and then its rate should be given by Eq. (4.30).
However, what exactly does one mean by a vanishing energy trans-

fer? The nucleons in the ambient medium constantly scatter with each
other so that their individual energies are uncertain to within about
1/τcoll with a typical time between collision τcoll . Therefore, one would
expect that the structure of Sσ (ω), which was calculated on the basis
of free nucleons which interact only once, is smeared out over scales
−1
of order ∆ω ≈ τcoll . In particular, this smearing-out effect naturally
regulates the low-ω behavior of the structure function (Raffelt and
Seckel 1991).
−1
Because Γσ ≈ τcoll is a typical nucleon-nucleon collision rate, a sim-
ple ansatz for a modified Sσ (ω) is a Lorentzian (Raffelt and Seckel 1991)
Γσ 1 γσ
Sσ (|ω|) → s(ω/T ) = s(x), (4.36)
ω 2 + Γ2σ /4 T x2 + γσ2 /4
)
+∞
with x = ω/T . For γσ ' 1 one has −∞ dω Sσ (ω) = 2π + O(γσ ) for the
modified Sσ (ω). Thus, for γσ ' 1 essentially Sσ (ω) = 2πδ(ω) so that
the total “inelastic” scattering rate Eq. (4.35) reproduces the elastic
scattering one. At the same time Sσ (ω) has wings which, for ω ) Γσ ,
give the correct inelastic scattering rate which is of order γσ .
This simple ansatz can be expected to give a reasonable approx-
imation to the true Sσ (ω) only in the limit γσ ' 1. For practical
applications in cooling calculations of young SN cores, however, one
has to confront the opposite limit γσ ) 1 causing the smearing-out ef-
fect by multiple nucleon collisions to be a dominating feature of Sσ (ω)
even for ω ) T . This implies that for typical thermal energies of
neutrinos and nucleons a reasonably clean separation between elastic
and inelastic scattering processes is not logically possible—there is only
one structure function Sσ (ω), broadly smeared out, which governs all
axial-vector scattering, emission, and absorption processes.
This observation has important ramifications not only for neutrino
scattering, but also for the bremsstrahlung emission of axions and neu-
trino pairs from a nuclear medium. Earlier it seemed that one did not
have to worry about details of the behavior of Sσ (ω) near ω = 0 be-
cause the low-ω part was suppressed by axion or neutrino phase-space
factors. However, since the notion of “low energy” presently means
ω< ∼ Γσ , and because T ' Γσ , all relevant energy transfers are low in
this sense, and even the emission processes are dominated by multiple-
scattering effects. Consequences of this behavior are explored in more
detail below after a formal introduction of the structure functions and
their general properties.
136 Chapter 4
4.6 Structure Functions

4.6.1 Formal Definition
To treat axion and neutrino pair emission and absorption on the same
footing it became useful in Sect. 4.3.1 to define the quantity S µν which
was the nuclear part of the squared matrix element of nucleon-nucleon
collisions, integrated over the nucleon phase space. The structure func-
tion thus obtained embodied the medium properties relevant for differ-
ent processes without involving the radiation phase space. Clearly, this
method is not limited to the bremsstrahlung process. For example, one
could include the interaction of nucleons with thermal pions or a pion
condensate, three-nucleon collisions, and so forth. Whatever the de-
tails of the medium physics, in the end one will arrive at some function
S µν (ω, k) of the energy and momentum transfer which embodies all of
its properties. High-density properties of the medium should also ap-
pear in the structure function so that multiple-scattering modifications
can be consistently applied to all relevant processes such as neutrino
scattering, pair emission, axion emission, and others.
A formal definition of the structure function without reference to
specific processes begins with a neutral-current interaction Hamiltonian
Hint = (gV V µ + gA Aµ ) Jµ , (4.37)
where gV and gA are (usually dimensionful) coupling constants. The
radiation or “probe” is characterized by a current J µ which for axions
is ∂ µ φ, for neutrino interactions ψ ν γ µ (1 − γ5 )ψν , and for photons the
electromagnetic vector potential. The medium is represented by the
vector and axial-vector currents V µ and Aµ . If the probe couples only
to one species N of nucleons, V µ = ψ N γ µ ψN and Aµ = ψ N γ µ γ5 ψN .
Next, one imagines that the medium properties are experimentally
investigated with a neutrino beam with fixed momentum k1 which is
directed at a bulk sample of the medium, and the distribution of final-
state momenta and energies are measured. The transition probability
W (k1 , k2 ) is proportional to (gV2 SVµν +gA2 SAµν +gV gA SVµνA )Nµν where Nµν
was defined in Eq. (4.17). A standard perturbative expansion (Sect. 9.3)
yields for the dynamical structure functions
1 ' +∞ 3 4
SVµν (ω, k) = dt eiωt V µ (t, k)V ν (0, −k) , (4.38)
nB −∞
and analogous expressions for SAµν in terms of *Aµ Aν + and for SVµνA
involving *V µ Aν + Aµ V ν +. The expectation values are to be taken with
respect to a thermal ensemble of medium states.
In addition, one frequently uses the static structure functions which

are functions of the momentum transfer alone. For example,
' +∞ dω µν 1 3 µ 4
SVµν (k) = SV (ω, k) = V (k)V ν (−k) . (4.39)
−∞ 2π nB
+∞ ) )
It was used that −∞ dω eiωt = 2πδ(t) so that dt in Eq. (4.38) is
trivially done and yields the operators at equal times. In Eq. (4.39)
V (k) is V (t, k) at an arbitrary time, for example t = 0. It only matters
that both V (k) and V (−k) are taken at equal times.
In an isotropic medium the tensorial composition of the dynamical
structure function can be obtained only from the energy-momentum
transfer K and the four-velocity U of the medium; U = (1, 0, 0, 0) in its
rest frame. The general form of the vector term is (Kirzhnits, Losyakov,
and Chechin 1990)
SVµν = S1,V U µ U ν + S2,V (U µ U ν − g µν )
+ S3,V K µ K ν + S4,V (K µ U ν + U µ K ν ). (4.40)
An analogous expression pertains to SAµν while the mixed term is
SVµνA = i SV A +µναβ Uα Kβ (4.41)
because of its transformation properties under parity. The functions
S(,V and S(,A (3 = 1, . . . , 4), and SV A depend on medium properties
and on the Lorentz scalars K 2 and U · K that can be constructed from
U and K; the third possibility U 2 = 1 is a constant. Instead of K 2 and
U · K one may use the energy and momentum transfer ω and k = |k|
measured in the medium rest frame.
The structure functions are defined for both positive and negative
energy transfers because the medium can both give or take energy from
a probe. Taking axion emission and absorption as an example, the rate
of change of the occupation number of an axion field mode k is given by
$ %2
CA nB 5 6
∂ t fk = (fk + 1) SAµν (−ω, k) − fk SAµν (ω, k) Kµ Kν .
2fa 2ω
(4.42)
If axions are trapped and reach thermal equilibrium, ∂t fk = 0 and fk =
(eω/T − 1)−1 , a Bose-Einstein distribution. This implies the detailed-
balance condition
SAµν (ω, k) = SAµν (−ω, k) eω/T . (4.43)
Recall that a positive energy transfer is energy given to the medium.
138 Chapter 4
4.6.2 Nonrelativistic Limit

In a nuclear medium one is interested primarily in the nonrelativistic
limit. The vector current is then dominated by V 0 = ρ = χ† χ where χ
is a nucleon two-spinor. Here, ρ is the operator for the nucleon number
density. The spatial component V is suppressed by a nonrelativistic
velocity factor v. The reverse applies to the axial-vector current where
A0 is suppressed; it is dominated by the spin density s = 12 χ† τ χ where
τ is a vector of Pauli matrices.
In Eqs. (4.40) and (4.41) all terms arise from correlators such as
*A0 V i + which are suppressed by v 2 , except for the first term of SVµν
which arises from *V 0 V 0 +, and the second term of SAµν which arises
from *Ai Ai +. Therefore, the only unsuppressed components are
SV00 (ω, k) = Sρ (ω, k) and SAij (ω, k) = Sσ (ω, k) δ ij , (4.44)
where the density and spin-density dynamical structure functions are

1 ' +∞ 3 4
Sρ (ω, k) = dt eiωt ρ(t, k)ρ(0, −k) ,
nB −∞
4 ' +∞ 3 4
Sσ (ω, k) = dt eiωt s(t, k) · s(0, −k) (4.45)
3nB −∞
(Iwamoto and Pethick 1982).
In order to determine the overall normalization in the nonrelativis-
tic limit consider the static structure function for an ensemble of NB
nucleons enclosed in a large volume V . At a given time the system is
characterized by a wavefunction Ψ which depends on the locations ri
7 B
of the nucleons. Then ρ(r))|Ψ+ = N i=1 δ(r − ri ) |Ψ+ while the Fourier-
−1 3 ik·r
transformed operator V d r e ρ(r) is
NB
1 !
ρ(k)|Ψ+ = eik·ri |Ψ+. (4.46)
V i=1
Therefore, a given configuration of nucleons yields

NB NB
1 ! NB 1 !
*Ψ|ρ(k)ρ(−k)|Ψ+ = eik·rij = + eik·rij , (4.47)
V i,j=1 V V i,j=1
i!=j
where rij ≡ ri −rj . When averaged over a thermal ensemble the second
term will disappear if there are no spatial correlations, and nB = NB /V
so that Sρ (k) = 1.
For the static spin-density structure function the same steps can be
performed with the inclusion of the spin operators 12 σ i for the individual
nucleons. This takes us to the equivalent of Eq. (4.47)
NB
1 !
*Ψ|s(k) · s(−k)|Ψ+ = ek·rij 14 *Ψ|σ i · σ j |Ψ+ =
V i,j=1
NB NB
1 ! 2 1 !
= 1
*Ψ|σ i |Ψ+ + eik·rij 14 *Ψ|σ i · σ j |Ψ+. (4.48)
V i=1 4 V i,j=1
i!=j
Noting that *( 12 σ i )2 + = 12 (1 + 12 ) = 34 the first term is 34 NB /V = 34 nB .

Therefore, in the absence of correlations one finds Sσ (k) = 1.
Even in a noninteracting medium there exist anticorrelations be-
tween degenerate nucleons. Standard manipulations yield in this case
(e.g. Sawyer 1989)
1 ' 2d3 p
Sρ,σ (k) = fp (1 − fp+k ), (4.49)
nB (2π)3
with the Fermi-Dirac occupation number fp for the nucleon mode p.

For small temperatures this result can be expanded to yield Sρ,σ (k) =
3k/2pF + 3T mN /p2F .
4.6.3 The f-Sum Rule

The structure functions have a number of general properties, indepen-
dently of details of the interactions of the medium constituents. We
have already seen that they must obey the normalization condition
' 8 NB
9
+∞ dω 1 !
Sρ (ω, k) = 1 + cos(k · rij ) ,
−∞ 2π nB i,j=1
i!=j
' 8 NB
9
+∞ dω 4 !
Sσ (ω, k) = 1 + σ i · σ j cos(k · rij ) . (4.50)
−∞ 2π 3nB i,j=1
i!=j
This can be referred to as a “sum rule” because the strength of Sρ,σ is

“summed” (integrated) over all frequencies ω. Usually we will assume
that the correlation expressions on the r.h.s. are negligible.
A more nontrivial sum rule obtains when a factor ω is included
under the integral. The definition of the density structure function
140 Chapter 4
then yields
' ' +∞
+∞ dω dω ' +∞
ωSρ (ω, k) = ω dt eiωt *ρ(t, k)ρ(0, −k)+
−∞ 2π −∞ 2π −∞
(4.51)
and a similar expression for Sσ . Under the integral, a partial integration
with suitable boundary conditions allows one to absorb the ω factor,
at the expense of ρ(t, k) → ρ̇(t, k). Because Heisenberg’s equation of
motion informs us that iρ̇ = [ρ, H] with H the complete Hamiltonian
of the system one finds (Sigl 1995b)
' +∞ dω 1 35 6 4
ω Sρ (ω, k) = ρ(k), H ρ(−k) ,
−∞ 2π nB
' +∞ dω 4 35 6 4
ω Sσ (ω, k) = σ(k), H · σ(−k) . (4.52)
−∞ 2π 3nB
)
Here it was used, again, that dω eiωt = 2πδ(t) and ρ(k) ≡ ρ(0, k) and
σ(k) ≡ σ(0, k).
In order to evaluate this sum rule more explicitly one must assume
a specific form for the interaction Hamiltonian. In the simplest case of
a medium consisting of only one species of nucleons one may assume
that H consists of the kinetic energy for each nucleon, plus a general
nonrelativistic interaction potential between all nucleon pairs which
depends on the relative distance and the nucleon spins, i.e.
NB NB
! p2i
1 !
H= + V (rij , σ i , σ j ), (4.53)
i=1 2mN 2 i,j=1
i!=j
where again rij ≡ ri −rj . One can then proceed to evaluate the commu-
tators in Eq. (4.52). By virtue of the continuity equation for the particle
number one can then show (Pines and Nozières 1966; Sigl 1995b)
' +∞ dω k2
ω Sρ (ω, k) = ,
−∞ 2π 2mN
' +∞ dω k2
ω Sσ (ω, k) =
−∞ 2π 2mN
8 NB 5
9
4 ! 6
+ σ(k), V (rij , σ i , σ j ) · σ(−k) . (4.54)
3nB i,j=1
i!=j
For the density structure function this exact relationship is known as

the f-sum rule.
For the spin-density structure function one can go one step further
by expressing the most general nonrelativistic interaction potential as
(Sigl 1995b)
V (rij , σ i , σ j ) = U0 (rij ) − US (rij ) σ i · σ j

5 6
− UT (rij ) 3(σ i · r̂ij )(σ j · r̂ij ) − σ i · σ j , (4.55)
where U0 is a spin-independent potential of the interparticle distance

rij = |rij |, US is the scalar, and UT the tensor part of the spin-dependent
potential. Of these terms, only the tensor part does not conserve the
total spin in nucleon-nucleon collisions and thus is the only part con-
tributing to spin fluctuations. With VijS ≡ US (rij ) σ i · σ j and VijT ≡
UT (rij ) [3(σ i · r̂ij )(σ j · r̂ij ) − σ i · σ j ] one finds
' +∞ dω k2
ω Sσ (ω, k) =
−∞ 2π 2mN
8 NB 5
9
4 ! 6 5 6
+ 1 − cos(k · rij ) VijS 1
+ 1 + cos(k · rij )
2
VijT . (4.56)
3nB i,j=1
i!=j
The f-sum of the spin-density structure function is thus closely related

to the average spin-spin interaction energy in the medium.
In order for the eigenvalues of the Hamiltonian to be bounded from
below one must require that the potentials U0,S,T (r) are not more sin-
gular than 1/r2 . In this case the r.h.s. of the spin-density f-sum rule
exists as a nondivergent expression. In our representation Sσ (|ω|) =
(Γσ /ω 2 ) s(ω/T ), the necessary existence of the l.h.s. of Eq. (4.56) im-
plies that s(x) must be a decreasing function of x for large x. This is
not the case for the s(x) derived from the OPE potential, indicating
that this interaction model is pathological in the sense that it is too
singular. Indeed, it corresponds to a dipole potential and thus varies
as 1/r3 . Real nucleon-nucleon interaction potentials have a repulsive
core and thus do not exhibit this pathology.
4.6.4 Long-Wavelength Properties
In calculations involving the emission or scattering of neutrinos or ax-

ions as in Sect. 4.3.1 one usually neglects the momentum transfer k
because the medium constituents are so heavy that recoil effects are
142 Chapter 4
small. In this “long-wavelength limit” one is only interested in the

small-k structure functions
Sρ,σ (ω) ≡ lim Sρ,σ (ω, k). (4.57)
k→0
It should be stressed that this quantity is not identical with Sρ,σ (ω, 0).
For example, in Eq. (4.47) for k = 0 the interference term does not
average to zero. The structure function becomes NB2 /V and thus co-
herently enhanced because the momentum transfer is so small that a
target consisting of many particles in a volume V cannot be resolved.
The limit k → 0 is understood such that |k|−1 remains much smaller
than the geometrical dimension V 1/3 of the system.
In the long-wavelength limit the normalization Eq. (4.50) and the
f-sum rule Eq. (4.56) yield for the spin-density structure function
' ' ∞ 8 NB
9
+∞ dω dω 4 !
Sσ (ω) = (1 + e−ω/T )Sσ (ω) = 1 + σi · σj ,
−∞ 2π 0 2π 3nB i,j=1
i!=j
' ' ∞ 8 NB
9
+∞ dω dω 4 !
ω Sσ (ω) = ω (1 − e−ω/T )Sσ (ω) = 3 T
V
2 ij
.
−∞ 2π 0 2π 3nB i,j=1
i!=j
(4.58)
These relations will be of great use to develop a general understanding
of the behavior of Sσ (ω) at high densities. The second column of expres-
sions follows from the first by detailed balance. Because Sσ (ω) ≥ 0 it is
evident that all of these expressions are always positive, independently
of details of the medium interactions.
In a noninteracting medium the operators ρ(t, k) and s(t, k) are
constant so that Sρ,σ (ω) = 2πδ(ω), allowing for scattering (zero energy
transfer), but not for the emission of radiation. This behavior is familiar
from a gas of free particles which can serve as targets for collisons, but
which cannot emit radiation because of energy-momentum constraints.
In an interacting medium the density correlator retains this prop-
erty )because in the long-wavelength limit it depends on ρ(t, k→0) =
V −1 d3 r ρ(t, r) which remains constant. Therefore, even in an inter-
acting medium one expects Sρ (ω) = 2πδ(ω), in agreement with the
finding that the neutrino vector current does not contribute to brems-
strahlung in the nonrelativistic limit relative to the axial-vector current
(Friman and Maxwell 1979). ) 7 B
The relevant quantity for the latter is V −1 d3 r s(t, r) = 12 N i=1 σ i
with σ i the individual nucleon spins. If the evolution of different spins
is uncorrelated one may ignore the cross terms in the correlator. With
the single-nucleon spin operator σ one finds then
' +∞ 3 4
Sσ (ω) = 1
3
dt eiωt σ(t) · σ(0) . (4.59)
−∞
With Eq. (4.21) one obtains

$ %2 '
dI˙a CN ω4 +∞ 3 4
= dt e−iωt σ(t) · σ(0) (4.60)
dω 2fa 12π 2 −∞
for the differential axion energy-loss rate (radiation power) per nucleon.
Because of collisions with other nucleons, and because of a spin depen-
dent interaction potential caused by pion exchange, the nucleon spins
evolve nontrivially so that the correlator has nonvanishing power at
ω 0= 0, allowing for axion emission.
4.6.5 Axion Emission in the Classical Limit

The correlator representation Eq. (4.60) of the axion emission rate is
extremely useful to develop a general understanding of its main proper-
ties without embarking on a quantum-mechanical calculation. To this
end the nucleon spin σ is approximated by a classical variable, basically
a little magnet which jiggles around under the impact of collisions with
other nucleons. With an ergodic hypothesis about the spin trajectory
on the unit sphere one may replace the ensemble average in Eq. (4.60)
by a time average. The radiation intensity (time-integrated radiation
power) emitted during a long (infinite) time interval is then
$ %2 :' :2
dIa CN ω2 :
:
+∞
iωt
:
= : dt e σ̇(t):: , (4.61)
dω 2fa 12π 2 −∞
where two powers of ω were absorbed by a partial integration with

suitable boundary conditions at t = ±∞.
In this form the energy-loss rate is closely related to a well-known
expression for the electromagnetic radiation power from a charged par-
ticle which moves on a trajectory r(t). The nonrelativistic limit of a
standard result (Jackson 1975) is
' : :
dIγ 2α :: +∞ iωt
:2
= dt e a(t):: , (4.62)
dω 3π : −∞
where a(t) = r̈(t) is the particle’s acceleration on its trajectory.
144 Chapter 4
The same result can be found with the above methods applied to the
spatial part of the vector current. Easier still, it can be obtained directly
from Eq. (4.61). To this end note that photon emission by an electron
involves nonrelativistically (e/me ) p = ev so that we must substitute
σ̇ → v̇ = a. The role of k (axions) is played by the polarization
vector # (photons) so that k2 = ω 2 must be replaced by #2 = 1. With
(CA /2fa ) → e, using α = e2 /4π, and inserting a factor of 2 for two
photon polarization states completes the translation.
In order to understand the radiation spectrum consider a single
“infinitely hard” collision with σ̇(t) = ∆σ δ(t). The radiation power is
$ %2
dIa CN ω2
= 2
|∆σ|2 . (4.63)
dω 2fa 12π
For photons one obtains the familiar flat bremsstrahlung spectrum
dIγ /dω = (2α/3π)|∆v|2 which is hardened, for axions, by the addi-
tional factor ω 2 from their derivative coupling.
In the form Eq. (4.63) the total amount of energy radiated in a sin-
gle collision is infinite. In practice, collisions are not arbitrarily hard,
and the backreaction of the radiation process on the emitter must be
included. This is not rigorously possible in a classical calculation, it re-
quires a quantum-mechanical treatment. Therefore, a classical analysis
is useful only for the soft part of the spectrum where backreactions can
be ignored, i.e. for radiation frequencies far below the kinetic energy
of the emitter. In a thermal environment, a classical treatment then
appears reasonable for ω < ∼ T.
Next, consider a large random sequence of n hard collisions with a
spin trajectory
n
!
σ̇(t) = ∆σ i δ(t − ti ). (4.64)
i=1
This yields the average radiation intensity per collision of

$ %2
dI˙a CN ω2
= Γcoll *(∆σ)2 + F (ω), (4.65)
dω 2fa 12π 2
where Γcoll is the average collision rate and *(∆σ)2 + is the average
squared change of the spin in a collision. Further,
n
1 ! ∆σ i · ∆σ j cos[ω(ti − tj )]
F (ω) = 1 + lim , (4.66)
n→∞ n
i,j=1 *(∆σ)2 +
i!=j
where the first term (the “diagonal” part of the double sum) gives the
total radiation power as an incoherent sum of individual collisions. The
second term takes account of interference effects between the radiation

emitted in different collisions.
The interference term yields a suppression of the incoherent sum-
mation because the ∆σ in subsequent collisions are anticorrelated. The
spin is constrained to move on the unit sphere whence a kick in one
direction is more likely than average followed by one in the opposite
direction. A similar argument pertains to the velocity; a ∆v in one di-
rection is more likely than average to be followed by one in the opposite
direction. The radiation spectrum with ω > ∼ Γcoll will remain unaffected
while for ω < Γ
∼ coll it is suppressed. The low-ω suppression of brems-
strahlung is known as the Landau-Pomeranchuk-Migdal effect (Landau
and Pomeranchuk 1953a,b; Migdal 1956; Knoll and Voskresensky 1995
and references therein).
The summation in Eq. (4.66) can be viewed as an integration over
the relative time coordinate ∆t = ti − tj with a certain distribution
function f (∆t). For a random sequence of “kicks” one expects an ex-
ponential distribution of the normalized form f (∆t) = 14 Γσ e−Γσ ∆t/2
where Γσ is some inverse time-scale. This implies the Lorentzian shape
(e.g. Knoll and Voskresensky 1995)
ω2
F (ω) = . (4.67)
ω 2 + Γ2σ /4
A Lorentzian model is familiar, for example, from the collisional broad-

ening of spectral lines. A comparison with Eq. (4.60) indicates that
' 3 4
Γcoll *(∆σ)2 + +∞
= dt eiωt σ(t) · σ(0) . (4.68)
ω 2 + Γ2σ /4 −∞
An integral over dω/2π reveals a normalization *σ 2 + so that
*(∆σ)2 +
Γσ = Γcoll . (4.69)
*σ 2 +
Therefore, Γσ is identified with a collisional spin-fluctuation rate.

For nucleons interacting by an OPE potential one may estimate Γσ
without much effort. The N N cross section is dimensionally απ2 /m2N .
A typical thermal nucleon velocity is v = (3T /mN )1/2 yielding for a
−5/2
typical collision rate Γcoll = *vσN N +nB ≈ απ2 T 1/2 mN nB . Because
|∆σ| ≈ 1 in a collision, Γσ ≈ Γcoll . This estimate agrees √ with the
detailed result of Eq. (4.7) apart from a numerical factor 4 π.
146 Chapter 4
4.6.6 Classical vs. Quantum Result

In order to make contact with the quantum calculation that led to the
axion emission rate in Eq. (4.6) it is useful to juxtapose it with the
classical result in terms of the spin-density structure function. The
classical calculation led to
Γσ ω2
Sσ (ω) = 2 × 2 (4.70)
ω ω + Γ2σ /4
while the quantum result is
2
Γσ 1 for ω > 0,
Sσ (ω) = 2 × s(ω/T ) × (4.71)
ω eω/T for ω < 0.
The function s(x) has the property s(0) = 1, it is even, and according
to the f-sum rule must decrease for large x. In Fig. 4.7 the classical and
quantum results are shown as dotted and dashed lines, respectively,
where for the purpose of illustration s(x) = (1 + x2 /4)−1/4 has been
assumed.
Fig. 4.7. Classical, quantum, and compound spin-density structure func-

tion in the nondegenerate limit according to Eq. (4.70) and Eq. (4.71) with
Γσ /T = 0.2 and s(x) = (1 + x2 /4)−1/4 .
This comparison highlights an important weakness of the classi-

cal result: it does not obey the detailed-balance requirement Sσ (ω) =
Sσ (−ω) eω/T . The classical correlators are invariant under time reflec-
tion t → −t and thus symmetric under ω → −ω. Again, the classical
result is adequate only for |ω| <
∼ T.
Equation (4.71) also highlights an important weakness of the quan-

tum result: It does not include the interference effect from multiple
collisions at |ω| <
∼ Γσ because the calculation was done assuming indi-
vidual, isolated collisions. The normalization condition Eq. (4.58) can
then be satisfied only by accepting a pathological infrared behavior of
Sσ (ω) like the one suggested by Sawyer (1995).
The classical and the quantum results thus both violate fundamental
requirements. The quantum calculation applies for |ω| > ∼ Γσ while the
<
classical one for |ω| ∼ T . If Γσ ' T (“dilute medium”) the regimes of
validity overlap for Γσ < <
∼ |ω| ∼ T and the results agree beautifully. In
this case the two calculations mutually confirm and complement each
other. The compound structure function, where the quantum result is
multiplied with ω 2 /(ω 2 +Γ2σ /4), fulfills the detailed-balance requirement
and approximately the normalization condition. Therefore, it probably
is a good first guess for the overall shape of Sσ (ω).
4.6.7 High-Density Behavior

In a SN core one is typically in the limit Γσ ) T so that the regime
of overlapping validity Γσ < <
∼ |ω| ∼ T between a classical and a per-
turbative quantum calculation no longer exists. Rather, one is in the
opposite situation where for T < <
∼ |ω| ∼ Γσ neither approach appears
directly justified. Because the structure function Sσ (ω) determines all
axial-vector interaction rates in the long-wavelength limit, it is rather
unclear what their high-density behavior might be.
Still, one has important general information about Sσ (ω). The de-
tailed-balance condition Sσ (ω) = Sσ (−ω) eω/T reveals that it is enough
to specify Sσ (ω) for positive energy transfers (energy given to the
medium). In the classical limit Sσ (|ω|) = Γσ /ω 2 , while the f-sum rule
informs us that the quantum version must fall off somewhat faster with
large |ω|. Finally, if )spin-spin correlations can be neglected, the nor-
malization condition 0∞ dω (1 + e−ω/T ) Sσ (ω) = 2π obtains.
In order to illustrate the overall impact of the high-density behavior
on axion or neutrino pair emission rates and on neutrino scattering
rates, it is enough to take the classical limiting case for large |ω|, even
though it does not have an integrable f-sum. Thus a simple ansatz is
Γσ
Sσ (|ω|) = 2 , (4.72)
ω + Γ2 /4
where Γ is to be determined by the normalization condition. In the
dilute limit (Γσ ' T ) this implies Γ ≈ Γσ while in the dense limit
(Γσ ) T ) one finds Γ ≈ Γσ /2.
148 Chapter 4
For axion bremsstrahlung,

)
the energy-loss rate of the medium is
given by +a = Qa /ρ ∝ 0∞ dω ω 4 Sσ (ω) so that with Eq. (4.72) one needs
to evaluate
' ∞ ω 4 e−ω/T
+ a ∝ Γσ dω (4.73)
0 ω 2 + Γ2 /4
with Γ determined from the normalization condition. In the dilute
limit (Γ ≈ Γσ ' T ) one may ignore Γ in the denominator, so that
+a ∝ Γσ . This is indeed what one expects from a bremsstrahlung process
for which the volume energy-loss rate is proportional to the density
squared, and thus +a proportional to the density which appears in the
spin-fluctuation rate Γσ .
In the high-density limit (Γ ≈ Γσ /2 ) T ) the denominator in
Eq. (4.73) is dominated by Γ because the exponential factor suppresses
the integrand for ω ) T so that one expects +a to be a decreasing
function of Γσ . In Fig. 4.8 the variation of +a with Γσ /T is shown
(solid line), taking Eq. (4.72) for the spin-density structure function.
The dashed line shows the “naive rate,” based on Γσ /ω 2 which ignores
multiple-scattering effects, and which violates the normalization con-
dition. Fig. 4.8 illustrates that even very basic and global properties
of Sσ (ω) reveal an important modification of the axion emission rate
at high density: they saturate with an increasing spin-fluctuation rate,
Fig. 4.8. Schematic variation of the axion emission rate per nucleon with
Γσ /T , taking Eq. (4.72) for the spin-density structure function. The dashed
line is the naive rate without the inclusion of multiple-scattering effects, i.e.
it is based on Sσ (|ω|) = Γσ /ω 2 .
and may even decrease at large densities, although such large values for
Γσ may never be reached in a nuclear medium as will become clear be-
low.
The high-density downturn of the axion emission rate can be in-
terpreted in terms of the Landau-Pomeranchuk-Migdal effect (Landau
and Pomeranchuk 1953a,b; Feinberg and Pomeranchuk 1956; Migdal
1956) as pointed out by Raffelt and Seckel (1991). The main idea is
that collisions interrupt the radiation process. The formation of a radi-
ation quantum of frequency ω takes about a time ω −1 according to the
uncertainty principle and so if collisions are more frequent than this
time, the radiation process is suppressed. Classically, this effect was
demonstrated in the language of current correlators in Sect. 4.6.5.
The impact of the high-density behavior of Sσ (ω) on the neutral-
current neutrino opacity is crudely estimated by the inverse mean free
path given in Eq. (4.35), averaged over a thermal energy spectrum of
the initial neutrino. Moreover, all expressions become much simpler if
one replaces the Fermi-Dirac occupation numbers with the Maxwell-
Boltzmann expression e−ωi /T ; the resulting error is small for nondegen-
erate neutrinos. The relevant quantity is then
3 4 ' ∞ ' ∞
−1
λ ∝ dω1 dω2 ω12 ω22 e−ω1 /T Sσ (ω1 − ω2 ). (4.74)
0 0
One integral can be done explicitly, leaving one with an integral over
the energy transfer alone. With Eq. (4.72) for the structure function
one finds
3 4 ' ∞
−1 Γσ (T 2 + T ω/2 + ω 2 /12) e−ω/T
λ ∝ dω . (4.75)
0 ω 2 + Γ2 /4
This expression is constant for Γσ ' T where Γ = Γσ and thus the
structure function is essentially 2πδ(ω). Indeed, the average scattering
cross section (or mean free path) is not expected to depend on the
density.
For dense media (Γσ ) T ), however, the broadening of Sσ (ω) be-
yond a delta function leads to a decreasing average scattering rate. This
means that at a fixed temperature the medium becomes more trans-
parent to neutrinos with increasing density, even without the impact
of degeneracy effects. This behavior is shown in Fig. 4.9 in analogy to
the axion emission rate Fig. 4.8.
A decreasing cross section is intuitively understood if one recalls
that the nucleon spin is typically flipped in a collision with other nucle-
ons because the interaction potential couples to the spin. Neutrino scat-
tering with an energy transfer ω implies that properties of the medium
150 Chapter 4
Fig. 4.9. Schematic variation of the neutrino scattering rate per nucleon
(axial-vector interaction only) with Γσ /T , taking Eq. (4.72) for the spin-
density structure function. The dashed line is the naive rate without the
inclusion of multiple-scattering effects, i.e. it is based on Sσ (ω) = 2πδ(ω).
that fluctuate on faster time scales cannot be resolved. Therefore, if

a nucleon spin flips many times within the time scale ω −1 , the probe
“sees” a vanishing average spin and the scattering rate is reduced ac-
cordingly.
For the vector current, the relevant zeroth component does not fluc-
tuate so that this contribution to neutrino scattering is not suppressed
(Sect. 4.6.4). This observation is intimately tied to the absence of
bremsstrahlung emission of neutrino pairs by the vector current. Both
of these effects are summarized in the statement that the vector-current
structure function Sρ (ω) is always given as23 2πδ(ω), allowing for scat-
tering (energy transfer ω = 0), but not for bremsstrahlung.
23
This statement is based on the assumption that there are no spatial correlations
among the nucleons—possibly a poor approximation in a dense medium. Moreover,
collective oscillations may occur so that it can be too simplistic to treat the medium
as consisting of essentially free, individual nucleons (Iwamoto and Pethick 1982).
Calculations of the long-wavelength structure factor by Sawyer (1988, 1989) in a
specific nucleon interaction model showed a substantial suppression of Sρ (0) and
thus, of the neutrino mfp. Other related works are those of Haensel and Jerzak
(1987) and of Horowitz and Wehrberger (1991a,b) who calculated the dynamic
structure functions within certain interaction models. Unfortunately, these works
do not shed much light on the high-density behavior of the quantities which are of
prime interest to this book such as the axion emission rate from a hot SN core.
The axion emission and the neutrino scattering rates begin to be

suppressed at high density if Γσ exceeds a few T . The numerical value
Eq. (4.34) that was derived in a simple OPE calculation indicates that
such large Γσ may be expected in a nuclear medium. Of course, one
may well ask if a perturbative calculation of the spin-fluctuation rate is
adequate if such a calculation fails for the neutrino scattering rate. Nu-
cleon spins fluctuate because of their spin-dependent interaction with
other nucleons and thus are themselves subject to spin averaging effects.
The question of the true value of Γσ in a nuclear medium (as opposed
to the OPE-calculated one) can be addressed empirically by virtue of
the SN 1987A neutrino signal and theoretically by the f-sum rule.
Because the axial-vector current contribution to the standard neu-
tral-current scattering rate dominates (the cross section is proportional
to CV2 + 3CA2 ), it would be much easier for neutrinos to diffuse out of
the hot SN core if the axial-current scattering rate were significantly
suppressed. The observed SN 1987A neutrino signal would have been
much shorter than predicted in standard cooling calculations. The
impact of reduced neutrino opacities on the SN 1987A signal has been
studied by Keil, Janka, and Raffelt (1995); see Sect. 13.6. The main
conclusion is that a suppression of the axial-vector current opacity by
more than about a factor of 2 is not compatible with the SN 1987A
signal duration. This result would indicate that the effective Γσ never
becomes much larger than a few T .
Theoretically, the f-sum rule Eq. (4.58) allows one to relate Γσ to
the average spin-spin interaction energy in the medium. According to
Sigl’s (1995b) estimate one concludes, again, that Γσ does not exceed
a few T in a nuclear medium.
These results imply that even in a dense medium the axial-vector
scattering rate is not suppressed as strongly as one may have expected
on the basis of a naive estimate of Γσ . It still remains impossible to cal-
culate its exact magnitude from first principles. Therefore, the neutral-
current neutrino opacity remains an adjustable function of density for
practical SN cooling calculations much as the equation of state.
4.7 Effective Nucleon Mass and Coupling

In a dense medium it is not necessarily possible to use the vacuum
masses and coupling constants to determine interaction rates. The
axial-vector neutrino couplings are probably suppressed somewhat (Ap-
pendix B). For the pion-nucleon coupling, Turner, Kang, and Steigman
152 Chapter 4
(1989) argued on the basis of the nonlinear sigma model that the combi-
nation of parameters απ2 αa /m2N should remain approximately constant.
Mayle et al. (1989) similarly found that this parameter should remain
somewhere in the range 0.3−1.5 of its vacuum value.
The nucleon effective mass m∗N deviates substantially from the vac-
uum value mN = 939 MeV. This shift has an impact on the kinematics
of reactions and the nucleon phase-space distribution. A calculation of
m∗N has to rely on an effective theory which describes the interaction
of nucleons and mesons. A typical result from a self-consistent rela-
tivistic Brueckner calculation including vacuum fluctuations is shown
in Fig. 4.10. For conditions relevant for a SN core, an effective value as
low as m∗N /mN = 0.5 is conceivable.
Fig. 4.10. Effective nucleon mass in a nuclear medium according to a self-

consistent relativistic Brueckner calculation (Horowitz and Serot 1987). Nu-
clear density corresponds to about ρ0 = 3×1014 g/cm3 .
4.8 The URCA Processes

Neutral-current weak processes are of prime interest for the topics stud-
ied in this book because they are closely related to “exotic” reactions
involving new particles such as axions. For completeness, however,
it must be mentioned that the charged-current reactions depicted in
Fig. 4.11 dominate the neutrino energy-loss rate of old neutron stars,
and also dominate the νe opacity in young SN cores. The processes
n → p e ν e (neutron decay) and e p → n νe are usually called the URCA
reactions after a casino in Rio de Janeiro. It must have been as easy

to lose money there as it is to lose energy in reactions which produce a
neutrino whether the electron is in the initial or final state.
Fig. 4.11. The URCA processes. For the modified versions, there are obvious
other graphs with the leptons attached to other nucleon lines, and exchange
amplitudes.
However, in a neutron star the direct URCA processes can be highly

suppressed by energy-momentum conservation. Because all participat-
ing degenerate fermions are close to their Fermi surface one must re-
quire pF,p + pF,e > pF,n . Because the proton and electron concentration
is small in a neutron star, it was thought that this “triangle condition”
could not be satisfied. However, the equilibrium proton concentration
depends on details of the equation of state—even in a naive model with
free fermions it is a sensitive function of the effective nucleon mass (Ap-
pendix D). Therefore, it is conceivable that the direct URCA processes
actually do take place in neutron stars (Boguta 1981; Lattimer et al.
1991), in which case they provide a cooling mechanism much faster
than all other proposed possibilities.
If the triangle condition is fulfilled, the energy-loss rate is (Lattimer
et al. 1991)
457π 2
QURCA = GF cos2 θC (1 + 3CA2 ) m2N pF,e T 6 , (4.76)
10080
where θC ≈ 0.24 is the Cabbibo angle and CA is the charged-current
axial-vector constant which is −1.26 in vacuum while in nuclear matter
it is suppressed somewhat (Appendix B).
154 Chapter 4
If the triangle condition is not satisfied the URCA processes re-

quire bystander particles to absorb momentum, leading to the “modi-
fied URCA process” shown in Fig. 4.11 (Chiu and Salpeter 1964). Of
course, as below in Sect. 4.9.1, the missing momentum can be provided
by pions or a pion condensate (Bahcall and Wolf 1965a,b).
An explicit result for the modified URCA rate in the OPE model
for the nucleon interactions is (Friman and Maxwell 1979)
11513π 2 2
Qmod. URCA = α G cos2 θC CA2 pF,e T 8 , (4.77)
120960 π F
ignoring factors of order unity to account for a nonzero mπ and nucleon
correlations. As in the νν processes, only the axial-vector coupling
contributes. Some details of the phase-space integration can be found
in Shapiro and Teukolsky (1983).24
The modified URCA reactions (Fig. 4.11) are closely related to νν
bremsstrahlung and the inelastic scattering process νN N → N N ν dis-
cussed earlier. It is interesting that the rate for the e p n → n n νe pro-
cess does not diverge, in contrast with νN N → N N ν where multiple-
scattering effects had to be invoked to obtain a sensible result. The
divergence was due to the intermediate nucleon going on-shell for a
vanishing energy transfer. In e p n → n n νe the electron has the energy
EF,e , the neutrino T , and because EF,e ) T (degenerate electrons!), the
minimum energy transfer to the leptons is EF,e and thus never zero. Of
course, this is the reason why the modified URCA reaction was in-
voked in the first place: if the intermediate nucleon line is cut, the two
sub-processes are suppressed by energy-momentum constraints.
The direct and modified URCA process should be expressed in terms
of a common structure function applicable to charged-current processes.
One may expect that at high density, spin and isospin fluctuations may
suppress these reactions in analogy to the neutral-current processes
discussed in the previous section. However, the URCA processes have
not been discussed in the literature from this particular perspective.
The equivalent of the modified URCA process for quark matter was
calculated by Iwamoto (1980, 1982), more recently by Goyal and Anand
(1990), and numerically by Ghosh, Phatak, and Sahu (1994) who claim
that Iwamoto’s phase-space approximations can lead to substantial er-
rors in the emission rate.
24
Shapiro and Teukolsky’s phase-space volume involves a factor p3F,e because in
their treatment of the nuclear matrix element there is no cancellation of p2F,e with
1/ω 2 from the nucleon propagator which occurs in a bremsstrahlung calculation.
4.9 Novel Phases of Nuclear Matter

4.9.1 Pion Condensate
Free particles cannot radiate because of energy-momentum constraints
which can be overcome by exchanging momentum with “bystander”
particles (bremsstrahlung). No bystander is required if the momentum
is taken up by the pion field directly which in Fig. 4.1 only mediated the
nucleon interaction (Bahcall and Wolf 1965a,b). This possibility is par-
ticularly important if a pion condensate develops so that the medium
is characterized by a macroscopic, classical pion field. The pion dis-
persion relation can be such that the lowest energy state involves a
nonvanishing momentum kπ (Baym 1973; Kunihiro et al. 1993; Migdal
et al. 1990). Nucleons can exchange pions with the condensate and
thereby pick up a momentum kπ which then allows for the radiation of
axions or other particles (Fig. 4.12).
Fig. 4.12. Axion emission by nucleons scattering off a pion condensate.

There is another amplitude with the axion attached to N1 .
The pion condensate causes a periodic potential for the nucleons

which thus must be described as quasi-particles or Bloch states with
a main momentum component p and admixtures p ± kπ . Because an
eigenstate of energy is no longer an eigenstate of momentum, these
quasi-particles can emit radiation without violating energy-momentum
conservation. Therefore, the process of Fig. 4.12 can be equally de-
scribed as a decay N; →N ; a.
1 2
The rate for this reaction was calculated by Muto, Tatsumi, and
Iwamoto (1994) who found that the dominant contribution was from a
π ◦ condensate. It forms a periodic potential A sin(kπ · r) where A is a
dimensionless amplitude which is small compared to unity for a weakly
developed condensate. The Bloch states are to lowest order in A
$ ± ± %
;± = N ± ∓ Aκ0
Np+k π
Np−k π
N p p + , (4.78)
2 Ep+kπ − Ep Ep−kπ − Ep
156 Chapter 4
where Np± is a plane-wave nucleon state with energy Ep and spin ori-
entation ± relative to kπ , and κ0 is a coupling constant.
In the nonrelativistic limit the axion energy-loss rate corresponding
to the decay N; →N ; a is written as
1 2
' '
d3 k d3 p1 ' d3 p2
Qa = ω f1 (1 − f2 )
2ω(2π)3 (2π)3 (2π)3
!
× 2π δ(E1 − E2 − ω) |M|2 , (4.79)
spins
where k is the axion momentum, p1,2 the nucleon momenta, and f1,2
their occupation numbers. The matrix element, averaged over axion
emission angles (the condensate is not isotropic!) is found to be
$ %2
3 ! 4 C0 − g<A C1
|M|2 = 4
3
A2 κ20
spins 2fa
5 6
× (2π)3 δ 3 (∆p + kπ ) + δ 3 (∆p − kπ ) , (4.80)
where ∆p = p1 − p2 − k. The isoscalar and isovector axion coupling

constants are C0 = 12 (Cp + Cn ) and C1 = 12 (Cp − Cn ) in terms of their
couplings to protons and neutrons. The isovector current carries a
renormalized axial-vector coupling constant for which Muto, Tatsumi,
and Iwamoto (1994) found g<A ≈ 0.43 × 1.26 = 0.54.
The phase-space integration was carried out by neglecting the axion
momentum in δ 3 (∆p ± kπ ) and taking the degenerate limit. Then,
$ %2
π C0 − g<A C1 A2 κ20 m2N T 4
Qa = , (4.81)
45 2fa |kπ |
an emission rate with a relatively soft temperature dependence.

For a numerical estimate Muto, Tatsumi, and Iwamoto (1994) found
that in cold neutron-star matter near the critical density (about 2.1
times nuclear) kπ ≈ 410 MeV, the neutron and proton Fermi momenta
are 410 and 150 MeV, respectively (corresponding to Yp = 0.04), and
the coupling strength to the condensate is κ0 ≈ 105 MeV. Therefore,
π 2 κ20 T 4
Q a = αa A2 = αa A2 1.03×1044 erg cm−3 s−1 T94 , (4.82)
45 |kπ |
where αa ≡ [2mN (C0 − g<A C1 )/2fa ]2 /4π and T9 ≡ T /109 K.

Instead of an axion one may also emit a neutrino pair in Fig. 4.12.
The corresponding energy-loss rate was worked out by Muto and Tat-
sumi (1988) who found
2π < 2 2 2 2 κ20 T 6
Qνν = C G m A . (4.83)
945 A F N |kπ |
Here, C<A is the renormalized neutron neutral-current coupling constant;

in vacuum CA = −1.26/2. A somewhat different emission rate was
found by Senatorov and Voskresensky (1987).
A process related to that shown in Fig. 4.12 is pionic Compton
scattering πN → N a involving thermal pions. The equivalent of the
axion emission rate was calculated by Turner (1992) and by Raffelt and
Seckel (1995). This process could be of some importance in the hot
nondegenerate medium of a SN core. However, a thermal population of
pions yields a rate which is always less important than bremsstrahlung.
For a charged pion or kaon condensate, the equivalent of the mod-
ified URCA process can occur by exchanging a meson with the con-
densate rather than with a bystander nucleon. Recent calculations of
such processes include Senatorov and Voskresensky (1987), Muto and
Tatsumi (1988), and Thorsson et al. (1995).
4.9.2 Quark Matter

It has been speculated that a “neutron” star may actually undergo a
phase transition where the nucleons dissolve in favor of a quark-gluon
plasma. Notably, it is possible that the true ground state of nuclear
matter consists of “strange quark matter” with about equal numbers
of up, down, and strange quarks. Such a system can emit axions by
virtue of quark-quark bremsstrahlung (Fig. 4.13).
Fig. 4.13. Axion emission by quark-quark bremsstrahlung. There is a total

of eight amplitudes, four with the axion attached to each quark line, and an
exchange graph each with q3 ↔ q4 .
158 Chapter 4
The gluon in Fig. 4.13 couples to the quarks with the strong fine-
structure constant αs and the gluon propagator involves an effective
mass for which Anand, Goyal, and Iha (1990) used m2g = (6αs /π) p2F
where pF is the Fermi momentum of the quark sea. The axion cou-
pling to quarks is the usual derivative form (Cq /2fa ) ψ q γµ γ5 ψq ∂ µ φ with
q = u, d, s. After a cumbersome but straightforward calculation Anand
et al. found for the energy-loss rate in the degenerate limit
=
62π 2 αs2 pF T 6 Cq2 (I1 + 2I2 ) for qq → qqa,
Qa = × (4.84)
945 (2fa )2 (Cq2 + Cq2" ) I1 for qq # → qq # a,
where the angular integrals are
1 ' π ' 2π s5θ (1 + c4φ )

I1 ≡ dθ dφ ,
2π 2 0 0 (s2θ s2φ + κ2 )2
1 ' π ' 2π s5θ
I2 ≡ dθ dφ , (4.85)
2π 2 0 0 (s2θ s2φ + κ2 )(s2θ c2φ + κ2 )
where sθ ≡ sin(θ/2), sφ ≡ sin(φ/2), cφ ≡ cos(φ/2), and κ2 ≡ m2g /2p2F =

3αs /2π. Numerically, I1 = 12.3, 4.23, and 2.25 for αs = 0.2, 0.4, and
0.6, while I2 = 2.34, 1.30, and 0.86, respectively.
The total emission rate involves three processes with equal quarks
(qq = uu, dd, ss), and three with different ones (qq # = ud, us, ds) so
that Qa is the prefactor of Eq. (4.84) times (Cu2 + Cd2 + Cs2 )(3I1 + 2I2 ),
to be compared with Eq. (4.10) for degenerate neutron matter. The
ratio between the rates is
Qqq
a Cu2 + Cd2 + Cs2 pF,q αs2 2π (3I1 + 2I2 )
= , (4.86)
Qnn
a Cn2 pF,n απ2 G(mπ /pF,n )
where the function G(u) was given in Eq. (4.12). The ratio of coupling
constants is model dependent (Sect. 14.3.3), the Fermi momenta are
approximately equal for equal densities.
For a typical value αs = 0.4 one has αs2 /απ2 = 0.6×10−3 while the last
factor is about 140 with G = 0.7 (Fig. 4.3) so that the last two factors
together are about 0.08. Hence, the emission rate from quark matter
is much smaller than that from neutron matter because απ is so large
relative to αs . In the inevitable presence of protons in a neutron star,
the np process will be even more important than nn (Fig. 4.4), further
enhancing the emission rate of nuclear matter relative to quark matter.
For neutrino pairs the ratio of the emission rates is about the same
as for axions and so quark matter is much less effective at νν emission
than nuclear matter (Burrows 1979; Anand, Goyal, and Iha 1990). This
observation has little effect on the neutrino cooling rate of a quark
star because the URCA processes, which are based on charged-current
reactions, dominate for both nuclear or quark matter.
For axions, however, it may seem that the emission rate is much
suppressed relative to nuclear matter. This is certainly true if “ax-
ion” stands for any generic pseudoscalar Nambu-Goldstone boson. The
QCD axion, however, which was introduced to solve the CP problem
of strong interactions (Chapter 14) necessarily has a two-gluon cou-
pling which allows for the gluonic Primakoff effect (Fig. 4.14) which is
analogous to the photon Primakoff effect discussed in Sect. 5.2.
Fig. 4.14. Axion emission by the gluon Primakoff effect.
Altherr (1991) has calculated the emissivity of a hot quark-gluon

plasma and found that it was similar to that of a nuclear medium
at the same temperature and density. Ellis and Salati (1990) found
a much smaller emission rate because they included only the gluonic
plasmon decay process gT → gL + γ, much in analogy to the corre-
sponding photonic process discussed in Sect. 5.2.2. This decay process,
however, is only part of the axion emissivity by the gluon field fluctu-
ations.
For highly degenerate quark matter corresponding to old neutron
stars the emission rate has not be calculated as far as I know.
4.9.3 Bubble Phase
In a hot lepton-rich neutron star the nuclear medium in the density

regime 12 ρ0 to ρ0 (nuclear matter density ρ0 = 3×1014 g cm−3 ) may
form a “bubble phase” with regions of low density embedded in the
medium. Leinson (1993) has estimated the bremsstrahlung rate for the
neutral-current reactions N + bubble → bubble + N + νν. Naturally,
he found that only the axial-vector current contributes. Also, it is easy
to translate his results into an axion emission rate. Apparently, the
bubble phase can be important for early neutron-star cooling.
160 Chapter 4
4.10 Emission of Right-Handed Dirac Neutrinos

So far in this chapter neutrinos were assumed to be massless parti-
cles which interact only by the standard left-handed weak current. It
is possible, however, that neutrinos have a Dirac mass in which case
any reaction with a final-state (anti)neutrino produces both positive
and negative helicity states. Typically, the “wrong-helicity” states will
emerge in a fraction (mν /2Eν )2 of all cases because of the mismatch
between chirality (eigenstates of γ5 ) and helicity. For Eν ) mν the
wrong-helicity states correspond approximately to right-handed chiral-
ity states and so their interaction-rate with the ambient medium is
weaker by an approximate factor (mν /2Eν )2 . This implies that for a
sufficiently small mass they would not be trapped in a SN core and
thus carry away energy in an “invisible” channel, allowing one to set
constraints on a Dirac neutrino mass from the observed neutrino signal
of SN 1987A (Sect. 13.8.1). Here, the relationship between the pro-
duction rate of left- and right-handed neutrinos is explored in some
detail because the simple scaling with (mν /2Eν )2 is not correct in
all cases.
When a neutrino interacts with a medium the transition probability
from a state with four-momentum K1 = (ω1 , k1 ) to one with K2 =
(ω2 , k2 ) is written as W (K1 , K2 ). The function W is defined for both
positive and negative energies. The emission probability for a pair
ν(K1 )ν(K2 ) is then W (−K1 , K2 ), the absorption probability for a pair
is W (K1 , −K2 ). The collisional rate of change of the occupation number
fk1 of a neutrino field mode k1 is then given by
: '
dfk1 :: d 3 k2 5
: = WK2 ,K1 fk2 (1 − fk1 ) − WK1 ,K2 fk1 (1 − fk2 )
dt :coll (2π)3
6
+ W−K2 ,K1 (1 − fk1 )(1 − f k2 ) − WK1 ,−K2 fk1 f k2 , (4.87)
where the variables of W were written as subscripts. The first term

corresponds to neutrino scatterings into the mode k1 from all other
modes, the second term is scattering out of mode k1 into all other
modes, the third term is pair production with a final-state neutrino k1 ,
and the fourth term is pair absorption of a neutrino of momentum k1
and an antineutrino of any momentum. f k is the occupation number
for the antineutrino mode k; (1−fk ) or (1−f k ) represent Pauli blocking
factors. This collision integral only includes (effective) neutral-current
processes while charged-current reactions were ignored.
According to the discussion of Sect. 4.6 the transition probability is

given as
G2F nB
W (K1 , K2 ) = S µν Nµν , (4.88)
8 2ω1 2ω2
where S µν is an effective structure function given in terms of the vector,
axial-vector and mixed structure function of the medium defined in
Eq. (4.38), and nB is the baryon density. The tensor N µν is the squared
matrix element of the neutrino current; it was given in Eq. (4.17) in
terms of K1 and K2 . In an isotropic medium the tensorial composition
of the dynamical structure function can be expressed in terms of the
energy-momentum transfer K and the four-velocity U of the medium;
U = (1, 0, 0, 0) in its rest frame. Thus, in analogy to Eqs. (4.40) and
(4.41) one may write
S µν = S1 U µ U ν + S2 (U µ U ν − g µν ) + S3 K µ K ν
+ S4 (K µ U ν + U µ K ν ) + i S5 +µναβ Uα Kβ , (4.89)
where the functions S( (3 = 1, . . . , 5) depend on medium properties
and on the energy-momentum transfer K = (ω, k). The transition
probability is then found to be (Raffelt and Seckel 1995)
G2F nB 5
W (K1 , K2 ) = (1 + cos θ)S1 + (3 − cos θ)S2
4
6
− 2(1 − cos θ)(ω1 + ω2 )S5 , (4.90)
where θ is the neutrino scattering angle. The terms proportional to S3
and S4 vanish identically.
Next, one may consider processes involving massive Dirac neutri-
nos with specified helicities. Gaemers, Gandhi, and Lattimer (1989)
showed that in this case the same expressions apply with Nµν as con-
structed in Eq. (4.17) if one substitutes Ki → 12 (Ki ± mν Si ), i = 1
or 2, where the plus sign refers to ν, the minus sign to ν, and Si is
the covariant spin vector. For relativistic neutrinos one may consider
a noncovariant lowest-order expansion in terms of mν . In this limit Ki
remains unchanged for left-handed states while
; = (m /2ω )2 (ω , −k )
Ki = (ωi , ki ) → K (4.91)
i ν i i i
for right-handed ones. After this substitution has been performed all
further effects of mν are of higher order so that one may neglect mν
everywhere except in the global “spin-flip factor.”
162 Chapter 4
The dispersion relation of neutrinos in a SN differs markedly from

the vacuum form; in the core the “effective mν ” is several 10 keV. How-
ever, mν in Eq. (4.91) is the vacuum mass which couples left-handed
to right-handed states and thus leads to spin flip while the medium-
induced “mass” only affects the dispersion relation of left-handed states.
This view is supported by a detailed study of Pantaleone (1991). Of
course, for nonrelativistic neutrinos the situation is more complicated
because an approximate identification of helicity with chirality is not
possible.
Next consider a specific process which involves a left- and a right-
handed neutrino such as “spin-flip scattering” νL (KL ) → νR (KR ).
Then, one needs to construct Nµν from K1 = (ωL , kL ) and K2 =
(mν /2ωR )2 (ωR , −kR ). The contraction with S µν leads to (Raffelt and
Seckel 1995)
2 - .2 5
; (K , K ) = GF nB
W
mν
(1 − cos θ)S1 + (3 + cos θ)S2
L R
4 2ωR
6
+ 4ωR2 (1 − cos θ)S3 − 4ωR (1 − cos θ)S4 + 2(ωR − ωL )(1 + cos θ)S5 .
(4.92)
Following Gaemers, Gandhi, and Lattimer (1989) it must be empha-
sized that this expression differs in more than the factor (mν /2ωR )2
from the nonflip case Eq. (4.90). This difference is due to the changed
angular momentum budget of reactions with spin-flipped neutrinos.
This angular-momentum difference between spin-flip and no-flip
processes is nicely illustrated by virtue of the pion decay process π ◦ →
νν. If both final-state neutrinos are left-handed, i.e. if ν has nega-
tive and ν positive helicity, this decay is forbidden by angular momen-
tum conservation. This is seen most easily if one recalls that it is the
pion current ∂µ π ◦ that interacts with the left-handed neutrino current.
The squared matrix element of the pion current is thus proportional
to Kπµ◦ Kπν◦ . For a pion decaying at rest only the 00-component con-
tributes. Contraction with Nµν leads to identically zero if one recalls
that for the final-state neutrinos ω1 = ω2 and k1 = −k2 . For the spin-
flip process one must use the reversed momentum instead so that in
Nµν one must use ω1 = ω2 and k1 = k2 , leading to a nonvanishing
contribution. The decay of thermal pions is an important process for
populating the right-handed states of massive Dirac neutrinos in the
early universe (Lam and Ng 1991). Contrary to a discussion by Na-
tale (1991), however, pion decays do not seem to provide a particularly
strong contribution in SN cores (Raffelt and Seckel 1991).
Right-handed neutrinos can be produced by the interaction with

all medium constituents. For a SN core, the production rate from the
interaction with charged leptons such as e+ e− → νL ν R or νL e− → e− νR
was explicitly studied by Pérez and Gandhi (1990) and by Lam and Ng
(1992). Still, in a SN core the dominant contribution to the interaction
between neutrinos and the medium is due to the nucleons.
Therefore, one may simplify the expression Eq. (4.92) by applying
the same approximations that were used earlier in Sect. 4.6 in the con-
text of purely left-handed neutrinos. Notably, one may use the nonrel-
ativistic and long-wavelength limits which reveal, again, that only the
terms proportional to S1 and S2 contribute, and that these structure
functions can be taken to be functions of the energy transfer ω alone.25
Moreover, the neutrino phase-space integration will always average the
cos θ term to zero. Then, the spin-flip reaction rate is indeed simply
the nonflip rate times the “spin-flip factor” (mµ /2Eν )2 .
Right-handed neutrinos can be produced by spin-flip scattering of
left-handed ones νL → νR , or by the emission of pairs νL ν R or νR ν L .
In a SN core, left-handed neutrinos are trapped while right-handed
ones can freely escape whence the quantity of interest is the energy-
loss rate of the medium in terms of right-handed states. The total
energy-loss rate in νR is then QR = Qscat + Qpair where “scat” refers to
spin-flip scattering and “pair” to the pair-emission process. The two
contributions are
'
d3 k L d 3 k R ;
Qscat = WKL ,KR fkL ωR ,
(2π)3 (2π)3
'
d3 k L d 3 k R ;
Qpair = W−KL ,KR (1 − f kL )ωR . (4.93)
(2π)3 (2π)3
An analogous expression pertains to the emission of ν R ; if the trapped
left-handed neutrinos are nondegenerate this contribution has the same
magnitude as that for the emission of νR .
Next, consider the limit where the transition probability W ; can be
represented in terms of a single structure function26 S(ω) = S1 (ω) +
3S2 (ω). Further, the possibility of neutrino degeneracy is ignored which
allows one to approximate the left-handed neutrino Fermi-Dirac distri-
bution by a Maxwell-Boltzmann one, (eω/T + 1)−1 → e−ω/T . Then the
25
In the notation of Sect. 4.6 and for a single species of nucleons one has in this
limit S1 = CV2 Sρ and S2 = CA 2
Sσ .
26
In the notation of Sect. 4.6 this is S = CV2 Sρ + 3CA
2
Sσ in a medium consisting
of a single species of nucleons.
164 Chapter 4
energy-loss rates are

>- . &
G2F m2ν nB T 4 ' ∞ ω 2 ω
Qscat = dω + 6 + 12 S(ω),
4π 4 0 T T
- .
G2F m2ν nB T 4 ' ∞ 1 ω 4
Qpair = dω S(ω), (4.94)
4π 4 0 12 T
where the detailed-balance condition S(−ω) = S(ω)eω/T was used.
These rates include a factor of 2 for the emission of νR and ν R .
As noted by Raffelt and Seckel (1995), unless the functional form
of S(ω) is very bizarre one has Qscat ) Qpair . This is easily under-
stood if one observes that in a scattering process the energy available
for the final-state νR is that of the initial-state νL plus energy that
is contributed by the medium. For pair emission, the energy for the
final-state ν L and νR both have to be provided by the medium. There-
fore, the spin-flip scattering process is favored by the neutrino phase
space.27 In a SN core the production of right-handed Dirac neutri-
nos is given essentially by the standard scattering rate times the fac-
tor (mν /2Eν )2 .
27
This conclusion is in contrast to a suggestion that the pair-emission rates
could be relatively important in a SN core (Turner 1992). This statement was
not based on a self-consistent treatment of the medium structure functions, but
rather on a perturbative treatment of the processes involved. The conflict between
Turner’s statement and our finding, which essentially was based on phase-space
considerations, highlights the inadequacy of naive perturbative results in a nuclear
medium.
Chapter 5
Two-Photon Coupling of
Low-Mass Bosons
Well-known particles such as neutral pions, or hypothetical ones such

as axions or gravitons, each have a two-photon interaction vertex. It
allows for radiative decays of the form π ◦ → 2γ as well as for the
Primakoff conversion π ◦ ↔ γ in the presence of external electric or
magnetic fields. The Primakoff conversion of photons into gravitons in
putative cosmic magnetic fields could cause temperature fluctuations
of the cosmic microwave background radiation. In stars, the Primakoff
conversion of photons leads to the production of gravitons and axions,
although the graviton luminosity is always negligible. The Primakoff
conversion of axions into photons serves as the basis for a detection
scheme for galactic axions, and was used in several laboratory axion
search experiments. The modified Maxwell equations in the presence
of very low-mass pseudoscalars would substantially modify pulsar elec-
trodynamics. These physical phenomena are explored and current lab-
oratory and astrophysical limits on the axion-photon coupling are re-
viewed.
5.1 Electromagnetic Coupling of Pseudoscalars

The idea that symmetries of the fundamental interactions can be bro-
ken by the “vacuum” or ground state of the fields plays an important
role in modern particle physics. The breakdown of a global symme-
try by the vacuum expectation value of a Higgs-like field leads to the
existence of massless particles, the “Nambu-Goldstone bosons” of the
broken symmetry. One well-known example are the pions which are
165
166 Chapter 5
the Nambu-Goldstone bosons of an approximate SU (2) symmetry of

the system of nucleons and pions in the sigma model (e.g. Itzykson
and Zuber 1983). Another example is the hypothetical axion which is
the Nambu-Goldstone boson of the Peccei-Quinn chiral U (1) symmetry
that would solve the CP problem of strong interactions (Chapter 14).
Both pions and axions carry a small mass because the underlying sym-
metry is not exact at low energies; they are sometimes called “pseudo
Nambu-Goldstone bosons.” An example for a true Nambu-Goldstone
boson is the hypothetical majoron which arises from the spontaneous
breakdown of a symmetry by a Higgs field which would give the neu-
trinos Majorana masses (Sect. 15.7).
In these examples the Nambu-Goldstone bosons are pseudoscalars
because the underlying symmetry is chiral. Unless the CP symmetry
is violated, their possible coupling to photons must be of the form
Lint = − 14 gaγ Fµν F! µν a = gaγ E · B a, (5.1)
where a is the pseudoscalar field; the axion will serve as a generic

example. Further, gaγ is a constant with the dimension (energy)−1 , F
is the electromagnetic field strength tensor, F! its dual, and E and B are
the electric and magnetic fields, respectively. Because E is a polar and
B an axial vector, E · B is a pseudoscalar under a CP transformation
and so Lint remains invariant.
New scalar particles φ almost inevitably would also couple to pho-
tons with a CP conserving structure Lint ∝ 14 Fµν F µν φ = 12 (E 2 + B 2 )φ.
Therefore, everything that will be said about axions applies mutatis
mutandis to scalar particles as well.
Gravitons would also have a two-photon vertex. Much of what is
said about axions also applies to them, except that their weak cou-
plings render most of the arguments irrelevant—an exception will be
mentioned in Sect. 5.5.5.
Returning to the case of pseudoscalar Nambu-Goldstone bosons,
by the very construction of the underlying models they interact with
certain fermions ψ by a pseudoscalar coupling of the form
Lint = ig ψγ5 ψ a, (5.2)
where g = m/f is a Yukawa coupling given in terms of the fermion

mass m and an energy scale f related to the vacuum expectation value
of the Higgs field which breaks the underlying symmetry—for axions
the Peccei-Quinn scale fa (Chapter 14), for pions the pion decay con-
stant fπ = 93 MeV. In the latter example the relevant fermions are
Two-Photon Coupling of Low-Mass Bosons 167
the nucleons, for axions they are the quarks, possibly the charged lep-
tons, and perhaps some exotic heavy quark state not contained in the
standard model. For majorons, such couplings exist to neutrinos, and
possibly to other fermions.
An interaction of the form Eq. (5.2) with charged fermions auto-
matically leads to an electromagnetic coupling of the form Eq. (5.1)
because of the triangle amplitude shown in Fig. 5.1. For one fermion
of charge e and mass m an explicit evaluation leads to the relationship
(e.g. Itzykson and Zuber 1983)
α g α
gaγ = = , (5.3)
π m πf
where m was taken to be much larger than the axion and photon en-
ergies. Remarkably, because g = m/f this coupling does not depend
on the fermion mass, but only on the scale f of symmetry breaking.
In general, one must sum over all possible fermions, taking account
of the appropriate charges which are fractional for quarks, and also of
the proper pseudoscalar coupling to the individual fermions which may
vary from m/f by model-dependent factors of order unity.
Fig. 5.1. Triangle loop for the coupling of a pseudoscalar a (axion) to two
photons.
For massive pseudoscalars the two-photon coupling allows for a de-

cay a → 2γ with a width
2
Γa→2γ = gaγ m3a /64π. (5.4)
For pions in the sigma model, the only charged fermion is the proton.
Then gπ◦ γ = α/πfπ and Γπ◦ →2γ = α2 m3π /64π 3 fπ2 = 7.6 eV, in close
agreement with the experimental value. (For subtleties of interpretation
of this result in the context of current algebra see the standard field
theory literature, e.g. Itzykson and Zuber 1983).
168 Chapter 5
The main point for the present discussion is that pseudoscalar mass-
less or low-mass bosons are a natural consequence of certain extensions
of the standard model, and that these particles couple to photons ac-
cording to Eq. (5.1) with a strength
α
gaγ = Caγ , (5.5)
πfa
where fa is the energy scale of symmetry breaking and Caγ is a model-
dependent factor of order unity. (For axions, model-dependent details
of the couplings are discussed in Chapter 14.) In the following I will
explore a variety of consequences arising from this interaction.
5.2 Primakoff Process in Stars

5.2.1 Screened Cross Section and Emission Rate
The two-photon coupling of pions or other pseudoscalars allows for the
conversion a ↔ γ in an external electric or magnetic field by virtue of
the amplitude shown in Fig. 5.2. This process was first proposed by
Primakoff (1951) to study the π ◦ -γ-coupling which is experimentally
difficult to measure in free decays π ◦ → 2γ. In stars, this process
allows for the production of low-mass pseudoscalars in the electric fields
of nuclei and electrons (Dicus et al. 1978).
Fig. 5.2. Primakoff conversion between axions or other pseudoscalars and

photons in an external electromagnetic field.
The Primakoff process turns out to be important for nonrelativistic

conditions where T % me so that both electrons and nuclei can be
treated as “heavy” relative to typical energies of the ambient photons.
Therefore, ignoring recoil effects one finds for the differential cross sec-
tion in this limit (target charge Ze)
dσγ→a g 2 Z 2 α |kγ × ka |2
= aγ , (5.6)
dΩ 8π q4
where q = kγ − ka is the momentum transfer; the axion and photon
energies are the same.
This cross section exhibits the usual forward divergence from the
long-range Coulomb interaction. For ma '= 0 it is cut off in vacuum by
the minimum necessary momentum transfer qmin = m2a /2ω (ma % ω);
2 1
the total cross section is then σγ→a = Z 2 gaγ [ 2 ln(2ω/ma ) − 14 ].
In a plasma, the long-range Coulomb potential is cut off by screening
effects; according to Sect. 6.4 the differential cross section is modified
with a factor q2 /(kS2 + q2 ). In a nondegenerate medium the screening
scale is given by the Debye-Hückel formula
4πα " # $
kS2 = nB Ye + Zj2 Yj , (5.7)
T j
where nB = ρ/mu (atomic mass unit mu ) is the baryon density while

Ye and Yj are the number fractions per baryon of the electrons and
various nuclear species j. With this modification the total scattering
cross section is easily calculated (Raffelt 1986a). Summing over all
targets one may derive an expression for the transition rate (“decay
rate”) of a photon of frequency ω into an axion of the same energy,
%& ' & ' (
g2 T k2 k2 4ω 2
Γγ→a = aγ S 1 + S2 ln 1 + 2 −1 , (5.8)
32π 4ω kS
where the plasma “mass” of the initial-state photon and the axion mass
were neglected relative to the energy ω. In the limit ω % kS this ex-
2
pression expands as Γγ→a = gaγ ω 2 T /16π which is entirely independent
of the density and chemical composition.
For a stellar plasma, however, this approximation is usually not
justified. Ignoring the plasma frequency for the initial-state photons,
the energy-loss rate per unit volume is
) 2
2 d3 kγ Γγ→a ω gaγ T7
Q= = F (κ2 ), (5.9)
(2π)3 eω/T − 1 4π
where κ ≡ kS /2T and

% & ' (
2 ) ∞ 2
κ x x
F (κ2 ) = 2 dx (x2 + κ2 ) ln 1 + 2 − x2 x , (5.10)
2π 0 κ e −1
with x = ω/T . This function is shown in Fig. 5.3. In a standard solar

model κ2 ≈ 12 throughout the Sun with a variation of less than 15%. In
the core of an HB star with ρ = 104 g/cm3 and T = 108 K it is κ2 ≈ 2.5.
One finds F = 0.98 and 1.84 for κ2 = 2.5 and 12, respectively.
170 Chapter 5
Fig. 5.3. Function F (κ2 ) according to Eq. (5.10).
5.2.2 Plasmon Decay and Coalescence

Several authors have used the plasma frequency ωP instead of the
Debye-Hückel wave number as a screening scale. In a nondegenerate
plasma kS2 /ωP2 ≈ me /T (Sect. 6.3) and so they overestimated the emis-
sion rate. Another source of confusion are statements that the decay
γ → γa was another axion emission process enabled by the 2γ coupling.
This issue can be easily clarified, but one needs to draw heavily on the
discussion of photon dispersion of Sect. 6.3.
A medium allows for both transverse and longitudinal electromag-
netic excitations. In a nondegenerate and nonrelativistic plasma the
dispersion relation of the former is ωT2 −kT2 = ωP2 (plasma frequency ωP )
while the latter oscillate essentially with a fixed frequency ωL = ωP ,
independently of their wave number. Therefore, the plasmon decay
process γT → γL a as well as the plasmon coalescence γT γL → a are
indeed kinematically possible for kL > ωP .
It is easy to calculate the inverse lifetime of transverse excitations
against these processes
)
d3 k L d3 k a 1
ΓγT →a = 3 3
ZT ZL |M|2 (2π)4
2ωL (2π) 2ωa (2π) 2ωT
% (
δ 4 (KT + KL − Ka ) δ 4 (KT − KL − Ka )
× + , (5.11)
eωL /T − 1 1 − e−ωL /T
where ZT,L are the vertex renormalization factors. Here, the first term
in square brackets corresponds to coalescence and so it involves a Bose-
Einstein occupation number for the initial-state γL . The second term
is from decay; it involves a Bose stimulation factor [1 + (eωL /T − 1)−1 ] =

(1−e−ωL /T )−1 for the final-state γL . For the axions, a stimulation factor
is not included because they are assumed to escape immediately.
Because a longitudinal excitation has no magnetic field, the elec-
tric field in the matrix element corresponding to Eq. (5.1) must be
associated with a longitudinal, the magnetic field with a transverse ex-
citation. In general B = ∇ × A and E = −∇A0 − ∂t A with the vector
potential A. For a propagating mode A ∝ e−i(ωt−k·x) , where , is a
polarization vector. This implies that B ∝ kT × !T where !T is a po-
larization vector transverse to kT with !2T = 1, and E ∝ ωL !L − kL ,0L
where ,L = (kL2 , ωL kL )kL−1 (ωL2 − kL2 )−1/2 according to Eq. (6.27) so that
E ∝ k̂L (ωL2 − kL2 )1/2 . Then one finds from the usual Feynman rules
|M|2 = gaγ
2
(ωL2 − kL2 ) |k̂L · (kT × !T )|2 . (5.12)
Note that |k̂L · (kT × !T )|2 = |(k̂L × kT ) · !T |2 . Averaging over the two
transverse polarization states yields 12 |k̂L × kT |2 .
If one writes ZL = Z!L ωL2 /(ωL2 − kL2 ) as in Sect. 6.3 and performs the
3
d kL integration in Eq. (5.11) one finds
2 )
gaγ ωa ωL
ΓγT →a = 2
dΩa dωa ZT Z!L |k̂L × kT |2
16 (2π) ωT
% (
δ(ωT + ωL − ωa ) δ(ωT − ωL − ωa )
× + . (5.13)
eωL /T − 1 1 − e−ωL /T
In a nondegenerate, nonrelativistic plasma typically ωP % T . Because

ωL = ωP one may expand the exponentials so that (eωL /T −1)−1 → T /ωL
and also (1 − e−ωL /T )−1 → T /ωL . If ωT = O(T ) + ωP one may ignore
ωL in the δ functions which are then trivial to integrate,
2
gaγ T )
ΓγT →a = 2
dΩa ZT Z!L |k̂L × kT |2 . (5.14)
8 (2π)
In this limit a has the same energy as γT while γL has only provided
momentum.
To finish up, note that in the nondegenerate, nonrelativistic limit
ZT = Z!L = 1. Because ωP % T and ωT = O(T ) we have ωP % ωT
so that kT ≈ ωT . Moreover, kL = kT − ka yielding |k̂L × kT |2 =
|ka × kT |2 /|ka − kT |2 = ωT2 (1 − z 2 )/2(1 − z) = 12 ωT2 (1 + z) where z is the
172 Chapter 5
cosine of the angle between ka and kT . The angular integration then

averages z to zero and leaves us with
2
gaγ T
ΓγT →a = . (5.15)
16π
Collective longitudinal oscillations only exist for kL <
∼ kD (Debye screen-
ing scale). Because a momentum transfer |kT −ka | = O(ωT ) is required,
this result applies only for ωT < ∼ kD . This conversion rate agrees with
the low-ω expansion of the Primakoff result Eq. (5.8).
It must be stressed that Eq. (5.15) is not an additional contribution
to the conversion rate, it is the same result derived in a different fashion.
Here, longitudinal plasmons in the static limit were used as the external
electric field in which the Primakoff effect takes place. Before, the static
limit was taken from the start; the collective behavior of the electron
motion was reflected in the screening of the Coulomb potential. These
two paths of performing the calculation in the end extracted the same
information from the electromagnetic polarization tensor which defines
both screening effects and the dispersion behavior of electromagnetic
excitations.
5.2.3 Axion Emission from Electromagnetic Plasma

Fluctuations
These simple calculations of the axion emission rate apply only in the
classical (nondegenerate, nonrelativistic) limit. Even though this is the
most relevant case from a practical perspective it is worth mentioning
how one proceeds for a more general evaluation. To this end note that
the 2γ interaction Eq. (5.1) corresponds to a source term for the axion
wave equation,
( + m2a ) a = gaγ E · B, (5.16)
where = ∂ µ ∂µ = ∂t2 − ∇2 . Axions are then emitted by the E · B fluc-
tuations caused by the presence of thermal electromagnetic radiation
as well as the collective and random motion of charged particles.
The Primakoff calculation in Sect. 5.2.1 used the (screened) electric
field of charged particles and the magnetic field of (transverse) electro-
magnetic radiation (photons) as a source. Actually, one can include
the magnetic field of moving charges for this purpose. Then axions are
emitted in the collision of two particles (Fig. 5.4), a process sometimes
referred to as the “electro Primakoff effect.” Unsurprisingly, the emis-
sion rate is much smaller because the magnetic field associated with
Fig. 5.4. Example for the electro Primakoff effect.
nonrelativistic moving particles is small. (For an explicit calculation

see Raffelt 1986a.) Put another way, the dominant contribution to
magnetic field fluctuations in a nonrelativistic plasma is from photons,
not from moving charges.
In order to illustrate the relationship between field fluctuations and
axion emission consider the amplitude for the conversion of a classical
transverse electromagnetic wave into a classical axion wave in the pres-
ence of an electric field configuration E(x). One finds from Eq. (5.16)
)
gaγ gaγ
f (Ω) = (! × k) · d3 x e−iq·x E(x) = (! × k) · E(q),
4π 4π
(5.17)
where k and ! are the wave and polarization vector of the incident
wave, respectively, q is the “momentum” transfer to the axion, and
E(q) is a Fourier component of E(x). The differential transition rate
is dΓ/dΩ = |f (Ω)|2 so that
dΓγ→a g2
= aγ 2 (! × k)i (! × k)j Ei (−q)Ej (q), (5.18)
dΩ (4π)
where it was used that E(x) is real so that E∗ (q) = E(−q). If E(x) is
a random field configuration one needs to take an ensemble average so
that the transition rate is proportional to -Ei Ej .q ≡ -Ei (−q)Ej (q)..
The electric and magnetic field fluctuations of a plasma are inti-
mately related to the medium response functions to electric and mag-
netic fields, i.e. to the polarization tensor. For a plasma at tempera-
ture T one can show on general grounds (Sitenko 1967)
) +∞ dω 2
-Ei Ej .q = ω/T
−∞ 2π e −1
% & ' (
qi qj Im ,L qi qj Im ,T
× 2 2
+ δij − 2 , (5.19)
q |,L | q |,T − q2 /ω 2 |2
where ,L,T (ω, q) are the longitudinal and transverse dielectric permit-
tivities of the medium (Sect. 6.3.3). A quantity such as Im ,L /|,L |2 is
known as a spectral density—here of the longitudinal fluctuations.
174 Chapter 5
One finds explicitly -Ei Ej .q = q̂i q̂j T /(1 + q2 /kS2 ) in the classi-
cal limit (Sitenko 1967) where kS2 is the Debye-Hückel wave number of
Eq. (5.7). With this result one easily reproduces the Primakoff transi-
tion rate Γγ→a (Raffelt 1988a).
The language of spectral densities for the electromagnetic field fluc-
tuations forms the starting point for a quantum calculation of the axion
emission rate in the framework of thermal field theory. This program
was carried out in a series of papers by Altherr (1990, 1991), Altherr and
Kraemmer (1992), and Altherr, Petitgirard, and del Rı́o Gaztelurrutia
(1994). Naturally, in the classical limit they reproduced the Primakoff
transition rate ΓγT →a of Eq. (5.8).
In the degenerate or relativistic limit their results cannot be rep-
resented in terms of simple analytic formulae. The most important
astrophysical environment to be used for extracting bounds on gaγ are
low-mass stars before and after helium ignition with a core temper-
ature of about 108 K (Sect. 5.2.5). Altherr, Petitgirard, and del Rı́o
Gaztelurrutia (1994) gave numerical results for the energy-loss rate for
this temperature as a function of density shown in Fig. 5.5 (solid line).
The dashed line is the classical limit Eq. (5.9); it agrees well with the
general result in the low-density (nondegenerate) limit. In the degen-
Fig. 5.5. Energy-loss rate of a helium plasma at T = 108 K by axion emission

with gaγ = 10−10 GeV−1 . The solid line is from transverse-longitudinal
fluctuations; the dashed line is the corresponding classical limit. The dotted
line is from transverse-transverse fluctuations, i.e. in the axion source term
gaγ E·B both fields are from transverse fluctuations. (Adapted from Altherr,
Petitgirard, and del Rı́o Gaztelurrutia 1994.)
erate limit, the emission rate drops precipitously, an important feature

which will be taken advantage of in Sect. 5.2.5 below.
5.2.4 Solar Axion Spectrum

As a first practical application it is easy to calculate the expected flux
of axions at Earth from the Primakoff conversion in the Sun where the
classical approximation is well justified. To this end van Bibber et al.
(1989) have integrated Eq. (5.9) over a standard solar model which
yields an axion luminosity
2
La = g10 1.7×10−3 L& , (5.20)
with L& the solar luminosity and g10 ≡ gaγ × 1010 GeV. (Recalling that
gaγ = (α/πfa ) Caγ this corresponds to fa /Caγ = 2.3×107 GeV.) The
differential flux at Earth is well approximated by the formula
dFa (ωa /keV)3

2
= g10 4.02×1010 cm−2 s−1 keV−1 ωa /1.08 keV (5.21)
dωa e −1
which is shown in Fig. 5.6. The average axion energy is -ωa . = 4.2 keV.
2
The total flux at Earth is Fa = g10 3.54×1011 cm−2 s−1 .
The “standard Sun” is about halfway through its main-sequence
evolution. Therefore, the solar axion luminosity must not exceed its
Fig. 5.6. Axion flux at Earth according to Eq. (5.25) from the Primakoff
conversion of photons in the Sun.
176 Chapter 5
photon luminosity; otherwise its nuclear fuel would have been spent
before reaching an age of 4.5×109 yr. This requirement yields a bound
gaγ < −9 −1
∼ 2.4×10 GeV . (5.22)
Detailed solar evolution calculations of Raffelt and Dearborn (1987)

showed that this bound was firm, but also that it could not be im-
proved easily (Sect. 1.3.2). The present-day properties of the Sun could
be obtained by a suitable adjustment of the unknown presolar helium
abundance.
5.2.5 Globular-Cluster Bound on gaγ

Armed with the Primakoff emission rate Eq. (5.9) it is an easy task
to derive a bound on gaγ from the energy-loss argument applied to
globular-cluster stars (Sect. 2.5). We need to require that at T ≈ 108 K
the axionic energy-loss rate is below 10 erg g−1 s−1 for a density of about
0.6×104 g cm−3 , corresponding to a classical plasma, and for about
2×105 g cm−3 , corresponding to degeneracy. From Fig. 5.5 it is evi-
dent that the emission rate is a steeply falling function of density when
degeneracy effects become important. Obviously, the more restrictive
limit is found from the low-density case which is based on the helium-
burning lifetime of HB stars (Sect. 2.5.1).
In order to calculate the average energy-loss rate of the core of an
HB star one needs -T 7 /ρ. if in Eq. (5.9) one uses a constant κ2 =
2.5 or F = 1.0. For a typical HB-star model (Fig. 1.4) one finds
-T87 /ρ4 . ≈ 0.3 where T8 = T /108 K and ρ4 = ρ/104 g cm−3 . There-
2
fore, -,a . ≈ g10 30 erg g−1 s−1 so that the criterion Eq. (2.40) yields a
constraint
gaγ < −10

∼ 0.6×10 GeV
−1
or fa /Caγ > 7
∼ 4×10 GeV. (5.23)
The temperature 108 K corresponds to 8.6 keV; a typical photon energy

is 3 T ≈ 25 keV. Therefore, this bound applies to pseudoscalars with a
mass ma < ∼ 30 keV while for larger masses it would be degraded.
5.3 Search for Cosmic Axions

One of the most interesting ramifications of the electromagnetic cou-
pling of pseudoscalars is the possibility to search for dark-matter axions.
It is briefly explained in Chapter 14 that axions would be produced in
the early universe by a nonthermal mechanism which excites classical
axion field oscillations, i.e. a highly degenerate axion Bose condensate

that would play the role of “cold dark matter,” and notably provide the
unseen mass necessary to explain the rotation curves of spiral galax-
ies such as our own (e.g. Kolb and Turner 1990). Because there are
uncertainties with regard to details of the primordial axion production
mechanism the exact value of the relevant axion mass is not known.
However, typically it is of order 10−5 eV so that it is a reasonable spec-
ulation that the mass of our galaxy is dominated by very low-mass
bosons. As these particles are bound to the galaxy they must be non-
relativistic; a typical velocity dispersion corresponding to the galactic
gravitational potential is around 10−3 in units of the speed of light.
Sikivie (1983) proposed to search for galactic axions by means of a
Primakoff-like method. The a → γ conversion of nonrelativistic axions
in the µeV mass range produces photons in the microwave (GHz) range.
Therefore, the idea is to place a microwave cavity in a strong magnetic
field and wait for cavity modes to be excited by the axion field. In
this context one may view the electromagnetic modes of the cavity and
the free axion field modes as oscillators which are coupled by virtue of
the interaction Eq. (5.1) where B is the external static field while E is
from an electromagnetic cavity mode. Then, power is transferred from
the axion field to the cavity excitations by virtue of the oscillator beats
induced by the coupling; detailed calculations of the conversion rate
were performed by Sikivie (1985) and Krauss et al. (1985).
It is worth noting that with a mass of 10−5 eV and a velocity of
10 a typical axion momentum is 10−8 eV which corresponds to a
−3
wave length of about 20 m. Thus on laboratory scales the axion field

is homogeneous. This does not apply to free microwaves—their en-
ergy and momentum are the same (ωγ = |kγ |); for 10−5 eV their wave
length is 2 cm. The role of the resonant cavity is to overcome this mo-
mentum mismatch: on resonance the fundamental cavity frequency is
degenerate with nonrelativistic axions of a certain mass for which the
energy transfer is maximized. In a search experiment the cavity must
be stepped through a range of resonant frequencies which defines the
range of axion masses to which a given experimental setup is sensitive.
Two pilot experiments of this sort were completed several years ago
(Wuensch et al. 1989; Hagmann et al. 1990). Assuming an axionic dark-
matter density at the Earth of 5×10−25 g cm−3 = 300 MeV cm−3 allowed
these groups to exclude the range of masses and coupling constants
shown in Fig. 5.7. The solid line indicates the relationship between gaγ
and ma in axion models where E/N = 8/3 or ξ = 1 in Eq. (14.24), i.e.
where gaγ = (ma /µeV) (0.69×1016 GeV)−1 .
178 Chapter 5
Most excitingly, the galactic axion search is going to be taken up

again with two new experimental setups. The one in Livermore (Cali-
fornia) has an increased detection volume and magnetic field (B 2 V =
14 T2 m3 ), and a refined microwave detection method (van Bibber et al.
1992, 1994). Within a running time of two or three years it will be pos-
sible to explore the axion mass range 1.3 − 13 µeV down to a coupling
strength gaγ which is only a factor of about 2.5 shy of the “axion line”
in Fig. 5.7. If axions interact with photons somewhat stronger than
indicated by this line, or if the local dark-matter density is somewhat
larger than assumed in Fig. 5.7, one may already be able to detect
axions in this round of measurements.
The second experiment (Kyoto, Japan) will use Rydberg atoms in a
novel scheme to detect single microwave quanta (Matsuki et al. 1995).
Because of the intrinsic low noise of this detector one can go to lower
physical temperatures, thereby reducing thermal noise, and thus allow-
ing one to use smaller cavities. This in turn permits one to search for
larger axion masses than is possible with the Livermore-type large cav-
ities. Together, the two experiments can probably cover two decades of
axion masses, between about 10−6 and 10−4 eV.
Fig. 5.7. Results of the galactic axion search experiments of the Rochester-
Brookhaven-Fermilab (RBF) collaboration (Wuensch et al. 1989) and of
the University of Florida (UF) experiment (Hagmann et al. 1990). The
hatched areas are excluded, assuming a local dark-matter axion density of
5×10−25 g cm−3 = 300 MeV cm−3 . The “axion line” is the relationship be-
tween axion mass and coupling strength for ξ = 1 or E/N = 8/3 according
to Eq. (14.24).
5.4 Axion-Photon Oscillations

5.4.1 Mixing Equations
The “axion haloscope” discussed in the previous section was based on
the Primakoff conversion between axions and photons in a macroscopic
magnetic or electric field. The same idea can be applied to other axion
fluxes such as that expected from the Sun (“axion helioscope,” Sikivie
1983). Typical energies of solar axions are in the keV range (Sect. 5.2.4)
so that a macroscopic laboratory magnetic field is entirely homogeneous
on the scale of the axion wavelength.
In this case the conversion process is best formulated in a way analo-
gous to neutrino oscillations (Anselm 1988; Raffelt and Stodolsky 1988).
This approach may seem surprising as the axion has spin zero while
the photon is a spin-1 particle. States of different spin-parity can mix,
however, if the mixing agent (here the external magnetic field) matches
the missing quantum numbers. Therefore, only a transverse magnetic
or electric field can mix a photon with an axion; a longitudinal field
respects azimuthal symmetry whence it cannot mediate transitions be-
tween states of different angular momentum components in the field
direction.
The starting point for the magnetically induced mixing between
axions and photons is the classical equation of motion for the system
of electromagnetic fields and axions in the presence of the interaction
Eq. (5.1). In terms of the electromagnetic field-strength tensor F and
its dual F! one finds, apart from the constraint ∂µ F! µν = 0,
∂µ F µν = J ν + gaγ F! µν ∂µ a,
( + m2a ) a = − 14 gaγ Fµν F! µν , (5.24)
where J is the electromagnetic current density.
In a physical situation with a strong external field plus radiation one
µν µν
may approximate gaγ F! µν ∂µ a → gaγ F!ext ∂µ a because a term gaγ F!rad ∂µ a
is of second order in the weak radiation fields. Moreover, if only an ex-
ternal magnetic field is present, the wave equation for the time-varying
part of the vector potential A and for the axion field are
A = gaγ BT ∂t a,
( − m2a ) a = −gaγ BT · ∂t A, (5.25)
where BT is the transverse external magnetic field. If one specializes to
a wave of frequency ω propagating in the z-direction, and denoting the
180 Chapter 5
components of A parallel and perpendicular to BT with A' and A⊥ ,

respectively, one finds (Raffelt and Stodolsky 1988)
    
n⊥ − 1 nR 0 A⊥
 2 2 2   
ω + ∂z + 2ω  nR n' − 1 gaγ BT /2ω   A'  = 0,
0 gaγ BT /2ω −m2a /2ω 2 a
(5.26)
where the off-diagonal terms were made real by a suitable global trans-
formation of the fields. Further, a photon index of refraction was in-
cluded because in practice one never has a perfect vacuum.
The refractive index is generally different for the two linear polar-
ization states parallel and perpendicular to BT (Cotton-Mouton effect).
Also, there may be mixing between the A⊥ and A' fields, i.e. the plane
of polarization may rotate in optically active media, an effect charac-
terized by nR . In general, any medium becomes optically active if there
is a magnetic field component along the direction of propagation (Fara-
day effect). Therefore, in general the refractive indices n⊥,' depend on
the transverse, the index nR on the longitudinal magnetic field.
Equation (5.26) is made linear by an approach that will be discussed
in more detail for neutrinos in Sect. 8.2. For propagation in the positive
z-direction and for very relativistic axions and photons one may expand
(ω 2 + ∂z2 ) = (ω + i∂z )(ω − i∂z ) → 2ω (ω − i∂z ). Then one obtains the
usual “Schrödinger equation”
    
∆⊥ ∆R 0 A⊥
    
ω +  ∆R ∆' ∆aγ  + i∂z   A'  = 0, (5.27)
0 ∆aγ ∆a a
where ∆',⊥ = (n',⊥ − 1) ω, ∆R = nR ω, ∆a = −m2a /2ω, and ∆aγ =
1
g B .
2 aγ T
If one ignores a possible optical activity or Faraday effect (nR = 0),
the lower part of this equation represents a 2 × 2 mixing problem. The
matrix is made diagonal by a rotation about an angle
1 ∆aγ gaγ BT ω
2
tan 2θ = = . (5.28)
∆' − ∆a (n' − 1)2ω 2 + m2a
In analogy to neutrino oscillations (Sect. 8.2.2) the probability for an
axion to convert into a photon after travelling a distance / in a trans-
verse magnetic field is
prob(a → γ) = sin2 (2θ) sin2 ( 12 ∆osc /), (5.29)
where ∆2osc = (∆' − ∆a ) +2
∆2aγ so that the oscillation length is /osc =
2π/∆osc .
By adjusting the gas pressure within the magnetic field volume one
can make the photon and axion degenerate and thus enhance the tran-
sition rate (van Bibber et al. 1989). This applies, in particular, to
solar axions which have keV energies so that the corresponding pho-
ton dispersion relation in low-Z gases is “particle-like” with the plasma
frequency being the effective mass.
If there is a gradient of the gas density, for example near a star, or
if the gas density and magnetic field strength change in time as in the
expanding universe, suitable conditions allow for resonant axion-photon
conversions in the spirit of the neutrino MSW effect (Yoshimura 1988;
Yanagida and Yoshimura 1988).
For the magnetic conversion of pseudoscalars in the galactic mag-
netic field one must worry about density fluctuations of the interstellar
medium which can be of order the medium density itself. In this case
Eq. (5.29) is no longer valid because it was based on the assumption of
spatial homogeneity of all relevant quantities. Carlson and Garretson
(1994) have derived an expression for the conversion rate in a medium
with large random density variations. They found that it can be sig-
nificantly suppressed relative to the naive result.
5.4.2 Solar Axions
An axion helioscope experiment was performed by Lazarus et al. (1992)

who used a vacuum pipe of 6)) diameter which was placed in the bore of
a dipole magnet of 72)) length (1.80 m); the field strength was 2.2 T. The
helioscope was oriented so that its long axis pointed along the azimuth
of the setting Sun. This provided a time window of approximately
15 min every day during which the line of sight through the vacuum
region pointed directly to the Sun. As a detector they used an x-ray
proportional chamber at the end of the pipe.
Data were taken on several days with He at different pressures in
the pipe. At 1 atm helium provides a plasma mass of about 0.3 keV to
x-rays. For vacuum, a 3σ bound of gaγ < 3.6×10−9 GeV−1 for ma <
0.050 eV was found. For a helium pressure of 55 Torr the limit was
3.9 in the same units, applicable to 0.050 < ma /eV < 0.086, and for
100 Torr it was 3.4, applicable to 0.086 < ma /eV < 0.110. These
bounds assume an axion flux as given by Eq. (5.21) which in turn
assumes an unperturbed Sun. Unfortunately, this assumption is not
consistent because the present-day age of the Sun already requires the
bound Eq. (5.22).
182 Chapter 5
However, an ongoing experimental project at the Institute for Nu-

clear Physics in Novosibirsk may be able to improve the helioscope
significantly. The conversion magnet has been gimballed so that it can
track the Sun, providing much longer exposure times. First results can
be expected for late 1995—see Vorobyov and Kolokolov (1995) for a
status report.
Another possibility would be to use the straight sections of the
beam pipe of the LEP accelerator at CERN as an axion helioscope.
Hoogeveen and Stuart (1992) have calculated the times and dates of
alignment with the Sun. They proposed an experimental setup that
might allow one to reach a sensitivity in gaγ down to 4×10−10 GeV−1 ,
which would be very impressive, but still far from the globular-cluster
bound Eq. (5.23).
Finally, Paschos and Zioutas (1994) proposed to use a single crystal
as a detector where the Primakoff conversion of solar axions is coher-
ently enhanced over the electric fields of many atoms. Put another way,
one would expect a strong enhancement via Bragg scattering. Even
with this improvement, however, it does not seem possible to beat the
bound from globular-cluster stars.
5.4.3 Shining Light through Walls
Instead of using the solar axion flux one can make one’s own by shining
a laser beam through a long transverse magnetic field region where it
develops an axion component. Then the laser beam is blocked while
the weakly interacting axions traverse the obstacle. In a second magnet
they are back-converted into photons so that one “shines light through
walls” (Anselm 1985; Gasperini 1987; van Bibber et al. 1987). Instead
of a freely propagating beam one may use resonant cavities on either
side of the wall which are coupled by the axion field (Hoogeveen and
Ziegenhagen 1991). Another possibility to improve the sensitivity is to
use squeezed light (Hoogeveen 1990).
An actual experiment was performed by Ruoso et al. (1992) who
used two superconducting magnets of length 440 cm each with a field
strength of 3.7 T. The light beam was trapped in a resonant cavity
in the first magnet, allowing for about 200 traversals; the incident
laser power was 1.5 W. At the end of the second magnet photons were
searched for by a photomultiplier. For an axion mass ma < −3
∼ 10 eV an
upper bound gaγ < 0.7×10−6 GeV−1 was found.
5.4.4 Vacuum Birefringence

Besides the conversion between photons and axions there is a more
subtle effect that can serve to search for the two-photon vertex of low-
mass pseudoscalars. In an external transverse E or B field the mix-
ing between the A' component with a leads to a backreaction on A'
(Fig. 5.8b) which amounts to a retardation of its phase. Put another
way, the / and ⊥ polarization states have different refractive indices in
vacuum with an E or B field (vacuum Cotton-Mouton effect). There-
fore, if one shines a light beam which is linearly polarized at 45◦ relative
to a transverse B field, the beam will develop a small degree of elliptic
polarization (Maiani, Petronzio, and Zavattini 1986).
A vacuum Cotton-Mouton effect is expected even in the absence of
axions from the QED amplitude shown in Fig. 5.8a; in the nonforward
direction it describes Delbrück scattering on a charged particle. More
generally, an electron loop mediates an effective γγ interaction which
for low energies can be described by the Euler-Heisenberg Lagrangian,
2α2 6 2 2 2 2
7
Lγγ = (E − B ) + 7(E · B) (5.30)
45m4e
(Heisenberg and Euler 1936; see also Itzykson and Zuber 1983).
Fig. 5.8. Vacuum birefringence in the presence of external fields. (a) QED
contribution according to the Euler-Heisenberg γγ interaction. (b) γa or
γπ ◦ oscillations in an external E or B field. (c) Photon birefringence in an
external axion field (axionic domain walls, cosmic axion field). (d) Axion-
mediated contribution in a strong E · B field, e.g. near a pulsar.
184 Chapter 5
The vacuum refractive index in an external magnetic field which

results from this interaction was studied by a number of authors—
see Tsai and Erber (1975, 1976) for references to the early literature.
Adler (1971) provided a comprehensive study and derived the correct
expression for the related photon splitting rate γ → γγ in an external
field.28 The refractive indices for the / and ⊥ polarization states are29
2α2 B 2 2α2 B 2
n' = 1 + 7 and n⊥ = 1 + 4 , (5.31)
45 m4e 45 m4e
where
(2α2 /45) B 2 /m4e = 1.32×10−32 (B/Gauss)2 . (5.32)
Note that 1 Gauss corresponds to 10−4 Tesla, and to 1.95×10−2 eV2 in

natural units (Appendix A).
Clearly one needs very strong magnetic fields for vacuum birefrin-
gence effects to become important. So far, no positive experimental
measurement exists. A proposal to measure the acquired elliptic polar-
ization of a laser beam was put forth by Iacopini and Zavattini (1979).
More recently Cantatore et al. (1991) proposed to use polarized light
scattered off an electron beam, a method which allows one to obtain
polarized GeV photons. As the relative phase shift is (n' − n⊥ )ω/ for
a distance of travel / in the magnetic field, high-energy photons show
a much stronger effect for otherwise equal conditions.
An experiment to search for the axion contribution of Fig. 5.8b
was recently performed (Semertzidis et al. 1990; Cameron et al. 1993).
Note that there are two axion-induced effects on a laser beam trapped
in an optical cavity. One is the birefringence effect analogous to the
QED effect which leads to a small amount of elliptical polarization.
Another is the loss of / photons into the axion channel which depletes
the amplitude of the / mode relative to the ⊥ one, which in turn leads
to a rotation of the plane of polarization. Both effects are of the same
order in the coupling constant; experimentally, the rotation effect led to
28
The photon-splitting box graph with one external field and three real photons
attached to an electron loop does not contribute. The lowest-order amplitude is
with the external field attached three times, and three real photons (hexagon dia-
gram). For references to the early literature and a discussion of the astrophysical
implications of the photon-splitting process see Baring (1991).
29
/, ⊥ refer to the electric field of the wave relative to the external transverse B
field while Adler (1971) refers with /, ⊥ to the magnetic field of the wave. Note
also that I use rationalized units where α = e2 /4π = 1/137 while in the literature
on photon refraction unrationalized units with α = e2 = 1/137 are often employed.
a more restrictive limit of gaγ < 3.6×10−7 GeV−1 for ma < −4

∼ 7×10 eV.
For a larger mass an a-γ oscillation pattern develops on the length scale
of the optical cavity, leading to an “oscillating limit” as a function of
ma . In this regime the ellipticity measurement was superior.
New experimental efforts in the birefringence category include a
proposal by Cooper and Stedman (1995) to use ring lasers. A laser
experiment which is actually in the process of being built is PVLAS
(Bakalov et al. 1994) which will be able to improve previous laboratory
limits on gaγ by a factor of 40, i.e. it is expected to be sensitive in
the regime gaγ > −8 −1 < −3
∼ 1×10 GeV as long as ma ∼ 10 eV. While such
strong couplings are astrophysically excluded it is intriguing that this
experiment should be able to detect for the first time the standard QED
birefringence effect of Fig. 5.8a.
5.5 Astrophysical Magnetic Fields

5.5.1 Transitions in Magnetic Fields of Stars
Certain stars have very strong magnetic fields. For example, neutron
stars frequently have fields of 1012 −1013 G (e.g. Mészáros 1992), and
even white dwarfs can have fields of up to 109 G. Therefore, one may
think that axions produced in the hot interior of neutron stars at a tem-
perature of, say, 50 keV would convert to γ-rays in the magnetosphere
(Morris 1986). However, the vacuum refractive term suppresses the
conversion rate because the photon momentum for a given frequency
is kγ = n',⊥ ω > ω with the refractive indices Eq. (5.31) while for the
axions ka = ω − m2a /2ω < ω. Therefore, in the presence of a magnetic
field axions and photons are less degenerate so that it is more difficult
for them to oscillate into each other.
In principle, the refractive index can be cancelled by the presence
of a plasma where the photon forward scattering on electrons induces a
negative n−1, i.e. something like a photon effective mass. In the aligned
rotator model for a magnetized neutron star a self-consistent solution of
the Maxwell equations with currents requires the presence of an electron
density of about ne = 7×1010 cm−3 B12 Ps−1 where B12 is the magnetic
field along the rotation axis in units of 1012 G and Ps is the pulsar period
in seconds (Goldreich and Julian 1969). The corresponding plasma
frequency is ωP2 = 4παne /me = 0.97×10−10 eV2 B12 Ps−1 . This implies
−1
k − ω = −ωP2 /2ω = −5×10−14 eV B12 Ps−1 ωkeV with ωkeV = ω/keV, to
be compared with ∆' = (n' − 1) ω = 0.92×10−4 eV B12 2
ωkeV which is
much larger, allowing one to ignore the plasma term.
186 Chapter 5
The off-diagonal term in the mixing matrix Eq. (5.27) is ∆aγ =

1
g B
2 aγ
= 0.98×10−9 eV g10 B12 with g10 = gaγ /(10−10 GeV−1 ). Ignoring
−2 −1
ma the oscillation length is 2π (∆2' + ∆2aγ )−1/2 ≈ 1.3 cm B12 ωkeV while
the geometric dimension of the dipole field is of order the stellar radius,
i.e. of order 10 km. Therefore, many oscillations occur within the mag-
netosphere, and the average transition probability between photons and
−1 −1
axions is θ2 with the mixing angle θ ≈ ∆aγ /∆' = 1.1×10−5 g10 B12 ωkeV .
Therefore, the transition rate is very small.
The vacuum refractive index is larger than unity, the plasma contri-
bution less than unity, and so near the stellar surface a crossover must
occur where axions and photons are degenerate. However, the length
scales do not work out to have a resonant MSW-type transition (Raffelt
and Stodolsky 1988; Yoshimura 1988).
At a pulsar, reducing B increases the mixing angle and thus the
transition rate while the oscillation length becomes larger. When /osc
far exceeds the geometric dimension R of the stellar magnetosphere,
and when the mixing angle is small, one can expand the sine functions
in Eq. (5.29) so that the transition probability is (θ∆osc R)2 ≈ (∆aγ R)2 .
This transition rate scales with B 2 as expected and becomes smaller for
smaller B. Therefore, the optimal situation is when the magnetic field
strength and geometric dimensions are matched such that R ≈ /osc .
This condition is approximately met in magnetic white dwarfs with,
say, B = 109 G, ω = 10 eV, and R = 103 km. In these systems one may
even expect a resonant level crossing if they have a dilute atmosphere
with an appropriate scale height (Raffelt and Stodolsky 1988; Gnedin
and Krasnikov 1992). However, a fortuitous combination of particle and
white-dwarf parameters is required, and, even then, observable effects
apparently have not been proposed.
Most recently, Carlson and Tseng (1995) have performed a study of
the conversion of very low-mass pseudoscalars in the magnetic field of
sunspots. They find that for certain parameters the x-ray flux from the
conversion process could be observable in solar x-ray telescopes such as
SXT and Yohkoh.
5.5.2 Birefringence in a Pulsar Magnetosphere

In the previous section it was shown that near a pulsar the QED vac-
uum Cotton-Mouton effect (Fig. 5.7a) induces a sizeable amount of
birefringence between the photon states which are linearly polarized
parallel or perpendicular to the transverse component of the magnetic
field. Recently, Mohanty and Nayak (1993) showed that in addition
there can be a strong and potentially observable circular birefringence

effect along the polar direction if massless pseudoscalars exist. The
masslessness of the pseudoscalars is crucial for this scenario and so ax-
ions (Chapter 14) which generically must have a mass do not fulfill this
requirement. Therefore, I use “arions” as a generic example which are
like axions in all respects except that they are true Nambu-Goldstone
bosons of a global chiral U (1) symmetry and thus strictly massless
(Anselm and Uraltsev 1982a,b; Anselm 1982).
The main idea of the pulsar birefringence scenario is that in the
oblique rotator model a strong E · B density exists in the pulsar mag-
netosphere which serves as a source for the arion field. Therefore, a
pulsar would be surrounded by a strong classical arion field density
which constitutes an optically active “medium,” causing a time delay
between the two circular polarization states of the pulsed radio emis-
sion from the polar cap region. As a Feynman graph, this situation is
represented by Fig. 5.8d.
In detail, Mohanty and Nayak (1993) considered the oblique ro-
tator model where the pulsar magnetic dipole axis is tilted with re-
gard to its rotation axis by an angle α. The instantaneous rotating
magnetic dipole field is B = (B0 R3 /r3 ) [3r̂ (r̂ · µ̂) − µ̂], where B0
is the magnetic field strength at the poles of the pulsar surface (ra-
dius R) and µ̂ is the instantaneous magnetic dipole direction with
the angle α relative to the angular velocity vector Ω. The time av-
erage of the electric field which is induced by the rotating magnetic
dipole, and which matches the boundary condition that the electric
field component parallel to the pulsar surface vanishes, is found to be
-E. = B0 R5 Ω cos α r−4 [3 (sin2 θ − 23 ) r̂ − 2 sin θ cos θ θ̂] where θ is the
polar angle relative to the rotation axis. The time-averaged value for
the pseudoscalar field density is
-E · B. = −B02 ΩR8 r−7 cos α cos3 θ. (5.33)
It appears as a source for the arion field on the r.h.s. of Eq. (5.24). Tak-
ing account of the relativistic space-time metric outside of the pulsar,
Mohanty and Nayak (1993) found for the resulting arion field
2 B02 R8 Ω cos α cos θ

a = −gaγ + O(r−3 ), (5.34)
575 (GN M)3 r2
where GN is Newton’s constant and M the pulsar mass. The en-
tire magnetosphere contributes coherently to this result. If the pseu-
doscalars had a mass, only the density E · B within a distance of about
188 Chapter 5
m−1
a would effectively act as a source for the local a field and so it would
be much smaller.
An inhomogeneous pseudoscalar field configuration represents an
optically active medium (Fig. 5.8c) as was noted, for example, in the
context of axionic domain wall configurations (Sikivie 1984). To low-
est order the dispersion relation for left- and right-handed circularly
polarized light is (e.g. Harari and Sikivie 1992)
k = ω ± 12 gaγ k̂ · ∇a, (5.35)
so that the momentum is shifted by a frequency-independent amount.
The corresponding refractive index is n = 1 ± 12 gaγ ω −1 k̂ · ∇a.
Taking account of the relativistic metric in the strong gravitational
field of a pulsar, Mohanty and Nayak (1993) then found for the time de-
lay between circularly polarized waves which propagate approximately
along the polar axis
8 9
2 2 2 4 B04 R11 Ω2 cos4 α
δt = gaγ . (5.36)
5 575 ω 2 (GN M )6
For the pulsar PSR 1937+21 a polarimetric analysis yields a time
delay 0.37 ± 0.67 µs and thus a 1σ upper limit of δt < 1µs (Klein
and Thorsett 1990). For typical pulsar parameters this allowed Mo-
hanty and Nayak (1993) to place a limit on the arion-photon coupling
of gaγ < −11
∼ 2×10 GeV .
−1
5.5.3 Conversion of Stellar Arions in the Galactic Field

Stars are powerful sources for pseudoscalars which can be produced in
the hot interior by the Primakoff process, i.e. by the conversion γ → a
in the electric fields of the charged medium constituents. Outside of
the star, the pseudoscalars can be converted back to photons in the
galactic magnetic field so that stars would appear to be sources of
x- or γ-rays, depending on the characteristic energy of the stellar core
(Carlson 1995).
In the galaxy, photons propagate with an effective mass given by
the plasma frequency ωP which for typical electron densities of order
0.1 cm−3 is of order 10−11 eV. As discussed in Sect. 5.4.1 the pseu-
doscalar to photon conversion process is an oscillation phenomenon
with a mixing angle given by Eq. (5.28); in the present context it is
1 gaγ BT ω
tan 2θ = 2 . (5.37)
2
ma − ωP2
A typical galactic field strength is 1 µG, a typical energy at most of or-
der 100 MeV for axions from supernovae. With gaγ < 0.6×10−10 GeV−1
one finds gaγ BT ω < −19 2

∼ 10 eV . For the allowed range of axion masses
this mixing angle is too small to yield a significant conversion effect.
Therefore, this entire line of argument is only relevant for massless
(or at least very low-mass) pseudoscalars which again shall be referred
to as “arions.” Carlson (1995) considered the star α-Ori (Betelgeuse),
a red supergiant about 100 pc away from us. He estimated its arion
luminosity from the Primakoff process, and compared the expected
x-ray flux with data from the HEAO-1 satellite. As a result, a new
limit of gaγ < −11 −1
∼ 2.5×10 GeV emerged which is more restrictive than
the above bound from globular-cluster stars.
Carlson’s argument yields an even more restrictive limit if applied
to SN 1987A. One may estimate the arion luminosity of the SN core on
the basis of the Primakoff process. If arions couple to quarks or elec-
trons, the luminosity can only be higher because existing axion limits
already indicate that arions cannot be trapped by these couplings. In
order to evaluate Eq. (5.9) an average temperature of 30 MeV and an
average density of 3×1014 g cm−3 with a proton fraction of 0.3 is used
(Sect. 13.4.2). The Debye screening scale by the protons is then found
to be 36 MeV so that κ2 = 1.41 in Eq. (5.10) leading to F = 0.72.
2
Therefore, the average energy loss rate is about g10 1.4×1016 erg g−1 s−1 .
Taking a core mass of 1M& = 2×1033 g and a duration of 3 s one ex-
pects about g102
1050 erg to be emitted in arions which is about 10−3 g102
of the energy emitted in each neutrino flavor.

Typical arion energies are 3T ≈ 100 MeV so that Eq. (5.37) together
with ωP ≈ 10−11 eV in the interstellar medium reveals that mixing is
nearly maximal for the relevant circumstances. Therefore, the oscil-
lation length is given by /osc = 4π/gaγ BT which is about 40 kpc for
BT = 1 µG and gaγ = 10−10 GeV−1 . Therefore, /osc far exceeds the
relevant magnetic field region which is of order 1 kpc as discussed in
Sect. 13.3.3b. The conversion rate is then
prob(a → γ) = ( 12 gaγ BT /)2

= 2.3×10−2 g10
2
(BT //µG kpc)2 , (5.38)
where / is the effective conversion region (distance to source or distance

within magnetic field region), and g10 ≡ gaγ /10−10 GeV−1 .
With Eq. (5.38) and an effective magnetic conversion region of / =
1 kpc the expected energy showing up as γ-rays at Earth corresponds
to about 10−5 g104
relative to the energy in one neutrino species. In
Sect. 12.4.3 the radiative decays of low-mass neutrinos from SN 1987A
was discussed. On the basis of the SMM data it was found that less
190 Chapter 5
than about 10−9 of a given neutrino species may show up in the form
of decay photons (Fig. 12.9) if the spectral distribution is taken to be
4 <
characterized by T ≈ 30 MeV. Therefore, one finds a limit of g10 ∼
10 or gaγ <
−4 −11 −1
∼ 10 GeV , applicable if the particle mass is below
about 10−10 eV. This is more restrictive than Carlson’s original limit,
and of the same order as the PSR 1937+21 birefringence limit quoted
after Eq. (5.36).
5.5.4 Polarimetry of Distant Radio Sources

Nambu-Goldstone bosons a are by definition the result of a sponta-
neously broken global symmetry. The cosmic evolution from a very
hot initial phase begins with the unbroken symmetry; as the universe
expands and cools a phase transition will occur where the field respon-
sible for the spontaneous breakdown must find its new minimum. As
this process occurs independently in each causally connected region of
the universe at that time, the universe today will be characterized by
different orientations of the ground state, i.e. by different values of a
classical background a field. If no inflation occurred in the universe
after the phase transition, and if the Nambu-Goldstone bosons remain
truly massless (in contrast with axions), the background field will not
have relaxed to a common ground state everywhere.
In this scenario a radio signal from a distant source travels through
regions with different values of the classical a field, and thus through
regions of gradients ∇a which act as an optically active medium accord-
ing to Eq. (5.35). Therefore, linearly polarized light will experience a
random rotation of its plane of polarization.
This effect is also expected from the Faraday rotation caused by
intervening magnetic fields which induce optical activity in the cosmic
background plasma. As this effect is frequency dependent it can be
removed by observing a given object at different wavelengths. The
effect induced by pseudoscalars, on the other hand, is independent of
frequency.
A systematic correlation between the geometric shape of distant
radio sources and the linear polarization of the emitted radiation has
been observed. This correlation proves that no random rotation of
the plane of polarization occurs over cosmic distances, except for the
Faraday effect which can be removed from the data. Therefore, the
maximum allowed coupling strength of photons to a random cosmic
Nambu-Goldstone field can be constrained. Harari and Sikivie (1992)
found that Caγ <∼ 50 in Eq. (5.5), independently of the symmetry break-
ing scale fa . This bound supersedes Sikivie’s (1988) previous scenario

where he tried to explain the polarization features of certain sources
by the conversion of cosmic-string-produced Nambu-Goldstone bosons
to photons in cosmic magnetic fields. This scenario would have re-
quired Caγ ≈ 105 .
5.5.5 Temperature Fluctuations in the Cosmic Microwave

Background
In the presence of large-scale magnetic fields in the universe, photons of
the cosmic microwave background radiation (CMBR) could convert into
arions. The angular variations of the CMBR temperature have been
measured by the COBE satellite and other instruments to be extremely
small; a typical value is δT /T ≈ 10−5 . Therefore, the conversion process
must not have been very efficient between the surface of last scattering
and us.
This argument has been studied in detail by Chen (1995) for photon-
graviton conversion which is a very similar effect due to the two-photon
coupling vertex which the massless gravitons must have. The coupling
constant involves the inverse Planck mass. Therefore, one may also
expect interesting effects for hypothetical arions which could couple to
photons more strongly than gravitons do.
5.6 Summary of Constraints on gaγ

The astrophysical and experimental bounds on the photon coupling
of arbitrary pseudoscalars are summarized in Fig. 5.9. “Haloscope”
refers to the search for galactic axions discussed in Sect. 5.3 and so
these constraints (Fig. 5.7) apply only if the pseudoscalars are the dark
matter in our galaxy. The dotted line is the search regime for the
ongoing experiment mentioned in Sect. 5.3.
“Helioscope” refers to the search for solar axion to x-ray conversion
(Sect. 5.4.2). It is shown as a dashed line because it is not self-consistent
in that it assumes an unperturbed Sun—the area enclosed by the dashed
line is already excluded by the solar age.
“Telescope” refers to the search for decay photons from the cosmic
axion background (Sect. 12.7.2, Fig. 12.23). It is assumed that the
pseudoscalars were in thermal equilibrium in the early universe.
“Laser” refers to the birefringence and shining-light-through-walls
experiments discussed above. The most restrictive such limit is from
the rotation of the plane of polarization of a laser beam trapped in
192 Chapter 5
an optical cavity in a strong transverse magnetic field (Sect. 5.4.4).

The dotted line marks roughly the expected range of sensitivity of the
PVLAS experiment (Sect. 5.4.4).
The solar limit (Eq. 5.22) is based on the Primakoff energy loss and
the requirement that axions must not exceed the photon luminosity;
otherwise the Sun could not have reached its present-day age.
The HB-star limit (Eq. 5.23) comes from the requirement that these
objects do not spend their nuclear fuel so fast that their observable
number in globular clusters is reduced by more than a factor of ≈ 2.
The “axion line” refers to models where E/N = 8/3 or ξ = 1 in
Eq. (14.24).
For very low-mass bosons (ma < −10
∼ 10 eV) the SN 1987A flux of
pseudoscalars would be efficiently converted into γ-rays, leading to a
limit gaγ < −11 −1
∼ 10 GeV (Sect. 5.5.3).
Fig. 5.9. Bounds on the photon coupling gaγ as a function of ma for arbitrary
pseudoscalars; see the text for details. (Adapted from Cameron et al. 1993.)
Chapter 6
Particle Dispersion and

Decays in Media
Dispersion effects in media have a significant impact on the propaga-

tion of some low-mass particles (photons, neutrinos) while others are
left unaffected (axions and other Nambu-Goldstone bosons). The rela-
tionship between forward scattering and refraction is derived, and the
dispersion relations for photons and neutrinos are thoroughly studied.
Modified particle dispersion relations allow certain decay processes to
occur in media that cannot occur in vacuum, notably the photon de-
cay γ → νν which dominates the neutrino emissivity in a wide range
of temperatures and densities (“plasma process”). Other examples are
the neutrino and majoron decay ν → νχ and χ → νν, respectively.
The rates for such processes are derived. The plasma process allows
one to derive the most restrictive limits on neutrino magnetic dipole
moments. Screening effects in reactions involving Coulomb scattering,
and neutrino electromagnetic form factors in media are discussed.
6.1 Introduction
Particles are the quantized excitations of certain fields—photons of the
electromagnetic field, electrons of the electron field, and so forth. It
is usually convenient to expand these fields in plane waves character-
ized by frequencies ω and wave vectors k; the excitations of these modes
then exhibit a temporal and spatial behavior proportional to e−i(ωt−k·x) .
The frequency for a given wave number is determined by the disper-
sion relation. Because (ω, k) is a four-vector, and because of Lorentz
invariance, in vacuum the quantity ω 2 − k2 = m2 is the same for all
193
194 Chapter 6
frequencies; m has the usual interpretation of a particle mass. One con-

sequence of this covariant dispersion relation is that decays of the sort
1 → 2 + 3 are only possible if m1 > m2 + m3 so that in the rest frame of
particle 1 there is enough energy available to produce the final states.
In media the dispersion relations are generally modified by the co-
herent interactions with the “background.” In the simplest case a parti-
cle acquires a medium-induced effective mass. For example, photons in
a nonrelativistic plasma acquire a dispersion relation ω 2 = ωP2 +k2 with
the plasma frequency given by ωP2 = 4πα ne /me (electron density ne ).
For ωP > 2mν this implies that the decay γ → νν becomes kinemati-
cally possible and occurs in stars because the ambient electrons mediate
an effective neutrino-photon interaction (Adams, Ruderman, and Woo
1963). In fact, this “plasma process” is the dominant neutrino source in
a wide range of temperatures and densities which covers, for example,
white dwarfs and red-giant stars (Appendices C and D).
Neutrinos may have nonstandard electromagnetic couplings, no-
tably magnetic dipole moments, which would enhance the plasma pro-
cess and thus the cooling of stars (Bernstein, Ruderman, and Feinberg
1963). Observational constraints on anomalous cooling rates derived
from white dwarfs and globular-cluster stars then provide the most
restrictive limits on neutrino electromagnetic couplings (Sect. 6.5.6).
Within the standard model all fermions are fundamentally massless;
they acquire an effective mass by their interaction with the vacuum
expectation value Φ0 of a scalar Higgs field (Sect. 8.1.1). Therefore,
even vacuum masses can be interpreted as “refractive” phenomena.
Because the scalar Φ0 is Lorentz invariant the dispersion relation thus
induced is of the standard form E 2 = m2 + p2 . “Normal” media,
however, single out a preferred Lorentz frame, usually causing E(p) to
be a more complicated function than (m2 + p2 )1/2 .
Notably, the dispersion relation can be such that the four-momen-
tum P = (E, p) is “space-like,” P 2 = E 2 − p2 < 0, which amounts
to a “negative mass-square” P 2 = m2eff < 0. There is nothing wrong
with such “tachyons” because the speed of signal propagation safely
remains below the speed of light (Sect. 6.2.2). The dispersion relation
in isotropic media is often expressed as k = |k| = nω in terms of a
refractive index n. Space-like excitations correspond to n > 1; examples
are photons in water or air. In this case the well-known decay process
e → eγ is kinematically allowed for sufficiently fast moving electrons
(Cherenkov radiation).
The dispersion relation can also depend on the spin polarization of
the radiation. In “optically active” media, the left- and right-handed
Particle Dispersion and Decays in Media 195
circular photon polarizations acquire different refractive indices. In this

sense all media are optically active for neutrinos where only the left-
handed states interact while the right-handed ones are “sterile.” For
Majorana neutrinos the helicity-plus states are equivalent to ν’s which
acquire an opposite energy shift from ν so that there is an energy gap
between ν(p) and ν(p). Therefore, in a medium the majoron decays
ν → νχ or χ → νν become possible where the majoron χ is a massless
particle (Sect. 6.8).
The interaction of νµ and ντ with a normal medium is different
from that of νe because of a charged-current νe -e− scattering ampli-
tude. Therefore, normal media are “flavor birefringent” in the sense
that the medium induces different dispersion relations for neutrinos
of different flavors. The importance of this effect for neutrino oscilla-
tions, which effectively measure relative phases in the propagation of
different-flavored neutrinos, cannot be overstated.
It must be stressed that usually all particles acquire nontrivial dis-
persion relations in media although it depends on the detailed circum-
stances whether or not the refractive effect is significant. For example,
until recently one found statements in the literature that in a sufficiently
dense medium where ωP > 2me photons were damped by electromag-
netic pair production γ → e+ e− . However, this is incorrect because the
charged leptons also acquire a medium-induced effective mass which is
so large that this decay never occurs (Braaten 1991). On the other
hand, the above majoron decay is only possible because the majorons
χ are Nambu-Goldstone bosons and thus remain massless even in a
medium, at least to lowest order (Sect. 6.8).
Besides modifying the dispersion relation of particles it is also possi-
ble that the presence of the medium allows for entirely new excitations.
The best known example is the longitudinal polarization state of the
electromagnetic field which exists in a plasma in addition to the usual
states with transverse polarization. These “plasmons” were first dis-
cussed by Langmuir (1926). Another example from electromagnetism
are the “plasminos,” spin- 12 excitations of a plasma that were discussed
for the first time only very recently (Klimov 1981; Weldon 1982b, 1989;
Pisarski 1989; Braaten 1992). For many purposes such (quantized) col-
lective modes play the same role as the usual particles. For example,
in a medium both photons and plasmons can decay into neutrinos and
thus contribute to the plasma process of neutrino emission.
In the present chapter I will follow up these questions in detail.
While the dispersion relations and couplings of particles in media are
formally best dealt with in terms of field theory at finite temperature
196 Chapter 6
and density, most of the results relevant for particle physics in stars
predate the development of this formalism; they were based on the
old-fashioned tools of kinetic theory. Indeed, for simple issues of dis-
persion or collective effects a kinetic approach seems often physically
more transparent while yielding identical results. At any rate, the fol-
lowing discussion is based entirely on kinetic theory.
6.2 Particle Dispersion in Media

6.2.1 Refractive Index and Forward Scattering
How does one go about to calculate the all-important dispersion rela-
tion for a given particle in a medium with known properties? Usually
it is enough to follow the elementary approach of calculating the for-
ward scattering amplitude of the relevant field excitations with the con-
stituents of the background medium, an approach which has the added
advantage of physical transparency over a more formal procedure.30
To begin, consider a scalar field Φ which may be viewed as repre-
senting one of the photon or electron polarization states. If a plane
wave excitation of that field with a frequency ω and a wave vector k in-
teracts with a scatterer at location r = 0 an additional spherical wave
will be created. The asymptotic form of the original plus scattered
wave is
! "
eikr
Φ(r, t) ∝ e−iωt eik·r + f (ω, θ) , (6.1)
r
where k = |k|, r = |r|, and f is the scattering amplitude. It was as-

sumed that it has no azimuthal dependence, something that will always
apply on average for a collection of randomly oriented scatterers. The
differential scattering cross section is dσ/dΩ = |f (ω, θ)|2 .
If there is a collection of scattering centers randomly distributed
in space, all of the individual scattered waves will interfere. However,
because of the random location of the scatterers, constructive and de-
structive interference terms will average to zero. Thus the total cross
section of the ensemble is the (incoherent) sum of the individual ones.
In the forward direction, however, the scattered waves add up co-
herently with each other and with the parent wave, leading to a phase
shift and thus to refraction. This is seen if one considers a plane wave in
the z-direction incident on an infinitesimally thin slab (thickness δa) at
30
The derivation below follows closely the exposition of Sakurai (1967).
z = 0 which contains n scattering centers per unit volume, and which is

infinite in the x- and y-directions. At a distance z from the slab, large
compared with k −1 , the asymptotic form of the parent plus scattered
wave is, ignoring the temporal variation e−iωt ,
# 2 2 1/2
iωz
∞ eik(ρ +z )
Φ(z) ∝ e + nδa f (ω, θ) 2πρ dρ , (6.2)
0 (ρ2 + z 2 )1/2
where ρ ≡ (x2 + y 2 )1/2 and θ = arctan(ρ/z). Moreover, it was assumed
that in vacuum the wave propagates relativistically so that k = ω.
The integral in Eq. (6.2) is ill defined because the integrand os-
cillates with a finite amplitude even for large values of ρ. It is made
convergent by substituting k → k + iκ with κ > 0 an infinitely small
real parameter. Integration by parts then yields
$ %
iωz 2πnδa
Φ(z) ∝ e 1+i f0 (ω) , (6.3)
ω
where a term of order (ωz)−1 was neglected which becomes small for
large z. Here, f0 (ω) ≡ f (ω, 0) is the forward scattering amplitude.
Turn next to a slab of finite thickness a. The phase change of the
transmitted wave is obtained by compounding infinitesimal ones with
δa = a/j and taking the limit j → ∞,
$ %j
2πna
lim 1 + i f0 (ω) = ei(2πn/ω)f0 a . (6.4)
j→∞ jω
Inserting this result in Eq. (6.3) reveals that over a distance a in the
medium the wave accumulates a phase einrefr ωa where
2π
nrefr = 1 + n f0 (ω) (6.5)
ω2
is recognized as the index of refraction.
If the relativistic approximation |nrefr − 1| & 1 is not valid one must
treat the wave self-consistently in the medium and distinguish carefully
between frequency and wavenumber. In this case one finds (Foldy 1945)
4π
n2refr = 1 + n f0 (k), (6.6)
k2
where the forward scattering amplitude must be calculated taking the
modified dispersion relation into account.
For the propagation of a field with several spin or flavor components
the same result applies if one remembers that “forward scattering” not
198 Chapter 6
only refers to scattering in the forward direction, but that all properties
of the wave and the scatterer are left unchanged. If the medium parti-
cles have a distribution of momenta, spins, etc. the forward scattering
amplitude must be averaged over those quantities, and different species
of medium particles must be summed over.
For a practical calculation it helps to recall that dσ/dΩ = |f (θ)|2
so that |f0 | is the square root of the forward differential cross section.
For example, the Thomson cross section for photons interacting with
nonrelativistic electrons is dσ/dΩ = (α/me )2 |, · ,$ |2 with the polariza-
tion vectors , and ,$ of the initial- and final-state photon. Forward
scattering implies |, · ,$ |2 = 1 so that |f0 | = α/me . The dispersion rela-
tion is then ω 2 = k 2 + ωP2 with the plasma frequency ωP2 = 4πα ne /me .
Of course, the absolute sign of f0 has to be derived from some other
information—for photon dispersion see Sect. 6.3.
The forward scattering amplitude and the refractive index are gen-
erally complex numbers. Physically it is evident that in a medium the
intensity of a beam is depleted as e−z/& . The mean free path is given by
-−1 = σnv where σ is the total scattering cross section, n is the number
density of scatterers, and v is the velocity of propagation. Thus the
amplitude of a plane wave varies as eikz−z/2& . Moreover, the derivation
of the refractive index indicates that the amplitude varies according to
einrefr ωz , yielding k = Re nrefr ω and (2-)−1 = Im nrefr ω. For relativistic
propagation (v = 1) the last equation implies σ(ω) = (4π/ω) Im f0 (ω),
a relationship known as the optical theorem.
For the applications discussed in this book specific interaction mod-
els between the propagating particles and the medium will be assumed
so that it is usually straightforward to calculate the dispersion relation
according to Eq. (6.5). One should keep in mind, however, that nrefr as
a function of ω has a number of general properties, independently of the
interaction model. For example, its real and imaginary part are con-
nected by the Kramers-Kronig relations (Sakurai 1967; Jackson 1975).
6.2.2 Particle Momentum and Velocity

The four-vector (ω, k) which governs the spatial and temporal behavior
of a plane wave can be time-like (ω 2 − k2 > 0) as for massive particles
in vacuum, it can be light-like (ω 2 − k2 = 0) as for photons in vacuum,
or it can be space-like (ω 2 − k2 < 0) as for visible light in water or air.
Because the quantized excitations of such field modes are interpreted
as particles, E = h̄ω is the particle’s energy. (I have temporarily re-
stored h̄ even though it is 1 in natural units.) Similarly one may be
tempted to interpret p = h̄k as the particle’s momentum. In vacuum a

particle’s velocity is p/E, a quantity which exceeds the speed of light
for space-like excitations. Occasionally one reads in the literature that
for this reason only those branches of a particle dispersion relation were
physical where |p| < E. Such statements are incorrect, however, and
the underlying concern about tachyonic propagation is unfounded.
The quantity p/E has no general physical relevance. Two signifi-
cant velocity definitions are the phase velocity and the group velocity of
a wave (Jackson 1975). The former is the speed with which the crest of
a plane wave propagates, i.e. it is given by the condition ωt − kz = 0 or
vphase = ω/k = n−1refr . For a massive particle in vacuum vphase > 1. How-
ever, the phase velocity can drop below the speed of light in a medium.
When this occurs for electromagnetic excitations in a plasma, electrons
can “surf” in the wave which thus transfers energy at a rate propor-
tional to the fine structure constant α (Landau 1946), an effect known
as Landau damping. As long as vphase > 1 the photon propagation is
damped only by Thomson scattering which is an effect of order α2 .
The group velocity vgroup = dω/dk is the speed with which a wave
packet or pulse propagates. In terms of the refractive index it is
−1
vgroup = nrefr (ω) + ω dnrefr /dω (6.7)
(Jackson 1975). For a massive particle in vacuum with ω 2 = (k 2 +m2 )1/2
it is vgroup = k/ω < 1, and also in a medium normally vgroup < 1. Near
a resonance it can happen that vgroup > 1, but there is still no reason
for alarm. The fast variation of nrefr (ω) as well as the presence of a
large imaginary part near a resonance imply that the issue of signal
propagation is much more complicated than indicated by the simple
approximations which enter the definition of the group velocity. For a
detailed discussion of electromagnetic signal propagation in dispersive
media see Jackson (1975).
Evidently a naive interpretation of h̄k as a particle momentum can
be quite misleading. Another example relates to the difficulty of sepa-
rating the momentum flow of a (light) beam in a medium into one part
carried by the wave and one carried by the medium. There was a long-
standing dispute in the literature with famous researchers on different
sides of an argument that was eventually resolved by Peierls (1976);
see also Gordon (1973). Experimentally, it was addressed by shining a
laser beam vertically through a water-air interface and measuring the
deformation of the surface due to the force which must occur because of
a photon’s change of momentum between the two media (Ashkin and
Dziedzic 1973).
200 Chapter 6
The problem of the physical momentum flow associated with a wave

will be of no concern to the issues addressed in this book. In microscopic
reactions the quantity which appears in the law of “energy-momentum
conservation” is the wave vector. For example, in the plasma process
γ → νν the momenta of the outgoing neutrinos must balance against
the wave vector of the decaying electromagnetic excitation. In this
book dispersion effects will be important only for pulse propagation
from distant sources, for particle oscillation effects, and for energy-
momentum conservation in microscopic reactions. In these cases the
naive interpretation of h̄k as a particle’s momentum is safe. For the
remainder of this book the wave number (or pseudomomentum) and
the momentum of a field excitation will not be distinguished.
6.2.3 Wave-Function Renormalization
In particle reactions the main impact of medium-induced modifications

of the dispersion relations is on the kinematics, notably if a threshold
condition is involved. One is thus tempted to proceed with the usual
Feynman rules and take account of the dispersion relations only in
the phase-space integration, notably in the law of energy-momentum
conservation. In most practical cases this approach causes no problems,
although an exception are interactions involving longitudinal plasmons
(Sect. 6.3). Therefore, one should be aware that the matrix element
also must be modified because of the subtle issue of what one means
with a “particle” in a medium.
After a spatial Fourier transform the equations of motion for the
Fourier components φk of a free field are those of a harmonic oscillator.
Interpreting the amplitude φk and its velocity φ̇k as conjugate variables,
the canonical quantization procedure leads to quantized energy levels
h̄ωk , where ωk is the classical frequency of the mode k according to its
dispersion relation. Conversely, a quantized excitation with energy h̄ωk
has a certain field strength which determines its coupling strength to a
source, e.g. the coupling strength of a photon to an electron.
In a medium, the energy associated with a frequency ωk is still h̄ωk .
However, because of the presence of interaction energy between φ and
the medium, the field strength associated with a quantized excitation
is modified. For example, photons with a given frequency couple with
a different strength to electrons in a medium than they do in vacuum. √
This modification can be lumped into a “renormalization factor” Z
of the coupling strength of external photon lines in a Feynman graph.
In order to determine this factor from the dispersion relation con-

sider a scalar field φ in the presence of a medium which induces a re-
fractive index. This means that the Klein-Gordon equation in Fourier
space, including a source term ρ, is of the form
[−K 2 + Π(K)]φ(K) = gρ(K), (6.8)
where K = (ω, k) is a four-vector in Fourier space and g is a coupling

constant. Π(K) is the “self-energy” which includes a possible vacuum
mass m2 and medium-induced contributions which are calculated from
the forward scattering amplitude. The homogeneous equation with
ρ = 0 has nonvanishing solutions only for K 2 = Π(K) which defines
the dispersion relation
ωk2 − k2 = Π(ωk , k). (6.9)
This equation determines implicitly the frequency ωk related to a wave

number k for a freely propagating mode.
A problem with Eq. (6.8) is the general dependence of Π on ω
and k which implies “dispersion,” i.e. in coordinate space it is not a
simple second-order differential equation. Otherwise the equation of
motion for a single field mode φk would be (∂t2 + k2 + Πk )φk = gρk .
Apart from the source term this is a simple harmonic oscillator with
frequency ωk2 = k2 + Πk . The canonical quantization procedure then
leads to quantized excitations with energy h̄ωk —the usual “particles.”
In a medium one follows this procedure in an approximate sense by
expanding Πk (ω) ≡ Π(ω, k) to lowest order around ωk ,
Πk (ω) = Πk (ωk ) + Π$k (ωk )(ω − ωk ), (6.10)
where Π$k (ω) ≡ ∂ω Πk (ω). To this order the Klein-Gordon equation is

& '
−ω 2 + ωk2 + Π$k (ωk )(ω − ωk ) φk (ω) = gρk (ω), (6.11)
where the dispersion relation Eq. (6.9) was used. To first order in ω−ωk
one may use 2ω = 2ωk = ω + ωk which allows one to write
Z −1 (ω 2 − ωk2 ) φk (ω) = gρk (ω), (6.12)
where
(
−1 2ωk − Π$k (ωk ) ∂Π(ω, k) ((
Z ≡ =1− ( . (6.13)
2ωk ∂ω 2 (ω2 −k2 =Π(ω,k)
202 Chapter 6
Because Z is a constant for a fixed k the approximate equation of

motion EE405 corresponds to a Hamiltonian
H = H0 + Hint = 12 Z −1 (φ̇2k + ωk2 φ2k ) + gφk ρk . (6.14)
The free-field term √ is of the standard harmonic-oscillator form if one

substitutes φk = Z φ)k , i.e. free particles are excitations of the field
φ)k which has a renormalized amplitude relative to φk .
In terms of the renormalized
√ field the interaction Hamiltonian is now
of the form Hint = Zg φ)k ρk which means that particles in√the medium
interact with an external source with a modified strength √Zg. There-
fore, in Feynman graphs one must include one factor of Z for each
external line of the φ field. Equivalently, the squared matrix element
involves a factor Z for each external φ particle.
For relativistic modes where |ωk2 − k2 | & ωk2 the modification is
inevitably small, |Z −1| & 1. For (longitudinal) plasmons, however, the
dispersion relation in a nonrelativistic plasma is approximately ω = ωP
with the plasma frequency ωP , i.e. they are far away from the light
cone, and then Z is a nonnegligible correction (Sect. 6.3). In the original
calculation of the plasma decay process γ → νν an incorrect Z was used
for the longitudinal excitations (Adams, Ruderman, and Woo 1963).
The correct factor was derived by Zaidi (1965).
6.3 Photon Dispersion

6.3.1 Maxwell’s Equations
For the astrophysical applications relevant to this book the photon re-
fractive index in a fully ionized plasma consisting of nuclei and electrons
will be needed. On the quantum level, this system is entirely described
by quantum electrodynamics (QED). It is sometimes referred to as a
QED plasma—in contrast with a quark-gluon plasma which is described
by quantum chromodynamics (QCD). The calculation of the refractive
index amounts to an evaluation of the forward scattering amplitude of
photons on electrons, a simple task except for the complications from
the statistical averaging over the electrons which are partially or fully
relativistic and exhibit any degree of degeneracy. Recently Braaten and
Segel (1993) have found an astonishing simplification of this daunting
problem (Sect. 6.3.4).
A more conceptual complication is the occurrence of a third photon
degree of freedom in a medium (Langmuir 1926), sometimes referred to
as Langmuir waves or plasmons.31 Still, one should not think of photons

as becoming literally massive like a massive vector boson which also
carries three polarization states. A photon mass is prohibited by gauge
invariance which remains intact. However, the medium singles out an
inertial frame and thus breaks Lorentz invariance, an effect which is
ultimately responsible for the possibility of a third polarization state.
It is useful, then, to begin with some general aspects of photon prop-
agation in a medium which are unrelated to specific assumptions about
the medium constituents. Notably, begin with the classical Maxwell
equations for the electric and magnetic fields
∇ · E = ρ, ∇ × B − Ė = J,
∇ · B = 0, ∇ × E + Ḃ = 0. (6.15)
An additional condition is that the electric charge density ρ and current

density J obey the continuity equation
∂ · J = ρ̇ − ∇ · J = 0, (6.16)
where J = (ρ, J) and ∂ = (∂t , ∇). Covariantly, Maxwell’s equations are
∂µ F µν = J ν , ,µνρσ ∂µ Fρσ = 0, (6.17)
where F µν is the antisymmetric field-strength tensor with the nonvan-

ishing components F 0i = −F i0 = −Ei , and F ij = −F ji = −,ijk Bk .
Applying ∂µ to the inhomogeneous equation and observing that F µν
is antisymmetric and ∂µ ∂ν symmetric under µ ↔ ν reveals that for
consistency J must obey the continuity equation.
An equivalent formulation arises from expressing the field strengths
in terms of a four-potential A = (Φ, A) by virtue of
F µν = ∂ µ Aν − ∂ ν Aµ , (6.18)
which amounts to E = −∇Φ − Ȧ and B = ∇ × A. This representation

is enabled by the homogeneous set of Maxwell equations which are then
automatically satisfied. The inhomogeneous set now takes the form
A − ∂ (∂ · A) = J, (6.19)
where = ∂ · ∂ = ∂µ ∂ µ = ∂t2 − ∇2 .
31
They are sometimes called “longitudinal plasmons” in contrast to “transverse
plasmons.” In this nomenclature the term “plasmon” refers to any excitation of the
electromagnetic field in a medium while “photon” refers to an excitation in vacuum.
204 Chapter 6
The field-strength tensor contains six independent degrees of free-

dom, the E and B fields. The redundancy imposed by the constraint of
the homogeneous equations was removed by introducing the vector po-
tential. There still remains one redundant degree of freedom related to
a constraint imposed by current conservation. The Maxwell equations
remain invariant under a “gauge transformation” A → A − ∂α where α
is an arbitrary scalar function. The modified A yields the same fields
E and B which are the physically measurable quantities.
The relationship to current conservation is easiest recognized if one
recalls that Maxwell’s equations can be derived from a Lagrangian
− 14 F 2 − J · A where F 2 = Fµν F µν . A gauge transformation introduces
an additional term J ·∂α which is identical to a total divergence ∂ ·(αJ)
if ∂ · J = 0 and thus leaves the Euler-Langrange equations unchanged.
Indeed, current conservation is a necessary and sufficient condition for
the gauge invariance of the theory (Itzykson and Zuber 1983).
A judicious choice of gauge can simplify the equations enormously.
Two important possibilities are the Lorentz gauge and the Coulomb,
transverse, or radiation gauge, based on the conditions
∂ · A = 0, Lorentz gauge,
∇ · A = 0, Coulomb gauge. (6.20)
Maxwell’s equations are then found to be (Jackson 1975)
Φ = ρ, A = J, Lorentz gauge,
−∇2 Φ = ρ, A = JT , Coulomb gauge, (6.21)
where JT is the transverse part of J characterized by ∇ · JT = 0.

In the absence of sources (ρ = 0 and J = 0) the potential Φ vanishes
in the Coulomb gauge while A obeys a wave equation. A Fourier trans-
formation leads to (−k2 + ω 2 )A = 0 whence the propagating modes
have the dispersion relation k2 = ω 2 corresponding to massless parti-
cles. Because of the transversality condition k · A = 0 there are only
two polarizations, the usual transverse electromagnetic waves. They
are characterized by an electric field E transverse to k and a magnetic
field of the same magnitude transverse to both.
6.3.2 Linear Response of the Medium

Maxwell’s equations allow one to calculate the electromagnetic fields
in the presence of prescribed external currents. However, the charged
particles which constitute the currents move themselves under the in-
fluence of electromagnetic fields. Therefore, the interaction between
fields and currents must be calculated self-consistently. If the fields are
sufficiently weak one may assume that the reaction of the currents to
the fields can be described as a linear response. (For a general review
of linear-response theory in electromagnetism see Kirzhnits 1987.)
In general this statement cannot be made locally in the sense that
the currents at space-time point (t, x) were only linear functions of
A(t, x). Within the restrictions imposed by causality the relation-
ship between fields and currents is nonlocal; for example, a solution
of Maxwell’s equations with prescribed currents requires integrations
over the sources in space and time. After a Fourier transformation,
however, the assumption of a linear response can be stated as
µ
Jind = −Πµν Aν . (6.22)
The polarization tensor Π(K) with K = (ω, k) is a function of the
medium properties.
Besides the induced current there may be an externally prescribed
one Jext which is unrelated to the response of the microscopic medium
constituents to the fields; the total current is J = Jind + Jext . Maxwell’s
equations (6.19) are then in Fourier space
µ
(−K 2 g µν + K µ K ν + Πµν )Aν = Jext . (6.23)
Invariance under a gauge transformation Aν → Aν + Kν α requires that
Πµν Kν = 0. Because the external and total currents are conserved
the induced current is conserved as well, leading to K · Jind = 0 or
Kµ Πµν = 0. Altogether
Kµ Πµν = Πµν Kν = 0 (6.24)
which is an important general property of the polarization tensor.
Considering the Maxwell equations in Coulomb gauge in the ab-
sence of external currents, the transversality of A still implies that it
provides only two wave polarization states, albeit with modified disper-
sion relations due to the presence of Π. With regard to the Φ equation
note that in an isotropic medium the induced charge density ρind must
be a spatial scalar and so can depend only on Φ and the combination
k · A = 0 which is the only available scalar linear in A. Therefore, the
homogeneous equation for Φ is
(k2 + Π00 )Φ = 0. (6.25)
00
Because Π is a function of ω and k this is a wave equation with the
dispersion relation k2 + Π00 (ω, k) = 0. The electric field associated
206 Chapter 6
with this third polarization degree of freedom is proportional to kΦ,

along the direction of propagation—hence the term longitudinal exci-
tation. There is no magnetic field associated with it. Physically, it
corresponds to a density wave of the electrons much like a sound wave.
Obviously this mode requires being carried by a medium, as opposed
to the transverse waves which propagate in vacuum as well.
The wave equation Eq. (6.23) corresponds to a Langrangian density
in Fourier space which involves a new term −V with V = 12 Aµ Πµν Aν
which plays the role of a medium-induced potential energy for the
field A. In vacuum Πµν can be constructed only from g µν and K µ K ν
which both violate the gauge condition ΠK = 0. Notably, this forbids a
photon mass term which would have to be of the form Πµν 2 µν
mass = m g .
In a medium an inertial frame is singled out, allowing one to construct
Π from the medium four-velocity U and to find a structure which obeys
the gauge constraint. Strictly speaking, however, the medium does not
induce an effective-mass term which would be of the form m2eff g µν and
which remains forbidden by gauge invariance.
6.3.3 Isotropic Polarization Tensor in the Lorentz Gauge

The Coulomb gauge is well suited to treat radiation in vacuum because
the propagating modes are neatly separated from the scalar potential,
and the gauge component is easily identified with the longitudinal part
of A. In a medium, however, the different appearance of Φ and A in
their respective wave equations is cumbersome. The Maxwell equations
in Lorentz gauge are symmetric between Φ and A which allows one to
treat all polarization states on the same footing.
In order to construct the most general Π(K) for an isotropic medium
it is useful to define four basis vectors for Minkowski space which are
adapted to the symmetry of the medium as well as to the Lorentz
condition (Weldon 1982a; Haft 1993). For that purpose one may use
the preferred directions in Minkowski space, namely K and the four-
velocity of the medium U which is (1, 0) in its inertial frame. Moreover,
the notation ω and k is used for the frequency and wave vector of K
in the medium frame; they are covariantly given by ω = U · K and
k 2 = k2 = (U · K)2 − K 2 .
In Lorentz gauge the physical A fields obey K · A = 0. Therefore,
one defines a basis vector for the gauge degree of freedom by
√
eg ≡ K/ K 2 . (6.26)
Next, one chooses a vector which is longitudinal relative to the spatial

part of K and which obeys K · eL = 0,

ωK − K 2 U (k 2 , ωk)
eL ≡ √ = √ , (6.27)
k K2 k K2
where the second expression refers to the medium rest frame. There
remain two directions orthogonal to eg and eL , or equivalently, to K
and U . If k is taken to point in the z-direction two possible choices are
the unit vectors ex and ey , respectively. However, in order to retain the
azimuthal symmetry around √ the k direction the “circular polarization
vectors” e± = (ex ± iey )/ 2 are needed. Then
e± ≡ (0, e± ) (6.28)
in the rest frame of the medium; a suitable covariant formulation is

also possible. The basis vectors obey e∗± = e∓ while eg,L are real for
K 2 > 0 (time-like) and e∗g,L = −eg,L for K 2 < 0 (space-like). They are
normalized according to e∗± · e± = e∓ · e± = −1 and e∗L · eL = ∓1 and
e∗g · eg = ±1, depending on K 2 being time- or space-like. Evidently, eg
and eL switch properties between a time- and space-like K.
This choice of basis vectors is only possible if K 2 ,= 0. If K is
light-like one must make some other choice, for example (1, 0) and
(0, k̂) in the rest frame of the medium. Because the goal is to describe
electromagnetic excitations in a medium, and because usually K 2 ,= 0
for such waves, this is no serious limitation. It will turn out that the
dispersion relation of (longitudinal) plasmons crosses the light-cone, i.e.
there is a wave number for which ω 2 − k 2 = 0. The degeneracy of eg
with eL at this single point will cause no trouble.
The most general polarization tensor compatible with the gauge
condition Eq. (6.24) must be constructed from e± and eL alone. Az-
imuthal symmetry about the k direction requires that they occur only
in the scalar combinations eµa e∗ν
a . Therefore, one defines the projection
operators on the basis vectors
Paµν ≡ −eµa e∗ν

a , (a = ±, L). (6.29)
The most general polarization tensor is then given as

*
Πµν = πa Paµν , (6.30)
a=±,L
where the πa are functions of the Lorentz scalars K 2 and U · K or

equivalently of ω and k in the medium frame. They represent the
medium response to circularly and longitudinally polarized A’s.
208 Chapter 6
The homogeneous Maxwell equations in Lorentz gauge in an iso-

tropic medium then have the most general form
+ * ,
−K 2 g µν + πa Paµν Aν = 0. (6.31)
a=±,L
The metric tensor is g = Pg + PL + P+ + P− where Pgµν = eµg e∗ν

g . Hence
2 2
one obtains decoupled wave equations [−ω + k + πa (ω, k)]Aa = 0 for
the physical degrees of freedom Aa = Pa A with a = ±, L. The corre-
sponding dispersion relation is
−ω 2 + k 2 + πa (ω, k) = 0. (6.32)
It yields the frequency ωk for modes with a given polarization and wave
number. The so-called effective mass is then m2eff = πa (ωk , k). This
expression is different for different polarizations and wave numbers,
and may even be negative.
Generally, an isotropic medium is characterized by three different
response functions because the left- and right-handed circular polariza-
tion states may experience different indices of refraction (Nieves and Pal
1989a,b). Such optically active media are not symmetric under a parity
transformation. For example, a sugar solution changes under a spatial
reflection because the sugar molecules have a definite handedness.
If the medium and all relevant interactions are even under parity the
circular polarization states have the same refractive index. Then one
needs to distinguish only between transverse and longitudinal modes;
one defines πT ≡ π+ = π− and PT = P+ + P− which projects on the
plane transverse to K and U in Minkowski space.
In macroscopic electrodynamics the medium effects are frequently
stated in the form of response functions to applied electric and mag-
netic fields instead of a response to A. The displacement induced by an
applied electric field is D = , E with , the dielectric permittivity. Simi-
larly, the magnetic field is H = µ−1 B for an applied magnetic induction
where µ is the magnetic permeability. For time-varying and/or inho-
mogeneous fields these relationships are understood in Fourier space
where the response functions depend on ω and k.
The magnetic field H and the transverse part of D, characterized
by k · DT = 0, do not have independent meaning (Kirzhnits 1987).
Therefore, among other possibilities one may choose H = B, DT =
,T ET , and DL = ,L EL . In this case ,L ≡ , is the longitudinal and
,T ≡ ,L + (1 − µ−1 ) k 2 /ω 2 the transverse dielectric permittivity.
The relationship to the transverse and longitudinal components of

the polarization tensor is (Weldon 1982a)
,L = 1 − πL /(ω 2 − k 2 ) and ,T = 1 − πT /ω 2 . (6.33)
This yields the well-known dispersion relations (Sitenko 1967)
,L (ω, k) = 0 and ω 2 ,T (ω, k) = k 2 (6.34)
for the longitudinal and transverse modes.
Calculations of the polarization tensor from the forward scattering
amplitudes on microscopic medium constituents are usually performed
in a cartesian basis and thus yield an expression for Πµν . The lon-
gitudinal and transverse components are projected out by virtue of
πL = e∗µ ν ∗µ ν
L Πµν eL and πT = e± Πµν e± , or explicitly in the medium frame
(Weldon 1982a)
πL = (1 − ω 2 /k 2 ) Π00 and πT = 12 (Tr Π − πL ), (6.35)
with Tr Π = gµν Πµν .
Recall that the dispersion relations are given by ω 2 −k 2 = πT,L (ω, k).
Thus the frequency ω(k) and the “effective mass” of a given mode are
generally complicated functions of k, notably in a medium involving
bound electrons where various resonances occur. It can be shown on
general grounds (Jackson 1975), however, that for frequencies far above
all resonances the transverse mode has a particle-like dispersion relation
ω 2 − k 2 = m2T where mT is the “transverse photon mass” which is a
constant independent of the wave number or frequency.
6.3.4 Lowest-Order QED Calculation of Π

On the level of quantum electrodynamics (QED) the potential V =
1
2
Aµ Πµν Aν which modifies the free Lagrangian is interpreted as the
self-energy of the photons in the medium. As a Feynman graph, it cor-
responds to an insertion of Πµν into a photon line of four-momentum K
and thus corresponds to forward scattering on the medium constituents
(Fig. 6.1), entirely analogous to the interpretation of the refractive in-
dex in terms of a forward scattering amplitude in Sect. 6.2.1. One then
concludes that Πµν (K) is the truncated matrix element for the forward
scattering of a photon with momentum K, i.e. it is the matrix ele-
ment of the medium constituents alone, uncontracted with the photon
polarization vectors ,µ and ,ν .
In general, the calculation of Π requires the methods of field the-
ory at finite temperature and density. To lowest order, however, this
210 Chapter 6
Fig. 6.1. Polarization tensor as photon self-energy insertion.
formalism is not required because the only contribution is from lowest-

order forward scattering on charged particles. Moreover, because the
scattering amplitude involves nonrelativistically the inverse mass of the
targets one may limit one’s attention to the electrons.
Then one takes the standard (truncated) Compton-scattering ma-
trix element (e.g. Bjorken and Drell 1964; Itzykson and Zuber 1983)
and takes an average over the Fermi-Dirac distributions of the elec-
trons. To lowest order in α = e2 /4π this yields (Altherr and Kraemmer
1992; Braaten and Segel 1993)
# - .
µν d3 p 1 1
Π (K) = 16πα +
2E(2π)3 e(E−µ)/T + 1 e(E+µ)/T + 1
(P · K)2 g µν + K 2 P µ P ν − P · K (K µ P ν + K ν P µ )
× ,
(P · K)2 − 14 (K 2 )2
(6.36)
where P = (E, p) and E = (p2 + m2e )1/2 , apart from refractive effects
for the electrons and positrons. The phase-space distributions represent
electrons and positrons at temperature T and chemical potential µ.
Over the years, the phase-space integration has been performed in
various limits (Silin 1960; Tsytovich 1961; Jancovici 1962; Klimov 1982;
Weldon 1982a; Altherr, Petitgirard, and del Rı́o Gaztelurrutia 1993).
The most comprehensive analytic result is that of Braaten and Segel
(1993) which contains all previous cases in the appropriate limits.
The main simplification occurs from neglecting the (K 2 )2 term in
the denominator of Eq. (6.36). For light-like K’s this is exactly cor-
rect, and in the nonrelativistic limit where me is much larger than all
other energy scales the approximation is also trivially justified. In the
relativistic limit it is only justified if one is interested in Π(K) near the
light cone (ω = k) in Fourier space. In the relativistic limit both trans-
verse and longitudinal excitations have dispersion relations which are
approximately (ω 2 − k 2 )1/2 ≈ e T or eEF in the nondegenerate and de-
generate limits, respectively. As detailed by Braaten and Segel (1993),
this deviation from masslessness is small enough to justify the approxi-
mation if one aims at the dispersion relations. Including 14 (K 2 )2 yields
an O(α2 ) correction—it can be ignored in an O(α) result.
In a higher-order calculation one has to include the proper e± disper-

sion relations which imply that electromagnetic excitations are never
damped by γ → e+ e− decay (Braaten 1991), in contrast with state-
ments found in the previous literature. Dropping the (K 2 )2 term in the
denominator of Eq. (6.36) prevents Π from developing an imaginary
part from this decay, even with the vacuum e± dispersion relations.
Therefore, the “approximate” integral actually provides a better repre-
sentation of the O(α) dispersion relations than the “exact” one.
With the approximation K 2 = 0 in the denominator of Eq. (6.36)
the angular integral is trivial.32 With Eq. (6.35) one finds
$ ! " %
4α ω 2 − k 2 # ∞ p2 ω ω + kv ω2 − k2
πL = dp fp log − 2 −1 ,
π k2 0 E kv ω − kv ω − k2v2
$ ! "%
4α ω 2 − k 2 # ∞ p2 ω2 ω ω + kv
πT = dp fp − log , (6.37)
π k2 0 E ω 2 − k 2 2kv ω − kv
where v = p/E is the e± velocity and fp represents the sum of their
phase-space distributions.
The remaining integration can be done analytically in the classical,
degenerate, and relativistic limits where one finds
& '
πL = ωP2 1 − G(v∗2 k 2 /ω 2 ) + v∗2 k 2 − k 2 ,
& '
πT = ωP2 1 + 12 G(v∗2 k 2 /ω 2 ) . (6.38)
Here, v∗ is a “typical” electron velocity defined by
v∗ ≡ ω1 /ωP . (6.39)
The plasma frequency ωP and the frequency ω1 are
2 4α # ∞
ωP ≡ dp fp p (v − 13 v 3 ),
π 0
2 4α # ∞
ω1 ≡ dp fp p ( 53 v 3 − v 5 ). (6.40)
π 0
The function G (Fig. 6.2) is defined by
$ ! √ "%
3 2x 1 − x 1+ x
G(x) = 1− − √ log √
x 3 2 x 1− x
∞
* xn
=6 . (6.41)
n=1 (2n + 1)(2n + 3)
Note that G(0) = 0, G(1) = 1, and G$ (1) = ∞.

32
In the degenerate limit, Jancovici (1962) has calculated analytically the full
integral without the K 2 = 0 approximation.
212 Chapter 6
Fig. 6.2. Function G(x) according to Eq. (6.41).
The most astonishing observation of Braaten and Segel (1993) is

that Eq. (6.38) is a good approximation for all conditions, not only for
the limiting cases for which it was derived. As the approximation is
much better than 1%, which is the approximate accuracy of an O(α)
result, these representations can be taken to be exact to this order.
6.3.5 Dispersion Relations

In order to determine the photon dispersion relation for specific condi-
tions one must determine ωP and v∗ corresponding to the temperature
T and chemical potential µ of the electrons. In Fig. 6.3 contours for
v∗ and γ ≡ ωP /T are shown in the T -ρ-plane of a plasma. Analytic
limiting cases are (Braaten and Segel 1993)

 1/2
 (5T /me ) Classical,
v∗ = vF Degenerate, (6.42)


1 Relativistic,
 - .

 4πα ne 5 T


 1 − Classical,



me 2 me

 4πα n
e 4α 2
ωP2 = = p F vF Degenerate, (6.43)
 E
 3π

 F




 4α + 2 1 2 2 ,
 µ + π T
3
Relativistic,
3π
where vF = pF /EF is the velocity at the Fermi surface, “classical” refers
to the nondegenerate and nonrelativistic limit, and “relativistic” is for
any degree of degeneracy.
Fig. 6.3. Contours for v∗ and γ = ωP /T as defined in Eqs. (6.39) and (6.40)
where Ye is the number of electrons per baryon.
Next, with Eq. (6.38) one must solve the transcendental equations
πT,L (ω, k) = ω 2 − k 2 which are explicitly
& '
ω2 − k2 = ωP2 1 + 12 G(v∗2 k 2 /ω 2 ) Transverse,
& '
ω 2 − v∗2 k 2 = ωP2 1 − G(v∗2 k 2 /ω 2 ) Longitudinal. (6.44)
In the classical limit this is to lowest order in T /me

! "
2 2 k2 T
ω =k + ωP2 1+ 2 Transverse,
ω me
! "
2 k2 T
ω = ωP2 1+3 2 Longitudinal. (6.45)
ω me
For small temperatures the longitudinal modes oscillate with an al-

most fixed frequency, independently of momentum, while the transverse
modes behave almost like massive particles (Fig. 6.4).
The general result Eq. (6.38) and the behavior of the function G(x)
reveal that for transverse excitations ω 2 − k 2 can vary only between ωP2
and 32 ωP2 . Also, ω 2 − k 2 > 0 so that K 2 is always time-like. For k . ωP
the transverse dispersion relation approaches that of a massive particle
with a fixed mass mT , the “transverse photon mass”. With Eq. (6.38)
and because k/ω → 1 for k . ωP one finds m2T = ωP2 [1 + 12 G(v∗2 )], or
214 Chapter 6
Fig. 6.4. Electromagnetic dispersion relations in the classical limit according

to Eq. (6.45) for v∗ = (5T /me )1/2 = 0.2. The shaded area indicates the
“width” of ω(k) in the longitudinal case due to Landau damping.
with ω = k directly from Eq. (6.37)

4α # ∞ p2
m2T = dp fp . (6.46)
π 0 E
Limiting cases are



1 Classical,

 $ - .%

m2T  3 1 − vF2 1 + vF
= 2
1 − log Degenerate, (6.47)
2
ωP 
 2v
 F
2vF 1 − vF


 3
2
Relativistic.
In Fig. 6.5 (ω 2 − k 2 ) is shown for several values of v∗ as a function

of k. It is quite apparent how the transverse mass is asymptotically
approached.
The dispersion relation for longitudinal modes is more interesting in
several regards. First, according to Eq. (6.44) the oscillation frequency
is only a function of v∗ k and so the natural scale for k is ωP /v∗ . In
Fig. 6.6 I show ω 2 − v∗2 k 2 as a function of v∗ k.
Fig. 6.5. Dispersion relation for transverse modes according to Eq. (6.44).
Fig. 6.6. Dispersion relation for longitudinal modes according to Eq. (6.44).
Second, for v∗ < 1 there is always a wave number k1 where ω(k)

“crosses the light cone” (ω/k = 1),
# 3 - . 4
4α ∞ p2 1 1+v
k12 = dp fp log −1
π 0 E v 1−v
3 - . 4
2 3 1 1 + v∗
= ωP 2 log −1 . (6.48)
v∗ 2v∗ 1 − v∗
The second identity (Braaten and Segel 1993) applies at the same level
of approximation as πT,L in Eq. (6.38). Some analytic limiting cases
216 Chapter 6
are


 1 + 3T /me Classical,

 3 - . 4
k12 
3 1 1 + vF
= log −1 Degenerate, (6.49)
ωP2 

 vF2 2vF 1 − vF


∞ Relativistic.
Then, for k > k1 the four-momentum is space-like, ω 2 − k 2 < 0.
As discussed in Sect. 6.2.2 there is nothing wrong with a space-like
four-momentum of an excitation. In media with electron resonances
such as water or air even (transverse) photons exhibit this behavior
which allows kinematically for their Cherenkov emission e → eγ or
absorption γe → e. In a plasma, transverse excitations are always
time-like and thus cannot be Cherenkov absorbed. Their lowest-order
damping mechanism is Thomson scattering γe → eγ which is not in-
cluded because it is an O(α2 ) effect. Longitudinal excitations with
k > k1 , in contrast, can and will be Cherenkov absorbed by the ambi-
ent electrons, leading to an O(α) damping rate. It corresponds to an
imaginary part of the dispersion relation (an imaginary part of πL ).
In the expression Eq. (6.36) this damping effect corresponds to a
vanishing denominator, essentially to P · K = 0, which occurs when the
intermediate electron in Compton scattering “goes on-shell.” Evidently,
P · K = Eω − p · k can never vanish for k < ω while for k > ω there are
always some electrons, even in a nonrelativistic plasma, which satisfy
this condition. When the phase velocity ω/k becomes of the order of
a typical thermal velocity the number of electrons which match the
Cherenkov condition becomes large, and then the damping of plasmons
becomes strong. Because v∗ measures a typical electron velocity this
occurs for k > ∼ ω/v∗ (Fig. 6.4). Therefore, while nothing dramatic
happens where the dispersion relation crosses the light cone, it fizzles
out near the “electron cone.” For k > ∼ ω/v∗ there are no organized
oscillations of the electrons—longitudinal modes no longer exist.
This damping mechanism of plasma waves was first discussed by
Landau (1946) and is named after him. A calculation in terms of Che-
renkov absorption was performed by Tsytovich (1961). In the classical
limit the Landau damping rate (the imaginary part of the frequency) is
5 ! "3 - .3/2 - .3 ω2
k2 √
ΓL π kD 5 ωP − 52 v∗2 kP2
e− 2k2 = π
D
= e , (6.50)
ωP 8 k 2 v∗ k
where kD = 4πα ne /T is the Debye screening scale. (Note that ωP2 /kD
2
=
T /me = v∗2 /5.) For a given wave number a plasmon must be viewed
as a resonance with a finite width ΓL . The approximate uncertainty

±ΓL (k) of the energy ω(k) is shown in Fig. 6.4 as a shaded area.
The damping rate Eq. (6.50) does not show a threshold effect at
k = ω because it was calculated nonrelativistically so that the high-
energy tail of the electron distribution contains particles with velocities
exceeding the speed of light. Tsytovich (1961) calculated a relativis-
tic result with the correct threshold behavior. However, because the
damping rate is exceedingly small for k & kD , the correction is very
small if ωP & kD , equivalent to T & me . For the degenerate case,
explicit expressions for the imaginary parts of πT,L (ω, k) were derived
by Altherr, Petitgirard, and del Rı́o Gaztelurrutia (1993).
The general expressions Eq. (6.38) for the real parts of πL and
πT were derived without the need to assume that K was time-like.
Therefore, they should also apply “below the light cone,” even though
Braaten and Segel (1993) confined their discussion to the region k < ω.
As expected, these expressions break down for k > ω/v∗ where Landau
damping becomes strong.
6.3.6 Renormalization Constants ZT,L

Armed with the dispersion relation one may determine the vertex renor-
malization constants ZT,L relevant for the coupling of external photons
or plasmons to an electron in the medium (Sect. 6.2.3),
(
−1 ∂πT,L (ω, k) ((
ZT,L =1− (
( 2 2
. (6.51)
∂ω 2 ω −k =πT,L (ω,k)
With the same approximations as before, Braaten and Segel (1993)

found an analytic representation accurate to O(α),
2ω 2 (ω 2 − v∗2 k 2 )
ZT = ,
ω 2 [3ωP2 − 2 (ω 2 − k 2 )] + (ω 2 + k 2 )(ω 2 − v∗2 k 2 )
2 (ω 2 − v∗2 k 2 ) ω2
ZL = . (6.52)
3ωP2 − (ω 2 − v∗2 k 2 ) ω 2 − k 2
In each case ω and k are “on shell,” i.e. they are related by the disper-
sion relation relevant for the T and L case, respectively.
Inspection of Eq. (6.52) reveals that ZT is always very close to unity,
as expected for excitations with only a small deviation from a massive-
particle dispersion relation. The contours in Fig. 6.7 confirm that ZT
never deviates from unity by more than a few percent.
218 Chapter 6
Fig. 6.7. Contours for the vertex renormalization factor ZT for transverse
electromagnetic excitations in a medium according to Eq. (6.52).
Fig. 6.8. Modified vertex renormalization factor Z)L for longitudinal electro-
magnetic excitations in a medium according to Eq. (6.53).
The longitudinal case is more complicated. ZL is a product of two

factors Z)L and ω 2 /(ω 2 − k 2 ) where the former,
2 (ω 2 − v∗2 k 2 )
Z)L ≡ , (6.53)
3ωP2 − (ω 2 − v∗2 k 2 )
is a function of the variable v∗ k alone because for plasmons ω is a
function of v∗ k alone. The function Z)L (v∗ k) is shown in Fig. 6.8; for
v∗ k . ωP it quickly drops to zero. However, in a relativistic plasma
with v∗ = 1 the complete factor is ZL = 2ω 2 /[3ωP2 − (ω 2 − k 2 )] and thus

rises quickly with k because ω ≈ k.
As discussed in the previous section, the dispersion relation crosses
the light cone at k = k1 , a point at which ZL diverges and changes sign.
The sign change is compensated by the change of the polarization vector
eL (Eq. 6.27) at the light cone where it becomes imaginary. Because
µ
in the squared matrix element a factor e∗L eνL appears, and because
this expression changes sign at the light cone, the expression ZL e∗µ ν
L eL
remains positive.
As for the divergence, it is harmless in reactions of the sort γL → νν
(plasmon decay), e → eγL (Cherenkov emission), γL e → e (Cherenkov
absorption), and γT γL → a (plasmon coalescence into axions) which are
of interest in this book. These reactions involving three particles are
constrained by their phase space to either time-like excitations (plas-
mon decay), or to space-like ones (Cherenkov and coalescence process).
Therefore, the threshold behavior of the phase space moderates the
divergence in these cases.
6.4 Screening Effects

6.4.1 Debye Screening
Scattering processes in the Coulomb field of charged particles such as
Rutherford scattering, bremsstrahlung, or the Primakoff effect typically
lead to cross sections which diverge in the forward direction because of
the long-range nature of the electrostatic interaction. In a plasma this
divergence is moderated by screening effects which thus are crucial for
a calculation of the cross sections or energy-loss rates.
Screening effects are revealed by turning to the static limit of Max-
well’s equations in a medium,
& '
k 2 + πL (0, k) Φ(k) = ρ(k),
& '
k 2 + πT (0, k) A(k) = J(k), (6.54)
where the current must be transverse in both Coulomb and Lorentz

gauge as ∂t ρ = 0 in the static limit. Notably, the equation for Φ in
vacuum is the Fourier transform of Poisson’s equation and thus gives
rise to a 1/r Coulomb potential if the source is point-like, ρ(r) = eδ(r).
In a QED plasma πL,T (ω, k) are given by the integrals Eq. (6.37).
In the static limit (ω = 0) one finds πT (0, k) = 0 because all terms
in the integrand involve factors of ω. Therefore, stationary currents
220 Chapter 6
(∂t J = 0) are not screened. The magnetic field associated with a sta-
tionary current is the same at a distance whether or not the plasma
is present.
Not so for the electric field associated with a charge. In the static
limit one finds
# ∞
4α
πL (0, k) = dp fp p (v + v −1 ). (6.55)
π 0
Because this expression does not depend on k it can be identified with

the square of a fixed wave number kS , leading to Poisson’s equation in
the form
+ ,
k 2 + kS2 Φ(k) = ρ(k). (6.56)
Because for a point-like source this gives a Yukawa potential
Φ(r) ∝ r−1 → r−1 e−kS r (6.57)
electric charges are screened for distances exceeding about kS−1 .

Evaluating Eq. (6.55) explicitly in the classical limit reproduces the
well-known Debye screening scale (Debye and Hückel 1923; for a text-
book discussion see Landau and Lifshitz 1958)
4πα ne me 2
kS2 = kD
2
= = ω . (6.58)
T T P
2
At this point one recognizes that kD is independent of the electron
mass, in contrast with the plasma frequency ωP2 . Therefore, it is no
longer justified to ignore the ions or nuclei; they contribute little to
dispersion because of their reduced Thomson scattering amplitude, but
they contribute equally to screening. Therefore, one finds kS2 = kD
2
+ ki2
with
4πα *
ki2 = nj Zj2 , (6.59)
T j
where the sum is over all species j with charge Zj e.

Comparing Eq. (6.55) with the corresponding expression for the
plasma frequency Eq. (6.40) reveals that in the relativistic limit (v = 1)
2
kD = 3ωP2 . It is clear that apart from a numerical factor they must
be the same because a relativistic plasma has only one natural scale,
namely a typical electron energy.
In the limit of degenerate electrons the integral is also easily solved

and leads to the familiar Thomas-Fermi wave number33 (Jancovici 1962)
4α 3ω 2
kS2 = kTF
2
= EF pF = 2P . (6.60)
π vF
2 2
However, kD always exceeds kTF so that in a medium of degenerate
electrons and nondegenerate ions the main screening effect is from the
latter. Recall that the Fermi momentum is related to the electron
density by ne = p3F /3π 2 and the Fermi energy is EF = (p2F + m2e )1/2 .
To compare the Thomas-Fermi with the Debye scale take the non-
relativistic limit (kTF /kD )2 = 32 T /(EF − me ). This is much less than 1
or the medium would not be degenerate whence kTF & kD . There-
fore, if the electrons are degenerate and the ions nondegenerate, a test
charge is mostly screened by the polarization of the ion “fluid” because
the electrons form a “stiff” background. Unfortunately, one often finds
calculations in the literature which include screening by the electrons
(screening scale kTF ) but ignore the ions. The resulting error need not
be large because the screening scale typically appears logarithmically
in the final answer (see below).
Screening effects in Coulomb processes are often found to be imple-
mented by a modified Coulomb propagator
1 1
→ , (6.61)
|q|4 (q2 + kS2 )2
where q is the momentum transfer carried by the intermediate photon.

This substitution arises if one considers Coulomb scattering from a
Yukawa-like charge distribution. It corresponds to the substitution
Eq. (6.57), i.e. to a single charge with an exponential screening cloud.
This picture is appropriate if the Coulomb scattering process itself is so
slow that the charged particles move around and rearrange themselves
so much that the probe, indeed, sees an average screening cloud.
In the opposite limit, a given probe sees a certain configuration, a
different probe a different one, etc., and one has to average over all of
these possibilities. In this case one needs to square the matrix element
first, and then take an average over different medium configurations.
For Eq. (6.61) one averages first, obtains an average scattering ampli-
tude or matrix element, and squares afterward.
33
For a textbook derivation from a Thomas-Fermi model see Shapiro and Teukol-
sky (1983). Note that they work in the nonrelativistic limit: their EF = p2F /2me .
222 Chapter 6
It depends on the physical circumstances which procedure is a better

approximation. If one considers bremsstrahlung processes with degen-
erate electrons scattering off nondegenerate nuclei, the crossing time of
an electron of a region the size kS−1 is short compared to the crossing
time of nuclei. Hence, the latter can be viewed as static, the probe sees
one configuration at a time, and one certainly should use the “square
first” procedure instead of Eq. (6.61) to account for screening. This is
achieved by the following consideration of correlation effects.
6.4.2 Correlations and Static Structure Factor

The screening of electric fields in a plasma is closely related to correla-
tions of the positions and motions of the charged particles. If a negative
test charge is known to be in a certain position, the probability of find-
ing an electron in the immediate neighborhood is less than average,
while the probability of finding a nucleus is larger than average. It is
this polarization of the surrounding plasma which screens a charge.
Take one particle of a given species to be the origin of a coordinate
system, and take their average number density to be n. The electro-
static repulsion of the test charge causes a deviation of the surrounding
charges from the average density by an amount
S(r) = δ 3 (r) + n h(r), (6.62)
where h(r) measures the particle correlations. They vanish in an ideal

Boltzmann gas: h(r) = 0. The Fourier transform
#
S(q) = d3 r S(r) e−iq·r (6.63)
is the static structure factor of the electron distribution. In the absence

of correlations (h = 0) one has trivially S(q) = 1.
In order to make contact with Debye screening consider the Yukawa
potential of Eq. (6.57) which represents a charge density
kS2 e−kS r
ρ(r) = δ 3 (r) − . (6.64)
4π r
The volume integral of ρ(r) vanishes, giving zero total charge, i.e. com-
plete screening at infinity. If one imagines that only one species of
charged particles is mobile on a uniform background of the opposite
charge, then Eq. (6.64) implies correlations between the mobile species
of n h(r) = −(kS2 /4πr) e−kS r . As expected, Debye screening corresponds
to spatial anticorrelations of like-charged particles. Fourier transform-

ing Eq. (6.64) yields the important result
q2
S(q) = (6.65)
q2 + kS2
for the structure factor.
The assumption of one mobile species of particles on a uniform
background corresponds to the model of a “one-component plasma.”
It is approximately realized in the interior of hot white dwarfs or the
cores of red-giant stars where the degenerate electrons form a “stiff”
background of negative charge in which the nondegenerate ions move.
In a nondegenerate situation, however, there are at least two mo-
bile species, ions of charge Ze and electrons. For this two-component
plasma Salpeter (1960) derived the structure functions
q2 + ZkD2
See (q) = 2
,
q2 + (1 + Z)kD
q2 + k D
2
Sii (q) = 2
,
q2 + (1 + Z)kD
2
kD
Sei (q) = 2
. (6.66)
q2 + (1 + Z)kD
The Fourier transform of the screening cloud around an electron is
q2
S(q) = See (q) − ZSei (q) = , (6.67)
q2 + kS2
with kS2 = kD2
+ ki2 = (1 + Z) kD
2
. (Note that for only one species of ions
2 2
ki = ZkD .) Hence one reproduces a screened charge distribution which
causes a Yukawa potential. However, the small-q behavior of See or Sii
is very different: See (0) = Z/(1 + Z) in a two-component plasma while
See (0) = 0 for only one component.
6.4.3 Strongly Coupled Plasma

For low temperatures, the screening will not be of Yukawa type and
the structure factor will deviate from the simple Debye formula. A
plasma can be considered cold if the average Coulomb interaction en-
ergy between ions is much larger than typical thermal energies. To
quantify this measure, one introduces the “ion-sphere radius” ai by
virtue of n−1
i = 4πa3i /3 where ni is the number density of the mobile
224 Chapter 6
particle species. Hence, a measure for the Coulomb interaction energy

is Z 2 α/ai , assuming the ions have charge Ze. One usually introduces
the parameter
Z 2α (ki ai )2
Γ≡ = (6.68)
ai T 3
as a measure for how strongly the plasma is coupled, where ki2 =
4πZ 2 α/T . For Γ & 1 it is weakly coupled and approaches an ideal
Boltzmann gas.
The Debye structure factor of a one-component plasma can be writ-
ten as
|ai q|2
SD (q) = . (6.69)
|ai q|2 + 3Γ
This result applies even for large Γ if |ai q| & 1. For Γ . 1, the plasma
is strongly coupled, and for Γ > ∼ 178 the ions will arrange themselves
in a body centered cubic lattice (Slattery, Doolen, and DeWitt 1980,
1982).
In Fig. 6.9 I show S and SD as functions of ai q = |ai q| for Γ = 2, 10
and 100 where S was numerically determined (Hansen 1973; Galam and
Hansen 1976). The emerging periodicity for a strongly coupled plasma
is quite apparent. It is also clear that for Γ <
∼ 1 the Debye formula gives
a fair representation of the structure factor while for a strongly coupled
plasma it is completely misleading. The interior of white dwarfs is in
the regime of large Γ, and old white dwarfs are believed to crystallize.
(See Appendix D for an overview over the conditions relevant for stellar
plasmas.)
6.4.4 Screened Coulomb Scattering

Armed with these insights one may turn to the issue of Coulomb scatter-
ing processes in a plasma. In the limit of nonrelativistic and essentially
static sources for the electric fields the relevant quantity entering the
matrix element is the Fourier component ρ(q) of the charge distribu-
tion ρ(r) where q is the momentum transferred by the Coulomb field
to the sources. The squared matrix element thus involves the quantity
ρ(q)ρ∗ (q) which is ρ(q)ρ(−q) because ρ(r) is real. Taking a statistical
average over all possible configurations of the charge distribution leads
to a rate proportional to
S(q) = 0ρ(q)ρ(−q)1. (6.70)
This is the static structure factor introduced earlier as a measure of the
correlation between the charged particles of the medium.
Fig. 6.9. Static structure factor for a one-component plasma according to

the numerical calculations of Hansen (1973) and Galam and Hansen (1976).
The dashed lines correspond to the Debye structure factor Eq. (6.69).
226 Chapter 6
Without correlations, the squared matrix element involves |q|−4

from the Coulomb propagator. In order to account for screening ef-
fects one should substitute
|q|−4 → |q|−4 S(q) (6.71)
which implies
1 1
→ 2 2 (6.72)
|q| 4 q (q + kS2 )
in the weak-screening limit (Debye screening).

The difference in a scattering cross section implied by Eq. (6.72)
relative to (6.61) is easily illustrated. Observe that a cross section
involving a Coulomb divergence is typically of the form
# +1 (1 − x) f (x)
dx , (6.73)
−1 (1 − x)2
where x is the cosine of the scattering angle of the probe. Here, f (x) is
a slowly varying function which embodies the details of the scattering
or bremsstrahlung process. If this function is taken to be a constant,
the two screening prescriptions amount to the two integrals
# ! "
+1 1 2 + κ2
dx = log ,
−1 (1 − x + κ2 ) κ2
# ! "
+1 (1 − x) 2 + κ2 2
dx = log − , (6.74)
−1 (1 − x + κ2 )2 κ2 2 + κ2
where κ2 ≡ kS2 /2p2 is the screening scale expressed in units of the initial-
state momentum of the probe. Usually, it far exceeds the screening scale
whence κ2 & 1. Then Eq. (6.72) yields a cross section proportional to
log(4p2 /kS2 ) while Eq. (6.61) gives [log(4p2 /kS2 ) − 1].
Thus, if one is only interested in a rough estimate, either screening
prescription and any reasonable screening scale yield about the same
result. For an accurate calculation, however, one needs to identify
the dominant source of screening (for example, the nondegenerate ions
in a degenerate plasma and not the electrons), and the appropriate
moderation of the Coulomb propagator, usually Eq. (6.72).
6.5 Plasmon Decay in Neutrinos34

6.5.1 Millicharged Neutrinos
Transverse and longitudinal electromagnetic excitations in a plasma
are both kinematically able to decay into neutrino pairs (Fig. 6.10) of
sufficiently small mass, namely 2mν < K 2 where K is the plasmon35
four-momentum. In the following, the neutrinos are always taken to
be massless relative to the plasma frequency and so K 2 > 0 is required
which restricts longitudinal excitations to k < k1 , the wave number
where their dispersion relation crosses the light cone.
Fig. 6.10. Plasmon decay in neutrinos.
In addition, a ν-γ-interaction is required which does not exist in

the standard model. Still, plasmon decays occur because the medium
itself mediates an effective coupling as will become clear below. As an
easy start, however, consider the hypothesis that neutrinos carry small
electric charges (“millicharges”). Interestingly, this possibility is not
excluded by the structure of the standard model and has received some
recent attention in the literature (Sect. 7.3.2).
With a neutrino millicharge eν the interaction with the electromag-
netic vector potential A is the standard expression
Lint = −ieν ψ ν γ α ψν Aα . (6.75)
The spin-summed squared matrix element is of the form
*
|M|2 = Mαβ P α P β (6.76)
spins
where explicitly
Mαβ = 4e2ν Z (gαβ + 2,∗α ,β ). (6.77)
Here, Z is the renormalization constant (Sect. 6.2.3 and 6.3.6), P and
P are the ν and ν four-momenta, and , is the plasmon polarization
vector for which one uses the basis vectors of Eqs. (6.27) and (6.28).
34
I closely follow Haft (1993).
35
In this section the term “plasmon” refers to both transverse and longitudinal
electromagnetic excitations in a medium.
228 Chapter 6
The decay width of a plasmon with four-momentum K = (ω, k) in

the medium frame, and with a definite polarization is
#
d3 p d3 p 4 4 1 *
Γ= 3 3
(2π) δ (K − P − P ) |M|2 .
2Ep (2π) 2Ep (2π) 2ω spins
(6.78)
Because of Eq. (6.76) one may use Lenard’s (1953) formula

#
d3 p d3 p α β 4 π + 2 αβ ,
P P δ (K − P − P ) = K g + 2K α K β .
2Ep 2Ep 24
(6.79)
With αν ≡ e2ν /4π this leads to
Γ = αν Z (ω 2 − k 2 )/3ω, (6.80)
where the normalization ,∗α ,α = −1 for transverse and time-like lon-

gitudinal plasmons was used as well as , · K = 0. Γ applies to both
transverse and longitudinal plasmons with the appropriate ZT,L . For a
chosen three-momentum k = |k| the quantities Z, ω, and K 2 = ω 2 − k 2
are all functions of k by virtue of the dispersion relation K 2 = πT,L (K).
In the classical limit transverse plasmons propagate like massive
particles with K 2 = ωP2 and ZT = 1. Then ΓT = 13 αν ωP3 (ωP /ω) where
the last factor is recognized as a Lorentz time-dilation factor. For a
general dispersion relation which is not Lorentz covariant it makes little
sense, of course, to express the decay rate in the plasmon frame. An
example is the classical limit for the longitudinal mode for which to
zeroth order in T /me the frequency is ω = ωP , ZL = ω 2 /K 2 , and then
ΓL = 13 αν ωP with the restriction k < ωP .
6.5.2 Neutrino Dipole Moments

Another direct coupling between neutrinos and photons arises if the for-
mer have electric or magnetic dipole or transition moments (Sects. 7.2.2
and 7.3.2)
*+ ,
Lint = 1
2
µab ψ a σµν ψb + ,ab ψ a σµν γ5 ψb F µν . (6.81)
a,b
Here, F is the electromagnetic field tensor, σµν = γµ γν − γν γµ , and ψa

with a = ν1,2,3 or νe,µ,τ are the neutrino fields.
The squared matrix element is of the form Eq. (6.76); for a magnetic
dipole coupling µ of a single flavor one finds
Mαβ = 4µ2 Z (2Kα Kβ − 2K 2 ,∗α ,β − K 2 gαβ ). (6.82)
This leads to a decay rate
Γ = (µ2 /24π) Z (ω 2 − k 2 )2 /ω, (6.83)
applicable to either plasmon polarization with the appropriate Z and

dispersion relation.
In the presence of electric and magnetic transition moments one
obtains the same result with
*+ ,
µ2 → |µab |2 + |,ab |2 . (6.84)
a,b
There is no interference term between the electric and magnetic cou-

plings. The presence of transition moments allows for plasmon decays
of the sort γ → νa ν b with different flavors a ,= b, doubling the final
states which may be νa ν b or νb ν a .
6.5.3 Standard-Model Couplings

In a medium there is an effective coupling between electromagnetic
fields and neutrinos mediated by the ambient electrons. For the purpose
of a plasmon-decay calculation it can be visualized with the Feynman
graph Fig. 6.11 which corresponds to the “photoneutrino process” or
Compton production of neutrino pairs (Sect. 3.2.4). However, the final-
state electron can have the same four-momentum as the initial state
which amounts to forward scattering of the electrons. In this case
the energy-momentum transfer to the electron vanishes, allowing for a
coherent superposition of these amplitudes from all electrons.
Fig. 6.11. Compton production of neutrino pairs (nonzero momentum trans-

fer to electrons) or plasmon decay (electron forward scattering).
230 Chapter 6
With the dimensionless coupling constants CV and CA given in Ap-

pendix B the neutral-current ν-e-interaction is
GF
Lint = − √ ψ e γα (CV − CA γ5 )ψe ψ ν γ α (1 − γ5 )ψν . (6.85)
2
The vector-current has the same structure that pertains to the elec-
tron interaction with photons, Lint = −ie ψ e γα ψe Aα . Therefore, af-
ter performing a thermal average over the electron forward scattering
amplitudes the plasmon decay is represented by the Feynman graph
Fig. 6.12 which is identical with Fig. 6.1 with one photon line replaced
by a neutrino pair. As far as the electrons are concerned, photon for-
ward scattering γ → γ is the same as the conversion γ → νν.
Fig. 6.12. Photon-neutrino coupling by the photon polarization tensor; the

second photon line in Fig. 6.1 was replaced by a neutrino pair.
Put another way, in a plasma the propagation of an electromagnetic

excitation is accompanied by an organized oscillation of the electrons.
This is particularly obvious for longitudinal modes which are the col-
lective oscillation of the electron gas, but it also applies to transverse
modes. The collective motion is the medium response Jind = ΠA to
an electromagnetic excitation, a response which is characterized by the
polarization tensor. The coherent electron oscillations serve as sources
for the neutrino current whence they emit neutrino pairs.
The electron collective motion is an oscillation of their location or
density while their spins remain unaffected apart from relativistic cor-
rections. Because the axial-vector current represents the electron spin
density, and because to lowest order no collective spin motion is ex-
pected as a response to an electromagnetic excitation, the axial-vector
neutrino coupling to electrons will contribute very little to plasmon de-
cay. This remains true in a relativistic plasma. For example, Koyama,
Itoh, and Munakata (1986) found numerically that the axial vector con-
tributes less than 0.01% to neutrino emission by the plasma process for
all conditions of astrophysical interest. A detailed study of the axial
response function can be found in Braaten and Segel (1993). For the
present discussion I will not worry any further about the axial-vector
contribution.
The matrix element for the interaction between neutrinos and pho-
tons can then be read from the effective vertex
C V GF
i √ Aα Παβ ψ ν γ α (1 − γ5 )ψν . (6.86)
e 2
The electric charge in the denominator removes one such factor con-
tained in Π which was calculated for photon forward scattering. In
the matrix element, Π is to be taken at the four-momentum K of the
photon. Besides plasmon decay γ → νν, this interaction also allows for
processes such as Cherenkov absorption γν → ν or emission ν → γν.
For the decay of a plasmon with a polarization vector , and four-
momentum K the squared matrix element has the form Eq. (6.76) with
G2F 2
Mαβ = 8 π (gαβ + 2,∗α ,β ). (6.87)
2e2 T,L
Because the plasmon is a propagating mode it obeys its dispersion
relation, i.e. πT,L = ω 2 − k 2 . The decay rate is then
CV2 G2F (ω 2 − k 2 )3
Γ= Z T,L . (6.88)
48π 2 α ω
This result was first derived by Adams, Ruderman, and Woo (1963),
the correct Z for the longitudinal case was first derived by Zaidi (1965).
This equation is understood “on shell” where ω depends on k through
the dispersion relation ω 2 − k 2 = πT,L (ω, k).
6.5.4 Summary of Decay Rates

In order to express the decay rates in a compact form, recall that on
shell the scale for K 2 is set by the plasma frequency ωP . Therefore, it
is useful to define
πT,L (ωk , k)
π̂T,L (k) ≡ , (6.89)
ωP2
where ωk is the frequency related to k as a solution of the dispersion

equation ω 2 − k 2 = π(ω, k). Recall that 1 ≤ π̂T < 32 while 0 ≤ π̂L ≤ 1
for a time-like K 2 (k < k1 ). At k = k1 the L dispersion relation crosses
the light cone.
232 Chapter 6
The decay rates of Eqs. (6.80), (6.83) and (6.88) of a plasmon with
three-momentum k are then expressed as


 ωP2 π̂k

 αν Millicharge,

 4π



 ! "2

 µ2 ωP2 π̂k
4π Zk
Γk = × Dipole Moment, (6.90)
3 ωk  

2 4π

 ! "3




 CV2 G2F ωP2 π̂k

 Standard Model,
α 4π
where for Zk and π̂k the T or L value appropriate for the chosen polar-
ization must be used.
6.5.5 Energy-Loss Rates

It is now an easy task to calculate stellar energy-loss rates for the
plasma process. An integration over the Bose-Einstein distributions of
the transverse and longitudinal plasmons yields for the energy-loss rate
per unit volume
2 #∞ ΓT ω
QT = 2 dk k 2 ω/T ,
2π 0 e −1
1 # k1 ΓL ω
QL = 2 dk k 2 ω/T . (6.91)
2π 0 e −1
In QT the factor 2 counts two polarization states. In QL the integration
can be extended only to the wave number k1 where the L dispersion
relation crosses the light cone—for k > k1 decays are kinematically
forbidden. In either case ΓT,L and ω are functions of k, the latter given
by the dispersion relation.
For the specific neutrino interaction models discussed in the previ-
ous section one obtains with the decay rates of Eq. (6.90)


 ωP2

 αν Q1 Millicharge,

 4π



 ! "2

 µ2 ωP2
8ζ3 3
Q= T × Q2 Dipole Moment, (6.92)
3π 


2 4π

 ! "3




 CV2 G2F ωP2

 Q3 Standard Model,
α 4π
where ζ3 ≈ 1.202 refers to the Riemann Zeta function. The dimension-
less emission rates Qn for the three cases are each a sum of a transverse
and longitudinal term, Qn = QL,n + QT,n where
1 # k1 n
2 ZL π̂L 1 # k1 2 )
2 ω ZL π̂L
n−1
QL,n = dk k = dk k ,
4ζ3 T 3 0 eω/T − 1 4ζ3 T 3 0 ωP2 eω/T − 1
1 #∞ n
2 ZT π̂T
QT,n = dk k . (6.93)
2ζ3 T 3 0 eω/T − 1
The second equation for QL,n relies on the definition Eq. (6.53), i.e.
ZL = Z)L ω 2 /(ω 2 − k 2 ), and ω 2 − k 2 = πL was used.
The normalization factors were chosen such that QT,n = 1 if the
plasmons are treated as effectively massless particles for the phase-space
integration. Then ZT = π̂T = 1 which is a reasonable approximation
in a nondegenerate, nonrelativistic plasma. In that limit to lowest
order k1 = ωP , Z)L = 1, and πL = ωP2 − k 2 . Therefore, in this limit
QL,n & QT,n . In fact, the longitudinal emission rate is of comparable
importance to the transverse one only in a narrow range of parameters
of astrophysical interest (Haft, Raffelt, and Weiss 1994).
These simple approximations, however, are not adequate for most of
the conditions where the plasma process is important. In Appendix C
the numerical neutrino emission rates are discussed; a comparison be-
tween Fig. C.1 and Fig. 6.3 reveals that the plasma process is important
for 0.3 < <
∼ ωP /T ∼ 30, i.e. transverse plasmons can be anything from
relativistic to entirely nonrelativistic. For a practical stellar evolution
calculation one may use the analytic approximation formula for the
plasma process discussed in Appendix C, based on the representation
of the dispersion relations of Sect. 6.3.
The main issue at stake in this book, however, is nonstandard neu-
trino emission from the direct electromagnetic couplings discussed in
Sect. 6.5. Instead of constructing new numerical emission rate formulae
one uses the existing ones for the standard-model (SM) couplings and
scales them to the novel cases. Numerically, one finds
! "4
Qcharge αν α (4π)2 Q1 10 keV Q1
= 2 2 4 = 0.664 e214 ,
QSM CV GF ωP Q3 ωP Q3
! "2
Qdipole µ2 α 2π Q2 10 keV Q2
= 2 2 2 = 0.318 µ212 , (6.94)
QSM CV GF ωP Q3 ωP Q3
where e14 = eν /10−14 e and µ12 = µ/10−12 µB with µB = e/2me .
Contours for Q1 /Q3 and Q2 /Q3 are shown in Fig. 6.13 according to
Haft, Raffelt, and Weiss (1994). Replacing these ratios by unity in a
practical stellar evolution calculation introduces only a small error.
234 Chapter 6
Fig. 6.13. Contours of Q1 /Q3 and Q2 /Q3 defined by Eq. (6.93) in the plane
defined by the plasma frequency ωP and a “typical” electron velocity v∗
discussed in Sect. 6.3. See Fig. 6.3 for contours of v∗ and ωP in the T -ρ-
plane. (Adapted from Haft, Raffelt, and Weiss 1994.)
6.5.6 Astrophysical Bounds on Neutrino Electromagnetic

Properties
The plasma decay process is the most important neutrino emission
process for a large range of temperatures and densities. Moreover,
it has an observable impact on the cooling of hot white dwarfs, and
on the core mass at helium ignition in low-mass red giants. For a
neutrino millicharge eν > −14 > −12
∼ 10 e and a dipole moment µ ∼ 10 µB the
nonstandard plasmon decay rates in Eq. (6.94) begin to compete with
the standard one if the plasma frequency is around 10 keV. Therefore,
neutrino millicharges or dipole moments of this magnitude will have
observable effects on these stars and thus can be excluded.
In Sect. 2.2.3 it has been discussed that neutrino emission cools

hot white dwarfs so fast that there is a clear depression of the white-
dwarf luminosity function at the hot and bright end relative to the
simple Mestel law which takes only surface photon emission into ac-
count. This depression would be enhanced by additional cooling caused
by dipole moments or millicharges, allowing one to derive a limit of
about µ12 < 10, although even for µ12 ≈ 3 a nonnegligible effect
is apparent.
A more restrictive and probably more reliable limit can be derived
from the properties of globular-cluster stars (Raffelt 1990b). To this
end one may use the simple criteria derived in Sect. 2.5 which state
that a novel energy-loss rate is constrained by 0,x 1 < −1 −1
∼ 10 erg g s for
the average core-conditions of a horizontal branch star, and for those of
a low-mass red giant before the helium flash. In both cases T ≈ 108 K.
For this temperature, the plasma loss rates are shown in Fig. 6.14 as
a function of density. The anomalous rates were obtained from the
standard one according to Eq. (6.94), taking Q1 /Q3 = Q2 /Q3 = 1 and
using the zero-temperature plasma frequency given in Eq. (D.12) as a
function of density.
The first criterion of Sect. 2.5, based on the helium-burning lifetime
of HB stars, requires calculating the energy loss rate at an average
density which is below 104 g cm−3 . Therefore, the medium is so dilute
Fig. 6.14. Neutrino energy-loss rate in helium at T = 108 K. Solid line:

Total standard rate. Long dashes: Standard plasma rate. Short dashes:
Plasma rate induced by a dipole moment µν = 2×10−12 µB . Dots: Plasma
rate induced by a neutrino “millicharge” eν = 10−14 e.
236 Chapter 6
that one may employ the simple analytic form Eq. (6.92) of the emission
rate with Qn = 1. The core of HB stars consists at first of helium, later
also of carbon and oxygen, for all of which Ye = 0.5. Then,


 5.0 e214 Millicharge,

,x = 1 erg g−1 s−1 × T83 ×  0.098 µ212 ρ4 Dipole Moment, (6.95)


0.0127 ρ24 Standard Model,
where T8 = T /108 K and ρ4 = ρ/104 g cm−3 . The core averages for a

typical HB star are 0T83 1 = 0.44, 0T83 ρ4 1 = 0.47 and 0T83 ρ24 1 = 0.57. The
requirement 0,x 1 < 10 erg g−1 s−1 then gives the limits
eν < −14
∼ 2×10 e and µν < −12
∼ 14×10 µB . (6.96)
Of course, for such large dipole moments the core would grow far be-
yond its standard value before helium ignites, causing an additional
acceleration of the HB lifetime. In fact, this indirect impact on the HB
lifetime would be the dominant effect as shown, for example, by the
numerical calculations of Raffelt, Dearborn, and Silk (1989).
From Fig. 6.14 it is clear that the dipole-induced emission rate is
larger for the conditions of the second criterion, based on the helium-
ignition argument where 0ρ1 ≈ 2×105 g cm−3 . According to Eq. (D.12)
the relevant plasma frequency is ωP = 8.6 keV so that Qcharge /QSM ≈
1.2 e214 and Qdipole /QSM ≈ 0.4 µ212 in Eq. (6.94). The average total emis-
sion rate is then given by the standard rate times Fν = 1 + Qj /QSM
where j stands for “charge” or “dipole.” In order to prevent the core
mass at helium ignition from exceeding its standard value by more
than 5% one must require Fν < 3. Then one finds eν < ∼ 1.3×10 e
−14
and µν < −12

∼ 2×10 µB . For the dipole case, a detailed numerical imple-
mentation yielded µν < −12
∼ 3×10 µB (Sect. 2.5.2), nearly identical with
this simple analytic estimate. The limit on the charge could also be
slightly degraded and so I adopt
eν < −14
∼ 2×10 e and µν < −12
∼ 3×10 µB (6.97)
as the final globular-cluster limits.

6.6 Neutrino Form Factors in Media

In a medium, neutrinos can interact with photons using electrons or
other charged particles as go-betweens. The basic idea is to consider the
Compton process of Fig. 6.11 with the initial- and final-state electrons
in the same state, i.e. forward scattering for the electrons. Then one
may sum over all electrons of the medium. This coherent superposition
of the amplitudes from all electrons was used in the previous section to
calculate the standard-model plasmon decay rate. There, only on-shell
(propagating) photons were considered. In general one may consider
other cases, for example electromagnetic scattering by the exchange of
a space-like photon, or the behavior of neutrinos in an external electric
or magnetic field.
The neutrino electromagnetic form factors in vacuum will be studied
in Sect. 7.3.2. They can be classified as a charge radius, an anapole
moment, and an electric and a magnetic dipole moment. They are
induced by intermediate (virtual) charged particles such as charged
leptons or W bosons. In the present case the form factors are induced
by the real particles of the ambient heat bath. The effective Lagrangian
Eq. (7.19) is fundamentally Lorentz covariant, a fact which reduces
the number of possible form factors to four. While in a medium the
couplings may also be written in a nominally Lorentz covariant form,
the medium singles out an inertial frame, leading to more complicated
structures. This is analogous to dispersion which is simple in vacuum
(a mass term is the only possibility) while in a medium the dispersion
relations can be excruciatingly complicated.
Limiting the couplings to the ones mediated by electrons and pro-
tons, the induced photon coupling to the neutrino is proportional to Aµ
because these fermions couple by the usual eψγµ ψAµ interaction. The
neutrinos
√ couple by their standard effective neutral-current interaction
(GF / 2) ψ ν γα (1 − γ5 )ψν ψ e γ α (CV − CA γ5 )ψe with the weak coupling
constants CV and CA given in Appendix B. Therefore, after summing
over all intermediate electron states the effective neutrino-photon in-
teraction may be written in the form
√
Leff = − 2 GF ψ ν γα 12 (1 − γ5 )ψν Λαβ Aβ , (6.98)
where Λαβ is a matrix which depends on the medium properties and

on the energy-momentum transfer, i.e. the energy momentum K of the
photon line. It consists of a symmetric piece Λαβ
V which is proportional
to CV , and an antisymmetric piece ΛαβA which is proportional to CA .
238 Chapter 6
Explicit expressions in terms of the electron phase-space integrals

were derived, e.g. by D’Olivo, Nieves, and Pal (1989) and by Altherr
and Salati (1994),
#
d3 p & '
Λαβ
V = 4e CV f e − (p) + f e + (p)
2E(2π)3
(P · K)2 g αβ + K 2 P α P β − P · K (K α P β + K α P β )
× ,
(P · K)2 − 14 (K 2 )2
#
d3 p & '
Λαβ
A = 2ie CA , αβµν
f e − (p) − fe+ (p)
2E(2π)3
K 2 P µ Kν
× , (6.99)
(P · K)2 − 14 (K 2 )2
where fe± (p) is the electron and positron phase-space distribution with
P = (E, p) the electron or positron four-momentum.
Instead of a neutrino pair, another photon can be thought of as
being coupled to the electron line in Fig. 6.11 or 6.12, a process which
represents photon forward scattering. Therefore, apart from overall
coupling constants Λαβ V is identical with the electronic contribution to
the photon polarization tensor Παβ studied earlier in this chapter
Λαβ αβ
V = (CV /e) Π . (6.100)
In an isotropic plasma, Παβ is characterized by the two medium charac-

teristics πT (ω, k) and πL (ω, k) which are functions of the photon four-
momentum K = (ω, k) with k = |k|. For the antisymmetric piece, in
an isotropic medium the phase-space integration averages the spatial
part of P α to zero so that
Λαβ
A = 2ieCA ,
αβµ0
Kµ a(ω, k) (6.101)
in terms of a single medium characteristic a.

The single most important application of the effective neutrino elec-
tromagnetic coupling is the photon decay process γ → νν that was
studied in the previous section. It turns out that Λαβ A contributes very
αβ
little to the decay process so that ΛV , or rather the polarization tensor
Παβ determines all aspects of the plasma process.
Another possible process is the Cherenkov emission of photons by
neutrinos. Of course, because in a plasma transverse photons acquire an
“effective mass,” i.e. their dispersion relation is time-like (ω 2 − k2 > 0),
this process is kinematically forbidden. However, longitudinal elec-

tromagnetic excitations (plasmons), which exist only in the medium,
propagate such that for some momenta ω 2 − k2 < 0 (space-like four-
momentum), allowing for Cherenkov emission. One finds statements
in the literature that the neutrino energy transfer to the medium by
this process exceeded the transfer by (incoherent) ν-e scattering (e.g.
Oraevskiı̆ and Semikoz 1984; Oraevskiı̆, Semikoz, and Smorodinskiı̆
1986; Semikoz 1987a). This is in conflict with the discussion of Kirzh-
nits, Losyakov, and Chechin (1990) who found on general grounds that
the energy loss of a neutrino propagating in a stable medium was al-
ways bounded from above by the collisional energy loss, apart from a
factor of order unity. Granting this, the Cherenkov process does not
seem to be of great practical importance.
If neutrinos have masses and mix, decays of the form ν2 → ν1 γ are
possible in vacuum, and can be kinematically possible in a medium if
the photon “effective mass” does not exceed the neutrino mass differ-
ence m2 − m1 . If kinematically allowed, this decay receives a contri-
bution from the medium-induced coupling which may far exceed the
vacuum decay rate. Explicit calculations were performed by a num-
ber of authors36 who unfortunately ignored the kinematic constraint
imposed by the photon dispersion relation. This is not a reasonable
approximation in view of the relatively small neutrino masses that re-
main of practical interest. Further, in order to judge the importance
of the medium-induced decay it is not relevant to compare with the
vacuum decay rate, but rather one should compare with the collisional
transition rate ν2 e → eν1 (mediated by photon exchange) which is the
process with which the coherent reaction directly competes.
The photon decay as well as the Cherenkov process and the medium-
induced neutrino decay all have in common that the neutrino couples
to an electromagnetic field which is a freely propagating wave, obeying
the dispersion relation in the medium which is ω 2 − k 2 = πT,L (ω, k)
for transverse and longitudinal excitations, respectively. However, one
may also consider the effect of a static external electric or magnetic
field.37 To this end, one must take the static limit ω → 0 of the
vertex functions Λαβ V,A (ω, k). For an external static electric field the
only nonvanishing component of the vector potential Aµ is A0 . Then
36
D’Olivo, Nieves, and Pal (1990); Kuo and Pantaleone (1990); Giunti, Kim, and
Lam (1991).
37
This issue has been investigated in many works, e.g. Oraevskiı̆ and Semikoz
(1985, 1987), Semikoz (1987a,b), Nieves and Pal (1989c, 1994), Semikoz and
Smorodinskiı̆ (1988, 1989), and D’Olivo, Nieves, and Pal (1989).
240 Chapter 6
Λαβ 00
A does not contribute so that only the ΛV component remains of
00
interest. In the static limit Π is simply given by πL (0, k) which in
turn can be identified with the square of the screening scale kS2 in a
medium (Sect. 6.4.1). This implies that the neutrino interacts with the
external electric field as if it had a charge
√
eν = −(CV /e) 2 GF kS2 . (6.102)
In a classical (nondegenerate, nonrelativistic) hydrogen plasma the
screening scale is given by the Debye scale through kS2 = 2kD
2
= 2e2 ne /T
with the
√ electron density ne so that the induced neutrino charge is eν =
CV e 2 2 GF ne /T . This induced charge is explained by the medium
polarization caused by the weak force exerted by the presence of the
neutrino.
While this induced charge is conceptually very interesting it does
not seem to have any immediate practical consequences. Notably, it
is not the relevant quantity for the interaction with a static magnetic
field, i.e. one may not infer that neutrinos move on curved paths in
magnetic fields. The presence of a “neutrino charge” was derived for
the interaction with a static electric field! The relevant form factor
for the interaction with a static magnetic field is identified by noting
that now only the spatial components of Aµ are nonzero. In the static
limit only the component Π00 of the polarization tensor survives when
contracted with Aµ . Then there is no contribution from Λαβ S for the
neutrino interaction with a magnetic field. The contribution from Λαβ A
can be interpreted as a “normal” or “Dirac magnetic moment” induced
by the medium (Semikoz 1987a; D’Olivo, Nieves, and Pal 1989)
√ # ∞ & '
µν = −eCA 2GF 4π dp fe− (p) − fe+ (p) . (6.103)
0
In the limit of a classical plasma this is µν = (eν /2me )(2CA /CV ) where
eν is the induced electric charge of Eq. (6.102).
This induced Dirac magnetic moment is to be compared with the
electron’s Dirac moment e/2me , not with an anomalous moment. The
former arises from the eψ e γµ ψe Aµ coupling, the latter is described by
1
µ ψ σ ψ F µν . This means that the induced dipole moment does
2 e e µν e
not lead to neutrino spin precession—it only couples to left-handed
states. It entails an energy difference between neutrinos moving in
opposite directions along a magnetic field. The transverse part of the
field has no impact on the neutrino—there is no spin precession, and no
curvature of the trajectory. (These conclusions pertain to the limit of
weak magnetic fields. For strong fields the modification of the electron
wavefunctions, i.e. Landau levels rather than plane waves, would have
to be used for a self-consistent treatment of the photon polarization
tensor and thus, for the neutrino coupling to a magnetic field. For a
first discussion see Oraevskiı̆ and Semikoz 1991.)
In summary, on the basis of the existing literature it appears that
the medium-induced electromagnetic form factors of neutrinos are of
practical importance only for the photon decay process that was dis-
cussed in the previous section.
6.7 Neutrino Refraction

6.7.1 Neutrino Refractive Index
When neutrinos propagate in a medium they will experience a shift of
their energy, similar to photon refraction, due to their coherent inter-
action with the medium constituents (Wolfenstein 1978). The neutrino
refractive index can be calculated in the same way as that for any
other particle which propagates in a medium, namely on the basis of
the forward scattering amplitudes as discussed in Sect. 6.2.1. As one
needs only forward scattering, and as the relevant medium constituents
are protons, neutrons, electrons, and possibly other neutrinos, only the
Feynman graphs of Fig. 6.15 need to be considered.38
In most situations of practical interest the energies of the neutrinos
and of the medium particles are much smaller than the W and Z mass
(80.2 and 91.2 GeV) so that the energy and momentum transferred by
the gauge bosons is always much less than their mass.39 This justifies
to expand their propagators (energy-momentum transfer Q) as
gµν Q2 gµν − Qµ Qν
Dµν (Q) = + + ... (6.104)
m2Z,W m4Z,W
and keep only the first term. (The second term is needed if the con-
tribution of the first one cancels as in a CP symmetric medium—see
38
In the formalism of finite temperature and density (FTD) field theory the am-
plitudes may be written in a more compact form so that the relevant Feynman
graphs reduce to a tadpole and a bubble graph (Nötzold and Raffelt 1988; Nieves
1989; Pal and Pham 1989). Apart from a more compact notation, however, the
FTD formalism leads to the same expressions as the “pedestrian” approach chosen
here.
39
See however Learned and Pakvasa (1995) as well as Domokos and Kovesi-
Domokos (1995) for a discussion of the oscillations of very high-energy cosmic neu-
trinos for which this approximation is not adequate.
242 Chapter 6
Fig. 6.15. Amplitudes contributing to forward scattering: (a) Neutral-

current scattering for any ν or ν on any f or f as target. (b) νe -ν e scattering.
(c) νe -e charged-current scattering. (d) νe -e+ charged-current scattering.
(e) Effective four-fermion vertex in the low-energy limit.
Sect. 6.7.2.) This amounts to reducing the weak interaction to the

low-energy Fermi effective Hamiltonian represented by graph (e) in
Fig. 6.15. For the neutral-current processes (a) and (b) it is explicitly
GF
Hint = √ ψ f γµ (CV − CA γ5 )ψf ψ ν! γ µ (1 − γ5 )ψν! , (6.105)
2
where ψν! is a neutrino field (- = e, µ, τ ) while ψf represents fermi-
ons of the medium (f = e, p, n, or even neutrinos ν&# ). Here, GF =
1.166×10−5 GeV−2 is the Fermi constant. The relevant values of the
vector and axial-vector weak charges CV and CA are given in Ap-
pendix B. In the low-energy limit the charged-current reactions (c)
and (d) can also be represented as an effective neutral-current interac-
tion of the same form with CV = CA = 1.
It is now straightforward to work out the forward scattering ampli-
tudes. The axial-vector piece represents the spin of f and so it averages
to zero if the medium is unpolarized. Then one finds for the refractive
index of a neutrino (upper sign) or antineutrino (lower sign) with en-
ergy ω
nf − nf
nrefr − 1 = ∓ CV$ GF √ , (6.106)
ω 2
where nf and nf are the number densities of fermions f and antifer-
mions f , respectively. The effective weak coupling constants CV$ are
identical with the CV given in Appendix B except for neutrinos as
medium particles40 which are left-handed and thus polarized. There-

fore, (1 − γ5 )ψν = 2ψν and CV$ = 2CV .
Because we are dealing with forward scattering where recoil effects
do not occur, the contributions from free or bound nucleons are the
same. Therefore, Eq. (6.106) allows one to determine the refractive
index of any normal medium. Because electric neutrality implies an
excess density of electrons over positrons which balances against the
protons, their neutral-current contributions cancel. (A possible excep-
tion is a π − condensate that may exist in neutron stars.) An excess
of νe over ν e appears to occur only in a young supernova core where
neutrinos have a large chemical potential for the first few seconds after
collapse.
All told, the dispersion relation for unmixed neutrinos, valid even in
the nonrelativistic limit (Chang and Zia 1988), can be written in terms
of a potential energy as
(ω − V )2 = k 2 + m2 , (6.107)
where V = −(nrefr − 1) ω. For all practical cases

√  (− 1 Yn + Ye + 2Yνe ) for νe ,
2
V = ± 2GF nB × (6.108)
 (− 1 Y + Y ) for νµ,τ ,
2 n νe
(upper sign ν, lower sign ν). Here, nB is the baryon density and
nf − nf
Yf ≡ (6.109)
nB
are the particle number fractions commonly used in astrophysics. Nu-
merically,
√ ρ
2 GF nB = 0.762×10−13 eV (6.110)
g cm−3
with the mass density ρ.
A remark concerning the absolute sign of V is in order. The relative
signs between the different CV ’s can be worked out easily from the
weak interaction structure of the standard model. Also, the relative
sign of the effective neutral-current amplitudes which follow from Z ◦
and W exchange follows directly, for example, from the FTD approach
(Nötzold and Raffelt 1988). Thus to fix the overall sign it is enough
40
In a supernova core or in the early universe it is not possible to distinguish
between a “test neutrino” and a “medium neutrino.” There, one has to study the
nonlinear evolution of the entire ensemble self-consistently (Sect. 9.3.2).
244 Chapter 6
to understand the absolute sign of neutrino-neutrino scattering. It is

a case where identically “charged” fermions scatter by the exchange of
a vector boson. The structure of this process is analogous to Coulomb
scattering of like-charged particles which experience a repulsive force.
(The exchange of a spin-0 or spin-2 boson leads to an attractive force.)
Thus a neutrino of given momentum in a region of space filled with
other neutrinos will have a positive potential energy V in addition to
its kinetic energy. The correct absolute sign was first pointed out by
Langacker, Leveille, and Sheiman (1983).
The deviation from relativistic propagation described by Eq. (6.107)
can be expressed as an effective refractive index which includes the
vacuum neutrino mass,
$- .2 %1/2
V m2 V m2
nrefr = 1− − 2 → 1− − 2 (6.111)
ω ω ω 2ω
(relativistic limit). Therefore, the effect of a medium can be expressed
as an effective mass
m2eff = m2 + 2ωV. (6.112)
A numerical comparison with the vacuum mass is achieved by

! "1/2 - .1/2
+ √ ,1/2
−4 ρ ω
2ω 2GF nB = 3.91×10 eV .
g cm−3 MeV
(6.113)
Of course, the term “effective mass” is a misnomer because meff depends

on the energy ω, and m2eff can be negative depending on the vacuum
mass, the medium composition, the flavor of the neutrino, and whether
it is ν or ν.
The phase velocity vphase = ω/k = n−1 refr can be larger or less than
the speed of light, depending on those parameters. However, the group
velocity vgroup = dω/dk remains at its vacuum value (1 + m2 /k 2 )−1/2
for a given momentum k because ω is shifted by a constant amount V ,
independently of k.
Because normal media contain about √ equal numbers of protons and
1 1
neutrons Yn ≈ Ye ≈ 2 and so Vνe ≈ 4 2GF nB ≈ −Vνµ,τ . Therefore, νe
and νµ,τ are shifted by almost exactly opposite amounts. An exception
is the proton-rich material of normal stars which initially contain about
75% hydrogen. Another exception is the neutron-rich matter of neutron
stars where even Vνe is negative.
The absolute shift of the neutrino “masses” is rather negligible be-

cause we are dealing with highly relativistic particles. Even in this limit,
however, the difference between the dispersion relation of different fla-
vors is important for oscillation effects. Hence the most noteworthy
medium effect is its flavor birefringence: νe and νµ,τ acquire different
effective masses because of the charged-current contribution from √ νe -e
scattering. The difference of their potentials is Vνe − Vνµ,τ = 2GF nL
with nL = YL nB the lepton-number density where the number frac-
tion of leptons is YL = Ye + Yνe . Of course, neutrinos as a background
medium contribute only in a young supernova core.
6.7.2 Higher-Order Effects

In the early universe one has nearly equal densities of particles and an-
tiparticles with an asymmetry of about 10−9 , leading to a near cancel-
lation of the refractive terms Eq. (6.106). One may think that the next
most important contribution is from ν-γ scattering, a process closely
related to the ν → ν $ γγ decay briefly discussed in Sect. 7.2.2. If one
approximates the weak interactions by an effective four-fermion cou-
pling the relevant amplitude is given by the graph Fig. 6.16 which on
dimensional grounds should be of order αGF . However, electromag-
netic gauge invariance together with the left-handedness of the weak
interaction implies that it vanishes identically (Gell-Mann 1961). For
massive neutrinos the amplitude is proportional to αGF mν , but even in
this case it vanishes in the forward direction (Langacker and Liu 1992).
Fig. 6.16. Neutrino-photon scattering with an effective four-fermion weak

interaction and a charged lepton % in the loop. This amplitude vanishes
entirely for massless neutrinos, and for massive ones it still vanishes in the
forward direction.
For the lowest-order ν-γ contribution to the refractive index one

must then use the full gauge-boson propagator and include all one-loop
amplitudes required by the standard model. Such calculations were
performed by Levine (1966) and by Cung and Yoshimura (1975) who
2
found that the scattering√ amplitude was proportional to αGF s/mW
(center of mass energy s). Recently this problem was revisited by
246 Chapter 6
Dicus and Repko (1993) who worked out explicitly the matrix elements
and cross sections. From their results one can extract the forward
scattering amplitude which leads to a refractive index for ν& (- = e, µ, τ )
in a photon bath,
$ ! "%
α GF 4 m2W
nrefr − 1 = 1 + ln 0Eγ 1nγ , (6.114)
4π m2W 3 m2&
where 0. . .1 means an average. Numerically, the term in square brackets

is 32.9 for - = e and thus not small, but with 4π in the denominator the
whole expression is still of order αGF /m2W , i.e. of order41 G2F . (See also
Nieves, Pal, and Unger 1983; Nieves 1987; Langacker and Liu 1992.)
Because the ν-γ term is so small a larger refractive index arises
if one includes the second term in the expansion Eq. (6.104) of the
gauge-boson propagators. Of course, in graphs (a) and (c) of Fig. 6.15
forward scattering implies Q = 0 so that only the first term contributes,
except when f = ν where the exchange graph has Q ,= 0. This case
and graphs (b) and (d) yield a second-order contribution (Nötzold and
Raffelt 1988). For a neutrino of flavor - which is either e, µ, or τ it is
√
8 2 GF + ,
nrefr − 1 = 0E ν! 1n ν! + 0E ν! 1n ν!
3m2
√ Z
8 2 GF + ,
+ 0E & − 1n&− + 0E&+ 1n&+ . (6.115)
3m2W
In this case the contributions from background fermions and antifermi-
ons add with the same sign, and the global sign remains the same for
ν and ν as test particles. In practice, only an electron-positron back-
ground is of relevance in the early universe so that ντ , for example,
only feels a second-order contribution from other ντ ’s and ν τ ’s. These
results are of order G2F /α and thus they are the dominant contribution
in a CP-symmetric plasma.
One-loop corrections to the amplitudes of Fig. 6.15 yield other
higher-order terms which are of order G2F like Eq. (6.114). They are
still interesting because the loops involve charged leptons with a mass
depending on their flavor. Therefore, the universality of the effective
neutral-current interaction is broken on this level, leading to different
refractive indices for different ν& . Between νe and νµ or ντ the medium
is already birefringent to lowest order from νe -e charged-current inter-
actions. Between νµ and ντ the one-loop correction dominates. Assum-
√ √
41
Note that m−2 2 2
Z = cos ΘW sin ΘW 2 GF /πα and m−2 2
W = sin ΘW 2 GF /πα.
ing electric neutrality it is (Botella, Lim, and Marciano 1987; see also
Semikoz 1992 and Horvat 1993)
$ ! " %
3G2F m2τ m2W Yn
Vντ − Vνµ = n B ln −1+ , (6.116)
2π 2 m2τ 3
with a sign change for Vν τ − Vν µ . The shift of the “effective mass” m2eff
is numerically
+ ,2 ρ ω 6.61 + Yn /3
2ω(Vντ − Vνµ ) = 2.06×10−6 eV −3
,
g cm MeV 7
(6.117)
much smaller than the corresponding difference between νe and νµ,τ .
6.7.3 The Sun a Neutrino Lens?

The most important consequence of neutrino refraction in media is its
impact on neutrino oscillations because different flavors experience a
different index of refraction. In optics, the most notable consequence
of refraction is the possibility to deflect light and thus to use lenses and
other optical instruments. In principle, the same is also possible for
neutrinos. The Sun, for example, could act as a gigantic neutrino lens.
One may easily calculate the deflection caused by a given body. If
s is a unit vector along the direction of a propagating wave, and if s
is a coordinate along the beam, the deflection is given by (Sommerfeld
1958)
|ds/ds| = n−1
refr |s × ∇nrefr |, (6.118)
where nrefr is the refractive index. One may equally write
|dα/ds| = n−1
refr |∇⊥ nrefr |, (6.119)
where α is the angle relative to the local tangential vector, i.e. dα/ds
is the local curvature of the beam, and ∇⊥ is the transverse gradient.
The total angle of deflection is
# +∞
|∆α| = ds |∇⊥ nrefr | (6.120)
−∞
if the curvature is small which is the case for |nrefr − 1| & 1.

If the beam hits a spherically symmetric body (radius R) at an
impact parameter b < R its angle against the radial direction at a
radius r from the center is sin β = b/r so that ∇⊥ nrefr = (b/r) ∂r nrefr .
248 Chapter 6
Moreover, if s is measured from the point of closest approach one has

s = (r2 − b2 )1/2 and so ds = (r2 − b2 )−1/2 r dr. Altogether one finds
# R b ∂r nrefr
∆α = −2 dr √ 2 . (6.121)
b r − b2
The refraction by the surface of the body was ignored: it is assumed
that it is “soft” with ∂r nrefr = 0 at r = R. The absolute sign was chosen
such that ∆α is positive if the spherical body acts as a “focussing lens.”
This is the case, for example, for light passing through the atmosphere
of the Earth where ∂r nrefr < 0 and so the Sun near the horizon appears
“lifted” (Sommerfeld 1958).
With the results for the neutrino refractive index of Sect. 6.7.1 one
has for a beam of energy ω
√
2 GF nB,c # R b ∂r [nB (− 12 Yn + Ye )]
∆α = 2 dr √ , (6.122)
ω b nB,c r2 − b2
where nB,c is the baryon density at the center of the lens. This ex-
pression applies to νe while for νµ or ντ the term Ye is absent, and for
antineutrinos the overall sign changes.
For the Sun with a central density of about 150 g cm−3 the overall
coefficient is 2.3×10−18 (10 MeV/ω). The integral expression is dimen-
sionless and thus of order unity. Therefore, the focal length of the
Sun as a neutrino lens is of order 1018 R) (solar radius) for 10 MeV
neutrinos, or about the radius of the visible universe!
6.8 Majoron Decay

As a first application for the neutrino dispersion relation I consider the
interaction of neutrinos with the hypothetical majorons (Sect. 15.7).
These particles are Nambu-Goldstone bosons of a symmetry which is
spontaneously broken by a Higgs fields which gives the neutrinos Ma-
jorana masses. For the present purposes it is enough to specify a pseu-
doscalar interaction
Lint = ih ψ ν γ5 ψν χ (6.123)
between the neutrinos of a given family and the massless majorons.
Here, ψν is a Majorana neutrino field while h is a dimensionless Yukawa
coupling constant.
Majorana neutrinos are fermions with only two degrees of freedom.
They correspond to a helicity-minus ν and a helicity-plus ν. In fact, if
they are massless there is no operational distinction between a Majo-

rana neutrino and the two active degrees of freedom of a Dirac neutrino.
Therefore, according to Eq. (6.107) the dispersion relation for the he-
licity ± states of a Majorana neutrino is
E± = (m2 + p2 )1/2 ∓ V, (6.124)
where the medium-induced

√ potential V was given in Eq. (6.108). For
νe it is V = 2GF (ne − 12 nn ) with the electron and neutron densities
ne and nn , respectively.
This dispersion relation implies that the medium is “optically ac-
tive” with regard to the neutrino helicities, just as some media are
birefringent with regard to the photon circular polarization. In the
optical case the left-right symmetry (parity) is broken by the medium
constituents which must have a definite handedness; sugar molecules
are a well-known example. In the neutrino case parity is broken by the
structure of the interaction; the medium itself is unpolarized.
Because there is an energy difference between the Majorana helicity
states ν± for a given momentum, decays ν− → ν+ χ are kinematically
allowed. For relativistic neutrinos the squared matrix element is found
to be |M|2 = 4h2 P1 · P2 with the four-momenta P1,2 of the initial and
final neutrino state. The differential decay rate is then
4h2 d3 p2 d3 k
dΓ = P1 · P2 (2π)4 δ 4 (P1 − P2 − K)
2E1 2E2 (2π)3 2ω(2π)3
(6.125)
with the majoron four-momentum K. Integrating out the d3 k variable

removes the momentum δ function. The remaining differential decay is
#
dΓ +1 P1 · P2 p22
= αχ dx δ(E1 − E2 − ω), (6.126)
dE2 −1 E 1 E2 ω
where x = cos θ for the angle between p1 and p2 and αχ ≡ h2 /4π

is the majoron “fine-structure constant.” In the δ function6
one must
use ω = k = |p1 − p2 | = (p21 + p22 − 2p1 p2 x)1/2 . With dx δ[f (x)] =
|df /dx|−1 = ω/p1 p2 and with energy-momentum conservation which
yields P1 − P2 = K and thus P1 · P2 = 12 (P12 + P22 ) one finds
dΓ (E12 + E22 − p21 − p22 ) p2

= αχ . (6.127)
dE2 2E1 E2 p1
250 Chapter 6
If one ignores the vacuum mass relative to V one has p1,2 = E1,2 ∓ V
so that to lowest order in V
dΓ V (E1 − E2 )
= αχ . (6.128)
dE2 E12
Therefore, the final-state neutrino spectrum has a triangular shape
where E2 varies between 0 and E1 . The integrated decay rate is
Γ = 12 αχ V (6.129)
as first shown by Berezhiani and Vysotsky (1987). Various subleties

were covered in the detailed discussion by Giunti et al. (1992).
Interestingly, under the same circumstances the decay χ → ν+ ν− is
equally possible (Rothstein, Babu, and Seckel 1993) and proceeds with
the same rate Eq. (6.129)—see Berezhiani and Rossi (1994).
In these calculations it was assumed that the majoron is massless
as it behooves a Nambu-Goldstone boson. Does this remain true in a
medium? The majorons could have a Yukawa coupling g to electrons
in which case one would expect on dimensional grounds that they de-
velop a medium-induced “mass” of order g(ne /me )1/2 in analogy to the
photon plasma mass. Indeed, if one works with a pseudoscalar cou-
pling analogous to Eq. (6.123) one finds such a result. Even if they did
not couple to electrons, in a supernova core there is a background of
neutrinos to which majorons couple by assumption.
However, a pseudoscalar coupling is not appropriate for a Nambu-
Goldstone boson as it is not invariant under a shift χ → χ + χ0 . The
pseudoscalar expression is only the lowest-order expansion of an expo-
nential coupling which respects the symmetry. Equivalently, a deriva-
tive coupling of the sort (1/2f ) ψγµ γ5 ψ ∂ µ χ can be used which satisfies
the symmetry explicitly (Sect. 14.2.3). Either way one finds that the
forward scattering amplitude between Nambu-Goldstone bosons and
fermions vanishes—there is no refractive index. The same conclusion
was reached by Flynn and Randall (1988) on more general grounds.
For a suitable choice of parameters the medium-induced decay of
electron neutrinos can deplete the solar neutrino flux before it leaves
the Sun. However, it is doubtful if this effect could explain all current
solar neutrino measurements (Sect. 10.8). For h > −6
∼ 10 a radical mod-
ification of the neutrino signal from a supernova collapse is expected
(Sect. 15.7.2).
Chapter 7
Nonstandard Neutrinos
The phenomenological consequences of nonvanishing neutrino masses

and mixings and of electromagnetic couplings are explored. The decay
channels and electromagnetic properties of mixed neutrinos are dis-
cussed. Experimental, astrophysical, and cosmological limits on neu-
trino masses, decays, and electromagnetic properties are summarized.
7.1 Neutrino Masses

7.1.1 The Fermion Mass Problem
In the physics of elementary particles one currently knows of two cat-
egories of apparently fundamental fields: the spin- 12 quarks and lep-
tons on the one-hand side, and the spin-1 gauge bosons on the other.
The former constitute “matter” while the latter mediate the electro-
magnetic, weak, and strong forces. The gauge-theory description of
the interactions among these particles is renowned for its elegance and
stunning in its success at accounting for all relevant measurements. At
the same time it has many entirely loose ends. Perhaps the most puz-
zling problem is that of fermion masses and the related issue of the
threefold replication of families: The electron and neutrino as well as
the up and down quarks (which make up protons and neutrons) each
come in two additional “flavors” or families which seem to differ from
the first one only in their masses (Fig. 7.1).
The standard model of particle physics holds that all fermions and
gauge bosons are fundamentally massless. The gauge symmetry for-
bids a fundamental mass for the latter while the masslessness of the
former is indicated by the handedness of the weak interaction: only
left-handed (l.h.) fermions feel this force while the right-handed (r.h.)
251
252 Chapter 7
Fig. 7.1. Mass spectrum of elementary fermions according to the Review

of Particle Properties (Particle Data Group 1994). For the top quark see
CDF Collaboration (1995) and D0 Collaboration (1995). For neutrinos the
experimental upper limits of Tab. 7.2 are shown.
ones are “sterile.” Handedness, however, is not a Lorentz-invariant

concept because a particle with a spin opposite to its momentum (l.h.
or negative helicity) is r.h. when viewed by a sufficiently fast-moving
observer. Only massless particles cannot be “overtaken” because they
move with the speed of light and so they can be classified into l.h. and
r.h. states without reference to a specific Lorentz frame.
Any deviation from relativistic propagation is thought to be a “re-
fractive” effect, much as a photon acquires a nontrivial dispersion in a
medium. Here, the “medium” is the spin-0 Higgs field Φ, a hypothetical
third category of fundamental objects, which is believed to take on the
nonzero classical value Φ0 ≈ 246 GeV in vacuum due to self-interactions
Nonstandard Neutrinos 253
(“spontaneous symmetry breaking”). Fermion fields ψ interact with the

Higgs field by virtue of a Lagrangian gΦψψ where g is a dimensionless
(Yukawa) coupling constant. In vacuum, this coupling leads to an in-
teraction term gΦ0 ψψ which has the form of a standard Dirac mass
term mψψ. Different fermion masses thus arise from different Yukawa
couplings. Φ0 does not depend on the Lorentz frame because of the
scalar nature of the Higgs field and so gΦ0 ψψ is the same in all frames,
unlike a refractive photon “mass” in a medium.
It is not known whether the Higgs mechanism is the true source
for the masses of the fundamental fermions. Experimentally, the Higgs
particle (excitations of the Higgs field) has not yet been discovered,
while theoretically the fermion masses are the least appealing aspect
of the standard model because they require a host of ad hoc coupling
constants which must be experimentally determined.
7.1.2 Dirac and Majorana Masses

Neutrinos break the pattern of Fig. 7.1 in that they are much lighter
than the other members of a given family, a discrepancy which is most
severe for the third family where cosmologically mντ < ∼ 30 eV, eight and
ten orders of magnitude less than mτ and mt , respectively! Moreover,
neutrinos are different in that their r.h. chirality states are sterile be-
cause of the handedness of the weak interaction. The r.h. states interact
with the rest of the world only by gravity and by a possible Yukawa
coupling to the Higgs field.
It is frequently assumed that neutrinos do not couple to the Higgs
field, and that the r.h. components do not even exist, assumptions which
are part of the particle-physics standard model. In this case there are
only two neutrino states for a given family as opposed to four states for
the charged leptons. Actually, one may interpret the two components
of such a neutrino as the spin states of a Majorana fermion which is de-
fined to be its own antiparticle. Fermions with four distinct states are
known as Dirac fermions. Naturally, a Majorana fermion cannot carry
a charge as that would allow one to distinguish it from its antiparticle.
A magnetic or electric dipole moment is equally forbidden: its orienta-
tion relative to the spin is reversed for antiparticles. For example, the
neutron cannot be a Majorana fermion among other reasons because it
carries a magnetic moment.
Because the r.h. neutrino components of a given family, if they exist,
are sterile anyway, there is no practical distinction between massless
Dirac and Majorana neutrinos except in a situation where gravitational
254 Chapter 7
interactions dominate. For example, even otherwise sterile neutrinos

should be thermally emitted from black holes which are thought to
emit blackbody radiation of all physical fields. Equally, they would have
been produced in the very early universe when quantum gravitational
effects dominate. However, their present-day cosmic density, like that
of primordial gravitons, would be very dilute relative to microwave
background photons.
If neutrinos were Majorana particles they could still have a mass
even though it could not arise from the usual Higgs field which in-
duces Dirac masses. However, a Majorana mass could arise from the
interaction with a second Higgs field which also develops a vacuum ex-
pectation value. Then the smallness of the neutrino masses could be
due to a small vacuum value of the new Higgs field while the Yukawa
couplings would not need to be anomalously small.
It is also possible that the r.h. components of the neutrinos do ex-
ist, but are themselves (sterile) Majorana fermions with large masses.
It should be noted that any Dirac fermion (four components) can be
viewed as a combination of two Majorana fermions (two components
each) with degenerate masses. Certain variations of such models (“see-
saw mass models”) predict for the light, interacting neutrinos
m1 : m2 : m3 = m2e : m2µ : m2τ or m2u : m2c : m2t . (7.1)
The smallness of the neutrino masses is then a suppression effect by

the large mass scale of the heavy sterile state whose mass would arise,
for example, at the grand unification scale of 1015 −1016 GeV.
For an elementary introduction to the most common models for neu-
trino masses see, for example, Mohapatra and Pal (1991). The dizzying
variety of such models alone attests to the fact that even very ba-
sic questions about the nature of neutrinos remain unanswered. While
some mass schemes like the see-saw relationship Eq. (7.1) are intriguing,
they have no predictive power because there are many other possibili-
ties. Therefore, it is best to remain open to all possibilities which are
not excluded by experimental or astrophysical arguments.
7.1.3 Kinematical Mass Bounds

Unsurprisingly, much experimental effort goes into attempts to measure
or narrow down the range of possible neutrino masses, an area where
astrophysics and cosmology have made their most renowned contribu-
tions to particle physics. Direct laboratory experiments rely on the
kinematical impact of a mass on certain reactions such as nuclear de-

cays of the form (A, Z) → (A, Z + 1) e− ν e where the continuous energy
spectrum of the electrons originally revealed the emission of another
particle that carried away the remainder of the available energy. The
minimum amount of energy taken by the neutrino is the equivalent of
its mass so that the upper endpoint of the electron spectrum is a sen-
sitive measure for mνe . Actually, the most sensitive probe is the shape
of the electron spectrum just below its endpoint, not the value of the
endpoint itself. The best constraints are based on the tritium decay
3
H → 3 He+ e− ν e with a maximum amount of kinetic energy for the
electron of Q = 18.6 keV. This unusually small Q-value ensures that a
large fraction of the electron counts appear near the endpoint (Boehm
and Vogel 1987; Winter 1991).
In Tab. 7.1 the results from several recent experiments are sum-
marized which had been motivated by the Moscow claim of 17 eV <
mνe < 40 eV (Boris et al. 1987). This range is clearly incompatible
with the more recent data which, however, find negative mass-squares.
This means that the endpoint spectra tend to be slightly deformed
in the opposite direction from what a neutrino mass would do. This
effect is particularly striking and significant for the Livermore exper-
iment where it is nearly impossible to blame it on a statistical fluc-
tuation. Therefore, at the present time one cannot escape the con-
clusion that this experimental technique suffers from some unrecog-
nized systematic effect. This problem must be resolved before it will
become possible to extract a reliable bound on mνe although it ap-
pears unlikely that an mνe in excess of 10 eV could be hidden by what-
Table 7.1. Summary of tritium β decay experiments.
Experiment m2νe ± σstat ± σsyst Reference

[eV2 ]
Los Alamos −147 ± 68 ± 41 Robertson et al. (1991)
Tokyo −65 ± 85 ± 65 Kawakami et al. (1991)
Zürich −24 ± 48 ± 61 Holzschuh et al. (1992)
Mainz −39 ± 34 ± 15 Weinheimer et al. (1993)
Livermore −130 ± 20 ± 15 Stoeffl and Decman (1994)
Troitsk −18 ± 6 Belesev et al. (1994)a
a See Otten (1995) for a published description.
256 Chapter 7
ever effect causes the “wrong” deformation of the end-point spectrum.

In spite of apparent systematic problems, the Troitsk group (Belesev
et al. 1994) claims a limit of mνe < 4.5 eV at 95% CL; see also Ot-
ten (1995).
One has attempted to determine mνµ by measuring the muon mo-
mentum from the decay of stopped pions, π + → µ+ νµ , leading to m2νµ =
m2π+ + m2µ − 2mπ+ (m2µ + p2µ )1/2 . A recent pµ measurement (Daum et al.
1991) implied a negative squared mass of m2νµ = −(0.154±0.045) MeV2 ,
probably due to large systematic uncertainties in the determination of
mπ+ . Hence, it seemed that the often-quoted bound of mνµ < 0.27 MeV
did not apply. An older experiment studied the in-flight decay of pi-
ons with a result m2νµ = −(0.14 ± 0.20) MeV2 , largely independent of
the pion mass (Anderhub et al. 1982). This implies a 90% CL upper
limit of mνµ < 0.50 MeV. Most recently, the mass of the negative
pion was reconsidered by Jeckelmann, Goudsmit, and Leisi (1994).
Their previous experiment allows for two mass assignments, mπ =
139.56782±0.00037 MeV or 139.56995±0.00035 MeV. The larger value
had previously been rejected on the basis of evidence which now appears
questionable. Together with a new pµ measurement (Assamagan et al.
1994) one finds m2νµ = −0.148 ± 0.024 MeV2 or −0.022 ± 0.023 MeV2 .
The first value is negative by 6.2 standard deviations and thus may
be rejected as unphysical. The second solution is compatible with zero
and gives a 90% CL upper limit of mνµ < 0.16 MeV.
For ντ the best bounds also come from limits on missing energy
in certain reactions, the only form in which ντ has ever been “ob-
served.” The ARGUS Collaboration (1988, 1992) studied the decay
τ − → 3π − 2π + ντ with a total of 20 events with good energy deter-
minations for all five pions, leading to mντ < 31 MeV at 95% CL.
A similar experiment by the CLEO Collaboration (1993) based on a
much larger data sample gave mντ < 32.6 MeV at 95% CL. Most re-
cently, the ALEPH Collaboration (1995) at CERN has reported a new
Table 7.2. Experimental neutrino mass limits.
Flavor Limit CL Reference

νe (5 eV) — See Tab. 7.1
νµ 0.16 MeV 90% Assamagan et al. (1994)
ντ 23.8 MeV 95% ALEPH Collaboration (1995)
95% CL mass limit of 23.8 MeV on the basis of 25 events of the form
τ → 5πντ and 5ππ ◦ ντ where π stands for a charged pion.
Another kinematical method to be discussed in Sect. 11.3.4 uses
the neutrino pulse dispersion from a distant supernova (SN). For νe the
observed neutrinos from SN 1987A gave mνe < ∼ 20 eV, less restrictive
than the tritium experiments. However, if the neutrino pulse from a
future galactic SN will be detected one may be able to probe even a ντ
mass down to the cosmologically interesting 30 eV range (Sect. 11.6)!
For Dirac neutrinos there is another essentially kinematical con-
straint from the SN 1987A neutrino observations. The sterile νDirac
components can be produced in scattering processes by helicity flips.
In a supernova core this effect leads to an anomalous energy drain,
limiting a Dirac mass to be less than a few 10 keV (Sect. 13.8.1).
7.1.4 Neutrinoless Double-Beta Decay

If neutrinos have Majorana masses, lepton number is not conserved as
one cannot associate a conserved “charge” with a Majorana particle.
One observable consequence would be the occurrence of neutrinoless
nuclear decay modes of the form (A, Z) → (A, Z+2) 2e− which would
violate lepton number by two units. There are several isotopes which
can decay only by the simultaneous conversion of two neutrons. Re-
cently it has become possible to observe the electron spectra from the
standard two-neutrino mode (A, Z) → (A, Z+2) 2e− 2ν e ; for a recent
review see Moe (1995). The decay 76 Ge → 76 Se 2e− 2ν e , for example, is
found to have a half-life of (1.43±0.04stat ±0.13syst )×1021 yr (Beck 1993).
The age of the universe, by comparison, is about 1010 yr.
In the 0ν decay mode, loosely speaking, one of the emitted Majorana
neutrinos would be reabsorbed as an antineutrino with an amplitude
proportional to mνe ,Majorana and thus a rate proportional to m2νe ,Majorana .
In a measurement of the combined energy spectrum of both electrons
the 0ν mode would show up as a peak at the endpoint. The best
current upper bound is from the Heidelberg-Moscow 76 Ge experiment
which yields mνe ,Majorana < 0.65 eV (Balysh et al. 1995), a number which
will likely improve to 0.2 eV over the next few years. This nominal limit
must be relaxed by as much as a factor of 2−3 for the uncertainty in the
nuclear matrix elements which are needed to translate an experimental
limit on the neutrinoless decay rate into a mass limit.
With neutrino mixing (Sect. 7.2) the other flavors also contribute so
!
that the bound is really on the quantity &mν ' ≡ j λj |Uej |2 mj where λj
is a CP phase equal to ±1, and the sum is to be extended over all two-
258 Chapter 7
component Majorana neutrinos. In this language a four-component

Dirac neutrino consists of two degenerate two-component Majorana
ones with λ = +1 and −1 so that their contributions cancel exactly,
reproducing the absence of lepton number violation for Dirac neutrinos.
If one takes the largest cosmologically allowed value mντ = 30 eV and
the largest experimentally allowed mixing amplitude (Sect. 8.2.4) of
|Ue,3 | ≈ 0.16 one may have a contribution as large as 0.8 eV from ντ .
7.1.5 Cosmological Mass Bounds

Cosmology arguably yields the most important neutrino mass bounds
(Kolb and Turner 1990; Börner 1992). In the framework of the big-bang
scenario of the early universe one expects about as many “blackbody
neutrinos” in the universe as there are cosmic microwave photons. In
detail, the cosmic energy density in massive neutrinos is found to be
3
"
3
ρν = 11
nγ mi , (7.2)
i=1
with nγ the present-day density of microwave background photons and

mi the neutrino masses. In units of the cosmic critical density this is
3
" mi
Ων h2 = , (7.3)
i=1 93 eV
where h is the Hubble constant in units of 100 km s−1 Mpc−1 . The

observed age of the universe together with the measured expansion
rate yields Ωh2 <
∼ 0.4 so that for any of the known families
mν <
∼ 30 eV . (7.4)
If one of the neutrinos had a mass near this bound it would be the main
component of the long-sought dark matter of the universe.
Certain scenarios of structure formation currently favor “hot plus
cold dark matter” where neutrinos with mνe + mνµ + mντ ≈ 5 eV play a
sub-dominant dynamical role but help to shape the required spectrum
of primordial density perturbations (Pogosyan and Starobinsky 1995
and references therein). Preferably, the three neutrino masses should
be degenerate rather than one dominating flavor.
If neutrinos were unstable and if they decayed so early that their
decay products were sufficiently redshifted by the expansion of the uni-
verse, the cosmological mass bound can be violated without running
into direct conflict with observations. The excluded range of masses
Fig. 7.2. Cosmological bounds on neutrino masses and lifetimes as described

in the text. The experimental limits are shown above the main panel. If the
dominant decay channel is the majoron mode ν → ν # χ the BBN-excluded
range extends between the dashed lines. The dotted line is τν |Ue3 |2 for
standard-model decays ν3 → ν1 according to Eq. (7.17).
and lifetimes according to Dicus, Kolb, and Teplitz (1977)42 is shown

in Fig. 7.2 as a shaded area marked “Mass Density.”
Decaying neutrinos would cause a second cosmic epoch of radiation
domination, suppressing the growth of density fluctuations and thus
the formation of structure (Steigman and Turner 1985; Krauss 1991;
Bond and Efstathiou 1991). Somewhat schematically, the area above
the shaded band in Fig. 7.2 marked “Structure Formation” is excluded
by this more model-dependent argument. For masses and lifetimes on
this band, neutrinos would actually have the beneficial effect of modify-
ing the primordial spectrum of density fluctuations such as to avoid the
problem of too much small-scale power in cold dark matter universes
(Bardeen, Bond, and Efstathiou 1987; Bond and Efstathiou 1991; Do-
delson, Gyuk, and Turner 1994; White, Gelmini, and Silk 1995).
42
Note that the corresponding limits discussed in the book by Kolb and Turner
(1990) are somewhat less restrictive because their treatment does not seem to be
entirely self-consistent (G. Gelmini, private communication).
260 Chapter 7
The expansion rate and thus the energy density of the universe are
well “measured” at the epoch of nucleosynthesis (T ≈ 0.3 MeV) by
the primordial light-element abundances (Yang et al. 1984). This big-
bang nucleosynthesis (BBN) argument has been used to constrain the
number of light neutrino families to Nν < ∼ 3.4 (Yang et al. 1984; Olive
et al. 1990). Even though the measured Z ◦ decay width has established
Nν = 3 (Particle Data Group 1994) the BBN bound remains of interest
as a mass limit because massive neutrinos contribute more than a mass-
less one to the expansion rate at BBN. For a lifetime exceeding about
100 s this argument excludes 500 keV < <
∼ mν ∼ 35 MeV (Kolb et al. 1991;
Dolgov and Rothstein 1993; Kawasaki et al. 1994), with even more re-
strictive limits for Dirac neutrinos (Fuller and Malaney 1991; Enqvist
and Uibo 1993; Dolgov, Kainulainen, and Rothstein 1995). In Fig. 7.2
the region thus excluded is hatched and marked “BBN.”
Kawasaki et al. (1994) have considered the majoron mode ν → ν # χ
(Sect. 15.7) as a specific model for the neutrino decay. Including the
energy density of the scalar χ they find even more restrictive limits
which exclude the region between the dashed lines in Fig. 7.2.
7.2 Neutrino Mixing and Decay

7.2.1 Flavor Mixing
One of the most mysterious features of the particle zoo is the threefold
repetition of families (or “flavors”) shown in Fig. 7.1. The fermions in
each column have been arranged in a sequence of increasing mass which
appears to be the only significant difference between them. There is no
indication for higher sequential families; the masses of their neutrinos
would have to exceed 12 mZ = 46 GeV according to the CERN and SLAC
measurements of the Z ◦ decay width (Particle Data Group 1994). If
the origin of masses is indeed the interaction with the vacuum Higgs
field, the only difference between the fermions of a given column in
Fig. 7.1 is their Yukawa coupling to Φ.
If the only difference between, say, an electron and a muon is the
vacuum refraction, any superposition between them is an equally legit-
imate charged lepton except for the practical difficulty of preparing it
experimentally. When such a mixed state propagates, the two compo-
nents acquire different phases along the beam exactly like two photon
helicities in an optically active medium, leading to a rotation of the
plane of polarization. Of course, now this “polarization” is understood
in the abstract flavor space rather than in coordinate space.
In three-dimensional flavor space one is free to choose any superpo-

sition of states as a basis. It is convenient and common practice to use
the mass eigenstates (vacuum propagation eigenstates) for each column
of Fig. 7.1. Thus by definition the electron is the charged lepton with
the smallest mass eigenvalue, the muon the second, and the tau the
heaviest, and similarly for the quarks.
All fermions interact by virtue of the weak force and thus couple to
the W ± and Z ◦ gauge bosons, the quarks and charged leptons in ad-
dition couple to photons, while only the quarks interact by the strong
force and thus couple to gluons. The W ± (charged current) interac-
tion has the important property of changing, for example, a charged
lepton into a neutrino as in the reaction p + e− → n + νe (Fig. 7.3). If
the initial charged lepton was an electron (the lightest charged lepton
mass eigenstate), the outgoing neutrino state is defined to be an “elec-
tron neutrino” or νe which in general will be a certain superposition of
neutrino mass eigenstates.
Fig. 7.3. Typical charged-current reaction.
This phenomenon of Cabbibo mixing is well established among the

quarks. For example, in the process of Fig. 7.3 the transition among the
quarks is between u and cos θC d + sin θC s where cos θC = 0.975 refers
to the Cabbibo angle. Kinematics permitting, the final-state hadron
will sometimes be uds which constitutes the Λ particle with a mass of
1.116 GeV compared with 0.934 GeV for the neutron (udd).
The superposition of quark states into which u transforms by a
charged-current interaction is commonly denoted by d# , charm couples
to s# , and top to b# while the unprimed states refer to the first, second,
and third mass eigenstates in the d-column of Fig. 7.1. Ignoring the
third family one has
# $ # $# $
d# cos θC sin θC d
= . (7.5)
s# − sin θC cos θC s
It is only by convention that the mixing is applied to the d-column of
the quarks rather than the u-column or both.
262 Chapter 7
Including the third generation, the mixing is induced by the three-

dimensional Cabbibo-Kobayashi-Maskawa (CKM) matrix V . After re-
moving all unphysical phases by an appropriate redefinition of the quark
fields this unitary matrix is given in terms of four significant parameters.
The Particle Data Group (1994) recommends a standard parametriza-
tion in terms of three two-family mixing angles θij < π/2 and one phase
0 ≤ δ < 2π. With Cij ≡ cos θij and Sij ≡ sin θij this standard form is
(Fritzsch and Plankl 1987)
   
1 0 0 C13 0 S13 e−iδ C12 S12 0
   
V =  0 C23 S23   0 1 0   −S12 C12 0
0 −S23 C23 −S13 eiδ 0 C13 0 0 1
 
C12 C13 S12 C13 S13 e−iδ
 
=  −C23 S12 − C12 S23 S13 eiδ C12 C23 − S12 S23 S13 eiδ C13 S23 
S12 S23 − C12 C23 S13 eiδ −C12 S23 − C23 S12 S13 eiδ C13 C23
 
1 S12 S13 e−iδ
 
≈ −S12 1 S23  . (7.6)
S12 S23 − S13 eiδ −S23 1
Experimentally one has the 90% CL ranges (Particle Data Group 1994)
0.218 < S12 < 0.224,

0.032 < S23 < 0.048,
0.002 < S13 < 0.005. (7.7)
The approximation in Eq. (7.6) is justified by the small mixing angles.

They and the CP-violating phase δ = 3.3×10−3 (Wolfenstein 1986) are
measured parameters of the standard model which, like the fermion
masses, are not theoretically accounted for at the present time.
If neutrinos have masses one naturally expects that they follow a
similar scheme and so the “weak interaction eigenstates” ν& (+ = e, µ, τ )
are thought to be given as linear superpositions of the mass eigenstates
νi by virtue of
3
"
ν& = U&i νi , (7.8)
i=1
where the unitary matrix U plays the role of the CKM matrix. Unless
otherwise stated ν1 will always refer to the dominant mass admixture
of νe and so forth. It seems plausible that m1 < m2 < m3 , a hierarchy
that is often assumed.
Flavor mixing is the only possibility for members of one family (one
row in Fig. 7.1) to transform into those of a different family. This phe-
nomenon is known as the absence of flavor-changing neutral currents;
it means that the Z ◦ coupling to quarks and leptons, like the photon
coupling, leaves a given superposition of fermions unaltered. For ex-
ample, muons decay only by the flavor-conserving mode µ− → νµ e− ν e
(Fig. 7.4); the experimental upper limits on the branching ratios for
µ− → e− γ and µ− → e− e+ e− are 5×10−11 and 1.0×10−12 , respectively.
In the absence of neutrino masses and mixing the individual lepton fla-
vor numbers are conserved: a lepton can be transformed only into its
partner of the same family, or it can be created or annihilated together
with an antilepton of the same family.
Fig. 7.4. Allowed and forbidded µ decays.
When neutrinos have masses and mixings, flavor-violating lepton

decays become possible, but their rate would be so small that their
experimentally observed absence does not yield interesting constraints
on neutrino parameters. Because neutrino masses must be very small if
they exist at all, the most significant observable effect is that of neutrino
oscillations.
7.2.2 Standard-Model Decays of Mixed Neutrinos

For massive neutrinos it is kinematically possible to decay according
to ν → ν # γ, ν → ν # γγ, or ν → ν # ν ## ν ## . In the absence of mixing, of
course, all of these modes are forbidden. Even in the presence of mixing,
Fig. 7.5. Allowed and forbidden tree-level decays of mixed neutrinos.

264 Chapter 7
however, the three-neutrino channel remains forbidden by the absence

of flavor-changing neutral currents because the flavor content of ν and
ν # in Fig. 7.5 must remain unaltered by the Z ◦ vertex. Therefore, in
the standard model low-mass neutrinos can decay only by higher-order
(radiative) amplitudes.
“Heavy” neutrinos νh with mh > 2me ≈ 1 MeV may decay at tree-
level through the channel νh → νe e+ e− at a rate
1 G2F
= |Ueh |2 m5 Φ(mh )
τe+ e− 3 (4π)3 h
= |Ueh |2 3.5×10−5 s−1 m5MeV Φ(mh ), (7.9)
where Ueh is the mixing amplitude between νh and νe , GF is the Fermi

constant, and mMeV ≡ mh /MeV. The phase-space factor is (Shrock
1981)
Φ(mh ) = (1 − 4a)1/2 (1 − 14a − 2a2 − 12a3 )

1 + (1 − 4a)1/2
+ 24a2 (1 − a2 ) ln (7.10)
1 − (1 − 4a)1/2
with a ≡ m2e /m2h ; it is shown in Fig. 7.6.
Fig. 7.6. Phase space factor for νh → νe e+ e− according to Eq. (7.10).
The one- and two-photon decay modes arise in the standard model
with mixed neutrinos from the amplitudes shown in Fig. 7.7. Turn first
to the one-photon decay νi → νj γ with the neutrino masses mi > mj .
Fig. 7.7. Feynman graphs for neutrino radiative decays. There are other
similar graphs with the photon lines attached to the intermediate W boson.
In general the matrix element can be thought of as arising from an

effective interaction Lagrangian of the form
Lint = 12 ψ i σαβ (µij + -ij γ5 )ψj F αβ + h.c., (7.11)
where F αβ is the electromagnetic field tensor, ψi and ψj are the neutrino

fields, and µij and -ij are magnetic and electric transition moments
which are usually expressed in units of Bohr magnetons µB = e/2me .
The decay rate is
+ ,3 + ,2
1 |µij |2 + |-ij |2 m2i − m2j −1 µeff 3
= = 5.308 s δm m3eV ,
τγ 8π mi µB
(7.12)
where µ2eff ≡ |µij |2 + |-ij |2 , meV ≡ mi /eV, and δm ≡ (m2i − m2j )/m2i .
An explicit evaluation of the one-photon amplitude of Fig. 7.7 yields
for Dirac neutrinos (Pal and Wolfenstein 1982)
- √
µD
ij e 2 GF "
= 2
(m i ± m j ) U&j U&i∗ f (r& ). (7.13)
-D
ij
(4π) &=e,µ,τ
For Majorana neutrinos one has instead µM D M

ij = 2µij and -ij = 0 or
M M D
µij = 0 and -ij = 2-ij , depending on the relative CP phase of νi and νj .
In Eq. (7.13) r& ≡ (m& /mW )2 where the charged-lepton masses are
me = 0.511 MeV, mµ = 105.7 MeV, and mτ = 1.784 GeV while the
W ± gauge boson mass is mW = 80.2 GeV. Thus for all charged leptons
r& + 1; in this limit
f (r& ) → − 32 + 34 r& . (7.14)
If one inserts the leading term − 32 into the sum in Eq. (7.13) one finds
that its contribution vanishes because the unitarity of U implies that its
rows or columns represent orthogonal vectors. Because the first nonzero
266 Chapter 7
contribution is from 34 r& , the transition moments are suppressed by

(m& /mW )2 , an effect known as GIM cancellation after Glashow, Il-
iopoulos, and Maiani (1970). Explicitly, the transition moments are
- # $2 " # $2
µD
ij /µB 3GF me mτ m&
= √ (mi ± mj ) U&j U&i∗
-D
ij /µB 2 (4π)2 mW &=e,µ,τ mτ
" # $2
−23 mi ± mj m&
= 3.96×10 U&j U&i∗ . (7.15)
1 eV &=e,µ,τ mτ
These small numbers imply that neutrino radiative decays are exceed-
ingly slow in the standard model.
Dirac neutrinos would have static or diagonal (i = j) magnetic
dipole moments while the electric dipole moments vanish according to
Eq. (7.13). Their presence would require CP-violating interactions.
Majorana neutrinos, of course, cannot have any diagonal electromag-
netic moments. For µD ii the leading term of Eq. (7.14) in Eq. (7.13)
does not vanish because the unitarity of U implies that the sum equals
unity for i = j. Therefore,
√
µD 6 2 G F me
ii
= mi = 3.20×10−19 meV , (7.16)
µB (4π)2
much larger than the transition moments because it is not GIM sup-
pressed.
The two-photon decay rate νi → νj γγ is of higher order and thus
may be expected to be smaller by a factor of α/4π. However, it is
not GIM suppressed so that it is of interest for a certain range of neu-
trino masses (Nieves 1983; Ghosh 1984). Essentially, the result in-
volves another factor α/4π relative to the one-photon rate, and f (r& )
in Eq. (7.13) is replaced by (mi /m& )2 .
As an example consider the different decay modes for ν3 → ν1 ,
assuming that m3 , m1 and that the mixing angles are small so that
ν3 ≈ ντ and ν1 ≈ νe . Then one has explicitly


 Φ(m3 ), ν 3 → ν 1 e+ e− ,



 # $
 27 α
 mτ 4
1 G2 m 5 , ν3 → ν1 γ,
≈ |Ue3 |2 F 33 ×  8 4π mW (7.17)
τ 3 (4π) 
 # $ # $

 1 α 2 m3 4


 , ν3 → ν1 γγ,
180 4π me
where Φ(mh ) was given in Eq. (7.10) and shown in Fig. 7.6. The γγ
decay dominates in a small range of m3 just below 2me .
The quantity τ |Ue3 |2 is shown as a dotted line in Fig. 7.2. Even

without nonstandard physics the decay rate is fast on cosmological
scales if m3 >∼ 2me . Because experimentally m3 may be as large as
24 MeV the cosmological mass bound of 30 eV does not automatically
apply. Experimental limits on Ue3 together with the BBN mass bound,
however, exclude a heavy standard ν3 . Even without reference to BBN
it can be excluded on the basis of the SN 1987A neutrino radiative
lifetime limits (Sect. 12.5.2).
7.3 Neutrino Electromagnetic Form Factors

7.3.1 Overview
When Wolfgang Pauli in 1930 first postulated the existence of neutri-
nos he speculated that they might interact like a magnetic dipole of a
certain moment µν . If that were the case they could be measured by
their ionizing power when they move through a medium; this ionizing
power was first calculated by Bethe (1935). Nahmias (1935) measured
the event rates in a Geiger-Müller counter in the presence and ab-
sence of a radioactive source and interpreted his null result as a limit
µν < −4
∼ 2×10 µB (Bohr magneton µB = e/2me ) on the neutrino dipole
moment. He concluded that “since this limit is already smaller than a
nuclear magneton, it seems probable that the neutrino has no moment
at all.” Subsequent attempts to measure ever smaller neutrino dipole
moments have consistently failed.
The main difference between then and now is the advanced the-
oretical understanding of neutrino interactions in the context of the
standard model of electroweak gauge interactions. A magnetic dipole
interaction couples l.h. with r.h. states so that the latter would not
be strictly sterile. This would be in conflict with the standard model
where neutrinos interact only by their l.h. coupling to W and Z gauge
bosons. Thus neutrino dipole moments must vanish identically because
weak interactions violate parity maximally.
This picture changes when neutrinos have masses because even the
r.h. components of a Dirac neutrino are then not strictly sterile as
they couple to the Higgs field—or else they would not have a mass.
Indeed, an explicit calculation in the standard model with neutrino
masses gave a magnetic dipole moment µν = 3.20×10−19 µB (mν /eV)
(Eq. 7.16). If neutrinos mix, they also obtain transition magnetic and
electric moments. However, they are even smaller because of the GIM
suppression effect—see Eq. (7.15).
268 Chapter 7
Much larger values would obtain with direct r.h. neutrino interac-
tions. For example, in left-right symmetric models there exist heavier
gauge bosons which mediate r.h. interactions; parity violation would
occur because of the mass difference between the l.h. and r.h. gauge
bosons. For a neutrino ν& (flavor + = e, µ or τ ) the dipole moment in
such models is (Kim 1976; Marciano and Sanda 1977; Bég, Marciano,
and Ruderman 1978)
2 + , + ,3
eGF m2W1 3 m2W1
µν = √ 2 m& 1− 2 sin 2ζ + mν"
4
1+ 2 ,
2 2π mW2 m W2
(7.18)
where ζ is the left-right mixing angle between the gauge bosons WL

and WR ; W1,2 are their mass eigenstates.
Because of the smallness of the mass-induced standard dipole mo-
ments any evidence for neutrino electromagnetic interactions would rep-
resent evidence for interactions beyond the standard model. Therefore,
the quest for neutrino electromagnetic interactions is more radical than
that for masses and mixings.
In a dense medium even standard massless neutrinos interact with
photons by an effective coupling which is mediated by the ambient elec-
trons. This coupling can be expressed in terms of an effective neutrino
charge radius. Presently I focus on neutrino interactions in vacuum,
leaving a discussion of their properties in media to Sect. 6.6.
7.3.2 Single-Photon Coupling

There are many possible extensions of the standard model which would
give sizeable neutrino dipole and transition moments by some novel r.h.
interaction. For the purposes of this book the underlying new physics
is of no concern; all we need is a generic representation of its observable
effects in terms of neutrino electromagnetic form factors.
The most general interaction structure of a fermion field ψ with the
electromagnetic field can be expressed as an effective Lagrangian
Lint = −F1 ψγµ ψAµ − G1 ψγµ γ5 ψ ∂µ F µν
− 12 ψσµν (F2 + G2 γ5 )ψF µν , (7.19)
where Aµ is the electromagnetic vector potential and F µν the field

strength tensor. The interpretation of the coupling constants is that
of an electric charge for F1 , an anapole moment for G1 , a magnetic
dipole moment for F2 , and an electric dipole moment for G2 . In the

matrix element derived from this Lagrangian these couplings should
be viewed as form factors which are functions of Q2 where Q is the
energy-momentum transfer to the fermion, i.e. the energy momentum
of the photon line attached to the fermion current. The interpretation
of a charge etc. then pertains to the Q2 → 0 limit.
It is usually assumed that neutrinos are electrically neutral, i.e. that
F1 (0) = 0 because electric charge quantization implies that elementary
particles carry only charges in multiples of 13 e where e is the electron
charge. In recent discussions of electric charge quantization43 it was
stressed, however, that the standard model of electroweak interactions
without grand unification requirements does allow neutrinos to carry
small electric charges. Their possible magnitude is thus an experimental
issue; existing limits are reviewed in Sect. 15.8. Because these limits
are very restrictive, i.e. because neutrino electric charges must be very
small, it appears likely that electric charge is quantized after all so that
neutrino electric charges vanish identically.
Even if neutrinos are electrically neutral, as shall be assumed hence-
forth, they can virtually dissociate into charged particles and so they
will have a form factor F1 (Q2 ) which does not vanish for Q2 -= 0. One
may visualize the neutral object as a superposition of two charge dis-
tributions of opposite sign with different spatial extensions. In terms
of a power series expansion of F1 (Q2 ) one usually defines the charge
radius by virtue of
4
2 ∂F1 (Q2 ) 44
&r ' = 6 4 (7.20)
e ∂Q2 4Q2 =0
where &r2 ' may be both positive or negative.

For neutrinos, the interpretation of the charge radius as an ob-
servable quantity is a rather subtle issue as it is probed by “off-shell”
photons (Q2 -= 0), i.e. by intermediate photons in processes such as
scattering by photon exchange. Because the form factor is proportional
to Q2 such scattering processes do not exhibit a Coulomb divergence.
The charge radius induces a short-range or contact interaction simi-
lar to processes involving Z ◦ exchange. Therefore, the charge radius
represents a correction to the standard tree-level electroweak scatter-
ing amplitude between neutrinos and charged particles. This tree-level
43
Babu and Mohapatra 1990; Babu and Volkas 1992; Takasugi and Tanaka 1992;
Foot, Lew, and Volkas 1993; Foot 1994. References to earlier works are given in
these papers.
270 Chapter 7
amplitude will receive radiative corrections from a variety of diagrams,

including photon exchange, which must be considered simultaneously
so that it is not at all obvious that one can extract a finite, gauge-
invariant, observable quantity that can be physically interpreted as a
charge radius.44
The charge radius, even if properly defined, represents only a cor-
rection to the tree-level weak interaction and as such it is best studied
in precision accelerator experiments. In the astrophysical context, weak
interaction rates involving standard l.h. neutrinos cannot be measured
with the level of precision required to test for small deviations from
the standard model. Indeed, a recent compilation (Salati 1994) reveals
that experimental bounds on &r2 ' are more sensitive than astrophysical
limits, except perhaps for ντ for which experimental data are scarce and
so its standard-model neutral-current interactions are not well tested.
The matrix element for the anapole interaction in the Lorentz gauge
is proportional to Q2 . Therefore, it vanishes in the limit Q2 → 0, i.e. for
real photons coupled to the neutrino current. The role of the anapole
form factor G1 is thus very similar to a charge radius: it represents a
correction to the standard tree-level weak interaction and as such does
not seem to be of astrophysical interest.
The form factors F2 and G2 are of much greater importance be-
cause they may obtain nonvanishing values even in the Q2 → 0 limit.
Henceforth I shall refer to µ ≡ F2 (0) as a magnetic dipole moment,
to - ≡ iG2 (0) as an electric dipole moment, respectively. This iden-
tification is understood if one derives the Dirac equation of motion
i∂t ψ = Hψ for a neutrino field ψ (mass m) in the presence of an exter-
nal, weak, slowly varying electromagnetic field F µν . From Eq. (7.19)
one finds for the Hamiltonian
5 6
H = −iα · ∇ + β m − (µ + i-γ5 )(iα · E + Σ · B) , (7.21)
where 12 σµν F µν = iα · E + Σ · B was used. In the Dirac representation

one has
# $ # $ # $
0 σ I 0 σ 0
α= , β= , Σ= , (7.22)
σ 0 0 −I 0 σ
where σ is a vector of Pauli matrices while I is the 2 × 2 unit matrix.
For a neutrino at rest the Dirac spinor is characterized by its large
44
For recent discussions of these matters see Lucio, Rosado, and Zepeda (1985),
Auriemma, Srivastava, and Widom (1987), Degrassi, Sirlin, and Marciano (1989),
Musolf and Holstein (1991), and Góngora-T. and Stuart (1992).
component, a Pauli two-spinor φ, which is then found to evolve as

i∂t φ = (µB0 + -E0 ) · σ φ, (7.23)
where the index 0 refers to quantities in the neutrino rest frame. A
neutrino polarized in or opposite to the field direction has the energy
±µB0 or ±-E0 , respectively, so that µ and - are indeed magnetic and
electric dipole moments, respectively.
The electromagnetic form factors obey certain constraints for Dirac
and Majorana neutrinos; detailed discussions were provided by a num-
ber of authors.45 For Dirac neutrinos, all form factors must be real
relative to each other (no relative phases) if CP invariance holds. For
the diagonal case (coupling to one neutrino species) all form factors
must be real, and CP invariance implies that the electric dipole moment
must vanish. For Majorana neutrinos, a magnetic transition moment
(F2 ) must be imaginary, an electric transition moment (G2 ) real. If
CP invariance holds, in addition one of them must vanish, i.e. there is
either a transition electric, or a transition magnetic moment, but not
both. Majorana neutrinos cannot have diagonal electric nor magnetic
moments, nor can they have a charge or charge radius; they may have
an anapole form factor.
7.3.3 Two-Photon Coupling

Discussions of neutrino electromagnetic form factors are usually re-
stricted to the effective neutrino coupling to an electromagnetic wave
or static field. However, a two-photon coupling of neutrinos is also pos-
sible and of some interest. Historically, it was thought for some time
that the process γγ → νν could be of great importance for the emis-
sion of neutrinos from stars until it was shown by Gell-Mann (1961)
that the amplitude for this process vanishes identically if neutrinos
have only l.h. local interactions with electrons. Several authors dis-
cussed the γγ → νν process when neutrinos are massive, or when they
have more general interaction structures (Halprin 1975; Fischbach et al.
1976, 1977; Natale, Pleitez, and Tacla 1987; Gregores et al. 1995). How-
ever, there does not seem to be a plausible scenario where this process
would be of serious astrophysical interest.
In the standard model, there is an effective two-photon coupling
to neutrinos because the interaction is not local; rather, it is medi-
ated by finite-mass gauge bosons. Early calculations of the effective
45
For example Nieves (1982), Kayser (1982), Shrock (1982), and Li and Wilczek
(1982).
272 Chapter 7
coupling in gauge theories were performed by Levine (1966) and Cung

and Yoshimura (1975); for more recent discussions in the framework of
the standard model see Nieves, Pal, and Unger (1983), Nieves (1987),
Dodelson and Feinberg (1991), Liu (1991), Langacker and Liu (1992),
Kuznetsov and Mikheev (1993), and Dicus and Repko (1993). This
two-photon coupling leads to a higher-order contribution to the neu-
trino refractive index in a bath of photons (Sect. 6.7.2). However, this
and other consequences do not seem to be important in any astrophys-
ical or laboratory setting of practical interest.
For neutrinos with masses and mixings there is a decay amplitude
ν → ν # γγ which can dominate for a small range of neutrino masses
below mν = 2me as discussed by Nieves (1983) and Ghosh (1984).
Explicit results were given in Sect. 7.2.2 above.
One of the photons may represent an external electric or magnetic
field, i.e. one may consider the neutrino decay ν → ν # γ in the pres-
ence of a strong external field which would modify the propagators
of the intermediate charged leptons. For very strong fields it appears
that a substantial decay rate can obtain (Gvozdev, Mikheev, and Vas-
silevskaya 1992a,b, 1993, 1994a,b). From this literature it does not
seem to become entirely clear under which if any circumstances these
results might be of practical interest for, say, the decay of supernova or
neutron-star neutrinos.
7.4 Electromagnetic Processes

The presence of electromagnetic dipole and transition moments implies
that neutrinos couple directly to the electromagnetic field, allowing for
a variety of nonstandard processes (Fig. 7.8). Most obviously, neutrinos
can scatter on electrons by photon exchange. The ν-e scattering cross
section (electrons at rest in the laboratory frame) was given, e.g. by
Vogel and Engel (1989)
2 # $2
dσ G2 m e T
= F (CV + CA )2 + (CV − CA )2 1 −
dT 2π Eν
3 2 3
me T 1 1
+ (CA2 − CV2 ) 2
+ αµ2ν − , (7.24)
Eν T Eν
where CV and CA are the weak coupling constants given in Appendix B,

T is the electron recoil energy with the limits 0 ≤ T ≤ 2Eν2 /(2Eν +me ),
and µν is the neutrino dipole moment.
Fig. 7.8. Important processes involving direct neutrino electromagnetic cou-

plings, notably magnetic dipole moments and magnetic or electric transition
moments.
Because an anomalous neutrino charge radius causes a contact in-

teraction
√ it would manifest itself by the modification CV → CV +
1 2
3
2 πα&r '/GF , i.e. there would be a small correction to the overall
cross section. A dipole moment, on the other side, modifies the energy
spectrum of the recoil electrons, notably at low energies, because of its
forward-peaked nature from the Coulomb divergence.
If neutrinos had electric dipole moments, or electric or magnetic
transition moments, these quantities would also contribute to the scat-
tering cross section. Therefore, the quantity measured in electron recoil
experiments involving relativistic neutrinos is (e.g. Raffelt 1989)
"
µ2ν = |µij − -ij |2 , (7.25)
j=νe ,νµ ,ντ
where i refers to the initial-state neutrino flavor. Therefore, in principle

there is a possibility of destructive interference between the magnetic
and electric transition moments of Dirac neutrinos. (Majorana neutri-
nos have only magnetic or electric transition moments, but not both if
CP is conserved.)
274 Chapter 7
A related process to scattering by photon exchange is the spin-

precession in a macroscopic magnetic or electric field into r.h. states of
the same or another flavor, i.e. electromagnetic oscillation into “wrong-
helicity” states. Such effects will be studied in Sect. 8.4 in the general
context of neutrino oscillations. In principle, this process can be impor-
tant at modifying the measurable l.h. solar neutrino flux as discussed
in Sect. 10.7 in the context of the solar neutrino problem. Spin oscilla-
tions can also be important in supernovae where strong magnetic fields
exist, although a detailed understanding remains elusive at the present
time (Sect. 11.4).
The most interesting process caused by dipole moments is the pho-
ton decay into neutrino pairs, γ → νν, which is enabled in media where
the photon dispersion relation is such that ω 2 − k2 > 0. It is the most
interesting process because it occurs even in the absence of dipole mo-
ments due to a medium-induced neutrino-photon coupling (Chapter 6).
This is the dominant standard neutrino emission process from stars for
a wide range of temperatures and densities. Details of both the stan-
dard and the dipole-induced “plasma process” depend on complicated
fine points of the photon dispersion relation in media that were taken
up in Chapter 6.
The salient features, however, can be understood in an approxima-
tion where photons in a medium (“transverse plasmons”) are treated as
particles with an effective mass equal to the plasma frequency ωP which
in a nonrelativistic medium is ωP2 = 4παne /me with the electron den-
sity ne and electron mass me . The decay rate of these electromagnetic
excitation in their own rest frame is then
ωP3 "7 8
Γγ = µ2ν with µ2ν = |µij |2 + |-ij |2 , (7.26)
24π i,j
while in the frame of the medium where the photon has the energy ω a
Lorentz factor ωP /ω must be included. The sum includes all final-state
neutrino flavors with mi + ωP ; otherwise phase-space modifications
occur, and even a complete suppression of the decay by a neutrino
mass threshold. In contrast with Eq. (7.25) relevant for the scattering
rate, no destructive interference effects between magnetic and electric
dipole amplitudes occur.
Transition moments would allow for the radiative decay νi → νj γ.
Again, because a neutrino charge radius or anapole moment vanish in
the Q2 → 0 limit relevant for free photons, radiative neutrino decays are
most generally characterized by their magnetic and electric transition
moments. The decay rate is found to be

m3i "7 8
Γνi = µ2ν with µ2ν = |µij |2 + |-ij |2 . (7.27)
8π j
The sum is extended over all νj with mj + mi , otherwise phase-space

corrections as given in Eq. (7.12) must be included. There is no destruc-
tive interference between electric and magnetic amplitudes. Because of
the long decay path available in the astronomical environment, the
search for decay photons from known astrophysical neutrino sources is
by far the most efficient method to set limits on radiative decays. This
method and its results are studied in detail in Chapter 12.
Neutrino radiative decays, the plasma process, and scatterings by
photon exchange all depend on the same electromagnetic form factors.
Therefore, a limit on the radiative decay time can be expressed as
a limit on neutrino transition moments, and a limit on a transition
moment from, say, a scattering experiment can be translated into a
limit on a radiative decay time. In much of the literature this simple
connection has not been made; the search for neutrino decays and that
for dipole moments were strangely dealt with as separate and unrelated
efforts.
There are a number of interesting neutrino electromagnetic pro-
cesses which I will not discuss in any detail because they either yield
very small rates or have not so far led to any new results. Among them
are the bremsstrahlung emission of photons in neutrino electron colli-
sions νe → νeγ (Mourão, Bento, and Kerimov 1990; Bernabéu et al.
1994). Another possibility is nuclear excitation due to neutrino dipole
moments (Dodd, Papageorgiu, and Ranfone 1991; Sehgal and Weber
1992) or the disintegration of deuterons (Akhmedov and Berezin 1992).
Also, Cherenkov radiation (Grimus and Neufeld 1993) or transition ra-
diation (Sakuda 1994) of neutrinos with magnetic moments have been
studied in the literature.
7.5 Limits on Neutrino Dipole Moments

7.5.1 Scattering Experiments
In principle, there are a number of possibilities to search for direct
neutrino electromagnetic couplings in the laboratory. In practice, the
only method that has so far yielded significant limits is a study of the
neutrino-electron scattering cross section that would receive a forward-
peaked contribution from a dipole moment.
276 Chapter 7
The best limits on µνe are based on the use of reactor neutrinos
as a source. The measurement of the ν e -e scattering cross section by
Reines, Gurr, and Sobel (1976) was interpreted by Kyuldjiev (1984)
to yield a bound of µνe < 1.5×10−10 µB . Since then, the reactor ν e
spectrum has been much better understood. Vogel and Engel (1989)
stressed that a literal interpretation of the old results by Reines, Gurr,
and Sobel (1976) would actually yield evidence for a dipole moment of
about 2−4×10−10 µB . There is, however, a more recent limit of
µνe < 2.4×10−10 µB (7.28)
from the Kurchatov Institute (Vidyakin et al. 1992).

When interpreting this bound as a limit on the νe -ντ transition mo-
ment, recall that for an experimentally allowed ντ mass in the 10 MeV
regime this limit would not apply as reactor neutrinos have relatively
low energies (below about 10 MeV). Therefore, it remains useful to con-
sider the bounds from the neutrino beam at LAMPF with a νe endpoint
energy of 52.8 MeV
µνe < 10.8×10−10 µB ,

µνµ < 7.4×10−10 µB (7.29)
(Krakauer et al. 1990). A similar limit of µνµ < 8.5×10−10 µB was

obtained by Ahrens et al. (1990).
An improvement of the bound on µνe by an order of magnitude
or more can be expected from a new reactor experiment currently in
preparation (“MUNU experiment,” Broggini et al. 1990). However,
results will not become available before a few years from now.
The transition moments between ντ and other sequential neutrinos
are bounded by the above experiments involving initial-state νe ’s or
νµ ’s or their antiparticles. The diagonal ντ magnetic dipole moment,
however, is much less constrained because no strong ντ sources are avail-
able in the laboratory; the ντ -e cross section has never been measured.
However, the calculated ντ flux produced in a proton beam dump from
the decay of Ds mesons can be used to derive an upper limit on the
ντ -e cross section which can be translated into a bound
µντ < 5.4×10−7 µB , (7.30)
(Cooper-Sarkar et al. 1992). Of course, in this range the ντ electromag-

netic cross section would far exceed a typical weak interaction one.
7.5.2 Spin-Flip Scattering in Supernovae
Neutrino scattering by photon exchange can also be important in astro-

physical settings, notably if neutrinos are Dirac particles. The magnetic
or electric dipole coupling is such that it flips the helicity of relativistic
neutrinos, i.e. the final state is r.h. for an initial l.h. neutrino. This
spin flip is of no importance in experiments where the electron recoil is
measured, but it can have dramatic consequences in supernovae where
l.h. neutrinos are trapped by the standard weak interactions. The spin-
flip scattering by a electromagnetic dipole interaction would produce
“wrong-helicity” states that could freely escape unless they scattered
again electromagnetically. The SN 1987A neutrino signal indicates that
this anomalous cooling channel cannot have been overly effective, yield-
ing a constraint of around µν < −12
∼ 3×10 µB on all Dirac diagonal or
transition moments in the sense of Eq. (7.25); see Sect. 13.8.3 for a
more detailed discussion.
One should keep in mind, however, that for dipole moments in this
range the spin precession in the strong macroscopic magnetic fields that
are believed to exist in and near SN cores could also cause significant
left-right transitions. Notably, the back conversion of r.h. neutrinos
could cause a transfer of energy between widely separated regions of the
SN core, and might even help at the explosion (Sect. 13.8.3). The role
of relatively large neutrino dipole moments in SN physics has not been
elaborated in enough depth to arrive at reliable regions of parameters
that are ruled out or ruled in by SN physics and the SN 1987A neutrino
signal.
7.5.3 Spin-Flip Scattering in the Early Universe
Neutrino spin-flip scattering has important consequences in the early

universe as it can bring some or all of the “wrong-helicity” Dirac neu-
trino degrees of freedom into thermal equilibrium. The usual big bang
nucleosynthesis (BBN) argument previously mentioned in Sect. 7.1.5
allows one to exclude this possibility because even one additional ther-
mally excited neutrino degree of freedom appears to be forbidden by the
spectacular agreement between the predicted and observed primordial
light-element abundances.
This argument was first advanced by Morgan (1981a,b). Unfortu-
nately, he used an unrealistically small cutoff for the Coulomb diver-
gence of the spin-flip scattering cross section, leading to an overestimate
of the efficiency by which r.h. Dirac neutrinos can be brought into ther-
278 Chapter 7
mal equilibrium. A reasonable cutoff by the Debye screening scale was

used by Fukugita and Yazaki (1987) who found
µν < −10
∼ 0.5×10 µB , (7.31)
about a factor of 3.5 looser than Morgan’s original constraint.
This limit can be avoided for the diagonal ντ magnetic moment
which would contribute to the ντ ν τ → e− e+ annihilation process. If
the ντ also had a mass in the 10 MeV regime the spin-flip excitation of
the r.h. degrees of freedom would be compensated by the annihilation
depletion of the ντ and ν τ population before nucleosynthesis (Giudice
1990). A detailed analysis (Kawano et al. 1992) reveals that a µντ larger
than about 0.7×10−8 µB is allowed in the mass range between a few and
about 30 MeV.
Like in a SN core, there could exist large magnetic fields in the
early universe that would allow for left-right transitions by the mag-
netic dipole induced spin precession (Sect. 8.4); for early discussions
of this possibility see Lynn (1981) and Shapiro and Wasserman (1981).
Arguments of this sort naturally depend on assumptions concerning the
primordial magnetic field distribution; such fields may be required as
seeds for the dynamo mechanism to create present-day galactic mag-
netic fields. A quantitative kinetic understanding of the process of
populating the r.h. neutrinos requires a simultaneous treatment of the
neutrino spin-precession and scattering much along the lines of Chap-
ter 9 where flavor oscillations are studied in an environment where neu-
trinos scatter frequently. The most recent investigation of primordial
neutrino magnetic oscillations is Enqvist, Rez, and Semikoz (1995); see
their work for references to the previous literature. In certain plausi-
ble scenarios of primordial magnetic field distributions neutrino Dirac
dipole moments as small as 10−20 µB seem to be in conflict with BBN.
7.5.4 Search for Radiative Neutrino Decays

The search for radiative decays of reactor, beam, solar, supernova, and
cosmic neutrinos will be discussed at length in Chapter 12. For electron
neutrinos, the effective transition moment in the sense of Eq. (7.27) will
be found to be limited by

−1 2
 0.9×10 µB (eV/mν )

 Reactors,


 0.5×10−5 µ (eV/m )2 Sun,
B ν
µν <
∼ −8
(7.32)



1.5×10 µB (eV/mν )2 SN 1987A,

 −10 2.3
3×10 µB (eV/mν ) Cosmic background,
where mν is the mass of the decaying parent neutrino. Of these limits,

all except the cosmic one are based on measured neutrino fluxes. For
nonelectron neutrinos, the cosmic and SN 1987A limits apply equally
because these sources emit neutrinos of all flavors.
The SN 1987A constraints apply in this form only for mν < ∼ 40 eV,
and for total lifetimes which in the laboratory exceed the transit time
between the SN and Earth. The cosmological limit assumes that the ra-
diative channel dominates and as such it requires mν < ∼ 30 eV; otherwise
the universe would be “overclosed” by neutrinos. Because mνe < ∼ 5 eV
these conditions are plausibly satisfied for νe . For νµ and ντ one may
contemplate masses in excess of 30 eV if one simultaneously contem-
plates novel interactions which allow for invisible fast decays. There-
fore, radiative decay limits for νµ and ντ must be derived as a func-
tion of the assumed mass and of the assumed total decay time. For
SN 1987A this exercise will be performed in Sect. 12.4.5, the results
are displayed in Fig. 12.17. For cosmic neutrinos, I do not know of a
published comparable contour plot.
7.5.5 Plasmon Decay in Stars

The last and most interesting constraint arises from the energy-loss
argument applied to globular cluster stars. The neutrino emissivity by
the plasma process γ → νν would be too large unless
µν < −12
∼ 3×10 µB . (7.33)
This bound, which applies for mν <

∼ 5 keV, has been derived in detail
in Sect. 6.5.6.
Chapter 8
Neutrino Oscillations
The phenomenon of neutrino oscillations in vacuum and in media as

well as in magnetic fields is studied. Experimental constraints on neu-
trino mixing parameters are reviewed.
8.1 Introduction
If neutrinos do not have novel interactions that allow them to decay
fast then they must obey the cosmological mass limit of mν < ∼ 30 eV.
This is even true for ντ although it could decay sufficiently fast into
the e+ e− νe channel if it had a mass in the 10 MeV range. However, the
absence of γ rays from SN 1987A in conjunction with the neutrino signal
(Sect. 12.5.2) and independently arguments of big-bang nucleosynthesis
(Fig. 7.2) exclude this option. If neutrino masses are indeed so small
then there is no hope for a direct experimental measurement at the
present time, with the possible exception of mνe which could still show
up in tritium β decay or neutrinoless ββ decay experiments as discussed
in Chapter 7.
Pontecorvo (1967) was the first to realize that the existence of sev-
eral neutrino flavors (two were known at the time) allows even very
small masses to become visible.46 A “weak-interaction eigenstate”
which is produced, say, in the neutron decay n → pe− ν e is in gen-
eral expected to be a mixture of neutrino mass eigenstates. The phe-
nomenon of particle mixing (Sect. 7.2.1) is familiar from the quarks
46
Pontecorvo’s (1957, 1958) original discussion referred to ν ↔ ν oscillations in
analogy to the experimentally observed case of K ◦ ↔ K ◦ . For a historical overview
see Pontecorvo (1983). In this book I will not discuss ν ↔ ν oscillations any
further—see Akhmedov, Petcov, and Smirnov (1993) for a recent reexamination of
“Pontecorvo’s original oscillations.”
280
Neutrino Oscillations 281
and hence does not appear to be an exotic assumption. Whatever the

physical cause of particle masses, it seems unrelated to their gauge in-
teractions! Expanding the neutrino state in plane waves, each mass
eigenstate propagates as e−i(ωt−ki ·x) where k2i = ω 2 − m2i . Therefore,
the different mass components develop phase differences, causing the
original superposition which formed a ν e to turn partially into other
flavors. Therefore, one can search for the disappearance of neutrinos
of a given flavor from a beam, or one can search for the appearance of
“wrong-flavored” states in a beam. The measured deficit of solar νe ’s
(Chapter 10) has long been attributed to the oscillation phenomenon
even though a definitive proof is still missing.
Neutrino oscillations effectively measure a phase difference between
two components of a beam, much as the rotation of the plane of polar-
ization of linearly polarized light represents a phase difference between
the circularly polarized components of a beam in an optically active
medium. This method is sensitive to small differences in the refractive
index of the two components. For example, the Faraday rotation ef-
fect can be used to measure very weak interstellar magnetic fields even
though the interstellar medium is quite dilute. Both for neutrinos and
photons, fine points of the dispersion relation have a significant impact
on the oscillation effect.
Wolfenstein (1978) was the first to recognize that the medium-
induced modification of the neutrino dispersion relations (Sect. 6.7.1)
is not an academic affair, but rather of immediate relevance for some
neutrino oscillation experiments. In Mikheyev and Smirnov’s (1985)
seminal paper it was shown that oscillations can be “resonant” when a
beam passes through such a density gradient that the flavor branches
of the dispersion relation cross. This Mikheyev-Smirnov-Wolfenstein
(MSW) effect is very important in astrophysics because neutrinos are
naturally produced in the interior of stars and stream through a density
gradient into empty space.
An adiabatic crossing of the dispersion relations has the effect of
interchanging the flavor content of the neutrino flux even if the mixing
angle is very small. This effect is one version of the oscillation solution
of the solar neutrino problem. For suitable parameters it is also signifi-
cant in supernovae where different-flavored neutrinos are thought to be
produced with different energy spectra. The MSW effect could swap
the spectral characteristics of the neutrinos emerging from a newborn
neutron star, allowing for a number of fascinating novel effects.
If neutrinos had large magnetic dipole moments they could spin-
precess in magnetic fields. This effect is completely analogous to flavor
282 Chapter 8
oscillations, except that here the two helicity components rather than
the flavor components get transformed into each other. Again, this is
a standard effect familiar from the behavior of electrons in magnetic
fields. In astrophysical bodies large magnetic fields exist, especially in
supernovae, so that magnetic helicity oscillations are potentially inter-
esting. However, much larger magnetic dipole moments are required
than are predicted for standard massive neutrinos. Thus, flavor oscil-
lations have rightly received far more attention.
Presently I will develop the theoretical tools for neutrino oscilla-
tions, and summarize the current experimental situation. In Chapter 9
I will discuss the more complicated phenomena that obtain when os-
cillating neutrinos are trapped in a supernova core. The story of solar
neutrinos (Chapter 10) is inextricably intertwined with that of neu-
trino oscillations, especially of the MSW variety. Finally, oscillations
may also play a prominent role for supernova neutrinos and the inter-
pretation of the SN 1987A signal (Chapter 11).
8.2 Vacuum Oscillations

8.2.1 Equation of Motion for Mixed Neutrinos
In order to derive a formal equation for the oscillation of mixed neu-
trinos I begin with the equation of motion of a Dirac spinor νi which
describes the neutrino mass eigenstate i. It obeys the Dirac and thus
the Klein-Gordon equation (∂t2 − ∇2 + m2i ) νi = 0. One may readily
combine all mass eigenstates in a single equation
(∂t2 − ∇2 + M 2 ) Ψ = 0, (8.1)
where
   
m21 0 0 ν1
2    
M ≡ 0 m22 0  and Ψ ≡  ν2  . (8.2)
0 0 m23 ν3
Eq. (8.1) may be written in any desired flavor basis, notably in the
basis of weak-interaction eigenstates to which one may transform by
virtue of Eq. (7.8),
   
νe ν1
   
 ν µ  = U  ν2  . (8.3)
ντ ν3
The mass matrix transforms according to M 2 → U M 2 U † and is no
longer diagonal. (Note that U −1 = U † because it is a unitary matrix.)
As usual one expands the neutrino fields in plane waves of the form
Ψ(t, x) = Ψk (t) eik·x for which Eq. (8.1) is
(∂t2 + k2 + M 2 ) Ψk (t) = 0 . (8.4)
In general one cannot assume a temporal variation e−iωt because there

are three different branches of the dispersion relation with ωi2 = k2 +m2i .
A mixed neutrino cannot simultaneously have a fixed momentum and
a fixed energy!
In practice one has always to do with very relativistic neutrinos for
which k = |k| ' mi . In this limit one may linearize Eq. (8.4) by virtue
of ∂t2 + k2 = (i∂t + k)(−i∂t + k). For each mass eigenstate i∂t → ωi ≈ k
and one needs to keep the exact expression only in the second factor
where the difference between energy and momentum appears. Thus
∂t2 + k2 ≈ 2k(−i∂t + k), leading to the Schrödinger-type equation
' (
M2
i∂t Ψk = Ωk Ψk where Ωk ≡ k + . (8.5)
2k
The vector Ψ originally consisted of neutrino Dirac spinors but it was

reinterpreted as a vector of (positive-energy) probability amplitudes.
For negative-energy states (antineutrinos) a global minus sign appears
in Eq. (8.5).
The Schrödinger equation (8.5) describes a spatially homogeneous
system with a nonstationary temporal evolution. In practice one usu-
ally deals with the opposite situation, namely a stationary neutrino
flux such as that from a reactor or the Sun with a nontrivial spatial
variation. Then it is useful to expand Ψ(t, x) in components of fixed
frequency Ψω (x)e−iωt , yielding
(−ω 2 − ∇2 + M 2 ) Ψω (x) = 0. (8.6)
In the relativistic limit and restricting the spatial variation to the

z-direction one obtains in full analogy to the previous case
' (
M2
i∂z Ψω = −Kω Ψω where Kω ≡ ω − . (8.7)
2ω
This equation describes the spatial variation of a neutrino beam prop-

agating in the positive z-direction with a fixed energy ω.
Ultimately one is not interested in amplitudes but in the observable
probabilities |Ψ$ |2 = Ψ∗$ Ψ$ with & = e, µ, or τ . One may derive an
284 Chapter 8
equation of motion for these quantities which is most compact in terms

of a density matrix
ρab = Ψ∗b Ψa . (8.8)
Then i∂t ρk = [Ωk , ρk ] or i∂z ρω = −[Kω , ρω ] where [A, B] = AB − BA

is a commutator of matrices. Therefore,
i∂t ρ = (2k)−1 [M 2 , ρ] or i∂z ρ = (2ω)−1 [M 2 , ρ] , (8.9)
where the indices ω or k have been dropped.

A beam evolves from the z = 0 state or density matrix as
Ψω (z) = eiKz Ψω (0) or ρω (z) = e−iKz ρω (0) eiKz . (8.10)
An analogous result applies to the case of temporal rather than spatial

oscillations. In the weak-interaction basis eiKz will be denoted by W ,
W (z) ≡ (eiKz )weak = U (eiKz )mass U † . (8.11)
If the neutrino is known to be a νe at the source (z = 0) its probability

for being measured as a νe at a distance z (“survival probability”) is
|Wee (z)|2 .
One may be worried that the simple-minded derivation and interpre-
tation of these results is problematic because the neutrino wave function
is never directly observed. What is observed are the charged leptons
absorbed or emitted in conjunction with the neutrino production and
detection. However, if one performs a fully quantum-mechanical calcu-
lation of the probability (or cross section) for the compound process of
neutrino production, propagation, and absorption, the naive oscillation
probability described by the elements of the W matrix factors out for
all situations of practical interest, and notably in the relativistic limit
(Giunti et al. 1993; Rich 1993).
8.2.2 Two-Flavor Oscillations

The neutrino mixing matrix U can be parametrized exactly as the
Cabbibo-Kobayashi-Maskawa (CKM) matrix in the quark sector in
Eq. (7.6). If one of the three two-family mixing angles is much larger
than the others (as for the quarks) one may study oscillations between
the dominantly coupled families as a two-flavor mixing problem. More-
over, because so far all experiments—with the possible exception of
solar and certain atmospheric neutrino observations—yield only upper
limits on oscillation parameters one usually restricts the analysis to a
two-flavor scenario. Of course, if one were to observe the appearance

of a certain flavor—rather than the disappearance of νe as for solar
neutrinos or of νµ in the atmospheric case—one would have to consider
the possibility that they arise from sequential transitions of the sort
νe → νµ → ντ . Note that in the bottom-left entry of the approximate
expression for the CKM matrix Eq. (7.6) the term S12 S23 had to be
kept because it is larger than the direct term S13 . The neutrino mixing
angles could show a similar hierarchy.
With these caveats in mind we turn to the two-flavor mixing case
where U has the 2×2 Cabbibo form Eq. (7.5),
U = cos θ I + i sin θ σ2 , (8.12)
with the mixing angle θ, the 2×2 unit matrix I, and the Pauli matrix47
σ2 . The mass matrix may be written in the form
M 2 /2ω = b0 − 12 B · σ, (8.13)
where b0 = (m21 + m22 )/4ω. In the weak-interaction basis
 
sin 2θ
2π  
B=  0 , (8.14)
&osc
cos 2θ
a vector which is tilted with regard to the 3-axis by twice the mixing
angle (Fig. 8.2). Further,
4π ω
&osc ≡ 2 (8.15)
m2 − m21
is the oscillation length. Its meaning will presently become clear.
In this representation it is straightforward to work out the spatial
behavior of a stationary neutrino beam. From K = ω − b0 + 12 B · σ
one finds
) * + * +* +,
πz πz − cos 2θ sin 2θ
W = ei(ω−b0 )z cos − i sin .
&osc &osc sin 2θ cos 2θ
(8.16)
Assuming that the oscillations are among the first two families, the
appearance probability for a νµ and the νe survival probability are for
an initial νe
prob (νe → νµ ) = |Weµ |2 = sin2 (2θ) sin2 (πz/&osc ),
prob (νe → νe ) = |Wee |2 = 1 − prob (νe → νµ ). (8.17)
The oscillation behavior is shown in Fig. 8.1.
* + * + * +
47 0 1 0 −i 1 0
The Pauli matrices are σ1 = , σ2 = , and σ3 = .
1 0 i 0 0 −1
286 Chapter 8
Fig. 8.1. Oscillation pattern for two-flavor oscillations (neutrino energy ω).
The flavor oscillations described by Eq. (8.16) are fully analogous to

the rotation of the plane of polarization in an optically active medium
or to the spin precession in a magnetic field. This analogy is brought
out more directly if one starts with the equation of motion for the
density matrix Eq. (8.9). Suppressing the index ω the matrices can be
expressed as
ρ = 12 (1 + P · σ) and K = ω − b0 + 12 B · σ, (8.18)
and a similar representation for Ω where B is expressed as a function

of k by virtue of ω → k to lowest order for relativistic neutrinos.
The vector P is a flavor polarization vector. In the weak-interaction
basis |Ψe |2 = 12 (1 + P3 ) and |Ψµ |2 = 12 (1 − P3 ) give the probability for
the neutrino to be measured as νe or νµ , respectively. P1 and P2 contain
phase information and thus reveal the degree of coherence between the
flavor states. For a pure state |P| = 1 while in general |P| < 1. For
P = 0 one has a completely incoherent equal mixture of both flavors.
In optics, P describes the degree of polarization of a light beam in
the Poincaré sphere representation of the Stokes parameters (Poincaré
1892; Born and Wolf 1959).
The equation of motion for the polarization vector in any flavor
basis is found to be (Stodolsky 1987; Kim, Kim, and Sze 1988)
∂z P = B × P or ∂t P = B × P. (8.19)
Here, B plays the role of a “magnetic field” and P that of a “spin

vector.” The precession of P for an initial νe where P(0) = (0, 0, 1) is
shown in Fig. 8.2.
Fig. 8.2. Flavor oscillation as a “spin precession.” (After Stodolsky 1987.)
8.2.3 Distribution of Sources and Energies

If the neutrino source region is not point-like relative to the oscilla-
tion length, one has to average the appearance or survival probabilities
accordingly. If the source locations z0 are distributed according to a
normalized function f (z0 ) the νµ appearance probability is
-
2 π (z − z0 )
prob (νe → νµ ) = sin 2θ dz0 f (z0 ) sin2 (8.20) .
&osc
2 2 √
For example, consider a Gaussian distribution f (z0 ) = e−z0 /2s /s 2π
of size s for which
. 2 (s/$ 2
/
prob (νe → νµ ) = 1
2
sin2 2θ 1 − e−2π osc )
cos(2πz/&osc ) .(8.21)
For s = 15 &osc this result is shown in Fig. 8.3. For s = 0 Eq. (8.21) is
identical with Eq. (8.17) while for s ' &osc it is 12 sin2 2θ which reflects
that the beam is an incoherent mixture: the relative phases between
different flavor components have been averaged to zero.
No source is exactly monochromatic; usually the neutrino energies
are broadly distributed. With a point source and a normalized distri-
bution g(ω) one finds
-
(m22 − m21 ) z
prob (νe → νµ ) = sin2 2θ dω g(ω) sin2 . (8.22)
4ω
As an example let g(ω) such √ that ∆ = 2π/&osc follows a Gaussian
−(∆−∆0 )2 /2δ 2
distribution e /δ 2π of width δ and with ∆0 = 2π/&0 . Then
. 2 z 2 /2
/
prob (νe → νµ ) = 1
2
sin2 2θ 1 − e−δ cos(2πz/&0 ) (8.23)
1
which is shown in Fig. 8.4 for δ = 10 ∆0 . For δ = 0 Eq. (8.23) reproduces
Eq. (8.17) while for z ' δ −1 it approaches 12 sin2 2θ.
288 Chapter 8
Fig. 8.3. Oscillation pattern for a Gaussian source distribution with s =

1
5 $osc according to Eq. (8.21).
Fig. 8.4. Oscillation pattern for a mixture of neutrino energies with δ =

1
10 π/$0 according to Eq. (8.23).
These phenomena are well described in the picture of a precessing

polarization vector which represents a density matrix and thus is de-
signed to deal with incoherent or partially coherent beams.0 Notably, for
a distribution of energies the polarization vector is P = dω g(ω) Pω .
Because the components Pω precess with different frequencies about a
common “magnetic field” direction the component of P transverse to B
disappears as the Pω approach a uniform distribution on the precession
cone in Fig. 8.2. The projection of P on B, however, is conserved so
that P → (P · B̂) B̂ for t → ∞. If originally P = (0, 0, 1) for initial νe ’s
the geometry of Fig. 8.2 indicates that P3 → cos2 2θ for t → ∞ and so
prob(νe → νµ ) = 12 sin2 2θ.
8.2.4 Experimental Oscillation Searches

Because neutrino masses must be very small the oscillation length in-
volves macroscopic scales. Numerically, it is
Eν 1 eV2
&osc = 2.48 m . (8.24)
1 MeV ∆m2
Therefore, the modulation of the flavor content of a neutrino beam can
occur on large, even astronomical length scales.
There exists a large number of oscillation searches using terrestrial
neutrino sources (reactors, accelerators); for detailed references see Par-
ticle Data Group (1994). In Fig. 8.5 (curves a–f ) I show the most
restrictive limits on oscillations between the known neutrinos where
Fig. 8.5. Experimental limits on neutrino masses and mixing angles. Re-
actors, νe disappearance: (a) Bugey 4 (Achkar et al. 1995), superseding the
Gösgen limits (Zacek et al. 1986); (b) Kurchatov Institute (Vidyakin et al.
1987, 1990, 1991). Accelerator experiments: (c) BNL Experiment 776, wide-
band beam, νe and ν e appearance (Borodovsky et al. 1992). (d) BNL Ex-
periment 734, measurement of νe /νµ ratio (Ahrens et al. 1985). (e) Fermilab
Experiment 531, ντ appearance (Ushida et al. 1986); similar constraints were
reported by the CHARM II Collaboration (1993). (f) CDHS Experiment,
νµ disappearance (Dydak et al. 1984). (g) Anticipated range of sensitivity
for the CHORUS and NOMAD experiments which are currently taking data
at CERN (DiLella 1993; Winter 1995).
290 Chapter 8
the analysis was always based on the assumption that two-flavor os-
cillations dominate. The disappearance experiments, of course, also
constrain oscillations into hypothetical sterile neutrinos.
Even though the experimental results look very impressive, a glance
on the CKM matrix Eq. (7.6) reveals that one could not yet have ex-
pected to see oscillations in the νe ↔ ντ or νµ ↔ ντ channel if the
neutrino mixing angles are comparably small. It is very encouraging
that the NOMAD and CHORUS experiments which are currently tak-
ing data at CERN (DiLella 1993; Winter 1995) anticipate a range of
sensitivity (curve g in Fig. 8.5) which is promising both in view of the
possible cosmological role of a mν in the 10 eV range and the small
mixing angles probed. Other future but less advanced projects for ter-
restrial oscillation searches were reviewed by Schneps (1993, 1995).
At the time of this writing the LSND Collaboration has reported a
signature that is consistent with the occurrence of ν µ → ν e oscillations
(Athanassopoulos et al. 1995). If this interpretation is correct, the
corresponding ∆m2 would exceed about 1 eV2 , while sin2 2θ would be a
few 10−3 . The status of this claim is controversial at the present time—
see, e.g. Hill (1995). No doubt more data need to be taken before one
can seriously begin to believe that neutrino oscillations have indeed
been observed.
8.2.5 Atmospheric Neutrinos

Besides reactors and accelerators, one may also use atmospheric neu-
trinos as a source to search for oscillations. Primary cosmic ray pro-
tons produce hadronic showers when interacting with atmospheric nu-
clei (A). Neutrinos are subsequently produced according to the simple
scheme
p + A → n + π/K + . . .
π/K → µ+ (µ− ) + νµ (ν µ )
µ+ (µ− ) → e+ (e− ) + νe (ν e ) + ν µ (νµ ).
(8.25)
Therefore, one expects twice as many νµ ’s as νe ’s, and equally many

neutrinos as antineutrinos of both flavors. At a detector, the neu-
trino flux is approximately isotropic except at energies below about
1 GeV where geomagnetic effects become important. Because the neu-
trinos come from anywhere in the atmosphere, from directly overhead
or from as far as the antipodes, oscillation lengths between about 10
Fig. 8.6. Limits on neutrino masses and mixing angles from atmospheric
neutrinos. (a) The shaded area is the range of masses and mixing angles
required to explain the νe /νµ anomaly at Kamiokande (Fukuda et al. 1994);
the star marks the best-fit value for the mixing parameters. The hatched
areas are excluded by: (b) νe /νµ ratio at Fréjus (Fréjus Collaboration 1990,
1995; Daum 1994). (c) Absolute rate and (d) stopping fraction of upward go-
ing muons at IMB (Becker-Szendy et al. 1992). Also shown are the excluded
areas from the experimental limits of Fig. 8.5.
and 13000 km are available.48 The energy spectrum and absolute nor-
malization of the flux must be determined by calculations and thus is
probably uncertain to within about ±30% while the νe /νµ flavor ratio
is likely known to within, say, ±5%.
Several underground proton decay experiments have reported mea-
surements of atmospheric neutrinos. The Fréjus detector (an iron
calorimeter) saw the expected νe /νµ flavor ratio and thereby excluded
the range of masses and mixing angles marked b in Fig. 8.6 for νe -νµ
and νµ -ντ oscillations (Fréjus Collaboration 1990, 1995; Daum 1994).
Instead of measuring the neutrinos directly one may also study the
flux of secondary muons produced by interactions in the rock surround-
48
The effect of matter must be included for νe -νµ atmospheric neutrino oscilla-
tions. For a recent detailed analysis see Akhmedov, Lipari, and Lusignoli (1993).
292 Chapter 8
ing the detector. (Electrons from νe interactions range out much faster
in the rock and so one expects mostly muons from νµ ’s.) This method is
sensitive to the high-energy spectral regime of the atmospheric νµ flux.
Moreover, one may select upward going muons which are produced from
νµ ’s which traversed the entire Earth and thus have a large oscillation
length available. The IMB detector excludes range c by this method
(Becker-Szendy et al. 1992). Also, one may determine the fraction of
muons stopped within the detector to those which exit, allowing one to
constrain a spectral deformation caused by the energy dependence of
the oscillation length. Range d is excluded by this method according
to the IMB detector (Becker-Szendy et al. 1992).
However, several detectors see a substantial deficit of atmospheric
νµ ’s relative to νe ’s, a finding usually expressed in terms of a “ratio of
ratios,” i.e. the measured over the expected ratio of e-like over µ-like
events (Fig. 8.7). While this procedure is justified because it is largely
free of the uncertain absolute flux normalization, one must be careful
at interpreting the significance of the flux deficit. The error of a mea-
sured ratio does not follow a Gaussian distribution; a representation
like Fig. 8.7 tends to overemphasize the significance of the discrepancy
(Fogli and Lisi 1995).
Fig. 8.7. Measured ratio of the atmospheric νµ /νe fluxes relative to the
expected value (“ratio of ratios”) in five detectors. Where two results are
shown they refer to different signatures or data samples. (See Goodman
1995 for references.)
Apparently, the anomaly observed at the Kamiokande water Che-

renkov detector (Hirata et al. 1992; Fukuda et al. 1994) can be ex-
plained in terms of oscillations for neutrino parameters in the shaded
area in Fig. 8.6; the best-fit value is indicated by a star. These results
are a combined fit for the sub-GeV and multi-GeV data as published
by the Kamiokande collaboration (Fukuda et al. 1994). The oscilla-
tion hypothesis appears to be buttressed by a zenith-angle variation of
the effect observed in Kamiokande’s multi-GeV data sample although
the claimed significance of this effect has been critiqued, e.g. by Fogli
and Lisi (1995) and by Saltzberg (1995).
The required large mixing angle as well as the exclusion regions
of the other experiments make it appear dubious that the anomaly is
caused by oscillations. Still, it is a serious effect that cannot be blamed
easily on problems with the Kamiokande detector. Also, the reliability
of some of the exclusion areas in Fig. 8.6 may be called into question,
notably because of their dependence on absolute flux normalizations.
The intuition against a large νµ -ντ mixing angle may be misguided.
In the future, it will be possible to test the relevant regime of mixing
parameters in long-baseline laboratory experiments (e.g. Schneps 1995).
At the time of this writing, the possibility that the atmospheric neutrino
anomaly may be revealing neutrino oscillations remains a lively-debated
possibility.
8.3 Oscillations in Media
8.3.1 Dispersion Relation for Mixed Neutrinos
The neutrino refractive index of a normal medium is extremely small

and so its only potentially observable effect occurs in neutrino oscilla-
tions. The refractive index is different for different flavors—the medium
is “flavor birefringent”—and so neutrinos from different families which
propagate with the same energy through the same medium acquire dif-
ferent phases. If in addition these flavors mix, the medium-induced
phase shift between them shows up in the interference between the
mixed states. Without a medium the phase difference between mixed
states arises from their mass difference. Hence, medium effects will
be noticable only if the induced “effective mass” is of the same order
as the vacuum masses. Therefore, medium refraction is important for
neutrino oscillations in certain situations because the vacuum masses
are very small.
294 Chapter 8
The dispersion relation for a single flavor in a medium was given

by Eq. (6.107). For three flavors which mix according to Eq. (8.3) the
Klein-Gordon equation in Fourier space is
  2

 3Ye − 1 0 0
 G n
F B  
 ω − √  0 Ye − 1 0 

 2 0 0 Ye − 1
   
m21 0 0  νe

  †  
−k 2 − U  0 m22 0  U  νµ  = 0, (8.26)


0 0 m23 ντ
where a possible neutrino background was ignored and Yn = 1 − Yp =

1−Ye was used. Also, the higher-order difference between νµ and ντ was
ignored. This equation has nonzero solutions only if det{. . .} vanishes,
a condition that gives us the dispersion relation for the three normal
modes. For unmixed neutrinos where U is the unit matrix one recovers
Eq. (6.107) for each flavor.
For all practical cases the neutrinos are highly relativistic so that
one may linearize this equation,
2
(ω − k − Meff /2k) Ψ = 0 (8.27)
2
where it is easy to read the matrix Meff from Eq. (8.26). Because to
2
lowest order ω = k one may equally use Meff /2ω in order to derive the
dispersion relation, depending on whether one wishes to write ω as a
function of k or vice versa.
For two-flavor mixing between νe and νµ or ντ (mixing angle θ0 ) the
effective mass matrix may be written in the same form as in vacuum
2
Meff /2ω = b0 − 12 B · σ, (8.28)
√
where b0 = (m21 + m22 )/4k + 2 GF nB (Ye − 12 ) and
     
sin 2θ sin 2θ0 √ 0
2π   m2 − m21    
B=  0 = 2  0  − 2 GF n e  0  .
&osc 2ω
cos 2θ cos 2θ0 1
(8.29)
This equation defines implicitly the mixing angle θ as well as the os-
cillation length &osc in the medium in terms of the masses, the vacuum
mixing angle θ0 , and the electron density ne .
Fig. 8.8. Mixing angle, oscillation length, and neutrino dispersion relation
as a function of the electron density. The medium was taken to have equal
numbers of protons and neutrons (Ye = 12 ), the ratio of neutrino masses was
taken to be m1 : m2 = 1 : 2, and sin2 2θ0 = 0.15.
296 Chapter 8
Explicitly one finds for the mixing angle and the oscillation length
in the medium the following expressions,
sin 2θ0
tan 2θ = ,
cos 2θ0 − ξ
sin 2θ0
sin 2θ = 2 , (8.30)
[sin 2θ0 + (cos 2θ0 − ξ)2 ]1/2
where
√
2 GF ne 2ω −7 Ye ρ ω eV2
ξ≡ = 1.53×10 , (8.31)
m22 − m21 g cm−3 MeV m22 − m21
and
4π ω sin 2θ
&osc = . (8.32)
m22 2
− m1 sin 2θ0
For m2 > m1 these functions are shown in Fig. 8.8; they exhibit a
“resonance” for cos 2θ0 = ξ.
The dispersion relation has two branches which in vacuum corre-
spond to ω1,2 = (k 2 − m21,2 )1/2 . In the relativistic limit they are
m21 + m22 √
ω1,2 − k = + 2 GF nB (Ye − 12 )
4k
m2 − m21 . 2 /1/2
± 2 sin 2θ0 + (cos 2θ0 − ξ)2 , (8.33)
4k
a result schematically shown in Fig. 8.8. The “resonance” of the mix-
ing angle corresponds to the crossing point of the two branches of the
dispersion relation. Of course, the levels do not truly cross, but rather
show the usual “repulsion.”
Because the medium effect changes sign for antineutrinos, a reso-
nance occurs between νe and νµ if m2 > m1 , while none occurs between
ν e and ν µ . If the mass hierarchy is the other way round, a resonance
occurs for ν e and ν µ , but not for νe and νµ .
8.3.2 Oscillations in Homogeneous Media

In a homogeneous medium the treatment of neutrino oscillations is
exactly as in vacuum except that one must use the effective medium
mixing angle and oscillation length given above. If the medium is suf-
ficiently dilute it will not affect the oscillations at all. This is the case
when the quantity ξ of Eq. (8.31) is much smaller than unity. In the
opposite limit one finds for the mixing angle in the medium
m2 − m21
sin 2θ = √ 2 sin 2θ0 . (8.34)
2 GF ne 2ω
The oscillation length becomes
2π g cm−3
&osc = √ = 1.63×104 km , (8.35)
2 GF ne Ye ρ
independent of the neutrino masses or energy.
Typically the effect of the medium is, therefore, to suppress the
mixing angle and thus the possibility to observe oscillations. Of course,
for normal materials with a density of a few g cm−3 and neutrino masses
in the eV range one needs TeV neutrino energies for the medium to be
relevant at all. For a review of the impact of oscillations in a medium
on neutrino experiments or the observation of atmospheric, solar, or
supernova neutrinos see, for example, Kuo and Pantaleone (1989).
8.3.3 Inhomogeneous Medium: Adiabatic Limit

In an inhomogeneous medium the oscillation problem is much more
complicated. Recall that the spatial variation of a stationary neutrino
beam in the z-direction is given by i∂z Ψ = −KΨ according to Eq. (8.7)
2
if one drops the index ω. The matrix K = ω−Meff /2ω is now a function
of z. Formally, the solution is Ψ(z) = W Ψ(0) with
* - z +
# #
W = S exp i K(z ) dz , (8.36)
0
where S is the space-ordering operator. An explicit solution is not

available because the matrices K(z) generally do not commute for dif-
ferent z. Of course, for a constant K one recovers the previous result
W = eiKz .
In certain limits one may still find simple solutions. The most in-
teresting case of neutrinos moving through an inhomogeneous medium
is the emission from stars, notably the Sun, where they are produced
in a relatively high-density region and then escape into vacuum. The
density at the center of the Sun is about 150 g cm−3 , the solar radius
6.96×1010 cm, yielding an extremely shallow density variation by ter-
restrial standards! Therefore, consider the adiabatic limit where the
density of the medium varies slowly over a distance &osc which is the
characteristic length scale for the oscillation problem.
298 Chapter 8
This case is best understood for two-flavor mixing if one studies

the (temporal) evolution of the neutrino flavor polarization vector P.
Recall that it evolves according to the spin-precession formula Ṗ =
B × P where the “magnetic field” is now a function of time. If B
varies slowly relative to the precession frequency the “spin” follows the
magnetic field in the sense that it moves on a precession cone which is
“attached” to B. If the spin is oriented essentially along the magnetic
field direction it stays pinned to that direction. Therefore, it can be
entirely reoriented by slowly turning the external magnet.
For the case of neutrino oscillations this means that in the adia-
batic limit a state can be entirely reoriented in flavor space, i.e. an
initial νe can be turned almost completely into a νµ even though the
vacuum mixing angle may be small. Consider θ0 , 1, an initial den-
sity so large that the medium effects dominate, begin with a νe , and
let m1 < m2 . This means that in Fig. 8.8 (lowest panel) begin on
the upper branch of the dispersion relation far to the right of the
crossover. Then let the neutrino propagate toward vacuum through
an adiabatic density gradient. This implies that it stays on the up-
per branch and ends up at ne = 0 (vacuum) as the mass eigenstate
of the upper eigenvalue m2 which corresponds approximately to a νµ .
This behavior is known as “resonant neutrino oscillations” or MSW ef-
fect after Mikheyev, Smirnov, and Wolfenstein. Mikheyev and Smirnov
(1985) first discovered this effect when they studied the oscillation of
solar neutrinos while Wolfenstein (1978) first emphasized the impor-
tance of refraction for neutrino oscillations. A simple interpretation of
resonant oscillations in terms of an adiabatic “level crossing” was first
given by Bethe (1986).
In order to quantify the adiabatic condition return to the picture
of a spin precessing around a magnetic field. For flavor oscillations
the precession frequency is 2π/&osc . The “magnetic field” is tilted with
an angle 2θ (medium mixing angle) against the 3-direction (Fig. 8.2)
and so its speed of angular motion is 2 dθ/dt. For spatial rather than
temporal oscillations the adiabatic condition is
|∇θ| , π/&osc . (8.37)
This translates into ∇θ = 12 ξ (sin2 2θ/ sin 2θ0 ) ∇ ln ne while &osc is given
by Eq. (8.32) and ξ by Eq. (8.31). Thus, the adiabatic condition is
ξ sin3 2θ |∇ ln ne | , sin2 2θ0 |m22 − m21 |/2ω. (8.38)
This condition must be satisfied along the entire trajectory.

If the neutrino crosses a density region such that a resonance oc-

curs, this part of the trajectory yields the most restrictive adiabaticity
requirement. On resonance ξ = cos 2θ0 and sin 2θ = 1. In this case one
defines an adiabaticity parameter
m22 − m21 sin 2θ0 tan 2θ0
γ≡ , (8.39)
2ω |∇ ln ne |res
where the denominator is to be evaluated at the resonance point. The
adiabatic condition is γ ' 1. It establishes a relationship between vac-
uum mixing angles and neutrino masses for which resonant oscillations
occur.
8.3.4 Inhomogeneous Medium: Analytic Results

For practical problems, notably the oscillation of solar neutrinos on
their way out of the Sun, one may easily solve the equation i∂z Ψ =
−KΨ numerically for prescribed profiles of the electron density and
neutrino production rates. The main features of such calculations, how-
ever, can be understood analytically because for certain simple density
profiles one can find analytic representations of Eq. (8.36) indepen-
dently of the adiabatic approximation.
Consider the situation where a νe is produced in a medium and
subsequently escapes into vacuum through a monotonically decreasing
density profile. Because in the adiabatic limit the production and de-
tection points are separated by many oscillation lengths, the oscillation
pattern will be entirely washed out and one may use average probabil-
ities for the flavor content. Initially, the projection of the polarization
vector on the “magnetic field” direction is cos 2θ where θ is the medium
mixing angle at the production point. Because the component of P in
the B direction is conserved when B changes adiabatically, the final av-
erage projection of P on the 3-axis is cos 2θ0 cos 2θ. Then the survival
probability is prob(νe →νe ) = 12 (1 + cos 2θ0 cos 2θ). This result applies
in the adiabatic limit whether or not a resonance occurs.
Next, drop the adiabatic condition but assume that the production
and detection points are many oscillation lengths away on opposite
sides of a resonance so that the oscillation pattern remains washed out;
then it remains sufficient to consider average probabilities. In this case
it is useful to write
1
prob(νe →νe ) = 2
+ ( 12 − p) cos 2θ0 cos 2θ, (8.40)
where the correction p to the adiabatic approximation is the probability
that the neutrino jumps from one branch of the dispersion relation to
300 Chapter 8
the other (Fig. 8.8, lowest panel) when it moves across the resonant
density region.
A linear density profile near the resonance region, which is always
a first approximation, yields the “Landau-Zener probability”
p = e−πγ/2 (8.41)
which was first derived in 1932 for atomic level crossings. The adia-
baticity parameter γ was defined in Eq. (8.39); in the adiabatic limit
γ ' 1 one recovers p = 0.
For a variety of other density profiles and without the assumption
of a small mixing angle one finds a result of the form
!
e−(πγ/2) F − e−(πγ/2) F
p= , (8.42)
1 − e−(πγ/2) F !
where F # = F/ sin2 θ0 and F is an expression characteristic for a given
density profile. For a linear profile ne ∝ r one has F = 1 so that for
a small mixing angle one recovers the Landau-Zener probability. The
profile ne ∝ r−1 leads to F = cos2 2θ0 / cos2 θ0 . Of particular interest is
the exponential ne ∝ e−r/R0 which yields
F = 1 − tan2 θ0 . (8.43)
All of these results are quoted after the review by Kuo and Pantaleone
(1989) where other special cases and references to the original literature
can be found.
Going beyond the Landau-Zener approximation requires assuming
one of the above specific forms for ne (r) for which analytic results ex-
ist. Recently, Guzzo, Bellandi, and Aquino (1994) used a somewhat
different approach which is free of this limitation. They derived an
approximate solution to the equivalent of Eq. (8.36) by the method of
stationary phases for the space-ordered exponential. They found
' (2
1 − γ# 2γ # . /
2 2 2
p= sin (θ0 − θ) + cos (θ 0 − θ) + cos (θ0 + θ) ,
1 + γ# (1 + γ # )2
(8.44)
where γ # ≡ πγ/16. It was assumed that a resonance occurs between

the production point (mixing angle θ) and the detection point which is
taken to lie in vacuum (mixing angle θ0 ). Eq. (8.44) can be used only
in the nonadiabatic regime as it works only for γ < 16/π.
8.3.5 The Triangle and the Bathtub

The most important application of resonant neutrino oscillations is the
possible reduction of the solar νe flux, an issue to be discussed more
fully in Chapter 10. Here, I will use the above simple analytic results to
discuss schematically the survival probability prob(νe →νe ) of neutrinos
produced in the Sun.
To this end I use an exponential profile for the electron density,
ne = nc e−r/R0 , which is a reasonable first approximation with R0 =
R$ /10.54 (Bahcall 1989). Then |∇ ln ne | = R0−1 is a quantity indepen-
dent of location so that there is no need to evaluate it specifically on
resonance. Thus |∇ ln ne |res = 3×10−15 eV is independent of neutrino
parameters. Therefore, the adiabaticity parameter is
∆m2 MeV
γ= 1
× 103 sin 2θ0 tan 2θ0 . (8.45)
6
meV2 Eν
An electron density at the center of nc = 1.6×1026 cm−3 yields
meV2 Eν
ξc = 40 . (8.46)
∆m2 MeV
The quantity ξ was defined in Eq. (8.31).
It is further assumed that all neutrinos are produced with a fixed
energy by a point-like source at the solar center. They will encounter
a resonance on their way out if ξc > cos 2θ0 . In this case one may
calculate the “jump probability” p according to Eq. (8.42) with F from
Eq. (8.43) for the exponential profile. The survival probability is then
given by Eq. (8.40) with the mixing angle at the solar center
cos 2θ0 − ξc
cos 2θ = . (8.47)
[(cos 2θ0 − ξc )2 + sin2 2θ0 ]1/2
If ξc < cos 2θ0 no resonance is encountered; the vacuum parameters

dominate throughout the Sun. In this case one may use p = 0.
In the framework of these approximations it is straightforward to
evaluate prob(νe →νe ) as a function of ∆m2 and sin2 2θ0 . In Fig. 8.9
contours for the survival probability are shown for Eν = 1 MeV. Be-
cause the energy always appears in the combination ∆m2 /2Eν one may
obtain an analogous plot for other energies by an appropriate vertical
shift. This kind of plot represents the well-known MSW triangle. The
dashed line marks the condition γ = 1 and thus divides the param-
eter plane into a region where the oscillations are adiabatic and one
302 Chapter 8
where they are not. The dotted line marks ξc = cos 2θ0 and thus indi-
cates for which neutrino parameters a resonance occurs on the way out
of the Sun.
Fig. 8.9. The MSW triangle for the simplified solar model discussed in the
text with Eν = 1 MeV.
Fig. 8.10. The MSW bathtub for the simplified solar model discussed in the
text with ∆m2 = 3×10−5 eV2 and sin2 2θ0 = 0.01.
It is also instructive to look at a vertical cut through this contour

plot, except that it is more useful to represent it as a function of Eν for
a fixed ∆m2 rather than the reverse. Both possibilities are equivalent
because what appears is the combination ∆m2 /2Eν . The result is the
MSW bathtub shown in Fig. 8.10 for ∆m2 = 3×10−5 eV2 and sin2 2θ0 =
0.01 which is one set of parameters that might solve the solar neutrino
problem (Chapter 10).
The main point of Fig. 8.10 is that resonant neutrino oscillations
modify the spectrum of observable νe ’s differently for different energies.
Low-energy neutrinos remain entirely unaffected, intermediate-energy
ones are strongly suppressed, and the effect on high-energy states varies
over a broad interval. There, the shape of the observable spectrum is
modified relative to the source spectrum.
8.3.6 Neutrino Oscillations without Vacuum Mixing

Neutrino oscillations in vacuum or in a medium arise because in the
2
weak-interaction basis the matrix of effective neutrino masses, Meff ,
has off-diagonal elements which induce transitions between, say, an
initial νe and a νµ . The effect of the medium is to modify the diagonal
elements of this matrix, possibly such that different flavor states become
degenerate, causing the effect of resonant oscillations. Conceivably,
2
oscillations could occur even if Meff = 0 in vacuum so that in the
standard model there are no off-diagonal terms.
2
The absence of off-diagonal medium contributions to Meff within
the standard model reflects the absence of flavor-changing neutral cur-
rents (Sect. 7.2). It cannot be excluded experimentally, however, that
for neutrinos such currents exist on some level, implying that a νe , for
example, sometimes emerges as a νµ from a collision with another parti-
cle. In the forward direction this would cause an “off-diagonal refractive
index” so that mixing and thus oscillations would be induced by the
medium (Valle 1987; Fukugita and Yanagida 1988). Certain supersym-
metric extensions of the standard model with R-parity breaking predict
such effects (Guzzo, Masiero, and Petcov 1991; Kapetanakis, Mayr, and
Nilles 1992). Therefore, if neutrino oscillations indeed explain the solar
neutrino problem (Chapter 10) this does not inevitably imply neutrino
vacuum masses and mixings—it could also point to the existence of
flavor-changing neutral currents. An analytic description of two-flavor
oscillations for this case was given by Guzzo and Petcov (1991) while a
detailed parameter study for the solution of the solar neutrino problem
was provided by Barger, Phillips, and Whisnant (1991).
304 Chapter 8
8.4 Spin and Spin-Flavor Oscillations

8.4.1 Vacuum Spin Precession
According to Sect. 7.2 neutrinos may have magnetic dipole moments.
In particular, if neutrinos have masses they inevitably have small but
nontrivial electromagnetic form factors. Novel interactions can induce
large dipole moments even for massless neutrinos. In the presence of
magnetic fields such a dipole moment leads to the familiar spin preces-
sion which causes neutrinos to oscillate into opposite helicity states.
If a state started out as a helicity-minus neutrino, the outgoing par-
ticle is a certain superposition of both helicities. The “wrong-helicity”
(right-handed) component does not interact by the standard weak in-
teractions, diminishing the detectable neutrino flux. As the Sun has
relatively strong magnetic fields it is conceivable that magnetic spin
oscillations could be responsible for the measured solar neutrino deficit
(Werntz 1970; Cisneros 1971)—see Sect. 10.7. Magnetic spin preces-
sions could also affect supernova neutrinos, and in the early universe it
could help to bring the “wrong-helicity” Dirac neutrino degrees of free-
dom into thermal equilibrium. Presently I will focus on some general
aspects of neutrino magnetic spin oscillations rather than discussing
specific astrophysical scenarios.
In the rest frame of a neutrino the evolution of its spin operator S =
1
2
σ (Pauli matrices σ) is governed by the Hamiltonian H0 = 2µB0 · S;
this follows directly from Eq. (7.23). Here, B0 refers to the magnetic
field in the neutrino’s rest frame as opposed to B in the laboratory
frame. The equation of motion of the spin operator in the Heisenberg
picture is then given by i∂t S = [S, H0 ] or
Ṡ = 2µB0 × S, (8.48)
which represents the usual precession with frequency 2µB0 around the
magnetic field direction. Equivalently, one may consider Eq. (7.23) for
the neutrino spinor which has the form of a Schrödinger equation.
Usually one will be concerned with very relativistic neutrinos so
that the magnetic field B0 in the rest frame must be obtained from the
electric and magnetic fields in the laboratory frame E and B by virtue
of the usual Lorentz transformation
B0 = (B · v̂)v̂ + γ[v̂ × (B × v̂) + E × v], (8.49)
where v is the velocity of the neutrino, v̂ the velocity unit vector, and
γ = (1 − v2 )−1/2 = Eν /m the Lorentz factor with Eν the neutrino en-
ergy and m its mass. The spin precession as viewed from the laboratory
system, however, involves a time-dilation factor γ −1 so that the Hamil-

tonian for the spin evolution in the laboratory system is H = µγ −1 B · σ
or explicitly
. /
H = µ (m/ω)(B · v̂)v̂ + v̂ × (B × v̂) + E × v · σ. (8.50)
If the spin is quantized along the direction of motion (helicity states)

and if there is only a magnetic field in the laboratory (E = 0), the
Hamiltonian is
* +
γ −1 cos θ e−iϕ sin θ
H = µB +iϕ , (8.51)
e sin θ −γ −1 cos θ
with θ the magnetic field direction relative to the neutrino direction of
motion and ϕ its azimuthal direction which may be chosen as ϕ = 0.
The bottom line is that very relativistic neutrinos (γ → ∞) spin-
precess like nonrelativistic ones, except that the relevant magnetic field
is the transverse component BT ; the effective laboratory Hamiltonian is
* +
0 1
H = µBT . (8.52)
1 0
Therefore, if one begins with left-handed neutrinos a complete preces-
sion into right-handed ones will always occur (assuming a sufficient
path length), independently of the direction between the laboratory
magnetic field and the neutrino direction of motion (Fujikawa and
Shrock 1980).
If there is only an electric field in the laboratory frame, magnetic
oscillations will nevertheless occur because Eq. (8.50) implies that the
neutrino sees an effective magnetic field BT = E × v (Okun 1986). The
precession thus takes place in the plane of E and v.
Contrary to flavor oscillations, the oscillation length does not de-
pend on the energy. Therefore, a broad energy spectrum would not
cause a depolarization of the neutrino flux from an astrophysical source
(Sun, supernovae). Rather, neutrinos of all energies would spin-precess
in step with each other.
Because the precession frequency is 2µBT the neutrinos will have
reversed their spin after a distance
π 10−10 µB 1 G
1
&
2 osc
= = 5.36×1013 cm , (8.53)
2µBT µ BT
with µB = e/2me the Bohr magneton. With a solar BT in the kG regime
the oscillation length is of order a solar radius R$ = 6.96×1010 cm. A
306 Chapter 8
substantial reduction of the left-handed solar neutrino flux then re-

quires magnetic dipole moments of order 10−10 µB . Typical galactic
magnetic fields are of order 10−6 G with coherence scales of order kpc.
Therefore, neutrinos from a galactic source such as a supernova could
be “flipped” before reaching Earth if their dipole moment is of order
10−12 µB . These values are in the neighborhood of what is experimen-
tally and observationally allowed so that there remains interest in the
possibility that magnetic spin oscillations could have a significant im-
pact on the neutrino fluxes from the Sun or from supernovae.
8.4.2 Spin Precession in a Medium

The picture of magnetic spin oscillations thus developed is incomplete.
If neutrinos propagate in a medium as in the Sun they are subject to
medium-induced refractive effects which shift the energy of left-handed
states relative to right-handed ones which propagate as in vacuum be-
cause they are sterile. Therefore, including medium effects the effective
Hamiltonian for the evolution of the two helicity states of νe is
* +
0 √ µBT
H= , (8.54)
µBT 2GF (ne − 12 nn )
where ne and nn are the electron and neutron densities, respectively.
Here, “spin up” refers to the direction of motion, i.e. to right-handed
states, “spin down” to left-handed ones. The two helicity states are
no longer degenerate, and the spin precession will no longer lead to
a complete reversal of the spin. One may write H = b + µB · σ so
that there is effectively a longitudinal magnetic field, i.e. parallel to the
direction of motion or z-direction
GF (ne − 12 nn ) 10−10 µB ρ (Ye − 12 Yn )
BL = √ = 66.6 kG . (8.55)
2µ µ g cm−3
Here, ρ is the mass density and Ye,n the electron and neutron number
per baryon. If BL ' BT the precession is around a direction close to
the direction of motion, and so the spin reversal will always be far from
complete. Hence, the presence of the medium suppresses magnetic spin
oscillations (Voloshin, Vysotskiı̆, and Okun 1986).
8.4.3 Spin-Flavor Precession

If neutrinos were Majorana particles they could not have magnetic
dipole moments, but they could still possess transition moments be-
tween different flavors. In this case the magnetic oscillation would be,
say, between a helicity-minus νe to a helicity-plus νµ which, because of

its assumed Majorana nature, is identical to a ν µ . Because typically the
νe and νµ masses will be different they must be included in the phase
evolution of the neutrinos. Thus, for relativistic states (momentum p)
one arrives at a two-level equation of motion of the form
* + * +* +
νe m2νe /2p µBT νe
i∂t = . (8.56)
νµ µBT m2νµ /2p νµ
The transition magnetic moment µ thus leads to simultaneous spin and
flavor oscillations (Schechter and Valle 1981). Moreover, neutrinos will
be converted to antineutrinos (and the reverse). If such spin-flavor
oscillations occurred, say, in the Sun one might actually be able to
measure a flux of antineutrinos from that source.
A transition magnetic moment implies that the individual flavor
lepton numbers are not conserved. Thus most likely there is also stan-
dard flavor mixing which allows for, say, νe ↔ νµ oscillations. If the
mass eigenstates are m1 and m2 , respectively, and if the vacuum mixing
angle is θ, the four-level equation of motion in vacuum is
    
νe ∆m c2θ ∆m s2θ 0 µBT νe
ν  ∆ s −∆m c2θ µBT 0   
     νµ 
i∂t  µ  =  m 2θ  , (8.57)
 νe   0 µBT ∆m c2θ ∆m s2θ   ν e 
νµ µBT 0 ∆m s2θ −∆m c2θ νµ
where ∆m ≡ (m22 − m21 )/4p, c2θ ≡ cos 2θ, and s2θ ≡ sin 2θ. Any of νe ,
νµ , ν e , and ν µ can oscillate into any of the others.
Including medium effects further complicates this matrix because
the energies of νe and νµ are shifted by different amounts, and each
is
√ shifted in 1the opposite directions
√ from its antiparticle. With Vνe =
2GF (ne − 2 nn ) and Vνµ = 2GF (− 12 nn ) the matrix becomes
 
∆m c2θ + Vνe ∆m s2θ 0 µBT
 ∆m s2θ −∆m c2θ + Vνµ µBT 0 
 
  . (8.58)
 0 µBT ∆m c2θ − Vνe ∆m s2θ 
µBT 0 ∆m s2θ −∆m c2θ − Vνµ
If there are density gradients (solar neutrinos!) one may have resonant
magnetic conversions between, say, νe and ν µ where the barrier to spin-
precessions caused by the mass difference is compensated by matter
effects, i.e. one may have resonant spin-flavor oscillations (Akhmedov
1988a,b; Barbieri and Fiorentini 1988; Lim and Marciano 1988).
For Dirac neutrinos, transitions to antineutrinos are not possible
and so one needs to consider oscillations separately in the four-level
308 Chapter 8
system (νeL , νµL , νeR , νµR ) and (ν Le , ν Lµ , ν R R

e , ν µ ). Moreover, one may have
both diagonal and transition magnetic moments so that for neutrinos
the r.h.s. of the equation of motion is
  
∆m c2θ + Vνe ∆m s2θ µee BT µµe BT νeL
 ∆m s2θ −∆m c2θ + Vνµ µeµ BT µµµ BT   νµL 
 
 
  , (8.59)
 µee BT µµe BT ∆m c2θ ∆m s2θ   νeR 
µeµ BT µµµ BT ∆m s2θ −∆m c2θ νµR
where the right-handed neutrinos do not experience a medium-induced
energy shift.
For certain values of the parameters the oscillations between two
states may dominate which are then described by a certain 2 × 2 sub-
matrix. Then the general treatment is much like the flavor mixing
problems studied earlier. In general, a much richer collection of solu-
tions obtains—for a review see Pulido (1992).
Electric transition moments can also play a role, and precessions
in electric fields may be important. Majorana neutrinos have either
electric or magnetic transition moments, but not both as long as CP
remains conserved. Still, even electric transition moments would lead
to spin-flavor oscillations in macroscopic magnetic fields.
8.4.4 Twisting Magnetic Fields

The problem of spin oscillations is further complicated by the possibility
that the magnetic field changes its direction along the neutrino trajec-
tory, i.e. that it may be “twisting” (Aneziris and Schechter 1991). In
this case the equation of motion of a two-level spin-precession problem
is based on the Hamiltonian
* +
BL BT e−iϕ
H=µ , (8.60)
BT eiϕ −BL
where the three parameters BL , BT , and ϕ vary along the neutrino
trajectory. The transverse field strength BT is a physical magnetic
field while the longitudinal one BL is an effective magnetic field which
represents the neutrino mass difference (spin-flavor oscillations!) and
refractive medium effects. As long as BT maintains its direction along
the trajectory, the phase ϕ can be globally chosen to be zero.
If ϕ varies along the trajectory so that the neutrino experiences it to
be a function of time, it is useful to transform the equation of motion to
a coordinate system which corotates with BT so that in the new frame
one is back to the old situation of a fixed direction for BT (Smirnov
1991). This transformation is achieved by ν # = U ν where ν represents

a two-level wave function and
* +
e−iϕ/2 0
U= , (8.61)
0 eiϕ/2
so that the spin-up and down states acquire opposite, time-dependent
phases. However, because the transformation does not mix spin up with
spin down, the transition rate between the two levels is the same in both
coordinate systems. The equation of motion i∂t ν # = H # ν # involves the
Hamiltonian
* + * +
# BL BT 1 −1 0
H =µ + ϕ̇ . (8.62)
BT −BL 2 0 1
Put another way, the effective longitudinal magnetic field in the rotating
frame is BL# = BL − 12 ϕ̇/µ while BT remains unchanged. |BL# | can be less
or larger than |BL | so that the transition rate between the two levels
can be increased or decreased by a twist. BL may even be cancelled
entirely, enabling resonant oscillations.
For a systematic discussion of the full four-level spin-flavor problem
see, for example, Akhmedov, Petcov, and Smirnov (1993).
Chapter 9
Oscillations
of Trapped Neutrinos
In a supernova (SN) core or in the early universe neutrinos scatter on

the background medium and on each other. Therefore, oscillations of
mixed neutrinos are frequently interrupted, leading to flavor equilib-
rium if there is enough time. A Boltzmann-type kinetic equation is
derived that accounts simultaneously for neutrino oscillations and col-
lisions for arbitrary neutrino degeneracy. It is applied to a SN core,
yielding an estimate of the time scale to reach flavor equilibrium. On
the basis of the SN 1987A neutrino observations a limit on the mixing
of sequential neutrinos with a hypothetical sterile flavor is derived.
9.1 Introduction
So far oscillations were discussed in the context of “beam experiments”
where neutrinos of a known flavor are produced at a certain location,
for example in a power reactor or in the Sun, then propagate over a
distance, and are then detected with a device that allows one to dis-
tinguish between different flavors. It was assumed that there were no
interactions between the production and detection point, with the pos-
sible exception of decays, allowing one to treat neutrino oscillations as
a simple propagation phenomenon fully analogous to light propagation
in an optically active medium.
This approach is not adequate for the early universe and a young SN
core. In both cases there are frequent neutrino collisions which affect
the free evolution of the phases. The impact of these collisions can
be understood if one assumes for the purpose of illustration that one
310
Oscillations of Trapped Neutrinos 311
neutrino flavor (say νe ) scatters with a rate Γ while the other (say νµ )
does not. An initial νe will begin to oscillate into νµ . The probability
for finding it in one of the two flavors evolves as previously discussed
and as shown in Fig. 9.1 (dotted line). However, in each collision the
momentum of the νe component of the superposition is changed, while
the νµ component remains unaffected. Thus, after the collision the two
flavors are no longer in the same momentum state and so they can no
longer interfere: each of them begins to evolve separately. This allows
the remaining νe to develop a new coherent νµ component which is made
incoherent in the next collision, and so forth. This process will come
into equilibrium only when there are equal numbers of νe ’s and νµ ’s.
This decoherence effect is even more obvious when one includes the
possibility of νe absorption and production by charged-current reactions
νe n ↔ pe. Because of oscillations an initial νe is subsequently found
to be a νµ with an average probability of 12 sin2 2θ (mixing angle θ)
and as such cannot be absorbed, or only by the reaction νµ n ↔ pµ
if it has enough energy. The continuous emission and absorption of
νe ’s spins off a νµ with an average probability of 12 sin2 2θ in each
collision! Chemical relaxation of the neutrino flavors will occur with
an approximate rate 12 sin2 2θ Γ where Γ is a typical weak interaction
rate for the ambient physical conditions. An initial νe population
turns into an equal mixture of νe ’s and νµ ’s as shown schematically
in Fig. 9.1 (solid line).
Fig. 9.1. Neutrino oscillations with collisions (solid line). In the absence
of collisions and for a single momentum one obtains periodic oscillations
(dotted line), while for a mixture of energies the oscillations are washed out
by “dephasing” (dashed line).
312 Chapter 9
A certain “damping” of flavor oscillations occurs even without colli-

sions when the neutrinos are not monochromatic because then different
modes oscillate with different frequencies. This “dephasing” effect was
shown in Fig. 8.4 and is repeated in Fig. 9.1 (dashed line). While the
dephasing washes out the oscillation pattern, it does not lead to flavor
equilibrium: the probability for νe ends at a constant of 1 − 12 sin2 2θ.
For the rest of this chapter “damping of oscillations” never refers to
this relatively trivial dephasing effect.
The interplay of collisions and oscillations leads to flavor equilibrium
between mixed neutrinos. In a SN core the concentration of electron
lepton number is initially large so that the νe form a degenerate Fermi
sea. The other flavors νµ and ντ are characterized by a thermal distri-
bution at zero chemical potential. However, if they mix with νe they
will achieve the same large chemical potential. In a SN core heat and
lepton number are transported mostly by neutrinos; the efficiency of
these processes depends crucially on the degree of neutrino degeneracy
for each flavor. Therefore, it is of great interest to determine the time
it takes a non-νe flavor to equilibrate with νe under the assumption
of mixing (Maalampi and Peltoniemi 1991; Turner 1992; Pantaleone
1992a; Mukhopadhyaya and Gandhi 1992; Raffelt and Sigl 1993).
Moreover, if νe mixes with a sterile neutrino species, conversion into
this inert state leads to the loss of energy and lepton number from the
inner core of a SN. The observed SN 1987A neutrino signal may thus
be used to constrain the allowed range of masses and mixing angles
(Kainulainen, Maalampi, and Peltoniemi 1991; Raffelt and Sigl 1993;
see also Shi and Sigl 1994).
These applications are discussed in Sects. 9.5 and 9.6 below. A
simple estimate of the rate of flavor conversion and the emission rate
of sterile neutrinos from a SN core requires not much beyond the ap-
proximate rate 12 sin2 2θ Γ. However, a proper kinetic treatment of the
evolution of a neutrino ensemble under the simultaneous action of os-
cillations and collisions is an interesting theoretical problem in its own
right. Notably, it is far from obvious how to treat degenerate neutri-
nos in a SN core because the different flavors will suffer different Pauli
blocking factors. Does this effect break the coherence between mixed
flavors in neutral-current collisions?
The bulk of this chapter is devoted to the derivation and discus-
sion of a general kinetic equation for mixed neutrinos (Dolgov 1981;
Rudzsky 1990; Raffelt, Sigl, and Stodolsky 1993; Sigl and Raffelt 1993).
This equation provides a sound conceptual and quantitative framework
for dealing with various aspects of coherent and incoherent neutrino
interactions with a medium, whether the neutrinos are degenerate or

not. As a free spin-off it will provide a proper formalism to deal with
the refractive effects of neutrinos propagating in a bath of neutrinos,
a problem that is of interest in the early universe where interactions
among neutrinos produce the dominant medium effect, and in the neu-
trino flow from a SN core where the density of neutrinos is larger than
that of nonneutrino background particles (Sect. 11.4).
Besides neutrino flavor oscillations, magnetically induced spin or
spin-flavor oscillations are also of potential interest because in super-
novae and the early universe strong magnetic fields are believed to exist.
Spin relaxation (the process of populating the r.h. degrees of freedom by
the simultaneous action of spin oscillations and collisions) is a very sim-
ilar problem to that of achieving chemical equilibrium between different
flavors which is discussed here. Therefore, similar kinetic methods can
be applied (Enqvist, Rez, and Semikoz 1995 and references therein).
9.2 Kinetic Equation for Oscillations and

Collisions
9.2.1 Stodolsky’s Formula
The loss of coherence between mixed neutrinos in collisions cannot be

properly understood on the amplitude level because it is not the am-
plitudes, but only their relative coherence that is damped—the flavor
states “decohere,” they do not disappear. This is different from the
decay of mixed particles where one of the amplitudes can be viewed
as decreasing exponentially so that the total number of particles is
not conserved. A natural description of decoherence is achieved by a
density matrix as in Eq. (8.8); for a two-flavor mixing problem it was
expressed in terms of a polarization vector P according to Eq. (8.18),
ρ = 12 (1 + P · σ). In the weak interaction basis its diagonal elements are
the probabilities for measuring ν in, say, the νe or νµ state, respectively,
while the off-diagonal elements contain relative phase information.
In a two-level system, the length of the polarization vector measures
the degree of coherence: length 1 corresponds to a pure state, shorter
P ’s to some degree of incoherence, and length zero is the completely
mixed or incoherent state (Stodolsky 1987). In this latter case ρ is
proportional to the unit matrix which is invariant under a transforma-
tion of basis. This state of “chemical” or “flavor equilibrium” has no
off-diagonal elements in any basis.
314 Chapter 9
Even coherent or partially coherent density matrices can be diago-

nalized in some basis. The interactions with the background medium
will also be diagonal in some basis; for neutrinos, this is the weak inter-
action basis. If the density matrix and the interactions are diagonal in
the same basis, there is no decoherence effect. For example, a density
matrix diagonal in the weak interaction basis implies that there is no
relative phase information between, say, a νe and a νµ and so collisions
which affect νe ’s and νµ ’s separately have no impact on the density
matrix. However, a density matrix which is diagonal in the mass basis,
assumed to be different from the weak interaction basis, will suffer a
loss of coherence in the same medium.
All told, the loss of coherence is given by a shrinking of the length
of P. More precisely, only the component PT is damped which repre-
sents the part “transverse” to the interaction basis, i.e. which in the
interaction basis represents the off-diagonal elements of ρ. Thus, in the
presence of collisions the evolution of P is given by (Stodolsky 1987)
Ṗ = V × P − DPT . (9.1)
The first part is the previous precession formula, except that here a
temporal evolution is appropriate since one has in mind the evolution
of a spatially homogeneous ensemble rather than the spatial pattern of
a stationary beam. The “magnetic field” V is
 
sin 2θ
2π  
V=  0 , (9.2)
tosc
cos 2θ
where θ is the mixing angle in the medium and tosc the oscillation
period. They are given in terms of the neutrino masses and momen-
tum, the vacuum mixing angle, and the medium density by Eqs. (8.30)
and (8.32) where strictly speaking the neutrino energy is to be replaced
by its momentum as we have turned to temporal rather than spatial
oscillations. The damping parameter D is determined by the scattering
amplitudes on the background.
The evolution described by Eq. (9.1) is a precession around the
“magnetic field” V, combined with a shrinking of the length of P to
zero. This final state corresponds to ρ = 12 where both flavors are
equally populated, and with vanishing coherence between them. For
1
D = t−1osc and sin 2θ = 2 the evolution of the flavors is shown in Fig. 9.1
(solid line). Ignoring the wiggles in this curve it is an exponential as
can be seen by multiplying both sides of Eq. (9.1) with P/P 2 which
leads to Ṗ /P = −D (PT /P )2 with P = |P| and PT = |PT |. For a

small collision rate relative to the oscillation frequency one may use
the precession-averaged PT which is found by taking the transverse
part of the projection of P on V. Elementary geometry in Fig. 8.2
yields $PT %/P = cos 2θ sin 2θ so that $Ṗ /P % = −D cos2 2θ sin2 2θ. If
one neutrino interacts while the other is sterile (for example νµ with
regard to charged-current absorption) D is half the collision rate Γ of
the active flavor (Stodolsky 1987). For small mixing angles one recovers
the previous intuitive relaxation rate 12 sin2 2θ Γ.
Equation (9.1) is based on a single-particle wave function picture
of neutrino oscillations and thus it is applicable if effects nonlinear in
the neutrino density matrices can be ignored. It does not allow one to
include the effect of Pauli blocking of neutrino phase space, which is
undoubtedly important in a SN core where the νe Fermi sea is highly
degenerate. In this case it is rather unclear what one is supposed to
use for the damping parameter D. If νe and νµ scatter with equal
amplitudes, there is no damping at all because the collisions do not
distinguish between flavors, preserving the coherence between them.
Does this remain true if νe collisions are Pauli blocked by their high
Fermi sea while those of νµ are not? Such and other related questions
can be answered if one abandons a single-particle approach to neu-
trino oscillations, i.e. if one moves to a field-theoretic framework which
includes many-body effects from the start.
9.2.2 Matrix of Densities

Which quantity is supposed to replace the previous single-particle den-
sity matrix as a means to describe a possibly degenerate neutrino
ensemble? For unmixed neutrinos the relevant observables are time-
dependent occupation numbers fp for a given mode p of the neu-
trino field. They are given as expectation values of number operators
np = a†p ap where ap is a destruction operator for a neutrino in mode p
and a†p the corresponding creation operator. The expectation value is
with regard to the state | % of the entire ensemble. For several flavors
it is natural to generalize the fp ’s to matrices ρp = ρ(p) of the form49
49
Strictly speaking ρ(p) is defined by $a†j (p)ai (p! )% = (2π)3 δ (3) (p − p! )ρij (p) and
similar for ρ(p). Therefore, the expectation values in Eq. (9.3) diverge because they
involve an infinite factor (2π)3 δ (3) (0) which is related to the infinite quantization
volume necessary for continuous momentum variables. In practice, this factor al-
ways drops out of final results so that one may effectively set (2π)3 δ (3) (0) equal to
unity.
316 Chapter 9
(Dolgov 1981)
' ( ' (
ρij (p) = a†j (p)ai (p) and ρij (p) = b†i (p)bj (p) , (9.3)
where ai (p) and a†i (p) are the destruction and creation operators for
neutrinos of flavor i in mode p while b is for antineutrinos which other-
wise are referred to by overbarred quantities. The reversed order of the
flavor indices in the definition of ρ(p) guarantees that both matrices
transform in the same way under a unitary transformation in flavor
space. Also, for brevity $ | . . . | % is always written as $. . .%.
The diagonal elements of ρp and ρp are the usual occupation num-
bers while the off-diagonal ones represent relative phase information.
In the nondegenerate limit, up to a normalization ρp plays the role of
the previously defined single-particle density matrix. Therefore, the
ρp ’s and ρp ’s are well suited to account simultaneously for oscillations
and collisions. In fact, one can argue that a homogeneous neutrino
ensemble is completely characterized by these “matrices of densities”
(Sigl and Raffelt 1993). It remains to derive an equation of motion
which in the appropriate limits should reduce to the previous preces-
sion equation, to a Boltzmann collision equation, and to Stodolsky’s
damping equation (9.1), respectively.
9.2.3 Free Evolution: Flavor Oscillations

The creation and annihilation operators which appear in the definition
of ρp are the time-dependent coefficients of a spatial Fourier expansion
of the neutrino field [notation dp ≡ d3 p/(2π)3 ]
) * +
Ψ(t, x) = dp ap (t)up + b†−p (t)v−p eip·x . (9.4)
More precisely, ap is an annihilation operator for negative-helicity neu-

trinos of momentum p while b†p is a creation operator for positive-
helicity antineutrinos. The Dirac spinors up and vp refer to mass-
less negative-helicity particles and positive-helicity antiparticles, re-
spectively; the spinor normalization is taken to be unity. For n flavors,
ap and b†p are column vectors of components ai (p) and b†i (p), respec-
tively. They satisfy the anticommutation relations {ai (p), a†j (p" )} =
{bi (p), b†j (p" )} = δij (2π)3 δ (3) (p − p" ).
In the massless limit and when only left-handed (l.h.) interactions
are present one may ignore the right-handed (r.h.) field entirely. How-
ever, in order to include flavor mixing one needs to introduce a n × n
mass matrix M which is nondiagonal in the interaction basis. Even

in this case the “wrong” helicity states will be ignored because in the
ultrarelativistic regime spin-flip reactions are suppressed by an approx-
imate factor (mν /2Eν )2 ' 1. In this limit lepton number violating
effects from possible Majorana masses are also ignored.
In the absence of interactions Ψ satisfies the free Dirac equation, im-
plying ap (t) = ap (0) exp(−iΩ0p t) and bp (t) = bp (0) exp(−iΩ0p t), where
, -1/2
Ω0p ≡ p2 + M 2 (9.5)
is a n × n matrix of “vacuum oscillation frequencies.” In the mass basis

it has only diagonal elements which are the energies Ei = (p2 + m2i )1/2 .
Therefore, one may use
) n *
. +
H0 = dp a†i (p)Ω0ij (p)aj (p) + b†j (p)Ω0ij (p)bi (p) (9.6)
i,j=1
as a free neutrino Hamiltonian.

In order to find the evolution of ρp and ρp one needs to study the
equations of motion of the n × n operator matrices
ρ̂ij (p, t) ≡ a†j (p, t)ai (p, t) and ρ̂ij (p, t) ≡ b†i (p, t)bj (p, t) . (9.7)
With a Hamiltonian H their evolution is given by Heisenberg’s equa-

tion,
i∂t ρ̂ = [ρ̂, H] , (9.8)
and similar for ρ̂p . With H = H0 from Eq. (9.6) one finds
i∂t ρ̂p = [Ω0p , ρ̂p ] and i∂t ρ̂p = −[Ω0p , ρ̂p ] . (9.9)
Taking expectation values on both sides yields equations of motion

for ρp and ρp . For two flavors one may use the representation Ω0p =
ωp0 + 12 Vp · σ and ρp = 12 fp (1 + Pp · σ). Then Eq. (9.9) leads to the
precession formulas Ṗ = V × P and P ˙ = −V × P .
p p p p p p
9.2.4 Interaction with a Background Medium

Interactions with a medium are introduced by virtue of a general in-
teraction Hamiltonian Hint (B, Ψ) which is a functional of the neutrino
field Ψ and a set B of background fields; specific cases will be discussed
in Sects. 9.3 and 9.4 below. The equation of motion for ρ̂p is found from
318 Chapter 9
Heisenberg’s equation with H = H0 +Hint . Taking an expectation value

with regard to the initial state yields
* + '* +(
ρ̇p (t) = −i Ω0p , ρp (t) + i Hint (B(t), Ψ(t)), ρ̂p (t) , (9.10)
and an analogous equation for ρp (t). These equations are exact, but
they are not a closed set of differential equations for the ρp and ρp . To
this end one needs to perform a perturbative expansion.
To first order one may set the interacting fields B(t) and Ψ(t) on the
r.h.s. of Eq. (9.10) equal to the free fields50 B0 (t) and Ψ0 (t). Under the
assumption that the original state contained no correlations between
the neutrinos and the background the expectation value factorizes into
a medium part and a neutrino part. With Wick’s theorem and ignoring
fast-varying terms such as b† b† it can be reduced to an expression which
contains only ρp ’s and ρp ’s. The result gives the forward-scattering or
refractive effect of the interaction.
To include nonforward collisions one needs to go to second order
in the perturbation expansion. At a given time t a general operator
ξ(t) = ξ(B(t), Ψ(t)) which is a functional of B and Ψ is to first order
) t * +
ξ(t) = ξ0 (t) + i dt" Hint
0
(t − t" ), ξ0 (t) , (9.11)
0
0
where ξ0 and Hint are functionals of the freely evolving fields B0 (t)
and Ψ0 (t). Applying this general iteration formula to the operator
ξ = [Hint (B, Ψ), ρ̂p ] which appears on the r.h.s. of Eq. (9.10) one
arrives at
* + '* +(
ρ̇p (t) = −i Ω0p , ρp (t) + i 0
Hint (t), ρ̂0p
) t '* * ++(
− dt" 0
Hint (t − t" ), Hint
0
(t), ρ̂0p , (9.12)
0
and similar for ρp (t). The second term on the r.h.s. is the first-order
refractive part associated with forward scattering. The second-order
term contains both forward- as well as nonforward-scattering effects.
50
These free operators are the solutions of the equations of motion in the absence
of Hint . However, internal interactions of the medium such as nucleon-nucleon
scattering are not excluded. Moreover, Ψ(0) = Ψ0 (0) etc. are taken as initial
conditions for the interacting fields. Also, the mass term is ignored in the definition
of Ψ0 ; its effect is included only in the first term on the r.h.s. of Eq. (9.10), the
“vacuum oscillation term.” Therefore, the free creation and annihilation operators
vary as a0j (p, t) = aj (p, 0)e−ipt etc. for all flavors with p = |p|. This implies that
the operators ρ̂0 (p) and ρ̂0 (p), which are constructed from the free a’s and b’s, are
time independent.
Because all operators on the r.h.s. of Eq. (9.12) are free the expec-
tation values in the first- and second-order term factorize between the
neutrinos and the medium. This leads one to equations for ρ̇p (t) and
ρ̇p (t) which on the r.h.s. involve only ρp (t) and ρp (t) as well as $ρ̂0p %
and $ρ̂0p % besides expectation values of B operators.
The interactions described by Hint are taken as individual, isolated
collisions where the neutrinos go from free states to free states as in
ordinary scattering theory. The duration of one collision (the inverse
of a typical energy transfer) is assumed to be small relative to the
time scale over which the density matrices vary substantially, i.e. small
relative to the oscillation time and the inverse collision frequency. Phys-
ically this amounts to the restriction that the neutrino collision rate is
small enough that multiple-scattering effects can be ignored. Further,
it is assumed that the medium is not changed much by the interactions
with the neutrino ensemble, allowing one to neglect evolution equations
for the medium variables which can thus be taken to be externally pre-
scribed, usually by conditions of thermal equilibrium. If the medium
is not stationary it is assumed that the time scale of variation is large
compared to the duration of typical neutrino-medium collisions.
One may then choose the time step of iteration t in Eq. (9.12)
both small relative to the evolution time scale and large relative to the
duration of one collision. Under these circumstances the time integral
can be extended to infinity while setting ρp (t) equal to ρp (0) = $ρ̂0p %.
This leads to
* + '* +(
ρ̇p (0) = −i Ω0p , ρp (0) + i 0
Hint (0), ρ̂0p
) +∞ '* * ++(
0 0
− 1
2
dt Hint (t), Hint (0), ρ̂0p , (9.13)
−∞
/ /
and similar for ρp . Here, 0∞ dt $. . .% was replaced by 12 −∞ +∞
dt $. . .%.
The difference between these expressions corresponds to a principle-
part integral which leads to a second-order correction to the refractive
term which is ignored. In the form of Eq. (9.13) the time integral leads
to energy conservation in individual collisions.
An explicit evaluation of the r.h.s. of Eq. (9.13) for a given interac-
tion model yields the desired set of differential equations for the ρp ’s
and ρp ’s at time t = 0. It will be valid at all times if the correlations
built up by neutrino collisions are “forgotten” before the next colli-
sion occurs. This assumption corresponds to “molecular chaos” in the
derivation of the usual Boltzmann equation.
320 Chapter 9
9.3 Neutral-Current Interactions

9.3.1 Hamiltonian
In order to make Eq. (9.13) explicit one must use a specific model for
the interactions between neutrinos and the medium. To this end I begin
with fermions which interact by virtue of an effective neutral-current
(NC) Hamiltonian,
GF . ) 3
HNC =√ d x Baµ (x) Ψ(x)Ga γµ (1 − γ5 )Ψ(x) . (9.14)
2 a
Here, Baµ typically is also a bilinear of the form ψ a γ µ ψa or ψ a γ5 γ µ ψa
with a Dirac field ψa which describes fermions of the medium. It is
assumed that all neutrino flavors scatter on a given species a in the
same way apart from overall factors which are given as a hermitian
n × n matrix Ga of dimensionless coupling constants. In the absence of
flavor-changing neutral currents it is diagonal in the weak interaction
basis.
As a concrete example consider the νe and νµ flavors in a medium
of ultrarelativistic electrons which may be classified into a l.h. and a
µ
r.h. “species,” a = L or R, so that BL,R = 12 ψ e γ µ (1 ∓ γ5 )ψe . With the
standard-model couplings given in Sect. 6.7.1 one finds in the weak
interaction basis GL = 2 sin2 θW + σ3 and GR = 2 sin2 θW .
For the calculations it is convenient to write Eq. (9.14) in momen-
tum space,
GF . )
HNC =√ dp dp" Baµ (p − p" ) Ψp Ga γµ (1 − γ5 )Ψp! , (9.15)
2 a
/
where Baµ (∆) = d3 x Baµ (x)e−i∆·x is the Fourier transform of Baµ (x)
and Ψp = ap up + b†−p v−p in terms of the annihilation and creation
operators of Eq. (9.4).
A special case of NC interactions are those among the neutrinos
themselves with a Hamiltonian that is quartic in Ψ. In momentum
space these “self-interactions” are given by
GF )
HS = √ dp dp" dq dq" (2π)3 δ (3) (p + q − p" − q" )
2
× Ψq GS γ µ (1 − γ5 )Ψq! Ψp GS γµ (1 − γ5 )Ψp! . (9.16)
In the standard model with three sequential neutrino families GS is the
3 × 3 unit matrix. For the evolution of a normal and a hypothetical
sterile flavor one would have GS = diag(1, 0).
9.3.2 Neutrino Refraction

As a first simple application one may recover neutrino refraction by
a medium that was previously discussed in Sect. 6.7.1. An explicit
evaluation of the second term of Eq. (9.13) with Hint = HNC yields
.
iρ̇ = [Ω0p , ρp ] + na [Ga , ρp ] , (9.17)
a
where na ≡ $Baµ %Pµ /P0 where P is the neutrino four momentum and
thus P/P0 their four velocity. In an isotropic medium the spatial parts
of $Baµ % vanish so that na is the number density of fermions a. If the
medium is unpolarized, axial currents do not contribute.
In the standard model with the coupling constants of Appendix B
one finds for an isotropic, unpolarized medium of protons, neutrons,
and electrons,
* √ +
iρ̇p = (Ω0p + 2 GF N# ), ρp ,
* √ +
−iρ̇p = (Ω0p − 2 GF N# ), ρp , (9.18)
where N# = diag(ne , 0, 0) in the flavor basis (electron density ne ). For

two flavors this is equivalent to the previous precession formula. No-
tably, the ν and ν oscillation frequencies are shifted in opposite direc-
tions relative to the vacuum energies.
Neutrino-neutrino interactions make an additional contribution to
the refractive energy shifts, i.e. to the first-order term in Eq. (9.13).
After the relevant contractions one finds51 (Sigl and Raffelt 1993)
√ ) 0 * +1
ΩSp = 2 GF dq GS (ρq − ρq )GS + GS Tr (ρq − ρq )GS .
(9.19)
ΩSp is given by the same formula with ρq and ρq interchanged. The

trace expression implies the well-known result that neutrinos in a bath
of their own flavor experience twice the energy shift relative to a bath
of another flavor.
The early universe is essentially matter-antimatter symmetric so
that higher-order terms to the refractive index must be included as
discussed in Sect. 6.7.2; see also Sigl and Raffelt (1993). In stars,
51
If the neutrino ensemble is not isotropic one has to include a factor (1 − cos θpq )
under the integral, where θpq is the angle between p and q.
322 Chapter 9
neutrinos are important only in young SN cores where one may ignore
antineutrinos. With the total neutrino matrix of densities
)
ρ≡ dp ρp , (9.20)
and with GS = 1 in the standard model, √ the neutrino contribution

S
to the refractive energy shift is Ωp = 2 GF ρ. The trace term was
dropped because it does not contribute to the commutator in the equa-
tion of motion. The diagonal entries of ρ are the neutrino densities.
However, in the presence of mixing and oscillations ρ also has off-
diagonal elements, i.e. there are “off-diagonal refractive indices” as first
realized by Pantaleone (1992b).
In a SN core the complete first-order equation of motion for ρp is
then
* + √ * +
iρ̇p = Ω0p , ρp + 2 GF (N# + ρ), ρp , (9.21)
which is intrinsically nonlinear. Interestingly, if one integrates both
sides over dp one obtains an equation for ρ which is linear as the neu-
trino term drops out from the commutator. Therefore, even though
individual modes of the neutrino field oscillate differently in the pres-
ence of other neutrinos, the instantaneous rate of change of the overall
flavor polarization is as if they were absent.
In a SN core the refractive effects are dominated by nonneutrino
particles, notably by electrons. However, above the neutrino sphere
the flow of neutrinos itself represents a particle density exceeding that
of the background medium (Sect. 11.4). Also, the medium of the early
universe is dominated by neutrinos so that self-interactions and the
corresponding nonlinearities of the neutrino flavor oscillations must be
carefully included. For recent studies of primordial neutrino oscillations
see Samuel (1993), Kostolecký, Pantaleone, and Samuel (1993), and
references to the earlier literature given there.
9.3.3 Kinetic Terms

The refractive term (first order) of Eq. (9.13) is just a sum over different
medium components, whereas the collision term (second order) in gen-
eral contains interference terms between different target species. How-
ever, if they are uncorrelated, corresponding to $Baµ Bbν % = $Baµ %$Bbν %
for a *= b, these interference terms only contribute to second-order
forward-scattering effects which are neglected. The collision term is
then an incoherent sum over all target species so that in the following
one may suppress the subscript a for simplicity.
After a lengthy but straightforward calculation one arrives at the

NC collision term (Sigl and Raffelt 1993)
) *
ρ̇p,coll = 1
2
dp" WP ! ,P Gρp! G(1 − ρp ) − WP,P ! ρp G(1 − ρp! )G
+
+ W−P ! ,P (1 − ρp )G(1 − ρp! )G − WP,−P ! ρp Gρp! G + h.c. ,
(9.22)
where P and P " are neutrino four momenta with physical (positive) en-
ergies P0 = |p| and P0" = |p" |. The nonnegative transition probabilities
WK ! ,K = W (K " , K) are Wick contractions of medium operators of the
form
W (K " , K) = 18 G2F S µν (K " − K) Nµν (K " , K) , (9.23)
where K and K " correspond to neutrino four-momenta with K0 and K0"
positive or negative. The “medium structure function” is
) +∞
µν
S (∆) ≡ dt ei∆0 t $B µ (t, ∆)B ν (0, −∆)% , (9.24)
−∞
where the energy transfer ∆0 can be both positive and negative. In the
ultrarelativistic limit the neutrino tensor can be written as
N µν = 12 (U µ U "ν + U "µ U ν − U · U " g µν − i,µναβ Uα Uβ" ), (9.25)
where U ≡ K/K 0 and U " ≡ K " /K 0 are the neutrino four velocities.
Therefore, N µν is an even function of K and K " . Note that the defini-
tion Eq. (9.25) differs slightly from the corresponding Eq. (4.17).
The first two terms of the collision integral Eq. (9.22) are due to
neutrino scattering off the medium. The positive term represents gains
from scatterings νp! → νp while the negative one is from losses by the
inverse reaction. The third and fourth expressions account for pair
processes, i.e. the creation or absorption of νp ν p! by the medium. The
pair terms are found by direct calculation or from the scattering ones
by “crossing,”
P → −P and ρp → (1 − ρp ) . (9.26)
For example, the reaction νp X → X " νp! transforms to X → X " ν p νp!
under this operation where X and X " represent medium configurations.
The collision integral for ρp is found by direct calculation or by
applying the crossing operation Eq. (9.26) to all neutrinos and antineu-
trinos appearing in Eq. (9.22). The neutrino gain terms then transform
to the antineutrino loss terms and vice versa.
324 Chapter 9
Equation (9.22) and the corresponding result for ρp were derived

by Sigl and Raffelt (1993) whose exposition I have closely followed. In
the nondegenerate limit where (1 − ρp ) → 1 it agrees with a kinetic
equation of Dolgov (1981) and Barbieri and Dolgov (1991). Moreover,
a similar equation was derived by Rudzsky (1990) which can be shown
to be equivalent to Eq. (9.22) in the appropriate limits.
The relatively complicated collision term that follows from the neu-
trino-neutrino Hamiltonian Eq. (9.16) has been worked out by Sigl and
Raffelt (1993). However, in a SN core the collisions of neutrinos with
each other are negligible relative to interactions with nucleons and elec-
trons.
In the limit of a single neutrino flavor, or several unmixed flavors,
the role of ρp is played by the usual occupation numbers fp while the
matrix G is unity, or the unit matrix. Then Eq. (9.22) is
) *
f˙p,coll = dp" WP ! ,P fp! (1 − fp ) − WP,P ! fp (1 − fp! )
+
+ W−P ! ,P (1 − fp )(1 − f p! ) − WP,−P ! fp f p! (9.27)
which is the usual Boltzmann collision integral. The main difference

to Eq. (9.22) is the appearance there of “nonabelian Pauli blocking
factors” which involve noncommuting matrices of neutrino occupation
numbers and coupling constants.
9.3.4 Recovering Stodolsky’s Formula

The damping of neutrino oscillations becomes particularly obvious in
the limit where a typical energy transfer ∆0 in a neutrino-medium in-
teraction is small relative to the neutrino energies themselves. This
would be the case for “heavy” and thus nonrelativistic background
fermions. Then pair processes may be ignored and neutrinos change
their direction of motion in a collision, but not the magnitude of their
momentum, which also implies W (P, P " ) ≈ W (P " , P ). If the neu-
trino ensemble is isotropic one then has ρp = ρp! under the integral
in Eq. (9.22). For the matrix structure of the collision term this leaves
2Gρp G − GGρp − ρp GG = −[G, [G, ρp ]] which puts the nature of the
collision term as a double commutator in evidence—see also Eq. (9.13).
One may define a total scattering rate for nondegenerate neutrinos of
momentum p by virtue of
)
Γp = dp" W (P " , P ) . (9.28)
In the present limit this yields a collision “integral”
ρ̇p,coll = − 12 Γp [G, [G, ρp ]] , (9.29)
where terms nonlinear in ρp have disappeared even though the neutri-

nos may still be degenerate.
Eq. (9.29) is more transparent in the case of two-flavor mixing where
one may write ρp = 12 fp (1 + Pp · σ) and G = 12 (g0 + G · σ). The
total occupation number fp is conserved while the polarization vector
is damped according to
Ṗp,coll = − 12 Γp G × (G × Pp ). (9.30)
The r.h.s. is a vector transverse to G, allowing one to write
Ṗp,coll = − 12 Γp |G|2 Pp,T . (9.31)
Thus one naturally recovers Stodolsky’s damping term Eq. (9.1) with
D = 12 Γp |G|2 . For νe and νµ and if one writes G = diag(gνe , gνµ ) in
the weak interaction basis D = 12 Γp (gνe − gνµ )2 . This representation
reflects that the damping of neutrino oscillations depends on the dif-
ference of the scattering amplitudes: D is the square of the amplitude
difference, not the difference of the squares. If one flavor does not
scatter at all, D is half the scattering rate of the active flavor.
Collisions thus lead to chemical equilibrium as discussed in the in-
troduction to this chapter and as shown in Fig. 9.1. However, the
simple exponential damping represented by Stodolsky’s formula can be
reproduced only in the limit of vanishing energy transfers in collisions,
an assumption which amounts to separating the neutrino momentum
degrees of freedom from the flavor ones. In a more general case the evo-
lution is more complicated. In particular, collisions usually lead to a
transient flavor polarization in an originally unpolarized ensemble if the
momentum degrees of freedom were out of equilibrium. Still, the neu-
trinos always move toward kinetic and chemical equilibrium under the
action of the collision integral Eq. (9.22) in the sense that the properly
defined free energy never increases (Sigl and Raffelt 1993).
9.3.5 Weak-Damping Limit

Even for two-flavor mixing the general form of the collision integral
Eq. (9.22) remains rather complicated. However, for the conditions
of a SN core one may apply two approximations which significantly
simplify the problem. First, one is mostly concerned with the evolution
326 Chapter 9
of νe ’s because initially they have a large chemical potential and thus

are far away from chemical equilibrium with the other flavors. Their
high degree of degeneracy implies that ν e ’s may be ignored and with
them all pair processes. Therefore, the evolution of νe ’s mixed with one
other flavor (standard or sterile) is given by
) *
ρ̇p = i[ρp , Ωp ] + 1
2
dp" WP ! P Gρp! G(1 − ρp )
+
−WP P ! ρp G(1 − ρp! )G + h.c. . (9.32)
The matrix of oscillation frequencies includes vacuum and first-order

medium contributions. With the momentum-dependent oscillation pe-
riod tosc it is
Ωp = (2π/tosc ) 12 vp · σ with vp ≡ (sp , 0, cp ). (9.33)
Here,
sp ≡ sin 2θp and cp ≡ cos 2θp (9.34)
with the momentum-dependent mixing angle in the medium θp .

The second approximation for the conditions of a SN core is the
weak-damping limit or limit of fast oscillations. It is easy to show that
for the relevant physical conditions 2π/tosc is typically much faster than
the scattering rate. Therefore, it is justified to consider density matrices
ρ2p averaged over a period of oscillation. While the ρp ’s are given by
four real parameters which are functions of time, the ρ2p ’s require only
two, for example the occupation numbers of the two mixed flavors. It
is straightforward to show that in the weak interaction basis
3 4 3 4
fpe 0 0 1
ρ2p = + 1
t
2 p
(fpe − fpx ) , (9.35)
0 fpx 1 0
where tp ≡ tan 2θp = sp /cp , fpe is the occupation number of νe , not of
electrons, and fpx refers to a standard or sterile flavor νx .
One way of looking at the weak-damping limit is that between col-
lisions neutrinos are best described by “propagation eigenstates,” i.e.
in a basis where the ρ2p are diagonal. Then the matrix of coupling con-
stants G is no longer diagonal and so flavor conversion is understood as
the result of “flavor-changing neutral currents” where “flavor” refers to
the propagation eigenstates. However, because in general the effective
mixing angle is a function of the neutrino momentum one would have
to use a different basis for each momentum, an approach that compli-
cates rather than simplifies the equations. Therefore, it is easiest and
physically most transparent to work always in the weak interaction

basis.
In order to derive an equation of motion for ρ2p one evaluates the
collision term in Eq. (9.32) by inserting ρ2p ’s under the integral. Ex-
panding the result in Pauli matrices leads to an expression of the form
1
(a + Ap · σ). In general, the polarization vector Ap produced by the
2 p
collision term is not parallel to vp = (sp , 0, cp ) because the collision
term couples modes with different mixing angles. However, the assumed
fast oscillations average to zero the Ap component perpendicular to vp .
Therefore, the r.h.s. of Eq. (9.32) is of the form 12 [ap +(vp ·Ap ) (vp ·σ)].
With a matrix of coupling constants in the weak basis of
3 4
ge 0
G= (9.36)
0 gx
the collision integral Eq. (9.32) becomes explicitly
) 0 * +
f˙px = 1
4
dp" w(νpx → νpx! ) (4 − s2p )gx2 + (2 − s2p )tp tp! ge gx
* +
+ w(νpe → νpx! ) −s2p gx2 − s2p tp tp! ge gx
* +
+ w(νpx → νpe ! ) s2p ge2 − (2 − s2p )tp tp! ge gx
* +1
+ w(νpe → νpe ! ) s2p ge2 + s2p tp tp! ge gx , (9.37)
where
w(νpa → νpb ! ) ≡ WP ! P fpb ! (1 − fpa ) − WP P ! fpa (1 − fpb ! ) (9.38)
(Raffelt and Sigl 1993). The equation for fpe is the same if one exchanges
e ↔ x everywhere. In the absence of mixing sp = tp = 0, leading to
the usual collision integral for each species separately.
If νx is neither νµ nor ντ but rather some hypothetical sterile species
its coupling constant is gx = 0 by definition. In this case the collision
integral simplifies to
) *
f˙px = 14 s2p ge2 dp" WP ! P fpe! (2 − fpe − fpx )
+
− WP P ! (fpe + fpx )(1 − fpe! ) . (9.39)
If the νe stay approximately in thermal equilibrium, detailed balance
yields
) * +
f˙px = 14 s2p ge2 dp" WP ! P fpe! (1 − fpx ) − WP P ! fpx (1 − fpe! ) . (9.40)
If in addition the mixing angle is so small that the νx freely escape one
may
/
set fpx = 0 on the r.h.s. so that the integral expression becomes
dp" WP ! P fpe! .
328 Chapter 9
9.3.6 Small Mixing Angle

In practice the mixing angle is usually small, allowing for substantial
further simplifications. In this limit the approach to flavor equilibrium
is much slower than that to kinetic equilibrium for each flavor separately
(νx is taken to be one of the active flavors νµ or ντ ). Therefore, each
flavor is characterized by a Fermi-Dirac distribution so that it is enough
to specify the total number density nνx rather than the occupation
numbers of individual modes. Integrating Eq. (9.37) over all modes,
using detailed balance to lowest order in s2p , and with tp = sp one finds
for the evolution of the νx number density
) *
ṅνx = 1
4
dp dp" WP P ! (gx sp − ge sp! )2 fpe (1 − fpx! )
+
− (ge sp − gx sp! )2 fpx (1 − fpe! ) , (9.41)
and a similar equation for ṅνe . Together with the condition of β equi-
librium, µn − µp = µe − µνe , that of charge neutrality, np = ne , and the
conservation of the trapped lepton number, d(ne + nνe + nνx )/dt = 0,
these equations represent differential equations for the chemical poten-
tials µνx (t) and µνe (t) if the temperature is fixed.
9.3.7 Flavor Conversion by Neutral Currents?

Next I turn to the conceptually interesting question whether flavor
conversion (or the damping of neutrino oscillations) is possible by NC
collisions alone. Considering only standard flavors the matrix of cou-
pling constants G is then proportional to the unit matrix. In this case
Stodolsky’s damping formula in the form Eq. (9.29) gives ρ̇p,coll = 0.
This formula applies in the limit when the neutrino energies do not
change in collisions (a medium of “heavy” fermions). If one lifts this
restriction the situation is more complicated, but it simplifies again
for weak damping and a small mixing angle. Then one may apply
Eq. (9.41) with ge = gx = 1,
) * +
ṅνx = 1
4
dp dp" WP P ! (sp − sp! )2 fpe (1 − fpx! ) − fpx (1 − fpe! ) .(9.42)
If in a collision |p| = |p" | and thus sp = sp! one recovers the previous
result ṅνx = 0. However, if the mixing angle is a function of the neutrino
momentum, NC collisions do lead to flavor conversion and thus to the
damping of oscillations.
Of course, if only true NC interactions existed, the mixing angle
in the medium would be fixed at its vacuum value and so no flavor
conversion could occur. The deviation of θ from its vacuum value in

a medium is entirely from charged-current interactions with electrons,
even though they may be written in an effective NC form. The coherent
neutrino energy shifts by an electron background are enough to allow
true NC collisions with, say, neutrons to achieve flavor equilibrium!
9.4 Charged-Current Interactions

9.4.1 Hamiltonian
Besides neutrino scattering or pair processes one must also include
charged-current (CC) reactions where neutrinos are absorbed or pro-
duced by the medium (converted into or from charged leptons) such
that the total lepton number of the neutrino ensemble changes by one
unit. The corresponding interaction Hamiltonian can be written in the
form
)
GF
HCC =√ d3 x Υ(x)Ψ(x) + h.c. , (9.43)
2
where the neutrino field Ψ is, again, a column vector in flavor space with
the entries Ψ# , . = e, µ, τ in the standard model. Further, Υ is a row of
Dirac operators representing the medium. In the interaction basis Υ#
carries the lepton number corresponding to the flavor .. For example,
in a medium of nucleons and electrons the field Υe corresponding to the
electron lepton number can be written for standard-model couplings as
Υe = γ µ (1 − γ5 )ψe ψ n γµ (CV − CA γ5 )ψp , (9.44)
where ψp , ψn , and ψe are the proton, neutron, and electron Dirac fields,
respectively, while CV = 1 and CA = 1.26 are the dimensionless CC
vector and axial-vector nucleon coupling constants.
9.4.2 Kinetic Terms

One may now insert HCC into Eq. (9.13) in order to derive the explicit
CC collision integral for the evolution of ρp and ρp . The operators
Υ# violate the lepton number L# corresponding to flavor .. Therefore,
$Υ# % = 0 at all times if the medium is in an eigenstate of L# (. = e, µ,
τ , or additional exotic flavors). This assumption implies that the CC
interaction Eq. (9.43) does not contribute to refractive effects given by
the first-order term in Eq. (9.13).
330 Chapter 9
In the second-order term Hint appears quadratic so that one obtains

expressions like $Υ# Υk %. However, because the medium is assumed to
be in an eigenstate of L# they do not contribute for . *= k. Thus,
in the final result the contributions of different flavors can be added
incoherently.
#
The rates of production P∆ and absorption A#∆ of a ν# are functions
of the energy-momentum transfer ∆ to the medium,
) +∞ ' (
#
P∆ = 1
2
G2F dt e−i∆0 t Υ# (∆, t)γµ ∆µ Υ# (∆, 0) ,
−∞
) +∞ ' * +(
A#∆ = 1
2
G2F dt e−i∆0 t Tr γµ ∆µ Υ# (∆, 0)Υ# (∆, t) . (9.45)
−∞
These expressions are defined for both positive and negative energy
#
transfer ∆0 because P−P plays the role of an absorption rate for an-
tineutrinos with physical (P0 > 0) four momentum while A#−P plays
that of a production rate. Put another way, A#∆ and P∆ #
represent the
rate of absorption or production of lepton number of type ., indepen-
dently of the sign of ∆0 .
It is useful to define a flavor matrix of production rates which in
the weak basis has the form
 
Pe 0 0
1 ∆ µ 
P∆ ≡  0 P∆ 0 . (9.46)
2 τ
0 0 P∆
An analogous definition pertains to A∆ . Then one finds for the CC
collision integrals (Sigl and Raffelt 1993)
ρ̇p,CC = {PP , (1 − ρp )} − {AP , ρp } ,
ρ̇p,CC = {A−P , (1 − ρp )} − {P−P , ρp } , (9.47)
where {·, ·} is an anticommutator. The kinetic term for ρp is related to
that for ρp by the crossing relation Eq. (9.26).
The r.h.s. of Eq. (9.47) for ρp is the difference between a gain and a
loss term corresponding to the production or absorption of a νp . For a
single flavor they take on the familiar form PP (1 − fp ) and AP fp where
(1 − fp ) is the usual Pauli blocking factor.
9.4.3 Weak-Damping Limit

The meaning of Eq. (9.47) becomes more transparent if one makes var-
ious approximations which are justified for the conditions of a SN core.
As discussed in Sect. 9.3.5 one may ignore the antineutrino degrees
of freedom, and one may use the weak-damping limit where neutrino
oscillations are much faster than their rates of collision or absorption.
Then one finds for two flavors (Raffelt and Sigl 1993)
f˙px = (1 − fpx ) PPx − fpx AxP

* +
+ 14 s2p (2 − fpx − fpe )(PPe − PPx ) − (fpx + fpe )(AeP − AxP ) ,
(9.48)
where fpe and fpx are the occupation numbers for νe and νx as in
Sect. 9.3.5. The corresponding equation for νe is found by exchang-
ing e ↔ x everywhere.
For νx = νµ flavor conversion can build up a nonvanishing muon den-
sity in a SN core because they are light enough to be produced initially
when the electron chemical potential is on the order of 200−300 MeV.
For νx = ντ or some sterile flavor, the direct production or absorption
is not possible, AxP = PPx = 0. This simplifies Eq. (9.48) considerably,
* +
f˙px = 14 s2p (2 − fpx − fpe ) PPe − (fpx + fpe ) AeP . (9.49)
If neither νe nor ντ are occupied because, for example, the medium

is transparent to neutrinos so that they escape after production, one
has f˙px = 12 s2p PPe so that the production rate of νx is that of νe times
1
2
sin2 2θp as one would have expected.
In a SN core where normal neutrinos are trapped Eq. (9.49) is
more complicated as backreaction and Pauli blocking effects must be
included. It becomes simple again if s2p ' 1, because then the produc-
tion of νx causes only a small perturbation of β equilibrium. Therefore,
one may use the detailed-balance condition (1 − fpe )PPe − fpe AeP = 0.
Inserting this into Eq. (9.49) leads to
* +
f˙px = 14 s2p (1 − fpx ) PPe − fpx AeP , (9.50)
so that now the νx follow a Boltzmann collision equation with rates

of gain and loss given by those of νe times 14 sin2 2θp . If the νx are
sterile they escape without building up so that their production rate is
1
4
sin2 2θp that of νe .
332 Chapter 9
9.5 Flavor Conversion in a SN Core

9.5.1 Rate Equation
As a first application of the formalism developed in the previous sec-
tions consider two-flavor mixing between νe and another active neutrino
species νx = νµ or ντ (vacuum mixing angle θ0 ). In a SN core immedi-
ately after collapse electron lepton number is trapped and so the νe ’s
have a large chemical potential on the order of 200 MeV. The trapped
energy and lepton number diffuses out of the SN core and is radiated
away within a few seconds. Will νx achieve equilibrium with νe on this
time scale and thus share the large chemical potential? For the µ fla-
vor this would also imply the production of muons by the subsequent
charged-current absorption of νµ so that the lepton number would be
shared between e, µ, νe , and νµ .
If the mixing angle in the medium were not small, flavor conversion
would occur about as fast as it takes to establish β equilibrium. In
this case a detailed calculation is not necessary so that one may focus
on the limit of small mixing angles. In addition the oscillations are
fast which allows one to use Eqs. (9.41) and (9.50). Moreover, the
medium properties are assumed to be isotropic so that the production
and absorption rates P e and Ae of νe ’s depend only on their energy E.
Also, the transition rate WP P ! for the scattering of a neutrino with four
momentum P to one with P " may be replaced by an angular average
which depends only on the energies E and E " . Altogether one finds a
rate of change for the νx density of
) * +
ṅνx = 1
4
dp s2p (1 − fpx ) PEe − fpx AeE
.) *
+ 1
4
dp dp" WEE
a a a 2 e x
! (gx sp − ge sp! ) fp (1 − fp! )
a +
− (gea sp − gxa sp! )2 fpx (1 − fpe! ) , (9.51)
where fpe and fpx are the occupation numbers of νe and νx which are
given by Fermi-Dirac distributions because kinetic equilibrium was as-
sumed for both flavors. Also, 14 s2p = 14 sin2 2θE = θE2 for small mixing
angles.
A summation over different species a of medium fermions was re-
stored; gea and gxa are dimensionless effective NC coupling constants of
νe and νx to fermion species a. For a = n or p these constants are
the same for all active neutrino species. Electrons as scattering targets
are very relativistic so that they may be classified into a l.h. and a
r.h. “species.” The scattering with νe and νx is then described by the

effective NC Hamiltonians
GF
HL,R = √ ψ e γµ (1 ∓ γ5 )ψe ΨGL,R γ µ (1 − γ5 )Ψ, (9.52)
2 2
where Ψ is again a neutrino column vector in flavor space. Further,
3 4
2 sin2 θW + 1 0
GL = ,
0 2 sin2 θW − 1
3 4
2 sin2 θW 0
GR = , (9.53)
0 2 sin2 θW
where sin2 θW ≈ 14 will be used. Hence the effective NC coupling con-
stants are different for νe and νx interacting with l.h. electrons while
they are the same for r.h. ones.
The νe Fermi sea is very degenerate. With regard to the neutrino
distributions one may thus use the approximation T = 0 so that neu-
trino occupation numbers are 1 below their Fermi surface, and 0 above.
Because the chemical potential µνe of the νe population exceeds µνx , and
because neutrinos can only down-scatter in the T = 0 limit, the term
proportional to fpx (1 − fpe! ) vanishes. Moreover, the detailed-balance
requirement (1 − fpe )PEe = fpe AeE implies PEe = 0 for E > µνe where
fpe = 0. Then altogether
) 5 6
µνeθE2 PEe E 2 ) E " . a (gxa θE − gea θE ! )2 E 2 E "2
ṅνx = dE + dE WEE ! .
µνx 2π 2 µνx a 4π 4
(9.54)
Because for degenerate neutrinos nν = µ3ν /6π 2 Eq. (9.54) can be written
as a differential equation for µνx .
According to Eq. (8.30) the mixing angle in a medium which is
dominated by protons, neutrons, and electrons is given by
sin 2θ0
tan 2θE = , (9.55)
cos 2θ0 − E/Eρ
where the density-dependent “resonance energy” is
∆m2
Eρ ≡ √ , (9.56)
2 2 GF ne
with the electron density ne . For E = µνe one finds
E (68 keV)2 4/3
= 2
Ye Yν1/3 ρ14 , (9.57)
Eρ ∆m e
where ρ14 is the density in units of 1014 g cm−3 and as usual Yj gives the
abundance of species j relative to baryons. The approximate parameter
334 Chapter 9
range where resonance is important is shown in Fig. 9.3 as diagonal

shaded band.
For conditions near resonance there is no need for a detailed cal-
culation because then the mixing angle is large. Therefore, one may
focus on the limiting cases where ∆m2 is either so small or so large that
θE2 ' 1,
7
1 if Eρ /E - 1 (large ∆m2 ),
|θE | = θ0 × (9.58)
Eρ /E if Eρ /E ' 1 (small ∆m2 ).
Thus, for large ∆m2 the νx production rate is

) 5 6
µνe PEe E 2 ) E " . a (gxa − gea )2 E 2 E "2
ṅνx = θ02 dE + dE WEE ! , (9.59)
µνx 2π 2 µνx a 4π 4
while for small ∆m2 it is

5 62 ) 5 ) 6
a "
θ0 ∆m2 µνe PEe E . a
a (gx E − ge E)
2
ṅνx = √ dE + dE " WEE ! .
2 2GF ne µνx 2π 2 µνx a 4π 4
(9.60)
9.5.2 Neutrino Interaction Rates

In order to evaluate these integrals one first needs the production rate
PEe of νe ’s with energy E due to the CC reaction p + e → n + νe .
The leptons are taken to be completely degenerate, the nucleons to
be completely nondegenerate. Because they are also nonrelativistic
the absorption of an e− produces a νe of the same energy. Therefore,
PEe = σE np where np is the proton density and σE = (CV2 +3CA2 )G2F E 2 /π
is the CC scattering cross section for electrons of energy E. Here,
CV = 1 and CA = 1.26 are the usual vector and axial-vector weak
couplings. Altogether one finds
CV2 + 3CA2 2 2
PEe = GF E np . (9.61)
π
In practice this rate is reduced by various factors. First, Pauli blocking
of nucleons cannot be neglected entirely. Second, the degeneracy of
electrons is not complete. Third, the axial-vector scattering rate may
be suppressed in a medium at nuclear densities (Sect. 4.6.7).
For NC scattering, nucleons may be neglected entirely. In Eq. (9.59)
their contribution vanishes identically because ge = gx . In Eq. (9.60) it
is suppressed because they are relatively heavy so that E " ≈ E.
For l.h. electrons the coupling strengths are different, and they are
relativistic so that recoil effects are not small. However, they are degen-
erate so that their contribution is expected to be smaller than the ep
process. It is not entirely negligible, however, especially if the nucleon
contribution is partly suppressed by many-body effects. The transition
rates were worked out in detail by Raffelt and Sigl (1993) for entirely
degenerate leptons. They found
R G2F µ2e (E − E " )E "

WEE ! ≈ ,
3π E2
G2 µ2 (E − E " )E
L
WEE ! ≈ F e , (9.62)
3π E "2
a result which is the lowest-order term of an expansion in powers of
E/µe (electron chemical potential µe ).
9.5.3 Time Scale for Flavor Conversion

Given enough time the ντ ’s will reach the same number density as the
νe ’s. Therefore, it is most practical to discuss the approach to chemical
equilibrium in terms of a time scale
5 6
−1 d nνe − nνx
τ ≡ − ln . (9.63)
dt nνe
If τ were a constant independent of nνx /nνe the difference between the

number densities would be damped exponentially.
Collecting the results of the previous section one then finds easily
for the case of a “large” ∆m2 of Eq. (9.59)
8 9
−1 3G2F µν
τ = θ02 np µ2νe (CV2 + 3CA2 ) Fp + e Fe , (9.64)
5π µe
where Fp and Fe give the contributions of protons (CC process) and

electrons (effective NC process). They are functions of µνx which is
parametrized by
η ≡ µνx /µνe . (9.65)
One finds (Fig. 9.2, left panel)
Fp = (1 − η 5 )/(1 − η 3 ),
Fe = ( 56 + 12 η + 14 η 2 + 1 3
12
η ) (1 − η)3 /(1 − η 3 ), (9.66)
336 Chapter 9
Fig. 9.2. Dependence of the flavor relaxation rate on η = µνx /µνe for the
case of a “large” ∆m2 Eq. (9.66) and a “small” ∆m2 Eq. (9.69). Fe is
the contribution of electron targets (νe → eν) while Fp is from protons
(ep ↔ nνe ).
and numerically
5 65/3 5 62
3G2F Yp ρ µνe
np µ2νe = 1.0×109 s−1 , (9.67)
5π 10 g cm−3
14 µe
which sets the time scale for flavor conversion.

For the case of a “small” ∆m2 one finds from Eq. (9.60) and from
the production and scattering rates of the previous section
8 9
−1 (∆m2 )2 µν
τ = θ02 (CV2 + 3CA2 ) Fp + e Fe , (9.68)
8πnp µe
where Fp = 1 and
9 −1 137
5
η − 40
+ 34 (1 + η 4 ) log η + 53 η + η 3 − 101 4
120
η − 15 η 5
Fe = 8 (9.69)
3
(1 − η 3 )
(Fig. 9.2, right panel). Numerically,

5 65 64
(∆m2 )2 1014 g cm−3 (∆m2 )1/2
= 1.4×102 s−1 . (9.70)
8πnp Yp ρ 1 keV
This time scale is independent of Fermi’s constant because a factor G2F

from the scattering rate cancels against G−2 2
F from θE .
Fig. 9.3. Contour plot for log(τ θ02 ) with τ in seconds according to Eqs. (9.64)
and (9.68), taking Fe = 0, Fp = 1, CV2 +3CA 2 = 4, and µ /µ = 1. (Adapted
νe e
from Raffelt and Sigl 1993.)
From Fig. 9.2 it is clear that effective NC scattering on electrons

slightly accelerates the initial rate of flavor conversion, but it does not
dramatically affect the overall time scale for achieving equilibrium. This
time scale is crudely estimated by ignoring Fe entirely in Eqs. (9.64)
and (9.68) and by setting Fp = 1 and CV2 + 3CA2 = 4. With µνe ≈ µe
one then finds results for τ θ02 shown as contours in Fig. 9.3. The diag-
onal band refers to the resonance condition of Eq. (9.57); there flavor
equilibrium would be established on a time scale nearly independent of
the vacuum mixing angle. However, the detailed behavior in this range
of parameters has not been determined.
Armed with these results it is straightforward to determine the range
of masses and mixing angles where νx would achieve flavor equilibrium
and thus would effectively participate in β equilibrium ep ↔ nν. The
initially trapped lepton number escapes within a few seconds. There-
fore, τ <∼ 1 s is adopted as a criterion for νx to have any novel impact
on SN cooling or deleptonization. The relevant density is about three
times nuclear while Yp ≈ 0.35 so that Yp ρ = 3×1014 g cm−3 is adopted.
Otherwise the same parameters are used as in Fig. 9.3. Then one finds
7
0.02 (keV2 /∆m2 )2 for ∆m2 <
∼ (100 keV)
2
sin 2θ0 >
2
∼ (9.71)
2×10−10 for ∆m2 >
∼ (100 keV)
2
as a requirement for νx to reach chemical equilibrium. This range of

parameters is shown as a hatched region in Fig. 9.4.
338 Chapter 9
Fig. 9.4. In the hatched parameter range νµ or ντ would achieve chemical

equilibrium with νe in a SN core within about one second after collapse.
The cosmological mass limit of about 30 eV prevents νµ or ντ from

playing any novel role in the cooling or deleptonization of a SN core.
However, only a very small mixing angle is required to achieve equi-
librium if one of the neutrinos defied either standard particle physics
or standard cosmology and had a mass in the keV range or above. Of
course, if the mass were of Dirac type the cooling effect from the pro-
duction of spin-flipped neutrinos would be too large to be compatible
with the SN 1987A neutrino signal, yielding a bound on Dirac neutrino
masses in the 10 keV range (Sect. 13.8.1). It is in this context that fla-
vor conversion was first discussed by Maalampi and Peltoniemi (1991),
Turner (1992), and Pantaleone (1992a).
9.6 Sterile Neutrinos and SN 1987A

If a hypothetical sterile neutrino νx existed, it would be produced in
the inner core of a SN by virtue of its assumed mixing with νe . The νx
would escape directly from the inner SN core, carrying away energy and
lepton number. If this process occurred too fast the observed neutrino
signal of SN 1987A would have been unduly shortened, allowing one
to exclude a certain range of νx masses and mixing angles with νe
(Kainulainen, Maalampi, and Peltoniemi 1991).
For small mixing angles one may use Eqs. (9.40) and (9.50) as a
starting point for the rate of change of the νx occupation numbers fpx .
Because the νx are assumed to escape freely one may set fpx = 0 on
the r.h.s. of these equations. Moreover, for the νx coupling constants
one may use gx = 0 because it is sterile. For an isotropic medium
and taking an angular average of the scattering rate as in the previous
section one finds
5 ) 6
.
f˙px = 1 2
s
4 p
PEe + (gea )2 dp "
WEa ! E fpe! . (9.72)
a
A summation over different target species for effective NC scattering

was restored. Taking the νe ’s to be completely degenerate one has
fpe = 1 for E = |p| < µνe and fpe = 0 otherwise so that
5 ) 6
. E "2 WEa ! E
f˙px = 1 2
s
4 p
PEe + (gea )2 dE "
. (9.73)
a 2π 2
With this result and
) µνe ) µνe
ṅL = − dE f˙px and Q̇ = − dE f˙px E (9.74)
0 0
the volume loss rates for lepton number and energy can be easily de-
termined.
Turn first to the case ∆m2 > 2
∼ (100 keV) so that one may use the
vacuum mixing angle in a SN core. One must now include both the CC
process ep → nνx as well as the NC process νe N → N νx which is not
suppressed because for a sterile νx there is no destructive interference
effect. For nondegenerate nucleons and using CV2 + 3CA2 ≈ 4 for both
CC and NC processes one easily finds from the results of Sect. 9.5.2
2G2F
ṅL = − 14 sin2 2θ0 nB (Yp + 14 ) µ5νe , (9.75)
5π 3
and the same for Q̇ with 15 µ5νe → 16 µ6νe . Electrons as NC scattering
targets may be neglected because of their degeneracy. Lepton number
is thus lost at a rate
ẎL = − sin2 2θ0 τ −1 (Ye + 14 ) Yν5/3

e
, (9.76)
where
5/3 5/3
τ −1 = 35 (36π)1/3 G2F nB = 7.7×1010 s−1 ρ15 . (9.77)
Here, Yp = Ye , YL ≡ nL /nB , and Yν nB = nν = µ3ν /6π 2 was used, and

ρ15 is ρ in units of 1015 g cm−3 .
340 Chapter 9
Because Yνe is about YL /4, lepton number and energy are lost at
about a rate of sin2 2θ0 1010 s−1 . Because the SN 1987A signal lasted
for several seconds, a conflict with these observations is avoided if
sin2 2θ0 <
∼ 10
−10
(9.78)
(Kainulainen, Maalampi, and Peltoniemi 1991; Raffelt and Sigl 1993).
If mνx > ∼ 1 MeV the assumed mixing with νe allows for decays νx →
νe e e and νx → νe e− e+ γ. The resulting γ signal from the SN 1987A
− +
νx flux (Sect. 12.4.7) does not allow for a dramatic improvement of

the bound Eq. (9.78). However, the decay argument does exclude the
possibility of a mixing angle so large that even the “sterile” νx would
be trapped by virtue of its mixing with νe .
The case of ∆m2 < 2
∼ (100 keV) is complicated because neutrinos in a
certain energy range below their Fermi surface encounter a νe -νx mixing
resonance. This implies that during the SN infall phase a large amount
of lepton number can be lost. The impact on the equation of state
can be strong enough to prevent a subsequent explosion. Shi and Sigl
(1994) found that this argument requires sin2 2θ0 < −8 2
∼ 10 keV /∆m for
2
2 > 2
∆m ∼ (1 keV) . They found additional constraints from the anoma-
lous contribution to the cooling by νx emission.
These are all limits on the mixing of a sterile neutrino with νe , the
only case that has been studied in the literature. Historically, this is
related to the now forgotten episode of the 17 keV neutrino which for
some time seemed to exist and which could have been a sterile neutrino
mixed with νe . However, similar limits can be derived for νµ -νx or ντ -νx
mixing. The main difference is that the nonelectron neutrinos would
not normally obtain a chemical potential so that only a thermal νµ and
ντ population can be converted. A typical temperature is 30 MeV or
more, the average energy of a thermal population of relativistic fermions
is about 3T > ∼ 100 MeV, so there will be a significant thermal muon
population (mµ = 106 MeV). Thus sterile states can be produced in
charged-current muon scatterings in analogy to the above discussion
of νx production involving electron scattering. The resulting limit on
the mixing angle may be slightly weaker than in the νe -νx case, but it
will be of the same general order of magnitude.52 For ντ -νx mixing the
situation is different in that there are no thermally excited τ leptons.
Still, any reaction that produces ντ ν τ pairs can also produce νx and ν x
particles.
52
This remark is relevant in the context of recent speculations about the existence
of a 34 MeV sterile neutrino (Barger, Phillips, and Sarkar 1995) as an explanation of
an anomaly observed in the KARMEN experiment (KARMEN Collaboration 1995).
Chapter 10
Solar Neutrinos
The current theoretical and experimental status of the Sun as a neu-

trino source is reviewed. Particle-physics interpretations of the appar-
ent deficit of measured solar neutrinos are discussed, with an emphasis
on an explanation in terms of neutrino oscillations.
10.1 Introduction
The Sun, like other hydrogen-burning stars, liberates nuclear binding
energy by the fusion reaction
4p + 2e− → 4 He + 2νe + 26.73 MeV (10.1)
which proceeds through a number of different reaction chains and cy-

cles (Fig. 10.2). With a total luminosity of L" = 3.85×1033 erg s−1 =
2.4×1039 MeV s−1 , the Sun produces about 1.8×1038 s−1 neutrinos, or
at Earth (distance 1.50×1013 cm) a flux of 6.6×1010 cm−2 s−1 . While
this is about a hundred times less than the ν e flux near a large nu-
clear power reactor it is still a measurable flux which can be used for
experimentation just like the flux from any man-made source.
The most straightforward application of the solar neutrino flux is
a search for radiative decays by measurements of x- and γ-rays from
the quiet Sun. Because of the long decay path relative to laboratory
experiments one obtains a limit which is about 9 orders of magnitude
more restrictive (Sect. 12.3.1).
A more exciting application is a search for neutrino oscillations.
In fact, the current measurements of the solar neutrino flux are nei-
ther compatible with theoretical predictions nor with each other (“solar
neutrino problem”); all discrepancies disappear with the assumption of
341
342 Chapter 10
neutrino oscillations. At the present time, however, this interpretation

is not established “beyond reasonable doubt”—a final verdict can be
expected from the new experiments currently in preparation. They may
be able to discover a characteristic distortion of the neutrino spectrum,
or they may actually measure the “wrong-flavored” neutrinos that were
produced by oscillations from the νe originating in the Sun.
Contrary to the ν e flux from a power reactor, the solar neutrino
spectrum arises from a small number of specific reactions (Fig. 10.1).
The main contribution (91%) is from the reaction p + p → d + e+ + νe
with a maximum neutrino energy 0.420 MeV (“pp neutrinos”). Second
at about 7% of the flux are the “beryllium neutrinos” from the electron-
capture reaction e− + 7 Be → 7 Li + νe with a fixed energy 0.862 MeV.
Finally, very small fraction (≈ 10−4 ) are the “boron neutrinos” from
8
B → 8 Be + e+ + νe . Still, they are of major importance because
their large energies of up to 15 MeV allow for a less difficult detection
procedure than is required for the soft part of the spectrum.
The first solar neutrino experiment is based on the nuclear reaction
νe + 37 Cl → 37 Ar + e− . With a threshold of 0.814 MeV it picks up
both beryllium and boron neutrinos, although the argon production
rate is dominated by the latter. The target consists of about 4×105
liters (615 tons) of perchloroethylene (C2 Cl4 ) in a huge tank which
is located in the Homestake Mine in South Dakota (U.S.A.), about
1.5 km underground for protection against the cosmic-ray background.
After an exposure of a few months to the solar neutrino flux a few
argon atoms have been produced (about 0.4 atoms per day). They
are chemically extracted and counted by their subsequent decays (half-
life 35.0 days).
When this pioneering experiment first produced data (Davis, Har-
mer, and Hoffman 1968) there appeared a deficit relative to the theoret-
ically expected flux, a discrepancy which has persisted ever since—the
experiment is still taking data today! On the theoretical side, the ever
refined predictions of Bahcall and his collaborators (e.g. Bahcall 1989)
were instrumental at establishing the notion that this discrepancy—a
factor of around 3—was to be taken seriously. However, in spite of the
acknowledged experimental care of Davis and his collaborators, doubt
has always lingered about the reliability of the data because this de-
tector has never been subject to an on-off test as the Sun is the only
available neutrino source powerful enough to cause a detectable signal.
On the solar side, the boron neutrino flux depends crucially on the
reaction rate p + 7 Be → 8 B + γ with a cross section that is relatively
poorly known.
Solar Neutrinos 343
Fig. 10.1. Solar neutrino flux at Earth according to the Bahcall and
Pinsonneault (1995) solar model. Upper panel: Continuum spectra in
cm−2 s−1 MeV−1 , line spectra in cm−2 s−1 . Solid lines are the sources of
dominating experimental significance. Above: Range of sensitivity of cur-
rent and near-future solar neutrino experiments. Lower panel: Cumulative
spectrum integrated from a given energy to infinity.
The situation changed radically when the Kamiokande detector,

originally built to search for proton decay, began in 1987 to mea-
sure the solar neutrino flux by virtue of the elastic scattering reaction
νe + e− → e− + νe which is detected by the Cherenkov light emitted by
the kicked electron. With a threshold of about 9 MeV (later 7 MeV) it
is exclusively sensitive to the boron neutrino flux. Because of its direc-
344 Chapter 10
tional sensitivity it is a true “neutrino telescope” and for the first time
established that indeed neutrinos are coming from the direction of the
Sun. It is perplexing, however, that the measured flux is less suppressed
relative to solar-model predictions than that found in the Homestake
experiment. This is the reverse from what would be expected on the
grounds that 37 Cl is sensitive to boron and beryllium neutrinos.
The situation changed yet again when the experiments SAGE (So-
viet-American Gallium Experiment) and GALLEX began to produce
data in 1990 and 1991, respectively. They are radiochemical experi-
ments using the reaction νe + 71 Ga → 71 Ge + e− which has a threshold
of 233 keV. Therefore, these experiments pick up the dominant pp neu-
trino flux which can be calculated from solar models with a precision of
a few percent unless something is radically wrong with our understand-
ing of the Sun. Therefore, a substantial deficit of measured pp neutrinos
would have been a “smoking gun” for the occurrence of neutrino os-
cillations. While the first few exposures of SAGE seemed to indicate
a low flux, the good statistical significance of the data that have since
been accumulated by both experiments indicate a flux which is high
enough so that no pp neutrinos are reported missing, but low enough
to confirm the existence of a significant problem with the high-energy
part of the spectrum (beryllium and boron neutrinos).
With four experiments reporting data, which represent three differ-
ent spectral responses to the solar neutrino flux, the current attention
has largely shifted from a comparison between experiments and theoret-
ical flux predictions to a “model-independent analysis” which is based
on the small number of possible source reactions each of which produces
neutrinos of a well-defined spectral shape. This sort of analysis cur-
rently indicates a lack of consistency among the experiments which can
be brought to perfect agreement if neutrinos are assumed to oscillate.
Even though the attention has currently shifted away from theo-
retical solar neutrino flux predictions it should be noted that in recent
years there has been much progress in a quantitative theoretical treat-
ment of the Sun. Independently of the interest in the Sun as a neutrino
source it serves as a laboratory to test the theory of stellar structure
and evolution. Particularly striking advances have been made in the
field of helioseismology. There are two basic vibration patterns for the
Sun, one where gravity represents the restoring force (g-modes), and
normal “sound” or pressure (p) modes. The former are evanescent in
the solar convection zone (depth about 0.3 R" from the surface) and
have never been unambiguously observed. The oscillation period of the
highest-frequency g-modes would be about 1 h.
Solar Neutrinos 345
There exist vast amounts of data concerning p-modes (periods be-

tween 2 min and 1 h) which can be measured from the Doppler shifts
of spectral lines on the solar surface. The main point is that one can
establish a relationship between the multipole order of the oscillation
pattern and the frequency. Because different vibration modes probe the
sound speed at different depths one can invert the results to derive an
empirical profile for the sound speed in the solar interior. The agree-
ment with theoretical expectations for the square of the sound speed
is better than about 0.3% except in the inner 0.2 R" which are not
probed well by p-modes (Christensen-Dalsgaard, Proffitt, and Thomp-
son 1993; Dziembowski et al. 1994). Such results make it very difficult
to contemplate the possibility that the Sun is radically different from
a standard structure. On the other hand, these results are not precise
enough to reduce the uncertainty of solar neutrino predictions which
arise from the uncertainty of the opacity coefficients.
The differences between the solar neutrino predictions of different
authors are minimal when identical input parameters are used. There-
fore, the expected error resulting from solar modelling is very small,
except that some key input parameters remain uncertain. The dom-
inant source of uncertainty for the boron neutrino flux is the cross
section for the reaction p + 7 Be → 8 B + γ which appears as a multi-
plicative factor for the solar flux prediction and thus is unrelated to
solar modelling.
A more astrophysical uncertainty are the opacity coefficients. Al-
though they are thought to be well known for the conditions in the
deep solar interior, even a relatively small error translates into a non-
negligible uncertainty of the boron flux prediction because of its steep
temperature dependence. Crudely, a 10% error of the opacity coeffi-
cients translates into a 1% error of the central solar temperature and
then into a 20% uncertainty of the boron flux. Indeed, a measure-
ment of the solar neutrino flux was originally envisaged as a method to
measure precisely the inner temperature of the Sun.
The Sun as a neutrino source naturally has received much attention
in the literature because of its outstanding potential to finally confirm
the existence of neutrino oscillations in nature. Because this book is
not primarily on solar neutrinos I will limit my discussion to what I
consider the most important features of this unique neutrino source,
and the role it plays for particle physics. A lot of key material can be
found in Bahcall’s (1989) book on solar neutrinos, in the more recent
Physics Report on the solar interior by Turck-Chièze et al. (1993), and
on the Sun in general in the book by Stix (1989).
346 Chapter 10
Fig. 10.2. Reaction chains PPI−PPIII and CNO tri-cycle. Nuclear reactions,
including decays, are marked with a bullet (•). Average (av) and maximum
(max) energies in MeV are given for the neutrinos. For photons (wavy
arrows) numbers in brackets refer to the total energy of a cascade; otherwise
it is the energy of a monochromatic γ line.
Solar Neutrinos 347
10.2 Calculated Neutrino Spectrum

10.2.1 Individual Sources
A hydrogen-burning star like the Sun liberates nuclear binding energy
by helium fusion from hydrogen which proceeds by virtue of the pp
chains and the CNO cycle (Fig. 10.2). The energy-generation rate for
the CNO cycle is a much steeper function of temperature whence it
dominates in hot stars. For a 2% mass fraction of CN elements, typical
for population I stars like the Sun, the crossover temperature is at
about 1.8×107 K (Clayton 1968). The central temperature of the Sun
is about 1.56×107 K and so the CNO reactions contribute only about
1% to the total energy budget. In fact, of the CNO tri-cycle shown in
Fig. 10.2, in practice only the first loop (the CN cycle) is of importance
in the Sun as the branching rates into the second or even third loop are
extremely low.
In the Sun, the pp chains terminate in about 85% of all cases via
PPI, i.e. by the fusion of two 3 He nuclei. In this case the only neutrino-
producing reactions are pp (pp → de+ νe ) and pep (pe− p → dνe ), the
latter occurring very rarely. In about 15% of all cases the termina-
tion is via PPII where 7 Be is formed from 4 He + 3 He. Because of the
small energy difference between the ground states of the 7 Be and 7 Li
Table 10.1. Source reactions for solar neutrinos.
Q (a ) Flux at Earthb Uncertaintyb

Reaction
[MeV] [cm−2 s−1 ] [%]
pp → 2 H e+ νe 0.420 c 5.9×1010 +1 −1
pe− p → 2 H νe 1.442 1.4×108 +1 −2
3
He p → 4 He e+ νe 18.77 c 1.2×103 Factor 6
7
Be e− → 7 Li νe 0.862 4.6×109 +6 −7
7
Be e− → 7 Li∗ νe 0.384 5.2×108 +6 −7
8
B → 8 Be∗ e+ νe ≈ 15 c 6.6×106 +14 −17
13
N → 13 C e+ νe 1.199 c 6.2×108 +17 −20
15
O → 15 N e+ νe 1.732 c 5.5×108 +19 −22
17
F → 17 O e+ νe 1.740 c 6.5×106 +15 −19
Total 6.5×1010 +1 −1
a Maximum νe energy for continuum (c) sources.
b Bahcall and Pinsonneault (1995).
348 Chapter 10
nuclei, the usual weak decay by e+ νe emission is not possible and so the
conversion proceeds by electron capture, leading to the emission of an
almost monochromatic neutrino. In about 10% of all cases the capture
reaction goes to the first excited state (478 keV) of 7 Li so that there are
two neutrino lines.
Instead of an electron, 7 Li very rarely captures a proton and forms
8
B which subsequently decays into 8 Be, a nucleus unstable against spon-
taneous fission into 4 He + 4 He. This PPIII termination occurs in about
0.02% of all cases, too rare to be of importance for nuclear energy gen-
eration. Its importance arises entirely from the high energy of the 8 B
neutrinos which are the ones least difficult to measure.
The neutrino-producing reactions are summarized in Tab. 10.1 with
their maximum energies, and with the resulting neutrino flux at Earth
found in the solar model of Bahcall and Pinsonneault (1995).
Apart from small screening and thermal broadening effects, the
spectral shape for each individual source is independent of details of
the solar model. The pp, 13 N, 15 O, and 17 F reactions are allowed or
superallowed weak transitions so that their spectra are
! "1/2
dN/dEν = A (Q + me − Eν ) (Q + me − Eν )2 − m2e Eν2 F ,
(10.2)
where Q is the maximum e+ kinetic energy and also the maximum
νe energy, A is a normalization constant, and F is a function of Ee+
which takes the e+ final-state interactions into account. For the low-Z
nuclei under consideration this correction is small for most of the neu-
trino spectrum. With F = 1 the normalization constants are given in
Tab. 10.2. In Fig. 10.3 the normalized spectra from the pp and the 15 O
processes are shown where the pp spectrum was taken from the tabula-
tion of Bahcall and Ulrich (1988). Using Eq. (10.2) instead would cause
a change so small that it would be nearly hidden by the line width of
the curve in Fig. 10.3.
Table 10.2. Normalization of the spectrum Eq. (10.2) with F = 1.
Source Q [MeV] A [MeV−5 ]

pp 0.420 193.9
13
N 1.199 3.144
15
O 1.732 0.668
Solar Neutrinos 349
Fig. 10.3. Normalized spectra of neutrino source reactions in the Sun.
It is much more difficult to obtain the spectrum from 8 B decay be-

cause the final-state 8 Be∗ nucleus is unstable against spontaneous fission
into two α particles. Even though several states of 8 Be contribute to
the transition, it is dominated by the 2.9 MeV excitation and so the
neutrino spectrum can be determined with relatively little ambiguity
by folding Eq. (10.2) with the experimental α spectrum (Kopysov and
Kuzmin 1968). A more recent and more detailed analysis was per-
formed by Bahcall and Holstein (1986). An analytic approximation to
their tabulated spectrum is
dN/dEν = 8.52×10−6 (15.1 − Eν )2.75 Eν2 , (10.3)
where the neutrino energies are in MeV. This normalized spectrum

is also shown in Fig. 10.3 where, again, the difference between the
tabulated values and the analytic approximation would be hidden by
the line width (maximum deviation less than 0.02 in units of the vertical
axis in Fig. 10.3).
Neutrinos from the hep reaction extend to the highest energies of
all solar sources, but their overall flux is very small and very uncertain
because of large uncertainties in the low-energy 3 He p cross section—
see Bahcall and Pinsonneault (1992) for a detailed discussion. The
tabulated spectrum (Bahcall and Ulrich 1988) can be represented by
dN/dEν = 2.33×10−5 (18.8 − Eν )1.80 Eν1.92 , (10.4)
where the quality of the fit is equally good as that for the 8 B neutrinos.
350 Chapter 10
In principle, any of the neutrino-producing reactions in the Sun

with a final-state positron can also occur with an initial-state electron
such as pe− p → d νe instead of pp → de+ νe . In most cases, however,
the electron-capture process is strongly suppressed relative to positron
emission because of an unfavorable phase-space factor, or equivalently
because of the relatively small electron density. (For a general com-
parison between these two reaction channels see Bahcall 1990.) The
only exception is pep which yields a nonnegligible contribution because
the Q value of the pp reaction and thus the available positron phase
space are rather small. Another exception are the beryllium neutri-
nos because positron emission is inhibited entirely by the small energy
difference between the ground states of 7 Be and 7 Li.
The shapes of the neutrino spectra from the individual source re-
actions are determined entirely by the matrix element and phase space
of the microscopic reactions apart from small broadening effects by the
thermal motion of the reaction participants. As typical thermal en-
ergies in the Sun are 1 keV the spectral modification by such effects
is entirely negligible. (For a detailed discussion of this point see Bah-
call 1991.) The high-energy part of the solar neutrino spectrum between
about 2−15 MeV is dominated by the boron neutrinos (Fig. 10.1), i.e.
by a single source reaction. Therefore, this energy range is a clean
example to test the spectral shape. The confirmation of the expected
spectral shape above 7 MeV by the Kamiokande II detector excludes
Fig. 10.4. Normalized spectra of the thermally and Doppler broadened beryl-
lium neutrino lines where E0 is the corresponding laboratory energy (Bahcall
1994). The line shapes involve an integral over a solar model.
Solar Neutrinos 351
a certain range of neutrino masses and mixing angles (Sect. 10.3.4).

Of course, the Superkamiokande detector with its much improved sen-
sitivity and lower threshold could still detect a deviation; this would
be a clear indication for neutrino oscillations or other neutrino-related
novel phenomena.
For the neutrino lines from e− + 7 Be → 7 Li + νe (861.8 keV) and
e + 7 Be → 7 Li∗ + νe (384.3 keV) thermal broadening effects are of some
−
interest because they dominate the line shape. A detailed discussion

is found in Bahcall (1994) who calculated the observable spectra at
Earth shown in Fig. 10.4 where E0 is the energy of the transition in
the laboratory. The peak of the lines is shifted to higher energies by
0.43 and 0.19 keV, respectively, while the average energy is shifted by
an even larger amount because of the asymmetric form. Perhaps in
some next-century detector this line shape could be used to measure
the central temperature of the Sun.
10.2.2 Standard Solar Models

In order to calculate the expected solar neutrino flux at Earth one
needs to construct a model of the Sun. A “standard solar model” is
obtained by solving the stellar structure equations discussed in Sect. 1.2
in several time steps to evolve it to the solar age of 4.5×109 yr. At this
point it must produce the observed present-day luminosity of the Sun
of 1 L" = 3.85×1033 erg s−1 (it is about 30% brighter than a zero-age
model). This agreement is enforced by tuning the unknown presolar
helium abundance Yinitial to the required value which is usually found
to be about 27%.
Another present-day boundary condition is the measured solar ra-
dius of 1 R" = 6.96×1010 cm which is adjusted by the mixing-length
parameter which enters the standard treatment of convection. In the
Sun, the outer layers (depth about 0.3 R" ) are found to be convective;
in lower-mass main-sequence stars convection reaches deeper, in higher-
mass ones it disappears entirely near the surface while the central region
becomes convective. The Sun calibrates the mixing-length parameter
which is then used in evolutionary calculations of other stars.
In order to calculate a standard solar model one needs a variety of
calculated or measured input information, notably the photon opaci-
ties, the equation of state, nuclear cross sections with an appropriate
screening prescription, diffusion coefficients, a prescription to treat con-
vection, the abundances of metals (elements heavier than helium), and
the solar age, luminosity, and radius. Because of the large number
352 Chapter 10
of details that have to be minded it is sometimes difficult to compare

exactly the solar models of different workers. However, wherever a
detailed comparison has been performed, different results for such out-
put quantities as predicted neutrino fluxes usually can be understood
in terms of different choices regarding some of the input physics (e.g.
Bahcall and Pinsonneault 1992; Turck-Chièze and Lopes 1993). Thus,
there is very little ambiguity in solar modelling for a common choice of
input physics.
Nonstandard solar models are ones where standard input parame-
ters have been chosen outside the range of recognized uncertainties (for
example the opacities), or which include entirely new physical effects
such as strong magnetic fields, fast rotation in the deep interior, non-
standard nuclear reaction rates (involving free quarks, for example),
energy transfer by a nonstandard mechanism (e.g. by trapped massive
weakly interacting particles), and others. Some such possibilities can
be excluded by the helioseismologically determined sound-speed profile
of the Sun. I will not discuss nonstandard solar models any further—
for an overview see Bahcall (1989). Naturally, it is always possible that
a new “nonstandard” effect is recognized which could be important for
the solar structure and neutrino fluxes.
One standard physical effect that has made its way into standard
solar models only recently is the gravitational settling of helium and
metals. This effect leads to a stronger concentration of helium in the
central region than is caused by nuclear burning alone so that the
present-day Sun is “more evolved.” Consequently, the central tem-
perature is slightly higher, leading to an increased flux of boron neu-
trinos. The settling of metals leads to an increased opacity, causing
a further increase of the temperature and of the neutrino fluxes by
a similar amount. Solar models with gravitational settling of helium
significantly improve the already good agreement with the helioseismo-
logically inferred sound-speed profile (Christensen-Dalsgaard, Proffitt,
and Thompson 1993) while the settling of metals does not seem to have
a strong additional impact on the p-mode frequencies (Proffitt 1994).
Bahcall and Pinsonneault (1992) were the first to include helium set-
tling in a standard solar model; in their 1995 paper they included metal
settling as well. They find that the compound effect is to change the 8 B
flux by +36%, the 7 Be flux by +14%, and the pp flux by −1.7%. Sim-
ilar changes (+31%, +13%, and −1.7%) were found by Proffitt (1994)
while Kovetz and Shaviv (1994), who included only helium settling,
obtained smaller effects relative to the corresponding case of Bahcall
and Pinsonneault (1992).
Solar Neutrinos 353
While helium and metal diffusion increases the neutrino fluxes, the
changes are roughly within the claimed errors of previous standard solar
model predictions. Many improvements of input physics over the years
have left the neutrino flux predictions of Bahcall and his collaborators
surprisingly stable over 25 years (Bahcall 1989, 1995). Bahcall (1994)
has compiled the central temperature predictions from a heterogeneous
set of 12 standard solar models without diffusion calculated by different
authors since 1988. The temperature predictions are almost uniformly
distributed on the interval 15.40−15.72 × 106 K, i.e. these authors agree
with each other on the value 15.56×106 K within ±1%.
Thus, in spite of differences in detail there exists a broad consensus
on what one means with a standard solar model. Therefore, the neu-
trino fluxes of the Bahcall and Pinsonneault (1995) model with element
diffusion (Tab. 10.1 and Fig. 10.1) can be taken to be representative.
The general agreement on a standard solar model does not guarantee, of
course, that there might not exist problems related to incorrect stan-
dard assumptions or incorrect input parameters common to all such
models.
10.2.3 Uncertainties of Standard Neutrino Predictions

a) Opacities
In order to compare the neutrino flux predictions with the experimental
measurements one needs to develop a sense for the reliability of the
calculations. Naturally, it is impossible to quantify the probability
for the operation of some hitherto unknown physical effect that might
spoil the predictions; an error analysis can only rely on the recognized
uncertainties of standard input physics. Very detailed error analyses
can be found for the standard solar models of Bahcall and Pinsonneault
(1992) and of Turck-Chièze and Lopes (1993).
As for solar modelling, the dominating uncertainty arises from the
radiative opacities which largely determine the temperature profile of
the Sun. A reduction of the Rosseland mean opacity in the central
region by 10% reduces the temperature by about 1%.
There is not a one-to-one correspondence between the central solar
temperature Tc and the neutrino fluxes because it is not possible to
adjust Tc and leave all else equal. Conversely, one must modify some
input parameters and evolve a self-consistent solar model. Still, if one
allows all input parameters to vary according to a distribution deter-
mined by their measured or assumed uncertainties one finds a strong
354 Chapter 10
correlation between Tc and the neutrino fluxes which allows one to un-
derstand the impact of certain modifications of a solar model on the
neutrino fluxes via their impact on Tc . Bahcall (1989) found
 −1.2

 T for pp,
 c
Neutrino Flux ∝  Tc8 for 7 Be, (10.5)


Tc18 for 8 B.
It is noteworthy that the pp flux decreases with increasing Tc because

of the constraint imposed by the solar luminosity. One concludes that
a 1% uncertainty in Tc translates roughly into a 20% uncertainty of the
boron flux.
There are two sources of uncertainty for the opacity. First, for an
assumed chemical composition, the actual opacity calculation which in-
volves complicated details of atomic and plasma physics. Second, there
is the uncertain metal content in the central region of the Sun (Z/X).
Because iron retains several bound electrons even for the conditions
at the solar center it contributes substantially to the Rosseland mean
opacity; removing iron entirely would reduce it by 25−30%.
For a recent calculation and detailed discussion of solar opacities see
Iglesias and Rogers (1991a). In view of the relatively small deviations
between different opacity calculations for the relevant conditions Turck-
Chièze and Lopes (1993) as well as Bahcall and Pinsonneault (1992)
agree that the radiative opacities are likely calculated with a precision
of better than a few percent.53
The actual amount of heavy elements in the Sun, notably iron, is
determined by spectroscopic measurements in the photosphere, and by
the abundance in meteorites which are assumed to represent the preso-
lar material. The results from both methods seem to agree essentially
on a common value—see Bahcall and Pinsonneault (1995) for a sum-
mary of the recent status. They believe that the uncertainty in the
8
B flux caused by the uncertainty of Z/X is about 8%. Of course, in
the central regions of the Sun the metal abundance is also determined
by gravitational settling as discussed above. The uncertainty of the 8 B
flux inherent in the treatment of gravitational settling is thought to
be 8% as well.
53
However, most recently Tsytovich et al. (1995) have claimed that relativistic
corrections to the electron free-free opacity as well as a number of other hitherto
ignored effects reduce the standard total opacity by as much as 5%.
Solar Neutrinos 355
Bahcall and Pinsonneault (1995) then find that the opacity-related

uncertainty of the 8 B neutrino flux is about 12%. For the 7 Be flux
the errors are thought to be roughly half as large, in agreement with
Eq. (10.5). However, as stressed by Bahcall and Pinsonneault (1995),
the interpretation as an effective 1σ error is misleading as the main
source of uncertainty is of a systematic and theoretical nature. Previ-
ously, these authors had stated “theoretical 3σ errors;” in that sense
an opacity-related uncertainty of the 8 B flux of about 30% was found
in Bahcall and Pinsonneault (1992). Turck-Chièze and Lopes (1993)
adopted 15% without a commitment to a specific number of sigmas.
None of these errors can be interpreted in a strict statistical sense.
Rather, they give one an idea of what the workers in that field consider
a plausible range of possibilities.
b) Beryllium-Proton Reaction
A dominating uncertainty for the important flux of boron neutrinos
arises from the cross section 7 Be + p → 8 B + γ which plays no role
whatsoever for the energy generation in the Sun because the PPIII
termination of the pp chain is extremely rare. Therefore, a modification
of this cross section has no impact on the structure of the Sun and
thus no other observable consequence but to modify the high-energy
neutrino flux.
The cross section for this reaction is parametrized for low energies
in the usual form with an astrophysical S-factor
σ(E) = S(E) E −1 e−2πη(E) , (10.6)
with the Sommerfeld parameter
η = Z1 Z2 e2 v −1 . (10.7)
Here, Z1,2 e are the charges of the reaction partners, v their relative
velocity, and E their CM kinetic energy. The S-factor is expected to
be essentially constant at low energies unless there is a resonance near
threshold.
The six “classical” measurements of the 7 Be + p → 8 B + γ reaction
are referenced in Tab. 10.3. In the Sun, the most effective energy range
is around E = 20 keV, far in the tail of the thermal distributions of
the reaction partners, but still far below the lowest laboratory energies
of around 120 keV in the experiments of Kavanagh et al. (1969) and
Filippone et al. (1983). Therefore, one must extrapolate the factor
356 Chapter 10
Table 10.3. Extrapolation of S17 to E = 0 for measurements of 7 Be p → 8 B γ

according to Johnson et al. (1992).
Experiment S17 (0) [eV b]

Kavanagh et al. (1960) 15 ± 6
Parker (1966, 1968) 27 ± 4
Kavanagh et al. (1969) 25.2 ± 2.4
Vaughn et al. (1970) 19.4 ± 2.8
Wiezorek et al. (1977) 41.5 ± 9.3
Filippone et al. (1983) 20.2 ± 2.3
S17 down to the astrophysically interesting regime. The most recent

comprehensive reanalysis was performed by Johnson et al. (1992) who
found the values listed in Tab. 10.3. They tend to be smaller by around
10% relative to previous extrapolations which did not take into account
that for laboratory energies there is a contribution from d-waves in the
entrance channel. Johnson et al. used two different interaction models
for the extrapolation which yielded identical results within 2%. This
does not necessarily imply that the theoretical extrapolation is known
with this precision; for example, Riisager and Jensen (1993) suggest
much smaller values for S17 (0).
As stressed by Johnson et al. (1992) it is problematic to combine the
results of Tab. 10.3 to a “world average” because they show systematic
discrepancies. In fact, the data of the two “low-energy” experiments
(Kavanagh et al. 1969; Filippone et al. 1983) agree very well with each
other over a wide range of energies with a systematic offset by a factor
1.34. A similar offset exists between the “high-energy” data of Parker
(1966, 1968) and Vaughn et al. (1970) with a factor 1.42 (Gai 1995).
Because it is unknown which of the experiments is right (if any)
Johnson et al. (1992) combined the data according to a prescription
adopted by the Particle Data Group (1994) for such cases. It amounts
to the usual weighted average, but increasing the error by a certain fac-
tor derived from the statistical significance of the discrepancies. John-
son et al. then arrive at a world average
S17 (0) = (22.4 ± 2.1) eV b. (10.8)
Bahcall and Pinsonneault (1992, 1995) used this value, i.e. they used
a 1σ uncertainty of 9%. Turck-Chièze and Lopes (1993) used (22.4 ±
1.3stat ± 3.0syst ) eV b, i.e. they adopted an uncertainty of ±15%.
Solar Neutrinos 357
Xu et al. (1994) have discussed a unique relationship between S17 (0)

and the nuclear vertex constant of the overlap wave function for the
virtual decay 8 B → 7 Be+p; the nuclear vertex constant can be predicted
from other nuclear data. These authors’ calculation of S17 (E) agrees
remarkably well with the data points of Filippone et al. (1983) at low
energies, and with Vaughn et al. (1970) at higher energies; these are
the experiments which gave a low S17 (0). Xu et al. (1994) find an even
lower value of S17 (0) ≈ 17.6 eV b.
On the other hand, Brown, Csótó, and Sherr (1995) studied the
relationship between the Coulomb displacement energy for the A = 8,
J = 2+ , T = 1 state and S17 . They found a high value S17 (20 keV) =
(26.5 ± 2.0) eV b.
The S17 (0) factor was recently determined from the Coulomb dis-
sociation of 8 B, i.e. by the Primakoff-type process 8 B → 7 Be + p in
the electric field of a 208 Pb nucleus (Motobayashi et al. 1994). These
authors find a very low preliminary value of S17 (0) = (16.7 ± 3.2) eV b.
Langanke and Shoppa (1994) think that the true value may be signif-
icantly lower still if allowance is made for a possible E2 amplitude in
the Primakoff reaction. However, this possibility is heavily disputed by
Gai and Bertulani (1995); see also the reply by Langanke and Shoppa
(1995).
All of these new results are somewhat preliminary at the present
time. Independently of their ultimate status it is evident that the S17
factor must be considered the weakest link in the prediction of the
boron neutrino flux.
10.3 Observations
10.3.1 Absorption Reactions for Radiochemical Experiments
The longest-running solar neutrino experiment is the Homestake chlo-
rine detector which is based on a reaction proposed by Pontecorvo
(1948) and Alvarez (1949),
νe + 37 Cl → 37 Ar + e (threshold 0.814 MeV). (10.9)
The two other data-producing radiochemical experiments use gallium
as a target according to the reaction
νe + 71 Ga → 71 Ge + e (threshold 0.233 MeV). (10.10)
Because of its low threshold, the gallium experiments can pick up the
solar pp neutrino flux. The absorption cross sections as a function of
358 Chapter 10
neutrino energy are shown in Fig. 10.5 according to the tabulation of

Bahcall and Ulrich (1988).
Fig. 10.5. Neutrino absorption cross section on 37 Cl and 71 Ga according to

Bahcall and Ulrich (1988).
Because the absorption cross sections are steeply increasing func-

tions of energy while the predicted solar neutrino spectrum (Fig. 10.1)
steeply decreases, the low flux of boron neutrinos yields the dominant
contribution to the expected counting rate for chlorine, and a sizeable
contribution to gallium. This is illustrated in Fig. 10.6 where the pre-
dicted counting rates from solar neutrinos, integrated between energy
Eν and infinity, are shown as a function of Eν . These plots correspond
to the lower panel of Fig. 10.1 if the differential flux is weighted with
the relevant absorption cross section. It is customary to express the
absorption rate per nucleus in “solar neutrino units”
1 SNU = 10−36 s−1 , (10.11)
not to be confused with 1 SNu, the supernova unit, which quantifies

the rate of supernova occurrences in a galaxy. Because one measures
a fully integrated flux one only needs the absorption cross sections
folded with the spectra of the individual source reactions. For 37 Cl and
71
Ga they are given in Tab. 10.4. Multiplying the predicted fluxes with
these cross sections and applying a factor 1036 gives the absorption rate
in SNUs.
Solar Neutrinos 359
Table 10.4. Spectrally averaged absorption cross sections on 37 Cl and 71 Ga

for different solar source reactions according to Bahcall (1989), except for
8 B neutrinos on 37 Cl which is according to Garcia et al. (1991). The unit is
10−46 cm2 .
7 8 13 15
pp pep Be B N O
37
Cl 0 16 2.4 10,900 1.7 6.8
71
Ga 11.8 215 73.2 24,300 61.8 116
Fig. 10.6. Absorption rate by 37 Cl and 71 Ga of the solar neutrino flux at

Earth, integrated from a given neutrino energy to infinity. The predicted
flux is according to Bahcall and Pinsonneault (1995), the absorption cross
sections according to Bahcall and Ulrich (1988)—see Fig. 10.5.
360 Chapter 10
10.3.2 Chlorine Detector (Homestake)

Solar neutrino observations were pioneered by the chlorine detector of
Davis (1964) which is located in the Homestake Gold Mine at Lead,
South Dakota (U.S.A.). The target consists of about 615 tons of per-
chloroethylene (C2 Cl4 ), a cleaning fluid, from which argon is extracted
every few months by an intricate chemical procedure. Because 37 Ar
has a half-life of 35.0 days the amount of neutrino-produced 37 Ar be-
gins to saturate after a couple of months whence a longer exposure
time is not warranted. The extracted argon—only a few atoms in a
small amount of carrier gas—is then viewed by a proportional counter
which registers the Auger electrons which are ejected when 37 Ar de-
cays by electron capture. The introduction in 1970 of an electronic
system which analyzes the pulse rise time greatly enhanced the sensi-
tivity. Therefore, usually only the results after 1970 are quoted, be-
ginning with run 18. Except for a period from May 1985 to Octo-
ber 1986 where the experiment was down due to successive electrical
failures of the circulation pumps, data have been taken continuously
since 1967.
The known backgrounds for the experiment are cosmic-ray produced
37
Ar atoms which correspond to (0.29±0.08) SNU, and an average neu-
tron background corresponding to (0.13 ± 0.13) SNU. Because these
average backgrounds as well as the backgrounds of the proportional
Fig. 10.7. Distribution of the counting rates of 99 runs (18−117). The

black bars represent the data shown in Fig. 10.7, the shaded histogram the
expected distribution (adapted from Lande 1995). The absorption rate in
SNU is obtained by multiplying the argon production rate by 5.31.
Solar Neutrinos 361
Table 10.5. Predicted absorption rate (in SNU) by 37 Cl for different solar
source reactions. BP92 give “theoretical 3σ” uncertainties.
7 8 13 15
Diffusion pp pep Be B N O Total
BP95 He, metals 0.0 0.2 1.2 7.2 0.1 0.4 9.3+1.2
−1.4
BP92 He 0.0 0.2 1.2 6.2 0.1 0.3 8.0 ± 3.0
BP92 — 0.0 0.2 1.2 5.5 0.1 0.2 7.2 ± 2.7
TL93 — 0.0 0.2 1.1 4.6 0.1 0.2 6.4 ± 1.4
BP92 = Bahcall and Pinsonneault (1992).

TL93 = Turck-Chièze and Lopes (1993).
Fig. 10.8. Solar neutrino flux at the Homestake experiment (νe 37 Cl →

37 Ar e) for the past quarter century. The shaded band indicates the 1σ
uncertainty of the global best-fit average neutrino flux (Lande 1995).
counters must be subtracted, the counting rate attributed to solar neu-

trinos in a given run is sometimes found to be formally negative. In
those cases, a zero counting rate is adopted. The distribution of count-
ing rates for 99 runs (18−117) is shown in Fig. 10.7. The global average
given in Eq. (10.12) is based on a maximum-likelihood analysis which
includes the background. It would not be correct to average the data
points with individual background subtractions to obtain a global av-
erage signal.
362 Chapter 10
The recognized systematic errors are 1.5% for the extraction ef-
ficiency, 3% for the proportional counter efficiency, 3% for the cos-
mic-ray background, 5% for the neutron background, and 2% for the
proportional counter background, which amounts to a total of 7%
or 0.18 SNU.
The measured counting rate for the individual runs at Homestake
is shown in Fig. 10.8. The current global best-fit average for the solar
neutrino flux measurement, derived from runs 18−124 is (Lande 1995)
(2.55 ± 0.17stat ± 0.18syst ) SNU = (2.55 ± 0.25) SNU (10.12)
where the errors were combined in quadrature. The average argon

production rate per day in the detector is obtained by dividing the
SNUs by 5.31 so that it is found to be 0.48 37 Ar/day. Eq. (10.12) is
to be compared with the predictions from different solar models shown
in Tab. 10.5.
10.3.3 Gallium Detectors (SAGE and GALLEX)

The gallium experiments involve far more complicated chemical extrac-
tion procedures for the neutrino-produced 71 Ge which is not a noble
gas. The Soviet-American (now Russian-American) Gallium Experi-
ment (SAGE) used at first 27 tons, later 55 tons of metallic gallium
while the European GALLEX collaboration uses 100 tons of an aque-
ous gallium chloride solution, corresponding to 30.3 tons of gallium.
The SAGE experiment is located in the Baksan Neutrino Observatory
in Mount Andyrchi, Caucasus Mountains (Russia) while GALLEX is
located in the Gran Sasso tunnel near Rome (Italy). SAGE has been
taking data since January 1990, GALLEX since May 1991. The cur-
rent results of SAGE were published by Abdurashitov et al. (1994)
and Gavrin (1995), those of GALLEX by the GALLEX Collaboration
(1994, 1995b) and Kirsten (1995).
The half-life of 71 Ge is 11.43 days so that one needs only about
a three-week exposure, allowing for frequent extractions. SAGE has
already accumulated a total of 21 analyzed runs, GALLEX a total
of 39. In Fig. 10.9 the individual counting rates are shown as well
as the global best-fit averages and the distribution of counting rates
in 25 SNU bins. For the SAGE data, formally negative rates after
background subtraction are forced to zero. In both cases, recognized
background signals of 7−8 SNU were subtracted. The average counting
Solar Neutrinos 363
rates attributed to solar neutrinos are found to be (in SNU)

SAGE: 74 + 13/−12stat + 5/−7syst = 74 ± 14,
GALLEX: 77 ± 9stat + 4/−5syst = 77 ± 10, (10.13)
where systematic and statistical errors were added in quadrature. These
results are to be compared with the predictions from standard solar
models shown in Tab. 10.6.
Fig. 10.9. Solar neutrino flux at the gallium detectors (νe 71 Ga → 71 Ge e).
The shaded bands indicate a 1σ uncertainty of the global best-fit average
neutrino fluxes (statistical and systematic errors added in quadrature). In
the SAGE data, formally negative fluxes are forced to zero. In the right
panels, the bin size is 20 SNU.
364 Chapter 10
Table 10.6. Predicted absorption rate (in SNU) by 71 Ga for different solar
source reactions. BP92 give “theoretical 3σ” uncertainties.
7 8 13 15
Diffusion pp pep Be B N O Total
BP95 He, metals 69.7 3.0 37.7 16.1 3.8 6.3 137+8
−7
BP92 He 70.8 3.1 35.8 13.8 3.0 4.9 131.5+21
−17
BP92 — 71.3 3.1 32.9 12.3 2.7 4.3 127+19
−16
TL93 — 71.1 3.0 30.9 10.8 2.4 3.7 122.5 ± 7

The GALLEX experiment was subjected to an “on-off test” by

virtue of a laboratory νe source strong enough to “outshine” the Sun
in its local neutrino flux (GALLEX collaboration 1995a). To this end
a container with activated chromium was inserted in the center of the
tank containing the target fluid. The relevant isotope is 51 Cr which
decays to 51 V with a half-life of 27.71 d by electron capture. The νe
spectrum consists of four monoenergetic lines of energies 426 keV (9%),
431 keV (1%), 746 keV (81%), and 751 keV (9%). The dominating line
is very close to the solar beryllium line. An analysis of the first seven
exposures reveals a ratio between the measured and expected counting
rate of 1.04 ± 0.12. At the present time, four further extractions are
still being analyzed. Meanwhile, the source is being reactivated at the
Siloé nuclear reactor in Grenoble (France) in order to perform further
exposures.
The GALLEX collaboration has interpreted the source experiment
as a global test of their detector efficiency. Most recently, Hata and
Haxton (1995) have advocated a somewhat different view. They argue
that the gallium absorption cross section for the 746 keV line is poorly
known because of excited-state contributions which have not been di-
rectly measured, and which are more uncertain than had been acknowl-
edged in the previous literature. According to Hata and Haxton the
source experiment should not be taken as measuring, say, the GALLEX
extraction efficiency but rather as measuring the excited-state contri-
butions to the absorption cross section for beryllium neutrinos.
Solar Neutrinos 365
10.3.4 Water Cherenkov Detector (Kamiokande)

A water Cherenkov detector like the one located in the Kamioka metal
mine (Gifu prefecture, Japan) is a large body of water, surrounded
by photomultipliers which register the Cherenkov light emitted by rel-
ativistic charged particles. Solar neutrinos are detected by virtue of
their elastic scattering on the electrons bound in the water molecules,
ν + e → e + ν. (10.14)
This process has no significant threshold, although in practice the detec-
tion of the electrons is background-limited to relatively large energies.
The Kamiokande detector began its measurement of solar neutrinos
in January 1987 with an effective analysis threshold of about 9 MeV
which was reduced to about 7 MeV in mid 1988. The much larger
Superkamiokande detector, which is scheduled to begin data-taking in
April 1996, will have a threshold of about 5 MeV.
In contrast with the radiochemical detectors which are effectively
based on the charged-current reaction νe + n → p + e, the elastic
scattering on electrons is sensitive to all (left-handed) neutrinos and
antineutrinos. The cross section is given by the well-known formula
(e.g. Commins and Bucksbaum 1983)
' (
dσνe G 2 m e Eν me
= F A + B (1 − y)2 − C y , (10.15)
dy 2π Eν
where GF is the Fermi constant and the coefficients A, B, and C are
tabulated in Tab. 10.7. Further,
Eν − Eν$ Te 2Eν
y≡ = and 0 < y < , (10.16)
Eν Eν 2Eν + me
where Eν and Eν$ are the initial- and final-state neutrino energies while
Te is the kinetic energy of the final-state electron.
Table 10.7. Coefficients in Eq. (10.15) for elastic neutrino electron scattering.
Flavor A B C
νe (CV + CA + 2)2 (CV − C A )2 (CV + 1)2 − (CA + 1)2
νe (CV − CA ) 2 (CV + CA + 2)2 (CV + 1)2 − (CA + 1)2
νµ,τ (CV + C A )2 (CV − C A )2 CV2 − CA2
ν µ,τ (CV − CA ) 2 (CV + CA ) 2 CV2 − CA2
CV = − 12 + sin2 ΘW ≈ −0.04, CA = − 12 .
366 Chapter 10
For Eν > 5 MeV, i.e. for energies above the detection threshold
of Kamiokande and Superkamiokande, the total cross section is well
approximated by


 1 for νe ,



 1 for ν e ,
G2F me Eν Eν 2.4
σνe ≈ (A + 13 B) = 9.5×10−44 cm2 × 1
2π 10 MeV 
 6.2 for νµ,τ ,


 1
7.1
for ν µ,τ .
(10.17)
Of course, for a flux of ν e ’s such as that from a supernova collapse, the

dominant signal in a water Cherenkov detector is from the charged-
current process ν e + p → n + e+ (Chapter 11).
The νe elastic scattering cross section is strongly forward peaked
for Eν ' me . One easily finds
dσ me (1 + me /Eν )2 cos θ dσ
=4 2 2 2
, (10.18)
d cos θ Eν [(1 + me /Eν ) − cos θ] dy
where dσ/dy was given in Eq. (10.15) with
2 (me /Eν ) cos2 θ
y= . (10.19)
(1 + me /Eν )2 − cos2 θ
Here, θ is the angle between the direction of motion of the final-state
electron relative to the incident neutrino; one finds 0 ≤ cos θ ≤ 1. For
Eν = 5 MeV and 10 MeV this cross section is shown in Fig. 10.10,
normalized to unity for θ = 0. The electron keeps the direction of the
incident neutrino within θ < 1/2
∼ (2me /Eν ) .
In a water Cherenkov detector, the direction of the charged par-
ticles can be reconstructed from the ring of Cherenkov light hitting
the photomultipliers. Therefore, the direction of the incident neutrino
is known within an uncertainty which is determined by the angular
distribution of Fig. 10.10, and by the random motion of the electron
due to multiple scattering in the Coulomb fields of the medium con-
stituents. At an electron energy of 10 MeV, the Kamiokande detec-
tor has an angular resolution of about 28◦ for the direction of the
electron. For this energy, the electron keeps the neutrino direction
within about 18◦ so that the uncertainty is dominated by the electron
Coulomb scattering.
Apart from these effects which blur the reconstruction of the in-
cident neutrino direction, a water Cherenkov detector is an imaging
Solar Neutrinos 367
Fig. 10.10. Differential scattering cross section for νe e → eνe according to

Eq. (10.18), normalized to unity for θ = 0, where θ is the angle between the
incident neutrino and final-state electron.
device and thus a true “neutrino telescope.” Thus, for the first time
the Kamiokande detector actually proved that neutrinos are coming
from the direction of the Sun. The first main publication of these
results was by Hirata et al. (1991) from a data sample of 1040 live de-
tector days, taken between 1987 and 1990; an update including data
until July 1993 (a total of 1670 live detector days) was given by Suzuki
(1995). The angular distribution of the registered electrons relative to
the direction of the Sun is shown in Fig. 10.11. Even though there
remains an isotropic background, probably from radioactive impurities
in the water, the solar neutrino signal beautifully shows up in these
measurements.
In elastic neutrino-electron scattering, the energy distribution of the
kicked electrons is nearly flat so that the spectral shape of the incident
neutrino flux is only indirectly represented by the measured electron
spectrum. In Fig. 10.12 the electron recoil spectrum is shown for an
incident spectrum of 8 B neutrinos. A water Cherenkov detector can
resolve the energy of the charged particles from the intensity of the
measured light which for low-energy electrons is roughly proportional
to the energy. Therefore, the electron recoil spectrum from the interac-
tion with solar neutrinos can be resolved at Kamiokande. The measured
shape relative to the theoretically expected one is shown in Fig. 10.13,
arbitrarily normalized at Eν = 9.5 MeV (Suzuki 1995). Within statis-
tical fluctuations the agreement is perfect.
368 Chapter 10
Fig. 10.11. Angular distribution of the solar neutrino events at Kamiokande

for 1670 live detector days (January 1987−July 1993), according to Suzuki
(1995). There remains an isotropic background, probably from radioactive
impurities in the water.
Fig. 10.12. Normalized spectra of 8 B neutrinos and of recoil electrons (kinetic

energy Te ) from the elastic scattering process νe e → eνe .
Even though the Sun is thought to produce only νe ’s one may spec-
ulate that some of them are converted into ν e ’s on their way to Earth
by spin-flavor oscillations in magnetic fields (Sect. 8.4) or by matter-
induced majoron decays (Sect. 6.8). The ν e ’s would cause a signal by
the reaction ν e p → ne+ with an isotropic angular distribution. The
remaining measured isotropic background shown in Fig. 10.11 can thus
Solar Neutrinos 369
Fig. 10.13. Measured over expected spectrum of recoil electrons from solar
neutrinos at Kamiokande, arbitrarily normalized to unity at Eν = 9.5 MeV
(Suzuki 1995).
Fig. 10.14. Limit on a solar ν e flux relative to 8 B solar νe ’s from the isotropic
background shown in Fig. 10.11. The data used for this result extend until
August 1992 (Suzuki 1993).
be used to constrain the ν e flux. At different energies a 90% CL up-

per limit on the solar ν e flux relative to 8 B solar neutrinos is shown
in Fig. 10.14 according to Suzuki (1993). The best relative limit is
at Eν = 13 MeV where the ν e flux is less than 5.8% of 8 B solar νe ’s
at 90% CL. If the cause of the isotropic background could be reliably
identified these upper limits could be improved accordingly.
370 Chapter 10
Table 10.8. Predicted flux of 8 B neutrinos at Earth. For BP92, the uncer-
tainty is a “theoretical 3σ error.”
Model Diffusion 8
B Flux [106 cm−2 s−1 ]
BP95 He, metals 6.6+0.9
−1.1
BP92 He 5.7 ± 2.4
BP92 — 5.1 ± 2.2
TL93 — 4.4 ± 1.1

Fig. 10.15. Measured 8 B solar neutrino flux at Kamiokande (Suzuki 1993).

The shaded band gives the average with 1σ statistical and systematic errors
added in quadrature.
Next, one may determine the absolute flux of solar 8 B neutrinos.

As a function of time it is shown in Fig. 10.15. Assuming that the
spectrum is indeed that of 8 B neutrinos, the measured flux from the
data of January 1987 until July 1993 is (units 106 cm−2 s−1 )
2.89 + 0.22/−0.21stat ± 0.35syst = 2.89 ± 0.41 (10.20)
where the errors were added in quadrature. This measurement is to be

compared with the predictions shown in Tab. 10.8
Table 10.9. Current and near-future solar neutrino experiments.
Experiment Operationa Detection Thresh. Solar Measurementc Prediction

from until [MeV] flux BP95 TL93
Homestake 1967 (?) Radiochemical 0.814 not pp 9.3+1.2
Solar Neutrinos
2.55 ± 0.25 −1.4 6.4 ± 1.4

νe 37 Cl → 37 Ar e SNU
SAGE 1990 (?) Radiochemical 0.233 all 74 ± 14 SNU 137+8

−7 123 ± 7
νe 71 Ga → 71 Ge e 77 ± 10 SNU ... ...
GALLEX 1991 (?) ... ... ...
8
Kamiokande II 1987 (1996) H2 O Cherenkov 7 B 2.89 ± 0.41 6.6+0.9
−1.1 4.4 ± 1.1
νx e → e ν x (b ) × 106 cm−2 s−1
Superkamiok. (1996) (?) ... 5 ... — ... ...
8
SNO (1996) (?) D2 O Cherenkov B — ... ...
νe d → p p e (1.44)
νx e → e ν x (b ) 5
νx d → p n ν x (b ) 2.225
7
BOREXINO (1997) (?) Scintillator 0.25 Be — 5.2+0.3
−0.4 —
νx e → e ν x (b ) × 109 cm−2 s−1
a In
brackets anticipated. b ν = ν , ν , or ν . c Statistical and systematic 1σ errors added in quadrature.
x e µ τ
BP95 = Bahcall and Pinsonneault (1995), with helium and metal diffusion.
TL93 = Turck-Chièze and Lopes (1993), no diffusion.
371
372 Chapter 10
10.3.5 Summary
The main features and results of the solar neutrino experiments dis-
cussed in this section, and of near-future experiments to be discussed
in Sect. 10.9, are summarized in Tab. 10.9. More technical aspects can
be found in the original papers quoted in this section and in previous
papers by the referenced authors. Overviews of many experimental as-
pects can be found in Bahcall (1989), Davis, Mann, and Wolfenstein
(1989), and Koshiba (1992). The two theoretical predictions shown
are somewhat extreme in that TL93 yields lowish neutrino fluxes when
compared with BP92 (no diffusion), while BP95 with the inclusion of
helium and metal diffusion is presently at the upper end of what is
being predicted on the basis of standard solar models. It is clear, of
course, that including diffusion would also increase the TL93 fluxes.
10.4 Time Variations

10.4.1 Day-Night Effect
It is commonly assumed that the Sun is in a stationary state so that
the solar neutrino flux should be constant in time. One may still ex-
pect certain temporal variations of the measured flux. Between day
and night the line of sight between the Kamiokande detector and the
Sun intersects with different parts of the Earth. If neutrinos oscil-
late, for certain masses and mixing angles the Earth’s matter would
alter the oscillation pattern such that the counting rate at Kamiokande
would be expected to vary between day and night. The day rate is
found to be 0.90 ± 0.10stat ± 0.12syst times the average, the night rate
1.04 ± 0.10stat ± 0.12syst , i.e. there is no significant difference (Suzuki
1995). The radiochemical detectors with their long exposure times can-
not resolve a possible day-night difference.
10.4.2 Seasonal Variation

Because of the ellipticity of the Earth’s orbit the distance to the Sun
varies during the year from a minimum of 1.471×1013 cm in January
to a maximum of 1.521×1013 cm in July, i.e. it varies by ±1.67% from
its average during the year. Therefore, the solar neutrino flux varies
by ±3.3% from average between January and July. This effect is too
small to be observed by any of the present-day experiments.
This variation can be amplified if neutrinos oscillate. Notably, if
the vacuum oscillation length of the monochromatic 7 Be neutrinos is
Solar Neutrinos 373
of order the annual distance variation, the 7 Be flux measured in the

chlorine and gallium detectors could vary between zero and its predicted
full rate. Moreover, at different times of the year the neutrinos have
to traverse on average a different amount of terrestrial matter which
could affect neutrino oscillations and thus lead to an annual variation.
There is no evidence for such an effect in any of the detectors.
GALLEX reports a counting rate of (82 ± 15stat ) SNU for October−
March and (78 ± 14stat ) SNU for April−September, i.e. there is no sig-
nificant difference (GALLEX collaboration 1994).
A semiannual variation could be caused by nonstandard neutrino in-
teractions with the solar magnetic field (Sect. 10.7). The solar equato-
rial plane is at an angle of 7◦ 15$ relative to the Earth’s orbital plane (the
ecliptic). Around 7 June and 8 December, the solar core is viewed from
Earth through the solar equator where the magnetic field is thought
to be weaker than at higher latitudes. The Kamiokande II data (1040
live detector days in 1987−1990) were subdivided into three-months
periods which include (rate ΓI ) or exclude (ΓII ) the intersection points,
respectively. The relative difference in counting rate was found to be
(ΓI − ΓII )/(ΓI + ΓII ) = −0.06 ± 0.11stat ± 0.02syst , i.e. there was no in-
dication for a time variation (Hirata et al. 1991). The Homestake data
also do not show any evidence for a semiannual variation.
10.4.3 Correlation with Solar Cycle at Homestake

The Sun shows a prominent magnetic activity cycle which is thought
to be due to dynamo action within the convective surface layers, driven
by the nonuniform rotation of the Sun (Stix 1989). One of the best-
known manifestations of this activity is the cycle of sunspots, measured
by their total number appearing on the solar disk.54 A given solar cy-
cle lasts for about 11 years from one minimum of sunspot number to
the following; the first recorded cycle begins with the minimum around
A.D. 1755. Sunspots are caused by magnetic flux tubes which break
through the surface; they are thought to be manifestations of a sub-
surface toroidal magnetic field with opposite directions between the
southern and northern hemisphere. In addition, the Sun has a poloidal
(dipole) field. The fields reverse polarity after 11 years so that the full
magnetic cycle lasts 22 years.
54
In the following, the “number of sunspots” refers to the international sunspot
index according to the Zürich system where both spots and spot groups are counted
and averaged from the reports of many solar observatories. The sunspot index is
regularly published in Solar Geophysical Data.
374 Chapter 10
The solar neutrino measurements of the chlorine experiment seem

to anticorrelate with solar activity as first suggested by Subramanian
(1979) and Bazilevskaya, Stozhkov, and Charakhch’yan (1982). These
latter authors also noted a correlation with the primary cosmic proton
flux which, however, is known to correlate with the solar magnetic
cycle because the solar magnetic field affects the flux of charged cosmic
rays hitting the Earth. The cosmic-ray flux is not expected to affect
the chlorine experiment directly; the average cosmic-ray induced 37 Ar
production rate is about 0.055 day−1 , compared with an average solar
neutrino signal of 0.48 day−1 . Figure 10.16 shows the temporal variation
from 1970−1989 of the Homestake argon production rate, the integral
number of sunspots, and the neutron counts at the McMurdo station in
Antarctica. The latter are indicative of the primary cosmic-ray proton
flux with energies above 0.4 GeV.
While a certain degree of anticorrelation between the number of
sunspots and the argon production rate is plainly visible in Fig. 10.16,
there are two important questions: How significant is the effect? If it
is significant, what are its physical origins?
As for the latter question, the only plausible explanation55 that has
been put forth over the years is the hypothesis that neutrinos possess a
small magnetic moment which allows them to spin-precess into sterile
right-handed states in the presence of magnetic fields (Voloshin and
Vysotskiı̆ 1986; Voloshin, Vysotskiı̆, and Okun 1986a,b), and variations
of this scheme which allow for simultaneous spin and flavor oscillations
(Sect. 10.7). In this case one would expect some degree of a semiannual
variation as explained above, but none is found in the Homestake data.
The issue of statistical significance was addressed on the basis of the
1970−1989 data by Filippone and Vogel (1990), Bieber et al. (1990),
and Bahcall and Press (1991). These latter authors argued for rank-
ordering as a statistical method because of the likely nonlinear relation-
ship (if any) with indicators of solar activity, and because of the possibly
non-Gaussian nature of the experimental errors. They produced pairs
of an observed argon production rate and the corresponding sunspot
index. Next, the Homestake data and sunspot indices were replaced by
the ordinal rank in their respective data sets, i.e. the smallest value has
rank 1, the second-largest rank 2, and so forth. In order to discover
correlations they applied two standard rank statistical tests, the Spear-
55
The refractive term in the MSW conversion probability from convective currents
could play a role and could be coupled to solar activity (Haxton and Zhang 1991).
However, extreme conditions are required to explain the observed effect.
Solar Neutrinos 375
Fig. 10.16. 37 Ar production rate at Homestake, integral number of sunspots,

and neutron counts at McMurdo (Antarctica) which are indicative of the
primary cosmic-ray proton flux above 0.4 GeV; annual bins each. In the
center panel, the number of the solar cycle is also indicated. (Adapted from
Bieber et al. 1990.)
man rank-order correlation coefficient and Kendall’s tau (e.g. Press

et al. 1986). They found a moderate statistical significance level of
1.3% (Spearman) and 0.9% (Kendall) for the argon-production rate to
be correlated with the sunspot index.56
Most recently, a correlation between the 1970−1991 argon produc-
tion rate and the solar activity cycle was investigated by Oakley et al.
(1994). They used the Mt. Wilson 150-ft. tower magnetograms as an
indicator for the solar magnetic surface flux, both for the full solar disk
56
The significance level gives approximately the probability that a random shuf-
fling of the data gives an equally good or better (anti)correlation. A significance
level smaller than 1% is considered “highly significant.”
376 Chapter 10
and for the central 14◦ × 14◦ which is the region most significant for the
neutrino flight path to us. The disk-centered magnetic-field cycle lags
full-disk indicators by about 1 y. Oakley et al. (1994) found a signifi-
cance level of 0.001% for an anticorrelation between the Homestake rate
and the disk-centered magnetic flux. Including the flux from increas-
ingly higher latitudes reduced the significance level of the correlation.
With only high-latitude magnetic information the correlation was lost.
Another test of time variation is a comparison as in Fig. 10.7 be-
tween the expected and measured distribution of argon production
rates. A time variation would broaden the measured distribution (black
histogram) relative to the expectation for a constant neutrino flux
(shaded histogram). A certain degree of broadening is certainly com-
patible with Fig. 10.7, but a precise statistical analysis does not seem
to be available at the present time. At any rate, the rank-ordering re-
sult does not specify the amplitude of a time varying signal relative to a
constant base rate while the width of the distribution in Fig. 10.7 would
be sensitive mostly to this amplitude. Therefore, the two methods yield
rather different information concerning a possible time variation.
The gallium experiments have not been running long enough to say
much about a time variation on the time scale of several years. Also,
there does not seem to be a significant time variation in the Kamiokande
data between 1987 and 1992 which covers a large fraction of the cur-
rent solar cycle No. 22. Notably, there is no apparent (anti)correlation
with sunspot number (Suzuki 1993). However, a correlation with disk-
centered magnetic indicators may yield a different result—according to
the analysis of Oakley et al. (1994) the use of a disk-centered indicator
with its inherent time-lag relative to full-disk spot counts is crucial for
the strong anticorrelation at Homestake.
The constancy of the Kamiokande rate, if confirmed over a more
extended period, severely limits a conjecture that the nuclear energy
generating region in the Sun was not constant (e.g. Raychaudhuri 1971,
1986). Because the Kamiokande detector is mostly sensitive to the
boron neutrinos with their extreme sensitivity to temperature, they
would be expected to show the most extreme time variations if the
solar cycle was caused by processes in the deep interior rather than by
dynamo action in the convective surface layers.
10.4.4 Summary
There is no indication for a day-night variation of the solar neutrino
flux at Kamiokande and none for a seasonal variation at Kamiokande,
Solar Neutrinos 377
GALLEX, or Homestake. The Homestake argon production rate ap-

pears to be strongly anticorrelated with solar disk-centered indicators
of magnetic activity, with no evidence for a long-term variation at
Kamiokande. However, the period covered there (1987−1992) may be
too short, and the Kamiokande rate has not been correlated with disk-
centered indicators. No homogeneous analysis of both Kamiokande and
Homestake data relative to solar activity exists at the present time.
It is very difficult to judge what the apparent Homestake anticorre-
lation with solar activity means. Is there an unrecognized background
which correlates with solar activity such as cosmic rays? Is there a
long-term drift in some aspect of the experiment which then naturally
correlates with other causally unrelated long-term varying phenomena
such as solar activity? Is it all some statistical fluke? Or is it an indica-
tion of nonstandard neutrino properties, notably magnetic dipole mo-
ments which spin precess in the solar magnetic field into sterile states?
This latter possibility will be elaborated further in Sect. 10.7.
10.5 Neutrino Flux Deficits
10.5.1 Boron Flux
Even if one ignores for the moment a possible time variation of the
Homestake neutrino measurements there remain several “solar neutrino
problems.” All of the solar neutrino flux measurements discussed in
Sect. 10.3 exhibit a deficit relative to theoretical predictions. The sim-
plest case to interpret is that of Kamiokande because it is sensitive only
to the boron flux. In this case the most uncertain input parameter is
the cross section for the reaction p7 Be → 8 Bγ which enters the flux pre-
diction as a multiplicative factor, independently of other details of solar
modelling. As discussed in Sect. 10.2.3, the astrophysical S-factor for
this reaction depends on theoretical extrapolations to low energies of
experimental data which themselves seem to exhibit relatively large sys-
tematic uncertainties. Therefore, one may turn the argument around
and consider the solar neutrino flux measurement at Kamiokande as
another determination of S17 (0).
According to Eq. (10.20) the measured flux is 2.89 × (1 ± 0.14) in
units of 106 cm−2 s−1 while the prediction is 4.4×(1±0.21)×S17 (0)/22.4
for TL93 (Turck-Chièze and Lopes 1993) where S17 is understood in
units of eV b. For the BP95 (Bahcall and Pinsonneault 1995) it is
6.6 × (1 ± 0.15) × S17 (0)/22.4 where I have used an average symmetric
378 Chapter 10
error for simplicity. This yields

)
14.7 ± 3.7 TL93,
S17 (0)/eV b = (10.21)
9.8 ± 2.0 BP95,
to be compared with the world average 22.4 ± 2.1 (Sect. 10.2.3). While
the discrepancy is severe, one would still be hard-pressed to conclude
with a reasonable degree of certainty that the solar neutrino problem
is not just a nuclear physics problem.
10.5.2 Beryllium Flux

The Kamiokande solar neutrino problem may be explained by a low
S17 (0) factor, perhaps in conspiration with lower-than-standard opaci-
ties that cause the central solar temperature to be lower than expected.
In this case the spectral shape would remain unchanged, leading to an
unambiguous prediction for the neutrino signal that must be caused by
8
B neutrinos in the chlorine and gallium detectors. This contribution
may be subtracted from the Homestake and SAGE/GALLEX data, re-
spectively, to obtain a measurement of the remaining neutrino sources
(Tab. 10.10)—see Kwong and Rosen (1994). Moreover, the predicted pp
and pep signals in these detectors do not seem to involve any significant
uncertainties so that they may be subtracted as well, effectively leading
to a measurement of the flux of 7 Be and CNO-neutrinos (Tab. 10.10).
These fluxes are then found to be formally negative; they are consistent
with zero within the experimental measurement errors.
These remaining fluxes inferred from the chlorine and gallium ex-
periments are to be compared with the predictions for the 7 Be and CNO
neutrinos. In Tab. 10.10 the predictions of BP95 are shown; those of
TL93 and other authors are similar. Bahcall (1994b) has compiled the
predictions for the 7 Be flux from a heterogeneous set of 10 solar models
by different authors whose predictions agree to within ±10%—there is
a broad consensus on the standard value of this flux. Therefore, the
errors of the predictions are much smaller than the experimental ones,
leaving one with a very significant discrepancy between the predicted
and observed flux both at Homestake and at SAGE/GALLEX, even if
one were to ignore the CNO contributions entirely.
In the analysis of Hata and Haxton (1995) which uses the GALLEX
source experiment as a constraint on the gallium absorption cross sec-
tion, it is found that at the 99% CL the measured 7 Be flux is less
than 0.3 of the BP95 prediction. This high significance of the beryl-
lium problem relies on the assumption that all solar neutrino experi-
Solar Neutrinos 379
Table 10.10. Measured vs. predicted 7 Be and CNO neutrino fluxes (in SNU).
Homestake GALLEX/
(37 Cl) SAGE (71 Ga)
Measurements:
Total 2.55 ± 0.25 77 ± 10
8
B (inferred from 3.15 ± 0.44 7±1
Kamiokande)
pp + pep 0.20 ± 0.01 74 ± 1
(calculated)
Remainder −0.8 ± 0.5 −4 ± 10
BP95 prediction:a
7
Be 1.24 37.7
CNO 0.48 10.1
Total 1.72 47.8
a Bahcall and Pinsonneault (1995).
ments are correct within the acknowledged uncertainties. If one ignored

Homestake the 99% CL limit would be at 0.5 of the BP95 prediction.
Therefore, the “new solar neutrino problem” consists in the mea-
sured absence of the beryllium neutrino flux. This is a far more serious
deficit than that of the boron flux which has been measured to be a
significant fraction of its expected value, and for which the prediction
is very uncertain anyway.
10.5.3 An Astrophysical Solution?

The beryllium problem is very significant because its predicted flux is
closely related to that of the boron flux. A glance at the solar reaction
chains (Fig. 10.2) reveals that both fluxes start with 7 Be,
7
Be + e → 7 Li + νe ,
7
Be + p → 8 B → 8 Be + e+ + νe . (10.22)
Therefore, the flux ratio of 7 Be/8 B neutrinos essentially measures the
branching ratio between the electron and proton capture on 7 Be, ad-
mittedly averaged over slightly different regions of the solar core. The
380 Chapter 10
boron and beryllium fluxes respond very differently to a modification

of the solar central temperature (Eq. 10.5). Because the boron flux
has been measured, even at a reduced strength, it is not possible to
explain the missing beryllium neutrinos by a low solar temperature un-
less one is willing to contemplate simultaneously an extremely reduced
temperature and an extremely enhanced S17 factor.
Many authors57 have recently investigated the question of the sig-
nificance of the solar neutrino problem, and if it can be solved by a
plausible, or even implausible, combination of erroneous nuclear cross
sections, solar opacities, and so forth. The consensus is that an as-
trophysical solution is not possible, even if one ignores part of the ex-
perimental data. There is no obvious uncertain input quantity into
solar modelling or the flux predictions that could be tuned to obtain
the measured fluxes. Therefore, something mysterious is wrong with
the calculation of the solar neutrino fluxes and/or the solar neutrino
experiments.
However, if our understanding of the solar neutrino source and of
the detection experiments is not totally wrong, an attractive alternative
is to contemplate the option that something happens to the neutrinos
as they propagate from the solar core to us.
10.6 Neutrino Oscillations

10.6.1 Which Data to Use?
The idea to test the hypothesis of neutrino oscillations by means of
the solar neutrino flux goes back to Pontecorvo (1957, 1958, 1967) and
Maki, Nakagawa, and Sakata (1962). After the first measurements
in 1968 it was revived by Gribov and Pontecorvo (1969) as well as
Bahcall and Frautschi (1969). If some of the νe ’s produced in the Sun
transformed into ν e ’s, into other sequential neutrinos (νµ , ντ ), or into
hypothetical sterile states νs , the measurable flux at Earth would be
depleted. Even the Kamiokande detector which responds to νµ and ντ
because of the possibility of neutral-current ν-e scattering would show
a reduced flux because of the smaller cross section. The measured flux
57
These include Bludman, Kennedy, and Langacker (1992a,b); Bludman et al.
(1993); Bahcall and Bethe (1993); Castellani, Degl’Innocenti, and Fiorentini
(1993); Castellani et al. (1994b); Hata, Bludman, and Langacker (1994); Shi and
Schramm (1994); Shi, Schramm, and Dearborn (1994); Kwong and Rosen (1994);
Bahcall (1994b); Berezinski (1994); Degl’Innocenti, Fiorentini, and Lissia (1995);
Parke (1995); Hata and Haxton (1995); Haxton (1995).
Solar Neutrinos 381
deficits in all detectors may give us the first indication for neutrino
oscillations and thus for nonvanishing neutrino masses and mixings.
Because of the statistically high significance of the anticorrelations
with solar activity of the Homestake results it is not entirely obvious
how to proceed with a quantitative test of the oscillation hypothesis
which can cause only a day-night or semiannual time variation. In the
present section a possible long-term variability of the solar neutrino
flux or its detection methods is ignored, i.e. the apparent anticorrela-
tion with solar activity at Homestake is considered to be a statistical
fluctuation. In Sect. 10.7 a possible explanation in terms of neutrino
interactions with the solar magnetic field is considered.
One may take the opposite point of view that the variability at
Homestake is a real effect, and that it is not related to neutrino mag-
netic moments because that hypothesis requires fairly extreme values
for the dipole moments and solar magnetic fields (Sect. 10.7). An in-
terpretation of the data in terms of neutrino oscillations then becomes
difficult because one admits from the start that unknown physical ef-
fects are either operating in the Sun, in the intervening space, or in
the detectors. One could argue perhaps that the apparent variability
of the Homestake data in itself was evidence that something was wrong
with this experiment. In this case one could still test the hypothesis
of neutrino oscillations under the assumption that the signal recorded
at Homestake was spurious in which case one must discard the entire
data set, not only parts of it as has sometimes been done. Ignoring
the Homestake data does not solve the solar neutrino problem, but its
significance is reduced.
Still, as no one has put forth a plausible hypothesis for a specific
problem with the Homestake experiment it is arbitrary to discard the
data. Admittedly, it is also arbitrary to consider the time variation
spurious even though standard statistical methods seem to reveal a
highly significant anticorrelation with solar activity.
10.6.2 Vacuum Oscillations

A formal treatment of vacuum neutrino oscillations was presented in
Sect. 8.3. If the observer is many oscillation lengths away from the
source, and if the neutrinos are produced with a broad spectrum of en-
ergies from an extended source, one will observe an incoherent mixture
of the mass eigenstates. If one considers two-flavor oscillations νe ↔ νa
with νa = νµ , ντ , or some hypothetical sterile flavor, the νe flux is re-
duced by a factor 1 − 12 sin2 2θ (vacuum mixing angle θ). Therefore,
382 Chapter 10
one needs a large mixing angle to achieve a substantial suppression of

the measurable flux. If one includes the possibility of three-flavor os-
cillations one can achieve, in principle, a maximum νe reduction by a
factor 13 .
In this picture the νe flux is reduced by the same factor for all ener-
gies, contrary to what is indicated by the observations. Therefore, one
needs an oscillation length of order the Earth-Sun distance (astronom-
ical unit 1 AU = 1.496×10−13 cm) because in this case the oscillation
pattern is not completely smeared out. Then one may measure a differ-
ent flux suppression for different energies; the wiggles in the reduction
factor as a function of neutrino energy can be resolved (Fig. 10.17).
For two-flavor oscillations the relationship Eq. (8.15) between neu-
trino energy Eν , the mass-square difference ∆m2ν , and the oscillation
length (osc can be written as
1 AU Eν
∆m2ν = 1.66×10−10 eV2 . (10.23)
(osc 10 MeV
Therefore, the neutrino masses are very small if these “long-wavelength
oscillations” are the explanation for the observed flux deficits.
Fig. 10.17. Survival probability of solar νe ’s for ∆m2ν = 0.8×10−10 eV2

and sin2 2θ = 0.8, parameters which reconcile all measurements with the
standard solar flux predictions. The solid line is for the minimum annual
distance to the Sun of 1.471×1013 cm (January), the dashed line for the
maximum distance 1.521×1013 cm (July). The right panel is an enlarged
section around the line of beryllium neutrinos. Between January and July,
their flux is suppressed by very different amounts.
Solar Neutrinos 383
The νe survival probability as a function of energy is shown in

Fig. 10.17 for ∆m2ν = 0.8×10−10 eV2 and sin2 2θ = 0.8. The Sun is
treated as a point-like source because the extension of its neutrino-
producing core is minute relative to the oscillation length of order the
Earth-Sun distance. The solid line is for the minimum, the dashed line
for the maximum annual distance to the Sun. The nearly monochro-
matic beryllium neutrinos (Eν = 0.862 MeV) easily retain their phase
relationship because they are distributed in energy only over a width
of about 2 keV (Fig. 10.4). Therefore, one would expect a pronounced
annual flux variation which is not observed, thereby reducing the al-
lowed range of masses and mixing angles where long-wavelength oscil-
lations could solve the solar neutrino problem. The analysis of Barger,
Phillips, and Whisnant (1992) and more recently of Krastev and Petcov
(1994) reveals that there remain a few small parameter pockets with
sin2 2θ around 0.8−1 and ∆m2ν around 0.6−0.9×10−10 eV2 . Krastev
and Petcov (1994) also conclude that oscillations into sterile neutrinos
are slightly disfavored.
While the spectral deformation of the 8 B neutrino spectrum is quite
dramatic, the spectrum of recoil electrons smears out most of this effect
(Fig. 10.18). Notably, above the Kamiokande detection threshold of
about 7 MeV the expected distortion is too small to be detected and
Fig. 10.18. Spectrum of 8 B neutrinos and of recoil electrons. Dotted lines:

Standard spectrum from Fig. 10.12. Solid lines: Surviving νe spectrum after
oscillations (suppression factor from Fig. 10.17) and modified electron recoil
spectrum. For oscillations into νµ,τ one would have to include their neutral-
current scattering on electrons with a cross section about 16 that of νe .
384 Chapter 10
so the spectral shape measured at Kamiokande (Fig. 10.13) does not

further constrain the vacuum oscillation hypothesis.
Barger, Phillips, and Whisnant (1992) as well as Krastev and Pet-
cov (1994) gave precise confidence contours in the sin2 2θ-∆m2ν -plane
where the solar neutrino problems are solved. However, some of the
input information involves large systematic uncertainties so that it is
well possible that, for example, a revised S17 factor would shift the
allowed regions beyond their stated confidence contours—see, for ex-
ample, Berezhiani and Rossi (1995). The main message is that there re-
main parameters in the quoted range where vacuum oscillations (“just
so oscillations”) could reconcile all solar flux measurements with the
standard flux predictions, except for the apparent anticorrelation of
the Homestake data with solar activity.
10.6.3 Resonant Oscillations (MSW Effect)

The solar neutrino problem has become tightly intertwined with the is-
sue of neutrino oscillations thanks to the work of Mikheyev and Smirnov
(1985) who showed that even for small mixing angles one can achieve
a large rate of flavor conversion because of medium-induced “resonant
oscillations.” This “MSW effect” was conceptually and quantitatively
discussed in Sect. 8.3.
For a practical application to the solar neutrino problem two main
features of the MSW effect are of great relevance. One is the bathtub-
shaped suppression function of Fig. 8.10 which replaces a constant
reduction factor (short-wavelength vacuum oscillations) or the wiggly
shape of Fig. 10.17 (long-wavelength vacuum oscillations). It implies
that νe ’s of intermediate energy can be reduced while low- and high-
energy ones are left relatively unscathed. This is what appears to be in-
dicated by the observations with the beryllium neutrinos more strongly
suppressed than the pp or boron ones.
Another important feature are the triangle-shaped suppression con-
tours in the sin2 2θ-∆m2ν -plane for a fixed neutrino energy (Fig. 8.9).
Each experiment produces its own triangular band of neutrino param-
eters where its measured rate is reconciled with the standard flux pre-
diction. As the signal in different detectors is dominated by neutrinos
of different energies, these triangles are vertically offset relative to one
another and so there remain only a few intersection points where all
experimental results are accounted for. Therefore, as experiments with
three different spectral responses are now reporting data the MSW so-
lution to the solar neutrino problem is very well constrained.
Solar Neutrinos 385
The allowed ranges from the different experiments as well as the

combined allowed range are shown in Fig. 10.19 according to the anal-
ysis of Hata and Haxton (1995). They used the Bahcall and Pinson-
neault (1995) solar model with helium and metal diffusion and its un-
certainties. The black areas represent the 95% CL allowed range if one
includes all the experimental information quoted in Sect. 10.3, includ-
ing the GALLEX source experiment. The center of the allowed area
for the “Large-Angle Solution” is roughly
sin2 2θ = 0.6,
∆m2ν = 2×10−5 eV2 , (10.24)
while for the “Nonadiabatic Solution” or “Small-Angle Solution” it is
sin2 2θ = 0.6×10−2 ,
∆m2ν = 0.6×10−5 eV2 . (10.25)
Both cases amount essentially to the same neutrino mass-square differ-

ence. Other recent and very detailed analyses are by Fiorentini et al.
(1994), Hata and Langacker (1994), and Gates, Krauss, and White
(1995) who find similar results. In detail, their confidence contours
differ because they used other solar models or a different statistical
analysis.
A certain region of parameters (dotted area in Fig. 10.19) can be
excluded by the absence of a day/night difference in the Kamiokande
counting rates which are expected for the matter-induced back oscilla-
tions into νe when the neutrino path intersects with part of the Earth
at night. (Even though the detector is deep underground in a mine
the overburden of material during daytime is negligible.) The excluded
range of masses and mixing angles is independent of solar modelling as
it is based only on the relative day/night counting rates.
Except for the Kamiokande day/night exclusion region one should
be careful at taking the confidence contours in a plot like Fig. 10.19 too
seriously. The solar model uncertainties are dominated by systematic
effects which cannot be quantified in a statistically objective sense. One
of the main uncertain input parameters is the astrophysical S-factor
for the reaction p 7 Be → 8 B γ. Speculating that all or most of the
missing boron neutrino flux is accounted for by a low S17 has the effect
of reducing the best-fit sin2 2θ of the small-angle solution by about a
factor of 3 while the large-angle solution disappears, at least in the
analysis of Krastev and Smirnov (1994). The best-fit ∆m2 , however,
386 Chapter 10
Fig. 10.19. Allowed range of neutrino masses and mixing angles in the neu-
trino experiments if the flux deficit relative to the Bahcall and Pinsonneault
(1995) solar model is interpreted in terms of neutrino oscillations. The ex-
perimental data include all summarized in Sect. 10.3. (Plot adapted from
Hata and Haxton 1995.)
is rather stable against variations of S17 (Krastev and Smirnov 1994;

Berezinsky, Fiorentini, and Lissia 1994).
Another approach is to allow S17 to float freely when performing a
maximum-likelihood analysis, i.e. to fit it simultaneously with the neu-
trino parameters from all solar neutrino experiments (Hata and Lan-
gacker 1994). The best-fit value is found to be 1.43+0.65
−0.42 times the stan-
dard 22.4 eV b. Of course, the 95% CL range for the best-fit neutrino
parameters is now much larger, allowing any sin2 2θ between about
10−3 and 0.8. Notably the range of allowed large-angle solutions is
vastly increased.
In summary, the hypothesis of neutrino oscillations can beautifully
explain almost all experimental results to date. Only the curious an-
ticorrelation of the Homestake data with solar activity remains unac-
counted for. It is interpreted as a statistical fluctuation.
Solar Neutrinos 387
10.7 Spin and Spin-Flavor Oscillations

If the anticorrelation between disk-centered magnetic indicators of so-
lar activity and the event rate at Homestake is taken seriously, the only
plausible explanation put forth to date is that of magnetically induced
neutrino spin or spin-flavor transitions that were discussed in Sect. 8.4.
It was first pointed out by Voloshin and Vysotskiı̆ (1986) that the vary-
ing magnetic-field strength in the solar convective surface layers could
cause a time-varying depletion of the left-handed solar neutrino flux.
A refined discussion was provided by Voloshin, Vysotskiı̆, and Okun
(1986a,b) after whom this mechanism is called the VVO solution to
the solar neutrino problem.
Strong magnetic fields may also exist in the nonconvective interior
of the Sun. In principle, they could also cause magnetic oscillations
and thus reduce the solar neutrino flux (Werntz 1970; Cisneros 1971).
However, the amount of reduction would not be related to the magnetic
activity cycle which is confined to the convective surface layers with an
approximate depth of 0.3R" ≈ 2×1010 cm = 200,000 km. The oscilla-
tion length is given by Eq. (8.53) so that over this distance a complete
spin reversal is achieved for
µν BT ≈ 3×10−10 µB kG, (10.26)
where µB = e/2me is the Bohr magneton and BT is the magnetic-

field strength perpendicular to the neutrino trajectory. Because the
magnetic field is mainly toroidal, the condition of transversality is au-
tomatically satisfied.
Magnetic spin oscillations in vacuum do not have any energy depen-
dence and so the solar neutrino flux would be reduced by a common
factor for the entire spectrum. However, a large conversion rate is
only achieved if the spin states are nearly degenerate which is not the
case in media where refractive effects change the dispersion relation of
left-handed neutrinos. In the context of spin-flavor oscillations, degen-
eracy can be achieved by a proper combination of medium refraction
and mass differences (resonant spin-flavor oscillations). In this case
the diagonal elements of the neutrino oscillation Hamiltonian involve
energy-dependent terms of the form (m22 − m21 )/2Eν , causing a strong
energy dependence of the conversion probability. Therefore, the mea-
sured signals in the detectors as well as the time variation at Homestake
and the absence of such a variation at Kamiokande can all be explained
by a suitable choice of neutrino parameters and magnetic-field profiles
388 Chapter 10
of the Sun.58 The simple estimate Eq. (10.26) remains approximately

valid. A neutrino mass-square difference m22 −m21 below about 10−5 eV2
is required because of the small matter densities encountered in the con-
vective layers.
The bounds on neutrino dipole and transition moments discussed
in Sect. 6.5.6 yield µν < −12
∼ 3×10 µB so that a magnetic-field strength
>
B ∼ 100 kG is required in the solar convection zone. Typical field
strengths measured in sunspots where the flux breaks through the sur-
face are of a few kG. While this may not be representative of the large-
scale toroidal field, several general arguments suggest that 10 kG is a
generous upper limit (Shi et al. 1993). If there were much stronger
fields they would have to be confined to flux ropes which would not
be effective at inducing neutrino spin oscillations. Therefore, the spin
oscillation scenario would require anomalously large magnetic fields, in
conflict with the arguments presented by Shi et al., or dipole moments
in excess of what is allowed by the bounds of Sect. 6.5.6. Even smaller
dipole moments than 3×10−12 µB require novel neutrino interactions.
The solar magnetic field is believed to consist of two opposite flux
tori, separated by the solar equatorial plane. In the course of a year the
line of sight from Earth to the solar core varies between ±7◦ 15$ solar
latitude and so the neutrinos measured here traverse a field configu-
ration which varies in the course of a year. The predicted semiannual
flux variation (Voloshin, Vysotskiı̆, and Okun 1986a), however, has not
been confirmed by any of the experiments.
In the spin-flavor oscillation scenario involving Majorana neutrinos
the Sun would be a source for antineutrinos, some of which could be
ν e ’s by a combination of spin-flavor (νe → ν µ ) and flavor (ν µ → ν e )
oscillations. However, the Kamiokande data already yield restrictive
limits on a solar ν e flux (Barbieri et al. 1991; see also Fig. 10.14).
In summary, the magnetic spin oscillation scenario requires new neu-
trino interactions to generate large magnetic dipole moments, and new
astrophysics to allow for sufficiently strong magnetic fields in the so-
lar convection zone. Neither a semiannual variation of the flux nor ν e ’s
have been observed; each would have been a smoking gun for the occur-
rence of this effect. Therefore, one is led to disfavor the magnetic spin
oscillation scenario. Then, of course, one is back to a statistical fluctu-
ation as an explanation for the time structure of the Homestake data.
58
Recent detailed investigations were performed by Akhmedov, Lanza, and Petcov
(1993, 1995), Krastev (1993), Nunokawa and Minakata (1993), Guzzo and Pulido
(1993), and Pulido (1993, 1994). For a review of earlier works see Pulido 1992.
Solar Neutrinos 389
10.8 Neutrino Decay
A deficit of solar neutrinos measured at Earth can be related to neutrino

decay. However, the in-flight decay of neutrinos does not provide the
required deformation of the spectrum because the decay rate in the
laboratory system involves a Lorentz factor mν /Eν so that low-energy
neutrinos decay faster. If νe is a mixture of mass eigenstates only the
heavier one decays. It is possible that the heavy admixture decays fast
so that the spectrum is reduced by a constant factor. Even this extreme
case does not provide a good fit to the data. A detailed analysis of
different cases was performed by Acker and Pakvasa (1994) who found
that the in-flight decay solution was ruled out at the 98% CL, even
when allowing for the solar model uncertainties.59
Neutrino decays can yield a solar ν e flux which is, in principle,
measurable. Such a flux can be produced if neutrinos are Majorana
particles, and if they couple to majorons χ (Sect. 15.7). Some fraction
of ν → ν $ + χ decays flip the helicity of the neutrino so that the ν $ is
effectively a ν $ . Thus after an MSW conversion νe → νµ,τ one could
have decays νµ,τ → ν e + χ (Raghavan, He, and Pakvasa 1988). Even
without oscillations one can have matter-induced decays of the form
νe → ν e + χ as discussed in Sect. 6.8. Detailed predictions for the
ν e flux for this type of scenario were worked out by Berezhiani et al.
(1992) and by Berezhiani, Moretti, and Rossi (1993). The Kamiokande
detector has already produced limits on solar ν e ’s (Fig. 10.14), with
much better limits to be expected from Superkamiokande. However,
in view of other limits on the neutrino-majoron coupling Berezhiani,
Moretti, and Rossi (1993) found that it seemed unrealistic to hope for
a detectable solar ν e signal.
Malaney, Starkman, and Butler (1994) showed that in decays of
the form ν → ν $ + boson, final-state stimulation effects (“neutrino
lasing”) could enhance the decay rate. However, the best-motivated
case is that of majoron decays which involve a γ5 coupling. The shape
of the resulting majoron spectrum is such that the crucial emission of
low-momentum bosons and thus the lasing effect is suppressed whence
Acker and Pakvasa’s conclusions remain valid. For models involving
scalar or vector bosons a detailed new analysis is required.
At the present time it looks rather unconvincing that the solar neu-
trino problem is related to some form of neutrino decays.
59
See Acker and Pakvasa (1994) for references to earlier discussions of neutrino
decay as a potential solution to the solar neutrino problem.
390 Chapter 10
10.9 Future Experiments

10.9.1 Superkamiokande
If some form of neutrino oscillations are the explanation for the mea-
sured solar flux deficits relative to standard predictions, how are we
ever going to know for sure? One needs to measure a signature which
is characteristic only for neutrino oscillations. The most convincing
case would be a measurement of the “wrong-flavored” neutrinos, i.e.
the νµ or ντ appearance rather than the νe disappearance. Other clear
signatures would be a deformation of the 8 B spectrum or a diurnal or
seasonal flux variation.
The latter cases can be very well investigated with the Superkamio-
kande detector which is scheduled to begin its operation in April of
1996. It is a water Cherenkov detector like Kamiokande, with about
20 times the fiducial volume. With about twice the relative coverage
of the surface area with photocathodes and a detection threshold as
low as 5 MeV it will count about 30 events/day from the solar boron
neutrino flux, as opposed to about 0.3 events/day at Kamiokande.
Fig. 10.20. Expected signal at night relative to the average daytime sig-
nal in a water Cherenkov detector as a function of the angle between the
Sun and the detector nadir. The large-angle example is for sin2 2θ = 0.7
and ∆m2ν = 1×10−5 eV2 , the small-angle case for sin2 2θ = 0.01 and
∆m2ν = 0.3×10−5 eV2 . Also shown are the existing Kamiokande measure-
ments and the expected Superkamiokande error bars after 1 month and 1 year
of running, respectively. (Adapted from Suzuki 1995.)
Solar Neutrinos 391
The small- and large-angle MSW solutions suggested by Fig. 10.19

would cause a day/night variation of the solar neutrino signal as indi-
cated in Fig. 10.20. The expected signal is shown in bins for the angle
between the Sun and the nadir of the detector, i.e. in bins of the inter-
section length of the neutrino flight path with the Earth. Also shown
are the current Kamiokande measurements, and the expected error bars
after 1 month and 1 year of Superkamiokande running time, respec-
tively. Shortly after Superkamiokande starts taking data one should be
able to decide whether the large-angle MSW solution applies!
The MSW solutions would also cause a spectral distortion of the re-
coil electron spectrum from the primary boron neutrinos. The expected
spectral shape relative to the standard one, arbitrarily normalized at an
electron kinetic energy of Te = 10 MeV, is shown in Fig. 10.21 for sev-
eral values of the assumed mixing angle. After several years of running,
Superkamiokande should be able to identify clearly the small-angle so-
lution if it applies.
If Superkamiokande measures neither a spectral distortion nor a
day/night effect, the deficiency of the boron flux probably would have
Fig. 10.21. Spectral distortion of the recoil electrons from the primary boron
neutrinos in a water Cherenkov detector. The ratio relative to the standard
spectrum is arbitrarily normalized at an electron kinetic energy of Te =
10 MeV. For the large-angle example (solid line) the assumed mass-square
difference is ∆m2ν = 2×10−5 eV2 , for the small-angle examples (broken lines)
it is ∆m2ν = 0.6×10−5 eV2 . The anticipated error bars after 5 years of
running Superkamiokande are also indicated for two energies. (Adapted
from Krastev and Smirnov 1994.)
392 Chapter 10
to be attributed to a small astrophysical S17 factor. In this case, the ex-

planation for the deficiency of beryllium neutrinos could not be resolved
by this detector.
10.9.2 Sudbury Neutrino Observatory (SNO)

The Sudbury Neutrino Observatory (SNO), also scheduled to take up
operation in 1996, is a heavy-water Cherenkov detector which is ex-
pected to be able to measure the appearance of “wrong-flavored” neu-
trinos if the MSW effect solves the solar neutrino problems (Sudbury
Neutrino Observatory Collaboration 1987; Lesko et al. 1993). The SNO
detector consists of 1000 tons of heavy water (D2 O) in a spherical acrylic
vessel of 12 m diameter, immersed in an outer vessel of ultrapure light
water (H2 O), surrounded by about 9600 photomultiplier tubes of 20 cm
diameter each. The detector is located 2000 m underground in the
Creighton mine, an operating Nickel mine, near Sudbury in Ontario
(Canada). Neutrinos can be detected by three different reactions in
this detector: by electron elastic scattering ν + e → e + ν and by the
deuterium dissociation reactions νe +d → p+p+e and ν +d → p+n+ν.
The electron elastic scattering reaction is analogous to the Kamio-
kande and Superkamiokande detectors: the recoiling electron is mea-
sured by the detection of its Cherenkov light. The effective detection
threshold is expected to be at 5 MeV as in Superkamiokande. This
reaction is sensitive to both νe and νµ,τ , albeit with a reduced cross
section for the latter (Eq. 10.17).
The charged-current deuterium dissociation νe d → ppe has a thresh-
old of 1.44 MeV, i.e. the final-state electron kinetic energy is essentially
Te = Eν − 1.44 MeV. The electron is detected by its Cherenkov light;
the effective energy resolution is about 20%. The angular distribution
relative to the incident neutrino is given by 1 − 13 cos Θ. The cross
section for this reaction is large. For an incident spectrum of boron
neutrinos one expects 9 times more electron counts above 5 MeV than
from electron elastic scattering; above 9 MeV even 13 times as many.
One can search for neutrino oscillations by a spectral distortion of
the electron spectrum, similar to Fig. 10.21 for Superkamiokande. In
fact, the spectral distortion is more pronounced as it is not washed out
by a broad final-state distribution of electron energies. Also, one can
search for a day/night effect.
The most important detection reaction, however, is the neutral-
current deuteron disintegration νd → pnν which has the same cross
section for all flavors and so it measures the total (left-handed) neutrino
Solar Neutrinos 393
flux above its threshold of 2.2 MeV, independently of the occurrence of

oscillations. The main problem here is the measurement of the final-
state neutron by the detection of γ rays from the subsequent neutron
capture, or by a neutron detector array in the heavy water. Ultrapure
water, acrylic, and other materials are needed to prevent an excessive
radioactive background that would spoil this measurement. One an-
ticipates to obtain the full unsuppressed 8 B flux with a precision of
about 1% after 5 years of operation. The measured ratio between the
charged-current and neutral-current deuterium disintegration will give
an immediate measure of the electron survival probability and thus of
the occurrence of neutrino oscillations.
If spin or spin-flavor oscillations occur such that a sizeable ν e flux
is produced, it can be detected by the reaction ν e + d → n + n + e+
which produces three detectable particles (Balantekin and Loreti 1992).
Of course, a possible ν e flux is already constrained by Kamiokande
(Fig. 10.14), and can be detected or constrained by Superkamiokande.
Suggestions for solar model independent methods of analyzing fu-
ture solar neutrino data were made, for example, by Spiro and Vignaud
(1990), Bilenky and Giunti (1993, 1994), and Castellani et al. (1994).
10.9.3 BOREXINO
Superkamiokande and SNO are both limited to a measurement of the
boron neutrino flux because of their relatively high detection thresholds.
If the boron flux is partly or mostly suppressed by a low S17 factor or
a low central solar temperature instead of neutrino oscillations these
experiments may have difficulties at identifying oscillations which would
still be indicated by the missing beryllium neutrino flux. Therefore, it is
interesting that another experiment (BOREXINO60 ) is being prepared
which would be sensitive dominantly to the beryllium neutrinos.
The main detection reaction is elastic ν-e scattering as in the light-
water Cherenkov detectors. However, the kicked electron is detected by
virtue of scintillation light rather than Cherenkov radiation, the former
60
The name of this experiment is derived from BOREX (boron solar neutrino
experiment), a proposed detector that was to use 11 B as a target (Raghavan, Pak-
vasa, and Brown 1986; see also Bahcall 1989). For a practical implementation it
was envisaged to use a boron loaded liquid scintillator (e.g. Raghavan 1990); be-
cause of the relatively small size of this detector the Italian diminutive BOREXINO
(baby BOREX) emerged. Ultimately, the idea of using a borated scintillator was
dropped entirely, leaving boron only in the name of the experiment. Confusingly,
then, BOREXINO is unrelated to a boron target, and also unrelated to the solar
boron neutrinos because the experiment is designed to hunt the beryllium ones.
394 Chapter 10
yielding about 50 times more light at the relevant energies below about
1 MeV. Naturally, as a target one needs to use an appropriate scintilla-
tor rather than water. The advantage of a lowered threshold is bought
at the price of losing all directional information. However, because of
the good energy resolution the monochromatic beryllium neutrinos at
862 keV should be clearly detectable as a distinct shoulder in the energy
spectrum of the recoil electrons. Optimistically, a scintillation detector
could have a threshold as low as Eν = 250 keV.
The main challenge at implementing this method is to lower the
radioactive contamination of the scintillator, its vessel, and the sur-
rounding water bath to an unprecedented degree of purity. For exam-
ple, the allowed mass fraction of 238 U of the scintillator is less than
about 10−16 g/g. The feasibility of this method is currently being stud-
ied at the CTF (Counting Test Facility) experiment, located in the
Gran Sasso underground laboratory. Assuming a positive outcome,
BOREXINO would be built, consisting of 300 tons of scintillator, sur-
rounded by 3000 tons of water. Optimistically, data taking with this
facility could commence in 1997.
10.9.4 Homestake Iodine Detector

Currently, a modular 100-ton iodine detector is under construction in
the Homestake mine. It is similar to the Homestake chlorine detector,
except that it uses the 127 I → 127 Xe transition to measure the νe flux. It
has an effective threshold of 0.789 MeV, similar to the chlorine detector.
However, for a standard solar neutrino flux the detection rate should
be about four times higher if the cross section calculations are correct.
The currently built detector should take up operation in mid-1995. It
may be expanded at a later time after running experience has been
obtained, and after the neutrino cross sections have been measured
(Bahcall et al. 1995; Engel, Krastev, and Lande 1995).
10.9.5 Summary
The hypothesis of neutrino flavor oscillations is strongly supported by
the results of all existing solar neutrino experiments. With the new
generation of detectors which will begin to take up operation in 1996
it looks plausible that nonstandard neutrino properties can be firmly
established on the basis of the solar neutrino flux before the mille-
nium ends.
Chapter 11
Supernova Neutrinos
The general physical picture of stellar collapse and supernova (SN)

explosions is described with an emphasis on the properties of the ob-
servable neutrino burst from such events. The measurements of the
neutrino burst from SN 1987A are reviewed. Its lessons for particle
physics are deferred to Chapter 13 except for the issue of neutrino
masses and mixings. Future possibilities to observe SN neutrinos are
discussed.
11.1 Stellar Collapse and Supernova Explosions

11.1.1 Stellar Collapse
A massive star (M > ∼ 8 M! ) inevitably becomes unstable at the end
of its life. It collapses and ejects its outer mantle in a SN explosion
as briefly described in Sect. 2.1.8. Within fractions of a second the
collapsing core forms a compact object at supranuclear density which
radiates its gravitational binding energy Eb ≈ 3×1053 erg within a few
seconds in the form of neutrinos. Gamow and Schoenberg (1940, 1941)
were the first to speculate that neutrino emission would be a major
effect in the collapse of a star. The only direct observation of neutrinos
from such an event occurred on 23 February 1987 when the blue su-
pergiant Sanduleak −69 202 in the Large Magellanic Cloud exploded in
what became known as SN 1987A. The neutrinos, and possibly other
low-mass particles, emitted from a collapsing star are the main topic of
this chapter, with SN 1987A playing a primary role.
Before an evolved massive star collapses, its core is a degenerate
configuration made up of iron-group elements. They cannot release
nuclear energy by fusion as they are already the most tightly bound
395
396 Chapter 11
nuclei so that no further nuclear burning stage can be ignited. The

precollapse inner “iron white dwarf” has a mass of about 1.5 M! , a
central density of about 3.7×109 g cm−3 , a central temperature of about
0.69 MeV, and a number fraction of electrons per baryon of Ye ≈ 0.42
(Brown, Bethe, and Baym 1982). As the mass of this object grows
and its radius shrinks it reaches its Chandrasekhar limit: relativistic
electrons cannot support a self-gravitating body.
In practice, a thermal pressure contribution cannot be neglected.
The collapse is triggered when the temperature has become so high
that the photodissociation of iron commences, γ + 56 Fe → 13α + 4n, a
reaction which consumes 124.4 MeV of energy. This energy loss reduces
the thermal contribution of the electron pressure. Therefore, compres-
sion yields a lesser pressure increase than would occur in the absence
of photodissociation. A star near its Chandrasekhar mass is close to
a point where the increased gravitational pull caused by a small con-
traction is no longer overcompensated by a large enough push from
the corresponding pressure increase. Therefore, a small reduction of
the adiabatic index Γ ≡ (∂ ln p/∂ ln ρ)s is enough to cause an insta-
bility. Once the collapse has begun, pressure support is also lost by
the capture of electrons on heavy nuclei which amounts to the reaction
e− + p → n + νe . It converts electrons to neutrinos which escape freely.
The inner part of the core (M ≈ 0.6 M! ) has Γ ≈ 43 . Its collapse is
homologous, i.e. it maintains its relative density profile (Goldreich and
Weber 1980). The collapse velocity is proportional to the radius with
v/r = 400 − 700 s−1 , yet it remains subsonic and so this part of the core
is in good communication with itself. The nearly free fall of the outer
part is supersonic.
At a certain density the neutrinos will no longer be able to stream
freely from the core. When their diffusion time exceeds a dynamical
collapse time scale they will be trapped (Mazurek 1974, 1975, 1976;
Sato 1975). Neutral-current scatterings on large nuclei are particularly
effective at trapping neutrinos because the cross section is coherently
enhanced (Freedman 1974).61 One finds a trapping density of around
1012 g cm−3 for 10 MeV neutrinos (Brown, Bethe, and Baym 1982). The
neutrino trapping radius as a function of time is shown as a dotted line
in Fig. 11.1.
61
Neutrinos with 10 MeV energies cannot “resolve” the nucleus, causing it to act
as a single scattering center. Because the neutral-current interaction with protons
is reduced by a factor 1 − 4 sin2 ΘW (weak mixing angle ΘW ) with sin2 ΘW ≈ 0.23
(Appendix B), the elastic scattering cross section of a nucleus scales with the square
of the neutron number.
Supernova Neutrinos 397
Fig. 11.1. Schematic picture of the core collapse of a massive star (M > ∼
8 M! ), of the formation of a neutron-star remnant, and the beginning of a
SN explosion. There are four main phases numbered 1−4 above the plot:
1. Collapse. 2. Prompt-shock propagation and break-out, release of prompt
νe burst. 3. Matter accretion and mantle cooling. 4. Kelvin-Helmholtz
cooling of “protoneutron star.” The curves mark the time evolution of several
characteristic radii: The stellar iron core (RFe ). The “neutrino sphere” (Rν )
with diffusive transport inside, free streaming outside. The “inner core”
(Ric ) which for t <∼ 0.1 s is the region of subsonic collapse, later it is the
settled, compact inner region of the nascent neutron star. The SN shock
wave (Rshock ) is formed at core bounce, stagnates for several 100 ms, and
is revived by neutrino heating—it then propagates outward and ejects the
stellar mantle. The shaded area is where most of the neutrino emission
comes from; between this area and Rν neutrinos still diffuse, but are no
longer efficiently produced. (Adapted from Janka 1993.)
Neutrino trapping has the effect that the lepton number fraction
YL is nearly conserved at the value Ye which obtains at the time of
trapping. However, electrons and electron neutrinos still interconvert
(β equilibrium), causing a degenerate νe sea to build up. The core of
a collapsing star is the only known astrophysical site apart from the
early universe where neutrinos are in thermal equilibrium. It is the
only site where neutrinos occur in a degenerate Fermi sea as the early
universe is thought to be essentially CP symmetric with equal numbers
of neutrinos and antineutrinos to within one part in 109 . When neutrino
trapping becomes effective, the lepton fraction per baryon is YL ≈ 0.35,
398 Chapter 11
not much lower than the initial iron-core value, i.e. not much of the
lepton number is lost during infall. The conditions of chemical and
thermal equilibrium dictate that neutrinos take up only a relatively
small part (about 14 ) of the lepton number (Appendix D.2). A typical
YL profile after collapse is shown in Fig. 11.2
The collapse is intercepted when the inner core reaches nuclear den-
sity (ρ0 ≈ 3×1014 g cm−3 ), a point where the equation of state stiffens.
Because the inner core collapse is subsonic, the information about the
central condition spreads throughout, i.e. the collapse of the entire ho-
mologous core slows down. However, this information cannot propagate
Fig. 11.2. Snapshots of the profiles of temperature T and lepton number

fraction YL in the collapsed core of a massive star. The indicated times are
in seconds after collapse. The shaded arrows in the upper panel indicate
the motion of the temperature maximum. For a soft equation of state the
maximum temperature can be up to about 70 MeV. (Adapted from Burrows
and Lattimer 1986.)
beyond the sonic point at the edge of the inner core which now encom-
passes about 0.8 M! . As material continues to fall onto the inner core
at supersonic velocities a shock wave builds up at the sonic point which
is at the edge of the inner core, not at its center. As more material
moves in, more and more energy is stored in this shock wave which
almost immediately begins to propagate outward into the collapsing
outer part of the iron core—see the thick solid line in Fig. 11.1. As-
suming that enough energy is stored in the shock wave it will eventually
eject the stellar mantle outside of what was the iron core. The rebound
or “bounce” of the collapse turns the implosion of the core into an
explosion of the outer star—a SN occurs.
This “bounce and shock” scenario of SN explosions was first pro-
posed by Colgate and Johnson (1960) and then elaborated by a number
of authors (see Brown, Bethe, and Baym 1982 and references therein).
In practice, however, the story of SN explosions appears to be more
complicated than this “prompt explosion scenario.” Neutrino losses
and the dissociation of the iron material through which the shock wave
propagates dissipate much of the shock’s energy so that in typical calcu-
lations it stalls and eventually recollapses. It is currently believed that
the energy deposition by neutrinos revives the shock wave, leading to
the “delayed explosion scenario” detailed in Sect. 11.1.3 below.
11.1.2 Deleptonization and Cooling

After core bounce and the formation of a shock wave the next dra-
matic step in the evolution of the core is when the outward propagating
shock breaks through the “neutrino sphere,” i.e. the shell within which
neutrinos are trapped, most effectively by the coherent scattering on
heavy nuclei. As the passage of the shock wave dissociates these nu-
clei, it is easier for neutrinos to escape. Moreover, the protons newly
liberated from the iron nuclei allow for quick neutronization by virtue
of e− + p → n + νe , causing a short νe burst which is often called the
“prompt νe burst” or “deleptonization burst” (phase No. 2 in Figs. 11.1
and 11.3). However, the material which is quickly deleptonized encom-
passes only a few tenths of a solar mass so that most of the leptons
remain trapped in the inner core (Fig. 11.2).
At this stage, the object below the shock has become a “protoneu-
tron star.” It has a settled inner core within the radius where the
shock wave first formed and which consists of neutrons, protons, elec-
trons, and neutrinos (lepton fraction YL ≈ 0.35). The protoneutron
star also has a bloated outer part which has lost a large fraction of
400 Chapter 11
Fig. 11.3. Schematic neutrino “lightcurves” during the phases of (1) core
collapse, (2) shock propagation and shock breakout, (3) mantle cooling and
accretion, and (4) Kelvin-Helmholtz cooling. (Adapted from Janka 1993.)
its lepton number during the νe burst at shock break-out. This outer
part settles within the first 0.5−1 s after core bounce, emitting most
of its energy in the form of neutrinos. Also, more material is accreted
while the shock wave stalls. As much as a quarter of the expected total
amount of energy in neutrinos is liberated during this phase (No. 3 in
Figs. 11.1 and 11.3).
Meanwhile, the stalled shock wave has managed to resume its out-
ward motion and has begun to eject the overburden of matter. There-
fore, the protoneutron star after about 0.5−1 s can be viewed as a star
unto itself with a radius of around 30 km which slowly contracts and
cools by the emission of (anti)neutrinos of all flavors, and at the same
time deleptonizes by the loss of νe ’s. After 5−10 s it has lost most
of its lepton number, and slightly later most of its energy. This is
the “Kelvin-Helmholtz cooling phase,” marked as No. 4 in Figs. 11.1
and 11.3. Afterward, the star has become a proper neutron star whose
small lepton fraction is determined by the condition of a vanishing
neutrino chemical potential (Appendix D.2), and whose further cooling
history has been discussed in Sect. 2.3.
Immediately after collapse the protoneutron star is relatively cold
(see the t = 0 curve in Fig. 11.2). Half or more of the energy to be ra-
diated later is actually stored in the degenerate electron Fermi sea with
typical Fermi momenta of order 300 MeV. The corresponding degener-
ate neutrino sea has its Fermi surface at around 200 MeV. The lepton
number profile (lower panel of Fig. 11.2) has a step-like form, with the
step moving inward very quickly when the shock breaks through the
neutrino sphere. The quick recession of the lepton profile during the
first 0.5 s represents deleptonization during the mantle cooling phase.
After this initial phase, however, the bloated outer part of the star
has settled; it is more difficult for neutrinos to escape from this compact
object. Still, the steep gradient of lepton number drives an outward dif-
fusion of neutrinos which move toward regions of lower Fermi momen-
tum and thus, of lower degeneracy energy. Therefore, they “downscat-
ter,” releasing most of the previous electron and neutrino degeneracy
energy as heat. Hence near the edge of the lepton number step the
medium is heated efficiently. In Fig. 11.2 it is plainly visible that the
temperature maximum of the medium is always in the region of the
steepest lepton number gradient.62 Therefore, the medium first heats
near the core surface, and then the temperature maximum moves in-
ward until it has reached the center. At this time the core is entirely
deleptonized; it continues to cool, the temperature maximum at the
center drops to obscurity.
The neutrino radiation leaving the star has typical energies in the
10 MeV range, compared with a 200 MeV neutrino Fermi energy in the
interior. Therefore, the loss of lepton number by itself is associated
with relatively little energy. Put another way, only a small excess of νe
over ν e is needed to carry away the lepton number. The total energy
is carried away in almost equal parts by each (anti)neutrino flavor.
11.1.3 Supernova Explosions

How do supernovae explode? For some time it was thought that the
outer layers of the star were ejected by the momentum transferred from
the outward neutrino flow which is released after the core collapse (Col-
gate and White 1966). This scenario had to be abandoned after the
discovery of neutral-current neutrino interactions which trap these par-
ticles so that they are released only relatively slowly. With the demise
of the neutrino explosion scenario the earlier suggestion of Colgate and
Johnson (1960) of a hydrodynamic shock wave driving the explosion be-
came the standard, the so-called “prompt explosion scenario” or “direct
mechanism.” It continued to malfunction, however, because in numer-
ical calculations the shock tended to stall because of energy dissipation
62
I thank David Seckel for explaining this point to me which is not usually stressed
in the pertinent literature.
402 Chapter 11
associated with the dissociation of the remaining iron shell it had to

work through before it could fly. Recall that the shock wave forms
relatively deep inside of the collapsing iron core of the progenitor star.
A possible solution is the “delayed mechanism” (Wilson 1983; Bethe
and Wilson 1985) where the shock wave lingers at a constant radius for
a few 100 ms and then takes off again (Fig. 11.1), powered both by the
accretion of material and by the energy deposition of the neutrino flow.
In this regard the dilute hot region (ρ of order 106 −108 g cm−3 , T of
order 1 MeV) below the stalled shock plays a major role at absorbing
neutrino energy.63 Unfortunately, it is not certain that the neutrino
flux can deposit enough energy to revitalize the shock as the energy
transfer is relatively inefficient. However, if one adjusts the amount of
neutrino energy transfer to a value above a rather well defined threshold
one obtains beautiful explosions (Fig. 11.4). It should be noted that
the spectacular explosion of a SN—which at the peak of its lightcurve
outshines an entire galaxy—is only a “dirt effect” relative to the release
of neutrino energy which equals the gravitational binding energy of
the newborn neutron star of about GN M/R ≈ 3×1053 erg with M ≈
1.4 M! and R ≈ 10 km. The total energy released in the kinetic energy
of the ejecta and in electromagnetic radiation is a few64 1051 erg, on the
order of 1% of the total neutron-star binding energy. Therefore, on
energetic grounds alone there is no problem at tapping the neutrino
flux for explosion energy.
A variety of schemes are currently being discussed to achieve a suc-
cessful shock revival. It is thought that convection may play a major
role at transporting energy to the surface of the bloated protoneutron
star which has formed after the break-out of the shock (e.g. Burrows
and Lattimer 1988; Wilson and Mayle 1988; Mayle and Wilson 1993;
Burrows and Fryxell 1992, 1993; Janka and Müller 1993b). It may be
that this mechanism can boost the effective neutrino luminosity for a
few hundred milliseconds, enough in some calculations to trigger an
explosion (see however Bruenn and Mezzacappa 1994). At any rate,
in the absence of a fundamental treatment of convection this method
is essentially one way of parametrizing the initial amount of neutrino
heating below the shock.
63
Goodman, Dar, and Nussinov (1987) proposed that the pair-annihilation pro-
cess νν → e+ e− might be the dominant mode of energy transfer. However, Cooper-
stein, van den Horn, and Baron (1987) critizised the neutrino emission parameters
of that study while Janka (1991) found that a proper treatment of the phase space
renders this process less efficient than had been originally thought.
64
This unit is sometimes referred to as 1 foe, for “ten to the fifty one ergs.”
Convection is also important at transporting energy within the di-

lute region between the protoneutron star and the stalling shock wave
(Herant, Benz, and Colgate 1992; Herant et al. 1994; Janka and Müller
1993a, 1994, 1995a,b; Sato, Shimizu, and Yamada 1993; Burrows,
Fig. 11.4. Unsuccessful (upper panel) and successful (lower panel) SN explo-
sion. In each case, the location of several mass shells is shown as a function
of time. The thick shaded line indicates the location of the shock. The only
difference between the two cases is the adjusted neutrino luminosity from a
central source; it was chosen as Lν = 2.10×1052 erg s−1 (upper panel) and
Lν = 2.20×1052 erg s−1 (lower panel), respectively. In the upper case, which
is just below threshold for a successful explosion, the shock displays an in-
teresting oscillatory behavior. (Curves courtesy of H.-T. Janka, taken from
Janka and Müller 1993a.)
404 Chapter 11
Fig. 11.5. Entropy contours between neutron star and shock wave in a 2-
dimensional calculation of a SN explosion. The entropy per nucleon is shown
in contours at equal steps of 0.5 kB between 5 and 16 kB , and in steps of 1 kB
between 16 and 23 kB . This snapshot represents model T2c of Janka and
Müller (1995b) at t = 377 ms after bounce. (Original of the figure courtesy
of H.-T. Janka.)
Hayes, and Fryxell 1995). For the first time 2- and 3-dimensional cal-
culations have become possible. They reveal a large-scale convective
overturn (Fig. 11.5) which helps at revitalizing the shock because it
brings hot material from depths near the neutrino sphere quickly up
to the region immediately behind the shock, and cooler material down
to the neutrino sphere where it absorbs energy from the neutrino flow.
Successful explosions can be obtained for amounts of neutrino heating
where 1-dimensional calculations did not succeed. The sharp transition
between failed and successful explosions as a function of neutrino heat-
ing that was found in 1-dimensional calculations (Fig. 11.4) is smoothed
out, but neutrino heating still plays a pivotal role at obtaining a suc-
cessful and sufficiently energetic explosion.
In summary, then, the current standard picture of SN explosions is
a modification and synthesis of the Colgate and Johnson (1960) shock-
driven and the Colgate and White (1966) neutrino-driven explosions.
At the present time there may still remain a quantitative problem at
obtaining enough neutrino energy deposition behind the shock wave to
guarantee a successful and sufficiently energetic explosion. It remains
to be seen if this scenario withstands the test of time, or if a novel
ingredient will have to be invoked in the future.
11.1.4 Nucleosynthesis
The universe began in a hot “big bang” which allowed for the formation
of nuclei from the protons and neutrons originally present in thermal
equilibrium with the ambient heat bath. The primordial abundances
“froze out” at about 22−24% helium, the rest hydrogen, and a small
trace of other light elements such as lithium. The present-day distri-
bution of elements was bred from this primeval mix mostly by nuclear
processes in stars; they eject some of their mass at the end of their lives
(Chapter 2), returning processed material to the interstellar medium
from which new stars and planets are born.
However, the normal stellar burning processes can produce elements
only up to the iron group which have the largest binding energy per
nucleon. Thus, the heavy elements must have been produced by dif-
ferent processes at different sites. It has long been thought that nu-
clei with A >∼ 70 were predominantly made by neutron capture, no-
tably the s- and r- (slow and rapid) processes (Burbidge et al. 1957;
Cameron 1957; Clayton 1968; Meyer 1994). The site for the occur-
rence of the r-process has remained elusive for the past three decades,
although many different suggestions have been made. The crux is that
406 Chapter 11
one needs to produce the heavy elements in the observed proportions,

and with a total amount compatible with a plausible galactic history.
It has long been held that the r-process elements were made in SN
explosions; in this case one needs a yield of about 10−4 M! of heavy
elements per SN.
Perhaps the first realistic scenario that appears to meet these re-
quirements is r-process nucleosynthesis in the hot bubble between a
protoneutron star and the escaping shock wave in a core-collapse SN
explosion at a time of a few seconds after core bounce (Woosley and
Hoffmann 1992; Meyer et al. 1992; Woosley et al. 1994; Witti, Janka,
and Takahashi 1994; Takahashi, Witti, and Janka 1994; Meyer 1995).
The material in this region is very dilute because of the successful ex-
plosion, yet very hot—around 109 K or 100 keV in the region where the
r-process is thought to occur. This hot bubble is not entirely empty
because of a neutrino-driven stellar wind. Therefore, one is talking
about a high-entropy environment (a few hundred kB per baryon), i.e.
a large number of photons per baryon (a few ten). For such con-
ditions the required neutron/proton ratio is achieved even for elec-
tron fractions of Ye ≈ 0.40 typical for the material outside of a col-
lapsed SN core.
This scenario appears to be qualitatively and quantitatively almost
perfect except that the necessary combination of entropy, electron frac-
tion Ye , and expansion time scale do not seem to be quite born out by
current numerical calculations. Whatever the explanation of this prob-
lem, it is fascinating that both the occurrence of a successful and suffi-
ciently energetic SN explosion as well as the occurrence of the r-process
in the high-entropy environment of the “hot bubble” seem to depend
crucially on the neutrino energy transfer which thus plays a dominant
role in this scenario. One may expect that r-process nucleosynthesis will
turn into a tool to calibrate the neutrino flux from a nascent neutron
star, and perhaps into a tool to study nonstandard neutrino proper-
ties (for a first example see Sect. 11.4.5). In effect, the distribution and
quantity of r-process elements gives us a measure of SN neutrino fluxes,
independent of direct observations! This is not unlike big-bang nucle-
osynthesis where the primeval light-element abundances have been an
extremely useful tool to study the properties of the primordial neutrino
heat bath (Kolb and Turner 1990). At the present time, of course, a
quantitative understanding of SN nucleosynthesis in conjunction with a
quantitative understanding of SN explosions is a field in its infancy—it
remains to be seen if it grows up to be as beautiful as big-bang nucle-
osynthesis.
11.2 Predicted Neutrino Signal

11.2.1 Overall Features
One of the most important aspects of SN physics relevant to particle
astrophysics is the immense flux of neutrinos liberated after the core
collapse. This flux has been measured from SN 1987A, and with luck
will be measured again from a galactic SN in the future. Therefore,
it is important to understand the neutrino signal to be expected from
this sort of event.
On a crude level of approximation one can understand the main fea-
tures of the overall neutrino signal on the basis of very simple physical
principles. The overall amount of energy to be expected is given by the
binding energy of the compact star that formed after collapse
! "2 ! "
3 GN M 2 M 10 km
Eb ≈ = 1.60×1053 erg . (11.1)
5 R M! R
It is reasonable to expect the energy to be equipartitioned among the
different neutrino flavors and so to expect about 16 Eb in each of the six
standard (anti)neutrino degrees of freedom. (Here and in the following
Newtonian physics is used; general relativistic corrections to energies,
temperatures, etc. as viewed from a distant observer can be as large as
several 10% due to gravitational redshifts.)
Neutrinos are trapped in the interior of the high-density neutron
star. Therefore, they are emitted from the relatively well defined surface
at a radius of 10−20 km, depending on the mass and the nuclear equa-
tion of state. As long as the material near the surface is nondegenerate
it must support itself against the local gravitational field by normal
thermal pressure. One may apply the virial theorem (Chapter 1) which
informs us that the average kinetic energy of a typical nucleon near
the neutron-star surface must be half of its gravitational potential, i.e.
2'Ekin ( ≈ GN M mN /R (nucleon mass mN ). With a neutron-star mass
of M = 1.4 M! and a radius R = 15 km one finds 'Ekin ( ≈ 25 MeV
or T = 23 'Ekin ( ≈ 17 MeV. Therefore, thermal neutrinos emitted from
the neutron-star surface are characterized by a temperature of order
10 MeV.
The duration of neutrino emission is a multiple of the neutrino dif-
fusion time scale over the dimension of the neutron-star radius,
tdiff ≈ R2 /λ, (11.2)
where λ is a typical mean free path. A typical neutral-current weak
408 Chapter 11
scattering cross section on nonrelativistic nucleons is given by

σ ≈ G2F Eν2 /π = 1.7×10−42 cm2 (Eν /10 MeV)2 . (11.3)
Nuclear density (ρ0 ≈ 3×1014 g cm−3 ) corresponds to a nucleon density
of about 1.8×1038 cm−3 so that λ ≈ 300 cm for 30 MeV neutrinos. This
yields a diffusion time scale tdiff = O(1 s).
In summary, one expects an energy of about 0.5×1053 erg to be
emitted in each (anti)neutrino degree of freedom over a time scale of
order 1 sec with typical energies of order several 10 MeV.
11.2.2 Energies and Spectra

These global properties of the expected neutrino signal are broadly
confirmed by detailed numerical calculations of neutrino transport.65
However, there are a number of important “fine points” to keep in
mind. First, the nonelectron neutrino degrees of freedom νµ,τ and ν µ,τ
have smaller opacities; their energies are too low for charged-current
reactions of the sort ν µ + p → n + µ+ because of the large masses of the
µ and τ leptons. These flavors decouple at higher densities and tem-
peratures than νe and ν e and so they are emitted with higher average
energies. Equally important, νe ’s have lower energies than ν e ’s because
the opacities are dominated by νe + n → p + e− and ν e + p → n + e+ ,
respectively, and because there are fewer protons than neutrons. Typ-
ically one finds (Janka 1993)

 10−12 MeV
 for νe ,
'Eν ( = 14−17 MeV for ν e , (11.4)

 24−27 MeV for ν
µ,τ and ν µ,τ ,
i.e. typically 'Eνe ( ≈ 23 'Eν e ( and 'Eν ( ≈ 53 'Eν e ( for the other flavors.
The number fluxes of the nonelectron flavors are smaller than those
of ν e because the energy is found to be approximately equipartitioned
between the flavors: the total Eνe +ν e lies between 13 and 12 of Eb . Simi-
larly, the number flux of νe is larger than that of ν e (the lepton number
is carried away in νe ’s!) so that, again, the energy is approximately
equipartitioned between νe and ν e . The total Eν e is found to lie between
1
6
and 14 of Eb . The SN 1987A observations were almost exclusively sen-
sitive to the ν e flux. The total Eν e inferred from these measurements
65
See, e.g., Burrows and Lattimer (1986); Bruenn (1987); Mayle, Wilson, and
Schramm (1987); Burrows (1988); Janka and Hillebrandt (1989a,b); Myra and
Bludman (1989); Myra and Burrows (1990). For reviews see Cooperstein (1988)
and Burrows (1990a,b).
must then be multiplied with a factor between 4 and 6 to obtain an

estimate of Eb .
Even though the emission of neutrinos is a quasithermal process
their energies are not set at the “neutrino sphere” which is defined
to be the approximate shell from where they can escape without sub-
stantial further diffusion. Of course, even the notion of a neutrino
sphere is a crude concept because of the Eν2 dependence of the scatter-
ing cross section on nonrelativistic nucleons which implies that there is
a separate neutrino sphere for each energy group. The scattering with
nucleons does not allow for much energy transfer apart from recoil ef-
fects. What is relevant for determining the neutrino energies is their
“energy sphere” where they last exchanged energy by the scattering on
electrons, by pair processes, and by charged-current absorption. Nat-
urally, this region lies interior to the neutrino sphere—see the shaded
areas in Fig. 11.1 as opposed to the dotted line which represents the
neutrino sphere.
The concept of an “energy sphere” (where neutrinos last exchanged
energy with the medium) and of a “transport sphere” (beyond which
they can stream off without further scattering) helps to explain the
apparent paradox that the ν µ spectrum is, say, twice as hard as that
of ν e , yet the same amount of energy is radiated. Both fluxes originate
from the same radius of about 15 km so that the Stefan-Boltzmann
law (L ∝ R2 T 4 ) would seem to indicate that the ν µ flux should carry
16 times as much energy. However, the place to which the Stefan-
Boltzmann law should be applied is the energy sphere, yet the neutrinos
cannot escape from there because the flow is impeded by neutral-current
scattering on an overburden of nucleons. One may crudely think of the
energy sphere being covered with a skin that does not allow the ra-
diation to stream off except through some holes. Thus the effectively
radiating surface is smaller than 4πR2 . (For a more technical elabora-
tion of this argument see Janka 1995a.)
Evidently, neutrino transport is a rather complicated problem, es-
pecially in the transition region between diffusion and free escape. The
most accurate numerical way to implement it would be a Monte Carlo
integration of the Boltzmann collision equation (Janka and Hillebrandt
1989a,b). In practice, this is not possible because of the constraints
imposed by the limited speed of present-day computers so that a va-
riety of approximation methods are used to solve this problem. While
there is broad agreement on the general features of the expected neu-
trino signal, there remain differences between the predicted spectra and
lightcurves of different authors.
410 Chapter 11
Apart from the highly nontrivial problem of neutrino transport,

the expected signal depends on a variety of physical assumptions such
as the nuclear equation of state which determines the stellar equilib-
rium configuration and the amount of energy that is liberated, on the
properties of the progenitor star (notably the iron core mass), on the
duration of the accretion phase while the shock stalls, on the treatment
of convection during the first few 100 ms, and others.
In the Monte Carlo integrations of Janka and Hillebrandt (1989a,b)
the neutrino spectra at a given time are found to be reasonably well
described by the Fermi-Dirac shape
dLν Eν3
∝ , (11.5)
dEν 1 + eEν /Tν −ην
where Tν is an effective neutrino temperature and ην an effective degen-
eracy parameter. This ansatz allows one to fit the overall luminosity by
a global normalization factor as well as the energy moments 'Eν ( and
'Eν2 (; finer details of the spectrum are probably not warranted anyway.
Throughout the emission process, ηνe decreases from about 5 to 3, ην e
from about 2.5 to 2, and ηνµ,τ ,ν µ,τ from 2 to 0. This effective degeneracy
parameter is the same for νµ,τ and ν µ,τ , in contrast with a real chemical
potential which changes sign between particles and antiparticles.
Fig. 11.6. Normalized neutrino spectral distribution according to a Maxwell-

Boltzmann distribution and a Fermi-Dirac distribution with an effective de-
generacy parameter η = 2, typical for the Monte Carlo transport calcula-
tions of Janka and Hillebrandt (1989a,b). The temperatures are Tν = 13 'Eν (
(Maxwell-Boltzmann) and Tν = 0.832 13 'Eν ( (Fermi-Dirac with η = 2).
Figure 11.6 shows a normalized Maxwell-Boltzmann spectrum and

a Fermi-Dirac spectrum with η = 2, both for the same average en-
ergy 'Eν (. This choice implies Tν = 13 'Eν ( (Maxwell-Boltzmann) and
Tν = 0.832 13 'Eν ( (Fermi-Dirac with η = 2). What is shown in each
case is the normalized number spectrum, i.e. dNν /dEν ∝ Eν2 e−Eν /Tν
(Maxwell-Boltzmann) and Eν2 /(1 + eEν /Tν −ην ) (Fermi-Dirac). It is ap-
parent that for a fixed average energy the Fermi-Dirac spectrum is
“pinched,” i.e. it is suppressed at low and high energies relative to the
Maxwell-Boltzmann case.
The most important difference between the two cases is the sup-
pressed high-energy tail of the pinched spectra which causes a signif-
icant reduction of neutrino absorption rate. This may be important
for neutrino detection in terrestrial detectors as well as for neutrino-
induced nuclear reactions in the SN mantle and envelope. However,
for the sparse SN 1987A signal the differences would not have been
overly dramatic. Therefore, in view of the many other uncertainties
at predicting the spectrum most practical studies of neutrino emission
and possible detector signals used simple Maxwell-Boltzmann spectra,
or Fermi-Dirac spectra with a vanishing chemical potential which differ
from the former only in minor detail.
11.2.3 Time Evolution of the Neutrino Signal

The schematic time evolution of the (anti)neutrino luminosities of the
different flavors was shown in Fig. 11.3. The prompt νe burst has rela-
tively high energies ('Eνe ( ≈ 15 MeV), but the total energy content of
a few 1051 erg renders it negligible relative to the integrated luminosity
of the subsequent emission phases.
The average energies and luminosities of the other flavors rise during
the first few 100 ms while the shock stalls, matter is accreted, and the
initially bloated outer region of the protoneutron star contracts. The
temperature of the region near the edge of the lepton number step
rises substantially during this epoch (Fig. 11.2). During the first 0.5 s
somewhere between 10% and 25% of the total binding energy is radiated
away; the remainder follows during the Kelvin-Helmholtz cooling phase
of the settled star.
Detailed parametric studies of the Kelvin-Helmholtz cooling phase
were performed by Burrows (1988), and more recently by Keil and
Janka (1995). In these works the expected SN 1987A detector signal
was studied as a function of the assumed nuclear equation of state
(EOS), the mass of the collapsed core at bounce, the amount of post-
412 Chapter 11
bounce accretion, and the temperature profile of the core after col-
lapse.
As expected, a soft EOS leads to a large amount of binding energy
and thus to large integrated neutrino luminosities; a large core mass
or large postbounce accretion rate has a similar effect. A soft EOS
leads to relatively high temperatures during deleptonization, causing
large neutrino opacities and thus long emission time scales. If the EOS
is too soft, or the core mass too large, the final configuration is not
stable and collapses, presumably to a black hole. This must not occur
too early to avoid conflict with the duration of the observed SN 1987A
Fig. 11.7. Luminosity and temperature of the ν e flux from the protoneutron
star model 55 of Burrows (1988) which is based on a “stiff equation of state,”
an initial baryonic core mass of 1.3 M! , and an accretion of 0.2 M! within
the first 0.5 s. The dotted line in the upper panel indicates a t−1 behavior,
in the lower panel it indicates e−t/4τ with τ ≈ 10 s. The neutrino spectral
distribution was taken to be thermal.
signal (Sect. 11.3). However, in some cases studied by Keil and Janka
(1995) with an EOS including hyperons this final collapse occurs so
late (at 8 s after bounce in one example) that black-hole formation is
difficult to exclude on the basis of the SN 1987A observations, notably
as no pulsar has yet been found there.
Needless to say, with so many parameters to play it is not difficult
to find combinations of EOS, core mass, accretion rate, and initial
temperatures which fit the observed SN 1987A signal well within the
statistical uncertainties of the observations. An example is model 55
of Burrows (1988) which is based on a “stiff EOS,” an initial baryonic
core mass of 1.3 M! , and an accretion of 0.2 M! within the first 0.5 s.
The evolution of the effective ν e luminosity and temperature is shown
in Fig. 11.7. After about 1 s the decay of the temperature is fit well by
an exponential e−t/4τ with τ ≈ 10 s while the decay of the luminosity
is poorly fit by an exponential; it decays approximately as t−1 after 1 s.
For later reference, the time-integrated flux (fluence) of this model is
shown in Fig. 11.8.
It must be stressed that the “cooling behavior” (decrease of the av-
erage ν e energy) shown in Fig. 11.7 may not be generic at early times.
Initially the star is quite bloated, and relatively cold. Therefore, the
Fig. 11.8. Expected ν e fluence (time-integrated flux) from SN 1987A, as-

suming a distance of 50 kpc, and taking Burrows’ (1988) model 55 neutrino
luminosity shown in Fig. 11.7. The solid line is for an assumed Maxwell-
Boltzmann energy spectrum for a given average neutrino energy, the dashed
line for a Fermi-Dirac spectrum with a degeneracy parameter η = 2 as
in Fig. 11.6.
414 Chapter 11
initial neutrino flux may be characterized by a decreasing radius (de-

creasing flux), yet increasing temperature. For an overview of various
model calculations see Burrows (1990b). One should be careful not to
take details of the time evolution of the temperature and luminosity of
any specific calculation too seriously.
11.3 SN 1987A Neutrino Observations

11.3.1 Supernova 1987A
Shelton’s (1987) sighting of a supernova (SN 1987A) in the Large Mag-
ellanic Cloud (LMC), a small satellite galaxy of the Milky Way at a
distance from us of about 50 kpc (165,000 ly), marked the discovery
of the closest visual SN since Kepler’s of 1604. It was close enough
that several underground detectors which were operational at the time
were able to measure the neutrino flux from the core collapse of the
progenitor star, the blue supergiant Sanduleak −69 202. The observed
neutrinos were registered within a few seconds of 7:35:40 UT (universal
Fig. 11.9. Early optical observations of SN 1987A according to the IAU

Circulars, notably No. 4316 of February 24, 1987. The times of the IMB,
Kamiokande II (KII) and Baksan (BST) neutrino observations (23:07:35)
and of the Mont Blanc events (23:02:53) are also indicated. The solid line
is the expected visual brightness, the dotted line the bolometric brightness
according to model calculations. (Adapted, with permission, from Arnett
et al. 1989, Annual Review of Astronomy and Astrophysics, Volume 27,
*c 1989, by Annual Reviews Inc.)
time) on 23 February 1987 while the first evidence for optical brighten-
ing was found at 10:38 UT on plates taken by McNaught (1987)—see
Fig. 11.9.
The main neutrino observations come from the Irvine-Michigan-
Brookhaven (IMB) and the Kamiokande II water Cherenkov detectors,
facilities originally built to search for proton decay, while a less sig-
nificant measurement is from the Baksan Scintillator Telescope (BST).
A likely spurious observation is from the Mont Blanc Liquid Scintilla-
tor Detector (LSD). It preceded the other observations by about 5 h,
with no contemporaneous signal at Mont Blanc with the other signals,
and no contemporaneous signal at the other detectors with the Mont
Blanc event. The Mont Blanc detector was built to search for neutrinos
from core collapse supernovae, except that it was optimized for galac-
tic events within a distance of about 10 kpc. The neutrino output of
a normal SN in the LMC could not have caused an observable signal
at Mont Blanc; the reported events probably represent a background
fluctuation.
Koshiba (1992) has given a lively account of the exciting and ini-
tially somewhat confusing story of the neutrino measurements and their
interpretation. Early summaries of the implications for astrophysics
and particle physics of the neutrino and electromagnetic observations
were written, for example, by Schramm (1987), Arnett et al. (1989),
and Schramm and Truran (1990). A more recent review of SN 1987A
is McCray (1993). A nontechnical overview was provided in a book by
Murdin (1990).
For the present purposes the bottom line is that SN 1987A broadly
confirmed our understanding of SN physics as outlined in Sect. 11.1.
A remaining sore point is the lack of a pulsar observation in the SN
remnant so that one may continue to speculate that a black hole has
formed in the collapse.
11.3.2 Neutrino Observations

In the IMB and Kamiokande water Cherenkov detectors neutrinos are
measured by the Cherenkov light emitted by secondary charged par-
ticles, e± for the relatively low-energy (anti)neutrinos emitted from a
stellar collapse. The IMB detector is now defunct while Kamiokande
continues to measure, for example, solar neutrinos until the much larger
Superkamiokande detector will take up operation in 1996. In the Bak-
san Scintillator Telescope (BST) one measures the scintillation light
produced by charged secondary particles.
416 Chapter 11
The relevant neutrino interaction processes in water are elastic scat-

tering on electrons (Sect. 10.3.4), and the charged-current reactions
ν e p → n e+ and νe 16 O → 16 F e− (Arafune and Fukugita 1987; Haxton
1987). The cross section for the ν e p reaction is given by
G2F ' (
σ= cos2 θC CV2 + 3CA2 pe Ee (1 + δ)
π
= 9.4×10−44 cm2 (1 + δ) pe Ee /MeV2 , (11.6)
where GF is the Fermi constant and cos2 θC ≈ 0.95 refers to the Cabibbo
angle. The charged-current vector and axial-vector weak coupling con-
stants are CV = 1 and CA = 1.26, and δ incorporates small corrections
from recoil, Coulomb, radiative and weak magnetism corrections (Vo-
gel 1984). Further, pe and Ee refers to the positron momentum and
energy. Ignoring recoil effects, the latter is Ee = Eν − mn + mp ≈
Eν − 1.3 MeV; the threshold is 1.8 MeV because the minimum Ee is me .
A general expression for the νe 16 O cross section is much more com-
plicated. A simple approximation, taking only the 2− state of the 16 F
nucleus, is
σ ≈ 1.1×10−44 cm2 (Eν /MeV − 13)2 (11.7)
(Arafune and Fukugita 1987).
Fig. 11.10. Total cross sections for the measurement of neutrinos in a water
Cherenkov detector according to Eqs. (10.17), (11.6), and (11.7). The curves
refer to the total cross section per water molecule so that a factor of 2 for
protons and 10 for electrons is already included.
All of the relevant cross sections per water molecule are shown in
Fig. 11.10 as a function of Eν . The curves incorporate a factor of 2 for
proton targets (two per H2 O), and a factor of 10 for electrons (ten per
H2 O). Above its threshold, the νe 16 O cross section rises very fast; it is
then the dominant detection process for νe ’s. Still, the ν e p reaction is
the absolutely dominant mode of observing SN neutrinos.
The BST detector is filled with an organic scintillator based on
“white spirit” Cn H2n+2 with n ≈ 9. The dominant detection reaction is
also ν e p → ne+ . In addition, elastic scattering on electrons is possible,
and the process νe 12 C → 12 N e− occurs for Eνe > ∼ 30 MeV.
The trigger efficiencies relevant for the three detectors are shown
as a function of the e± energy in Fig. 11.11. Analytic fit formulae to
these curves were given by Burrows (1988) for IMB and Kamiokande.
IMB reports a dead time of 13% during the SN burst (Bratton et al.
1988); the IMB curve includes a factor 0.87 to account for this effect.
The fiducial volume of Kamiokande II relevant for the SN 1987A obser-
vations was 2,140 tons, for IMB 6,800 tons, and for BST 200 tons. It
corresponds to a target of 1.43×1032 protons at Kamiokande, 4.6×1032
at IMB, and 1.88×1031 at BST.
With the ν e p cross section of Eq. (11.6) and the efficiency curves
of Fig. 11.11 one may compute a prediction for the number of events
Fig. 11.11. Trigger efficiency for electron (positron) detection at the

Kamiokande (Hirata et al. 1988) and IMB (Bratton et al. 1988) water Che-
renkov detectors, and the Baksan scintillator telescope (Alexeyev et al. 1988),
relevant for the SN 1987A neutrino observations. In the IMB curve a factor
0.87 is included to account for their reported dead time of 13%.
418 Chapter 11
per energy interval due to the dominant ν e p reaction in the detectors.

For Kamiokande and IMB an example is shown in Fig. 11.12, based
on Burrows’ (1988) model 55 flux calculation which was tuned to fit
the data. The predicted fluence at Earth per unit energy was shown
in Fig. 11.8 where a distance of 50 kpc was adopted. The solid lines
are for the case when the instantaneous ν e spectra are assumed to be
Maxwell-Boltzmann. Of course, the time-integrated flux is then no
longer thermal as it is a superposition of Maxwell-Boltzmann spectra
at different temperatures. The solid line leads to a total expectation
of 13.1 events at Kamiokande and 6.3 at IMB. The dashed lines cor-
respond to spectra which are instantaneously pinched (Fig. 11.6) with
a degeneracy parameter η = 2. Again, the time-integrated spectra are
not necessarily pinched. This case leads to 11.6 events at Kamiokande
and 3.9 at IMB.
Fig. 11.12. Expected number of events per energy interval at Kamiokande

and IMB from the SN 1987A ν e flux on the basis of the ν e p → ne+ reaction.
The flux prediction is based on Burrows’ (1988) model 55 which was tuned to
fit the SN 1987A data. The distance is taken to be 50 kpc, and the detector
efficiency curves of Fig. 11.8 are used. The dashed lines refer to neutrino
spectra which are instantaneously “pinched” as in Fig. 11.6.
Water Cherenkov detectors, as opposed to scintillation ones, are

imaging devices in that they can resolve the direction of the electron
(positron) because of the directionality of the emitted Cherenkov light.
The main limitation is multiple Coulomb scattering of low-energy e±
in the medium. A typical path length in water is only a few cm so that
hard collisions are relatively unlikely. The rms angular deviation due
to multiple scattering varies from about 34◦ at Ee = 5 MeV to 22◦ at

20 MeV (Hirata 1991).
The direction of motion of the charged lepton relative to the inci-
dent neutrino is well preserved in νe collisions (Sect. 10.3.4) so that the
main limitation to a reconstruction of the primary neutrino direction is
multiple Coulomb scattering. For the ν e p reaction, the angular distribu-
tion is isotropic, apart from a small (about 10%) backward asymmetry
(e.g. Boehm and Vogel 1987). The νe 16 O reaction yields a distribution
approximately proportional to 1 − 13 cos Θ, i.e. it is also nearly isotropic
with a backward bias (Haxton 1987). As a SN is expected to produce all
(anti)neutrino flavors in about equal numbers, the signal is dominated
by the isotropic ν e p reaction.
Table 11.1. Neutrino burst at the Kamiokande detector (Hirata et al. 1988).
The time is relative to the first event at 7:35:35 ± 0:01:00 UT, 23 Feb. 1987.
The energy refers to the detected e± , not to the primary neutrino.
Event Time Angle Energy

[s] [degree] [MeV]
1 0.000 18 ± 18 20.0 ± 2.9
2 0.107 40 ± 27 13.5 ± 3.2
3 0.303 108 ± 32 7.5 ± 2.0
4 0.324 70 ± 30 9.2 ± 2.7
5 0.507 135 ± 23 12.8 ± 2.9
6a 0.686 68 ± 77 6.3 ± 1.7
7 1.541 32 ± 16 35.4 ± 8.0
8 1.728 30 ± 18 21.0 ± 4.2
9 1.915 38 ± 22 19.8 ± 3.2
10 9.219 122 ± 30 8.6 ± 2.7
11 10.433 49 ± 26 13.0 ± 2.6
12 12.439 91 ± 39 8.9 ± 1.9
13a,b 17.641 ... 6.5 ± 1.6
14a,b 20.257 ... 5.4 ± 1.4
15a,b 21.355 ... 4.6 ± 1.3
16a,b 23.814 ... 6.5 ± 1.6
a Usually attributed to background.
b Quoted after Loredo and Lamb (1995).
420 Chapter 11
Table 11.2. Neutrino burst at the IMB detector (Bratton et al. 1988).
The time is relative to the first event at 7:35:41.374 ± 0:00:00.050 UT, 23
Feb. 1987. The energy refers to the detected e± , not to the primary neutrino.
Event Time Angle Energy

[s] [degree] [MeV]
1 0.000 80 ± 10 38 ± 7
2 0.412 44 ± 15 37 ± 7
3 0.650 56 ± 20 28 ± 6
4 1.141 65 ± 20 39 ± 7
5 1.562 33 ± 15 36 ± 9
6 2.684 52 ± 10 36 ± 6
7 5.010 42 ± 20 19 ± 5
8 5.582 104 ± 20 22 ± 5
Table 11.3. Neutrino burst at the Baksan detector (Alexeyev et al. 1987,
1988). The time is relative to the first event at 7:36:06.571+02.000
−54.000 UT, 23
Feb. 1987. The energy refers to the detected e± , not to the primary neutrino.
Event Time Energy

[s] [MeV]
0a 0.000 17.5 ± 3.5
1 5.247 12.0 ± 2.4
2 5.682 18.0 ± 3.6
3 6.957 23.3 ± 4.7
4 12.934 17.0 ± 3.0
5 14.346 20.1 ± 4.0
a Usually attributed to background.
The energy of the electron (positron) can be reconstructed from

the total amount of Cherenkov or scintillation light emitted. For small
energies it is roughly proportional to the number of photomultipliers hit
in a given event. Because of the reaction threshold and recoil effects,
the energy of the primary neutrino in the ν e p → ne+ reaction is about
2 MeV larger than the measured e+ energy. For the rare νe collisions,
the final-state electron energy distribution is broad so that one can infer
only a lower limit to the ν energy.
The measured events at the Kamiokande (Hirata et al. 1987, 1988),
IMB (Bionta et al. 1987; Bratton et al. 1988), and BST (Alexeyev et al.
1987, 1988) detectors are summarized in Tabs. 11.1, 11.2, and 11.3.
The absolute timing at IMB is accurate to within ±50 ms while at
Kamiokande only to within ±1 min. At BST, the clock exhibited an
erratic behavior which led to an uncertainty of +2/−54 s. Within the
timing uncertainties the three bursts are contemporaneous and may
Fig. 11.13. SN 1987A neutrinos at Kamiokande, IMB, and Baksan. The

energies refer to the secondary positrons, not the pimary neutrinos. In the
shaded area the trigger efficiency is less than 30%. The detector clocks have
unknown relative offsets; in each case the first event was shifted to t = 0. In
Kamiokande and Baksan, the events marked with open circles are usually
attributed to background.
422 Chapter 11
thus be simultaneously attributed to SN 1987A. The events are shown

in the t-E-plane in Fig. 11.13 and for the Cherenkov detectors in the
cos Θ-E-plane in Fig. 11.14 where Θ is the angle relative to the opposite
direction of the SN, i.e. relative to the direction of the neutrino flux.
In principle, the bursts of events observed in the detectors could be
due to rare background fluctuations rather than due to SN 1987A. The
Kamiokande group has performed a detailed analysis of this possibility
by analyzing the multiplicity of events in 10 s time intervals, i.e. the
number of chance events in an arbitrarily chosen 10 s time interval.
They found a probability of about 0.6×10−7 that the observed burst is
a random fluctuation of a constant background.
However, there are time-correlated backgrounds, notably the spal-
lation of oxygen induced by primary muons which can cause clusters of
events with large multiplicities; the Kamiokande group found one clus-
Fig. 11.14. SN 1987A neutrinos at Kamiokande and IMB, excluding the ones
which likely are due to background.
ter of 53 events! Therefore, any cluster following a high-energy muon

would be very suspicious. Performing a cut on the data for this back-
ground leaves no burst with multiplicity 3 or larger in several data sets
of several hundred days each (Hirata 1991). Therefore, it is extremely
unlikely that the event cluster 10−12 in the Kamiokande data has been
caused by background.
The BST detector has a relatively large background rate. Event
clusters of multiplicity 5 or more within 9 s occur about once per day.
Thus the probability for such a background cluster to fall within a
minute of the IMB and Kamiokande events is about 5×10−4 .
11.3.3 Analysis of the Pulse

Many authors have studied the distribution of energies and arrival times
of the reported events. Probably the most significant work is that of
Loredo and Lamb (1989, 1995) who performed a maximum-likelihood
analysis, carefully including the detector backgrounds and trigger ef-
ficiencies. Loredo and Lamb also gave detailed references to previous
works, and in some cases offered a critique of the statistical method-
ology employed there. Their more extensive 1995 analysis supersedes
certain aspects of the earlier methodology and results.
Because of the small number of neutrinos observed, a relatively
crude parametrization of the time-varying source is enough. Among a
variety of simple single-component emission parametrizations, Loredo
and Lamb (1989, 1995) found that an exponential cooling model was
preferred. It is characterized by a constant radius of the neutrino
sphere, R, and a time-varying effective temperature
T (t) = T0 e−t/4τ , (11.8)
so that τ is the decay time scale of the luminosity which varies with
the fourth power of the temperature according to the Stefan-Boltzmann
law. It should be noted, however, that numerical cooling calculations
do not yield exponential lightcurves. For example, the model shown in
Fig. 11.7 displays an exponential decline of the effective temperature,
but a power-law decline of the neutrino luminosity. Other calculations
even yield early heating and a constant temperature for some time (see
Burrows 1990b for an overview).
Of course, for the time-integrated spectrum the exponential cooling
law is just another assumption concerning the overall spectral shape.
For example, one easily finds that the average ν e energy of the time-
integrated spectrum is 'Eν e ( = 2.36 T0 if Fermi-Dirac distributions with
424 Chapter 11
η = 0 are taken for the instantaneous spectra. The time-integrated

spectrum of the exponential cooling model looks quite similar to the
time-integrated spectrum shown in Fig. 11.8.
Loredo and Lamb also used α ≡ (R/10 km) (50 kpc/D) g 1/2 as a fit
parameter where D is the distance to SN 1987A and g a statistical
weight factor which is unity if only left-handed, massless or low-mass
neutrinos of the three sequential flavors are emitted. The registration
time of the first neutrino in each detector is taken as a free parameter
relative to the arrival time of the first neutrinos. In the 1995 analy-
sis, Loredo and Lamb included the Baksan signal without “event 0”
which is attributed to background because it precedes the main bunch
by 5 s.
The following six parameters are then allowed to float freely in order
to achieve a maximum-likelihood result: T0 , τ , α, toff (IMB), toff (KII),
and toff (BST). All best-fit offset times are found to be zero. The other
best-fit values are α = 4.02, τ = 4.37 s, and T0 = 3.81 MeV. This
initial temperature of the exponential cooling model corresponds to an
average neutrino energy of the time-integrated flux of 'Eν e ( = 9.0 MeV.
In Fig. 11.15, the 68% and 95% credible regions are shown in the T0 -τ -
plane where T0 has been translated into 'Eν e ( which is of greater direct
relevance.
Fig. 11.15. Two-dimensional marginal distribution for the parameters τ and

'Eν e ( = 2.36 T0 of the exponential cooling model. (Curves courtesy of Tom
Loredo, taken from Loredo and Lamb 1995.)
Given these parameters one infers that the number of expected

Kamiokande events is 16.9 plus 5.6 background, 4.0 events at IMB,
and 1.8 plus 1.0 background at Baksan. The inferred best-fit neutrino-
sphere radius is 40.2 km, and the inferred total emitted ν e energy is
0.84×1053 erg which corresponds to a total binding energy of the neu-
tron star of 5.02×1053 erg if exact equipartition of the energy among
the neutrino flavors is assumed.
The inferred average neutrino energy, the luminosity-decay time
scale, radius of the source, and total energy emitted all agree reasonably
well with what one expects from a core collapse SN, even though 'Eν e (
is somewhat low, the radius and inferred binding energy somewhat
large. Of course, all of the inferred quantities carry large uncertainties
because of the sparse data.
Loredo and Lamb (1995) have also considered two-component cool-
ing schemes where the neutrino signal is modelled to consist of Kelvin-
Helmholtz cooling, plus a low-energy component which mimics the neu-
trinos emitted by the accreting matter during the stalled-shock phase
in the delayed-explosion scenario. With more parameters they nat-
urally find a better fit to the data. More interestingly, the inferred
Fig. 11.16. Two-dimensional marginal distribution for the parameters τ and

'Eν e ( = 2.13 T0 of the Kelvin-Helmholtz component of the Loredo and Lamb
(1995) best-fit two-component cooling model. The parameters τ and T0 are
those of a displaced power law as described in the text. (Curves courtesy of
Tom Loredo, taken from Loredo and Lamb 1995.)
426 Chapter 11
neutrino-sphere radius of 18 km and the inferred total binding energy

of 3.08×1053 erg correspond much better to theoretical expectations.
Perhaps this finding can be taken as a hint that the delayed-explosion
scenario with a significant matter accretion phase is favored by the SN
1987A data over a prompt-explosion picture.
In Loredo and Lamb’s best-fit two-component model the Kelvin-
Helmholtz signal is described by a “displaced power law” cooling model
with a neutrino sphere of fixed radius R and thermal neutrino emission
with T (t) = T0 /(1 + t/3τ ). It turns out that the luminosity, which is
proportional to T 4 , follows a surprisingly similar curve to the exponen-
tial e−t/τ so that the parameter τ has practically the same meaning
as before. 'Eν e ( for the time-integrated flux is given by 2.13 T0 , very
similar to 2.36 T0 for the exponential. In Fig. 11.16 the 68% and 95%
credible regions are shown in the T0 -τ -plane. While the best-fit value
is not too different from the single-component exponential model of
Fig. 11.15, the 95% credible region is much larger, including the lowest
values of the typical theoretical 'Eν e ( predictions quoted in Eq. (11.4).
11.3.4 Neutrino Mass and Pulse Duration

Zatsepin (1968) was the first to point out that the ν e burst expected
from stellar collapse offers a possibility to measure or constrain small
neutrino masses. Because a neutrino with mass mν travels slower than
the speed of light its arrival at Earth will be delayed by
! " ) *2 ) *2
D 10 MeV mν
∆t = 2.57 s . (11.9)
50 kpc Eν 10 eV
Because the measured ν e ’s from SN 1987A were registered within a few
seconds and had energies in the 10 MeV range, the mνe mass is limited
to less than about 10 eV.
A detailed study must proceed along the lines of the maximum-
likelihood analysis of Loredo and Lamb (1989, 1995) quoted in the
previous section where the detector background is included, and such
parameters as the unknown offset times between the detectors are left
unconstrained. Including the possibility that some of the registered
events are due to background is particularly important because the
neutrino mass limit is very sensitive to the early low-energy events at
Kamiokande which have a relatively high chance of being due to back-
ground. Loredo and Lamb (1989) found a vanishing best-fit neutrino
mass and a 95% CL upper limit of mνe < 23 eV. This bound is less
restrictive than limits found by previous authors on the basis of less
thorough statistical analyses. In their 1995 paper, Loredo and Lamb

have not studied neutrino mass limits which likely would change some-
what because of corrections to their previous approach.
An analysis by Kernan and Krauss (1995) on the basis of a similar
method yields a limit 19.6 eV at 95% CL. Apparently, the reduction
of the limit is due to their inclusion of the 13% dead-time effect in
the IMB detector. The difference to the Loredo and Lamb (1989) limit
illustrates that changing a relatively fine point of the analysis procedure
can significantly change a so-called 95% CL limit. Therefore, instead
of quoting a specific confidence limit it is at present more realistic to
state qualitatively that a violation of the mass limit
mν e <
∼ 20 eV (11.10)
would have caused a significant and perhaps intolerable modification of

the SN 1987A signal.
This limit is weaker than the current bounds from the tritium β
decay endpoint spectrum (Sect. 7.1.3). Therefore, the above analysis
can be turned around in the sense that the observed neutrino signal
duration is probably representative of the duration of neutrino emission
at the source. The observed long time scale of Kelvin-Helmholtz cooling
which is indicated by the late IMB and Kamiokande events cannot be
blamed on neutrino dispersion effects. In this context it is interesting
to observe that Loredo and Lamb (1989) also performed a maximum-
likelihood analysis with mνe held fixed at their 95% CL upper limit
23 eV. The best-fit time scale in an exponential cooling model changed
from 4.15 to 2.96 s. Therefore, even assuming a large value for mνe did
not allow one to contemplate a significantly shorter Kelvin-Helmholtz
cooling phase than implied by massless neutrinos.
11.3.5 Anomalies in the Signal?

The distribution of the Kamiokande and IMB events shows a number of
puzzling features. The least worrisome of them is a certain discrepancy
between the neutrino energies observed in the two detectors which point
to a harder spectrum at IMB. The maximum-likelihood analysis in the
exponential cooling model of Loredo and Lamb (1989) was also per-
formed for the two detectors separately. The 95% confidence volumes
projected on the T0 -Eb -plane are shown in Koshiba (1992); a similar re-
sult is found in Janka and Hillebrandt (1989b). There is enough overlap
between the confidence contours to allow for a joint analysis. Still, the
best-fit value for Kamiokande lies outside the 95% CL volume of IMB.
428 Chapter 11
Including the pinching effect discussed in Sect. 11.2.2 would enhance

the discrepancy between the signals in the two detectors. Either way,
the IMB detector with its high energy threshold is mostly sensitive to
the high-energy tail of the neutrino spectrum. Therefore, the IMB-
inferred 'Eν e ( depends sensitively on the assumed spectral shape and
is thus a poor indicator of the true average energies.
A more conspicuous anomaly is the 7.3 s gap between the first 9
and last 3 events at Kamiokande. Ideas proposed to explain the alleged
pulsed structure of the signal range from the occurrence of a phase
transition in the nuclear medium (pions, quarks) to a secondary collapse
to a black hole. It should be noted, however, that the gap is partially
filled in by the IMB and Baksan data, thus arguing against a physical
cause at the source. The random occurrence of a gap exceeding 7 s with
three or more subsequent events can be as high as several percent, but
naturally it is sensitive to the expected late-time signal (Lattimer and
Yahil 1989).
The most significant and thus the most troubling anomaly is the
remarkable deviation from isotropy of the events in both detectors, in
conflict with the expected signature from ν e p → ne+ which actually
predicts a slight (about 10%) backward bias. LoSecco (1989) found
a probability of about 1.5% that the combined Kamiokande and IMB
data set was drawn from an isotropic distribution. Kie&lczewska (1990)
analyzed the expected signal from standard SN cooling calculations
and found agreement only at the 0.8% CL with the measured angu-
lar distribution. The combined set of IMB plus those Kamiokande
data which are above the IMB threshold, i.e. the combined set of
“high-energy” events is consistent with isotropy only at the 0.07% level
(van der Velde 1989); the four relevant events at Kamiokande are all
very forward.
The IMB collaboration claims that their reconstruction of the event
direction was not seriously impeded by the outage of about a quarter of
their phototubes due to the failure of a high-voltage supply. They con-
ducted a detailed calibration of their detector to investigate this point
(Bratton et al. 1988). Therefore, one must accept that the forward-
peaked angular distribution shown in Fig. 11.14 is not a problem of the
detectors or the event reconstruction.
A forward-peaked distribution is expected from νe elastic scattering
which has a much lower cross section than the ν e p process (Fig. 11.8).
One expects far less than one event due to νe scattering from the cool-
ing phase, although the first Kamiokande event has sometimes been
interpreted as being a scattering event due to the prompt νe burst.
The forward events in both detectors have relatively large energies

compared with the isotropic ones at Kamiokande,66 contrary to what
would be expected from νe → eν where some of the energy is carried
away by the secondary neutrino. Assuming a larger-than-standard flux
of νe ’s with larger-than-standard energies (LoSecco 1989) does not solve
the problem because above 30−35 MeV the process νe 16 O → 16 F e−
takes over (Fig. 11.8) which has a backward bias.
Anomalously large fluxes of νµ,τ or ν µ,τ are difficult to arrange on
energetic grounds—the binding energy of the neutron star is limited.
Even allowing for extreme values of Eb and extreme temperature differ-
ences between (anti)electron neutrinos and the other flavors improves
the agreement only marginally; one can achieve an agreement at the
5% CL with the observed angular distribution (Kie&lczewska 1990). The
simple problem with elastic νe scattering to explain the data is that
this process is too strongly forward peaked, especially for high-energy
neutrinos, hence it does not fit the data very well either. This is espe-
cially true for the IMB events which are selected for high energies by
the detector threshold, and yet are very broadly distributed around the
forward direction.
A very speculative idea was put forth by van der Velde (1989) who
proposed the existence of a new neutral boson X ◦ which could pro-
duce photons when interacting with nucleons. These MeV photons
would look very similar to charged particles in the detectors. In the
forward direction, the X ◦ cross section on 16 O would be coherently en-
hanced, causing the observed forward bias for high-energy events while
low-energy ones would naturally follow a more isotropic distribution.
However, the opposite process γ + 4 He → 4 He + X ◦ would then con-
tribute to the energy-loss of horizontal-branch stars (Raffelt 1988b).
The resulting bound on the interaction cross section (Tab. 2.5), valid
at an energy of about 10 keV, excludes van der Velde’s scenario unless
σ increases with energy at least as E 2 , a scaling which could bring it
up to the requisite level for E in the 10 MeV range, relevant for the SN
detection. The Primakoff conversion of axions or similar particles on
oxygen is much too inefficient in view of the restrictive limits on the
axion-photon interaction strength.
In summary, the angular and energy distributions of the IMB and
Kamiokande events appear to indicate a low-energy isotropic and a
66
The Kamiokande events have an obvious correlation between energy and di-
rection. The application of Spearman’s rank-ordering test (e.g. Press et al. 1986)
gives a confidence level of 0.06% where event No. 6 was excluded as background.
Therefore, the Kamiokande data alone show a fairly significant “angular anomaly.”
430 Chapter 11
high-energy forward component; they are not fit well by the assumed
dominant detection process ν e p → ne+ which is supposed to yield an
isotropic positron distribution with no directional correlation with en-
ergy. Elastic νe scattering, however, does not fit the data well either
because it is too forward peaked, and anyhow it is disfavored by a small
cross section unless the flux of νµ,τ or ν µ,τ was extremely high.
However, because no plausible and/or viable nonstandard cause for
the observed events has been proposed one has settled for the interpre-
tation of a statistical fluctuation for the apparent anomalies. After all,
it is difficult to imagine a small sample drawn from any distribution
without some “anomalies” which are easy to overinterpret. Still, if a
reasonable alternative to the standard interpretation of the signal were
to come forth this topic would have to be reconsidered.
11.4 Neutrino Oscillations
11.4.1 Overview
The expected neutrino signature from a stellar collapse and conversely

the inferred protoneutron star properties from the SN 1987A neutrino
signal both depend on the assumption that “nothing happens” to the
neutrinos on their way to us. One simple modification of the expected
signal is a dispersion of the ν e burst caused by a nonvanishing mνe in
the 10 eV range (Sect. 11.3.4). Dispersion effects could also be caused
by novel interactions with the galactic magnetic field, dark matter, the
neutrino background, or simply by decays. All of these scenarios require
relatively exotic particle-physics assumptions which can be constrained
by the SN 1987A signal (Chapter 13). The assumption of small neu-
trino masses and mixings, however, fits into the standard model with
minimal extensions, and may already be implied by the solar neutrino
observations (Sect. 10.6). Therefore, it is prudent not to ignore the
possible impact of oscillations on SN neutrinos.
The most obvious consequence is that the prompt νe burst could os-
cillate into another flavor which then would be much harder to observe
because of the reduced ν-e cross section for non-νe flavors (Fig. 11.10).
Notably, if the solar νe flux is depleted by resonant oscillations one may
expect the same in the SN mantle and envelope where a large range of
densities and density gradients is available. It will turn out, however,
that the small-angle MSW solution to the solar neutrino problem leaves
an observable νe burst (Sect. 11.4.2).
Another interesting possibility is a partial swap νe ↔ νµ,τ and ν e ↔

ν µ,τ by oscillations. Because the energy spectrum of the non-νe flavors is
much harder than that of νe or ν e , a number of interesting consequences
obtain. First, the detected ν e ’s could have larger average energies than
expected. Conversely, the SN 1987A-implied emission temperature and
neutron-star binding energy could be an overestimate of the true values.
It will turn out that one seriously needs to worry about these effects,
for example, if the large-angle MSW solution or the vacuum solution
to the solar neutrino problem obtain, or if the atmospheric neutrino
anomaly is caused by oscillations (Sect. 11.4.3).
Even more importantly, a swap νe ↔ νµ or νe ↔ ντ would cause
a more efficient energy transfer from the neutrino flux to the matter
behind the stalled shock after core bounce but before the final explo-
sion. As enhanced neutrino heating actually appears to be required to
obtain successful and sufficiently energetic explosions, neutrino oscil-
lations may help to explode supernovae! This scenario works only if
the spectral swap occurs inside of the stalled shock wave. In view of
the relevant medium densities, resonant transitions obtain for neutrino
masses in the cosmologically interesting range of 10−100 eV. A mixing
angle as small as sin2 2θ > −8
∼ 3×10 would be enough (Sect. 11.4.4).
Hardening the νe spectrum by a swap with νµ or ντ , however, can
suppress r-process nucleosynthesis just outside of the nascent neutron
star a few seconds after core bounce. Normally the ν e spectrum is
harder than the νe spectrum, driving β equilibrium in the hot bub-
ble to the required neutron-rich phase. The oscillation scenario can
cause the reverse. For this effect the oscillations would need to oc-
cur close to the protoneutron star surface and so again a relatively
large neutrino mass-square difference is required which falls into the
cosmologically interesting range. However, the required mixing angle
is larger (sin2 2θ > −5
∼ 10 ), leaving ample room for, say, small νe -ντ
mixing angles where ντ ’s with cosmologically relevant masses could
help explode supernovae without disturbing r-process nucleosynthe-
sis (Sect. 11.4.5).
Finally, neutrino oscillations could allow the non-νe flavors to par-
ticipate in β equilibrium in the inner core and thus build up their own
degenerate Fermi seas. This possibility has been studied in Sect. 9.5
where it turned out that a significant flavor conversion obtains only
for large neutrino masses (keV range and above). Such large masses
are cosmologically forbidden unless neutrinos decay fast into invisible
channels, a hypothesis that would require novel neutrino interactions
beyond masses and mixings (Sect. 12.5.2).
432 Chapter 11
11.4.2 Prompt νe Burst

The prompt νe burst from a core collapse SN can be detected, in princi-
ple, by the forward-peaked signal from the elastic νe e → eνe scattering
in a water Cherenkov detector. In the Kamiokande SN 1987A observa-
tions, the first event could have been caused by the prompt νe burst,
but naturally one event contains no statistically significant information.
It could have been caused by ν e p → ne+ and simply happen to point
in the forward direction. Therefore, the main interest in the prompt νe
burst is the possibility that it could be observed from a future galactic
SN by the Superkamiokande or SNO detectors which would yield sta-
tistically significant signatures. A possible oscillation of the νe burst
into other flavors would reduce the number of forward events because
of the reduced ν-e scattering cross section of non-νe flavors (Fig. 11.10).
If the neutrino mass hierarchy is normal where the lightest mass
eigenstate is the dominant νe admixture, the medium-induced neutrino
refractive index in the stellar mantle and envelope can cause a “mass
inversion” and thus level crossing between, say, νe and νµ in analogy
to the solar MSW effect. Therefore, one may expect resonant flavor
conversion of the prompt νe burst in a collapsing star as shown by a
number of authors;67 I follow the analysis of Nötzold (1987).
When the shock wave breaks through the neutrino sphere and lib-
erates the prompt νe burst, the overlaying part of the progenitor star
has not yet noticed the collapse of its core so that the density profile is
given by that of the progenitor star. The electron density is reasonably
well approximated by a simple power law for which Nötzold (1987) used
ne ≈ 1034 cm−3 r7−3 , (11.11)
where r7 ≡ r/107 cm. Note that 107 cm = 100 km is the approximate

radius of the shell from where the νe ’s originate. According to the
discussion in Sect. 6.7.1 the electron
√ density causes −3 an energy shift
between νe and νµ or ντ of ∆V = 2GF ne = 1.3×10 eV r7−3 . Com-
paring this with the energy shift ∆m2ν /2pν of neutrinos with momentum
pν one finds an effective medium-induced effect of ∆m2eff = 2pν ∆V =
3×104 eV2 p10 r7−3 where p10 = pν /10 MeV.
However, because the prompt νe burst itself constitutes a large local
νe density, the neutrino-induced refractive index may be more impor-
tant than the standard electron-induced contribution. The total num-
67
Mikheev and Smirnov (1986), Arafune et al. (1987a,b), Lagage et al. (1987),
Minakata et al. (1987), Nötzold (1987), Walker and Schramm (1987), Kuo and
Pantaleone (1988), Minakata and Nunokawa (1988), and Rosen (1988).
ber of νe ’s in the burst is of order 1056 , its duration of order 50 ms. Thus,
while it passes it represents a νe density of about 1032 cm−3 r7−2 which
exceeds the local electron density for r > 9
∼ 10 cm. However, the phase-
space distribution of the neutrinos is locally far from isotropic and so
the energy shift involves a factor '1 − cos Θ( where Θ is the angle be-
tween the “test neutrino” and a “background neutrino;” the average is
to be taken over all background neutrinos (Sect. 9.3.2). A typical angle
between two neutrinos moving within the burst at the same location
is given by the angle subtended by the neutrino sphere as viewed from
the relevant radial position, i.e. Θ ≈ R/r with R the radius of the neu-
trino sphere. For a large r one thus finds '1 − cos Θ( ≈ Θ2 ≈ (R/r)2 .
Then, with R ≈ 107 cm the effective neutrino density is approximately
1032 cm−3 r7−4 , a value which is always smaller than the electron density
Eq. (11.11). Therefore, in the present context one may ignore neutrino-
neutrino interactions. This will not be the case for the issue of r-process
nucleosynthesis (Sect. 11.4.5).
One may proceed to determine the MSW triangle as in Fig. 8.9
for solar neutrinos, except that there an exponential electron density
profile was used while now Eq. (11.11) pertains. Nötzold (1987) found
a conversion probability in excess of 50% if
∆m2ν sin3 2θ > −9 2

∼ 4×10 eV Eν /10 MeV, (11.12)
assuming that ∆m2ν < 4 2

∼ 3×10 eV Eν /10 MeV so that a resonance can
occur outside of the neutrino sphere. The region in the ∆m2ν -sin2 2θ-
plane (mixing angle θ) with a conversion probability exceeding 50% for
Eν = 20 MeV is shown as a shaded area in Fig. 11.17, together with
the MSW solutions to the solar neutrino problem. For orientation, the
Kamiokande-allowed range for solar neutrinos is also indicated.
The solar small-angle MSW solution would seem to have a small
impact on the prompt νe burst from a collapsing star. Thus, the first
Kamiokande SN 1987A event may still be interpreted as a prompt νe ,
and one may well observe nearly the full νe burst from a future SN.
It is interesting that the MSW triangle for the prompt νe burst
reaches to relatively large neutrino masses. In the 3−30 eV regime neu-
trino masses would play an important cosmological role as dark matter
and for the formation of structure in the universe. Such massive neu-
trinos likely would mix with νe . Unless the mixing angle is very small
the appearance of an unoscillated prompt νe burst from a stellar col-
lapse would be in conflict with a cosmological role of massive neutrinos
(Arafune et al. 1987b).
434 Chapter 11
Fig. 11.17. MSW triangle for the prompt νe burst from a stellar collapse. In
the shaded area the conversion probability exceeds 50% for Eν = 20 MeV,
assuming the electron density profile of Eq. (11.11). The MSW solutions
to the solar neutrino problem and the Kamiokande allowed range for solar
neutrinos are indicated (see Fig. 10.19).
11.4.3 Cooling-Phase ν e ’s
Neutrino oscillations would cause a partial swap νe ↔ νµ,τ and ν e ↔

ν µ,τ so that the measured ν e flux at Earth could be a mixture of the
original ν e and ν µ or ν τ source spectra (Wolfenstein 1987). The energy
spectra of the neutrinos emitted during the Kelvin-Helmholtz cooling
phase are flavor dependent (Eq. 11.4); typically, one finds 'Eν µ ( =
(1.3 − 1.7) × 'Eν e (. From the SN 1987A measurements one infers a low
value of roughly 'Eν e ( ≈ 10 MeV. While the lowest typical predictions
are about 14 MeV, this discrepancy is not a serious problem. However,
it is probably not tolerable that a significant fraction of the observed
events were due to oscillated ν µ ’s.
One may contemplate an “inverted” mass hierarchy where the pre-
dominant mass component of νe is larger than that of, say, ντ . In this
case one would obtain resonant oscillations and thus a complete spectral
swap in the shaded triangle of mixing parameters shown in Fig. 11.17.
Even if 'Eν τ ( is only 1.3 × 'Eν e ( this would be in contradiction with the
soft ν e energies observed from SN 1987A. Therefore, an inverted mass
scheme looks excluded for a large range of masses and mixings.
A “normal” mass hierarchy prevents a level crossing among an-

tineutrinos. Still, if the large-angle or the vacuum solution to the so-
lar neutrino problem obtain, the mixing angle would be so large that
the spectral swapping could still be uncomfortably large. Smirnov,
Spergel, and Bahcall (1994) have considered in detail the probability
for swapping the ν e with the ν µ or ν τ spectrum. For vacuum oscilla-
tions, relevant for small ∆m2ν , the fractional exchange of the spectra
is p = 12 sin2 2θ (vacuum mixing angle θ) with a maximum of p = 0.5,
i.e. the observed spectrum could be as much as an equal mixture of the
primary ones. In general, medium refractive effects must be included,
although for a large range of ∆m2 one may still take p to be indepen-
dent of energy. Contours for p in the sin2 2θ-∆m2ν -plane are shown in
Fig. 11.18. In the shaded area (A) the Earth effect is important so that
the amount of conversion is energy dependent and differs between the
Fig. 11.18. Contours for the “swap fraction” p between the ν e and the ν µ
or ν τ fluxes from the protoneutron star cooling phase. The matter effect of
the stellar envelope and of the Earth are included. In (A) the Earth effect is
important; in this area p is an average over neutrino energies while otherwise
it does not depend on the energy. In (B) the exact contours depend on the
detailed matter distribution of the stellar envelope. Black areas indicate
the approximate mixing parameters which would explain the solar neutrino
problem. (Adapted from Smirnov, Spergel, and Bahcall 1994.)
436 Chapter 11
two detectors. In this case the contours refer to an average value 'p(.
In the shaded area (B) the contours are not independent of the detailed
matter distribution in the stellar envelope.
In Fig. 11.18 the large-angle MSW and the vacuum oscillation solu-
tions of the solar neutrino problem are indicated. If either one of them
is correct the measurable ν e spectrum in a detector is a substantial mix-
ture of different-flavor source spectra. Smirnov, Spergel, and Bahcall
(1994) argued on the basis of a joint analysis between the SN 1987A
signals at the Kamiokande and IMB detectors that p < 0.17−0.27 at
the 95% CL, depending on the assumed primary neutrino spectra. If
p were any larger, the expected spectra would be much harder than
has been observed. This analysis excludes the vacuum oscillation solu-
tion to the solar neutrino problem. Kernan and Krauss (1995) arrive
at the opposite conclusion that all mixing angles are permitted by the
SN 1987A signal, and that sin2 2θ = 0.45 is actually a favored value.
11.4.4 Shock Revival

The “swap fraction” of the ν e with the more energetic ν µ or ν τ spec-
trum discussed in the previous section is always very small unless the
mixing angle is very large because of the assumed normal mass hier-
archy which prevents resonant conversions. By the same token a swap
of the νe with the νµ or ντ spectrum will be resonant for certain mix-
ing parameters and so it can be almost complete even for very small
mixing angles. If the resonance is located between the neutrino sphere
and the stalling shock wave after bounce, but before the final explo-
sion, the shock would be helped to rejuvenate because the higher-energy
νµ,τ ’s are more efficient at transferring energy once they have converted
into νe ’s (Fuller et al. 1992). Of course, approximately equal luminosi-
ties in all flavors have been assumed.
A typical density profile for a SN model 0.15 s after bounce is shown
in Fig. 11.19; the step at a radius of about√400 km is due to the shock
front. A resonance occurs if ∆m2ν /2E √ ν = 2GF1/2 ne . On the right scale
of the plot the quantity mres ≡ ( 2GF ne 2Eν ) is shown for Eν =
10 MeV and Ye = 0.5. A resonance occurs inside of the shock wave
only for neutrino masses in the cosmologically interesting regime of
order 10 eV and above. Fuller et al. (1992) have performed a detailed
numerical calculation of the additional heating effect for (∆m2ν )1/2 =
40 eV; they found a 60% increase of the energy of the shock wave.
The conversion probability between neutrinos is large if the adia-
baticity parameter γ defined in Eq. (8.39) far exceeds unity. Accord-
Fig. 11.19. Typical density profile for a SN model at 0.15√ s after the core
bounce (Fuller et al. 1992). The right-hand scale is mres = ( 2GF ne 2Eν )1/2
which indicates the resonance value for (∆m2ν )1/2 for Eν = 10 MeV, assuming
an electron number fraction of Ye = 0.5.
ing to the Landau-Zener formula Eq. (8.41) the swap probability is

1 − e−πγ/2 . It exceeds 86% if one requires γ > 4/π or
8 Eν
sin2 2θ > |∇ ln ne |res , (11.13)
π ∆m2ν
where the vacuum mixing angle θ was assumed to be small. For
(∆m2ν )1/2 = 40 eV the density scale height at the resonance region is
|∇ ln ne |−1
res ≈ 50 km so that
sin2 2θ > −8
∼ 10 Eν /10 MeV. (11.14)
Therefore, unless the mixing angle with νe is very small a cosmologically

interesting neutrino mass for, say, the ντ may help to explode super-
novae! According to the discussion in Sect. 11.4.2 this would imply that
the prompt νe burst would also oscillate. Then the first Kamiokande
event could not be associated with the prompt νe burst.
11.4.5 R-Process Nucleosynthesis

a) Basic Picture
A partial swap of the νe cooling flux with the more energetic νµ or ντ
flux can prevent the synthesis of heavy nuclei by the r-process neutron
438 Chapter 11
capture. As discussed in Sect. 11.1.4, the hot bubble between the set-
tled protoneutron star and the escaping shock wave at a few seconds
after core bounce might be an ideal high-entropy environment for this
process for which no other site is currently known that could reproduce
the observed galactic heavy element abundance and isotope distribu-
tion. Naturally, the r-process can only occur in a neutron-rich medium
(Ye < 12 ). The p/n ratio in the hot bubble is governed by the β reac-
tions νe n ↔ pe− and ν e p ↔ ne+ . Because the neutrino number density
is much larger than the ambient e+ e− population the proton/neutron
fraction is governed by the neutrino spectra and fluxes. The system
is driven to a neutron-rich phase because normally the ν e ’s are more
energetic than the νe ’s. They emerge from deeper and hotter regions of
the star because their opacity is governed by the same reactions, and
because the core is neutron rich, yielding a larger opacity for νe .
If an exchange νe ↔ νµ,τ occurs outside of the neutrino sphere the
subsequent νe flux is more energetic than the ν e flux which did not
undergo a swap. (A normal mass hierarchy has been assumed.) Even
a partial swap of a few 10% is enough to shift the medium to a proton-
rich state, i.e. to Ye > 12 , to be compared with the standard values of
0.35−0.46. Therefore, the occurrence of such oscillations would be in
conflict with r-process nucleosynthesis in supernovae (Qian et al. 1993).
Fig. 11.20. Density profile for a SN model at 6 s after the core bounce,
typical for √
the “hot bubble phase” (Qian et al. 1993). The right-hand scale
is mres = ( 2GF ne 2Eν )1/2 which indicates the resonance value for (∆m2ν )1/2
for Eν = 10 MeV, assuming Ye = 0.5. (Note that out to a few km above the
neutrino sphere Ye . 0.5.)
The approximate parameter range for which this effect is important

can be estimated from the density profile of a typical SN core a few
seconds after bounce (Fig. 11.20). Again, one expects resonant conver-
sions for the cosmologically interesting neutrino mass range as in the
above shock revival scenario. However, the present effect reaches down
to mν of about 3 eV; for smaller masses the oscillations would occur at
radii too large to have an impact on nucleosynthesis.
With regard to the required mixing angle there is an important dif-
ference to the previous case because the neutron star has already settled
so that the neutrino sphere is now at a radius of about 11 km rather
than at 50 km. Therefore, the relevant length scales in Fig. 11.20 are
reduced relative to Fig. 11.19 which corresponds to a postbounce but
preexplosion configuration. For example, at mres ≈ 40 eV the density
scale height is now |∇ ln ne |−1
res ≈ 0.3 km, about two orders of magni-
tude smaller than before so that the lower limit on sin2 2θ is about
two orders of magnitude larger than it was for the shock revival sce-
nario. From a more detailed analysis Qian et al. (1993) found the
hatched area in Fig. 11.21 where Ye would be driven beyond 0.5 and so
this area would be in conflict with r-process nucleosynthesis in super-
novae.
Fig. 11.21. Mass difference and mixing angle of νe with νµ or ντ where a spec-
tral swap would be efficient enough to help explode supernovae (schemat-
ically after Fuller et al. 1992), and where it would prevent r-process nu-
cleosynthesis (schematically after Qian et al. 1993; Qian and Fuller 1994).
440 Chapter 11
b) Impact of Neutrino-Neutrino Interactions

The discussion so far has been relatively simplistic because the role of
neutrino-neutrino interactions has been ignored. During the hot-bubble
phase the neutrino refractive index caused by other neutrinos is not nec-
essarily negligible (Pantaleone 1995; Qian and Fuller 1995). In terms of
the neutrino luminosity the number flux and thus the density of neutri-
nos of a given species at a radius r is given by nν = Lν 'Eν (−1 (4πr2 )−1 .
Moreover, as in the discussion in Sect. 11.4.2 one must include an aver-
age of the factor (1 − cos Θ) to account for the anisotropy of the neu-
trino phase space distribution (angle Θ between test and background
neutrino). At a given distance r the neutron star (neutrino sphere
radius R) subtends an angle given by sin ΘR = R/r. For a radially
moving test neutrino one finds (for a more rigorous treatment see e.g.
Qian and Fuller 1995)
+ 1 ,+ 1
'1 − cos Θ( = (1 − cos Θ) d cos Θ d cos Θ
cos ΘR cos ΘR
- . /
1
= 2
1− 1− (R/r)2 , (11.15)
which for large r approaches 14 (R/r)2 . Therefore, the effective neutrino

density nν '1 − cos Θ( varies as r−4 at large distances.
The electrons and positrons
√ cause a refractive energy shift between
νe and, say, νµ of ∆V = 2GF (ne− − ne+ ) while the effect of the
neutrinos is
√
∆V ≈ 2GF (nνe − nν e − nνµ + nν µ )'1 − cos Θ(
√ ! "
2GF '1 − cos Θ( Lνe Lν e Lνµ Lν µ
≈ − − + .
4πr2 'Eνe ( 'Eν e ( 'Eνµ ( 'Eν µ (
(11.16)
The tau-flavored neutrino contribution has not been included because
it cancels exactly between ντ and ν τ . The same is true for the mu-
flavored terms before oscillations have taken place. For a test-neutrino
of momentum pν one finds numerically
2pν ∆V ≈ 420 eV2 (10 km/r)2 '1 − cos Θ( ×
! "
pν Lνe Lν e Lνµ Lν µ
× 51 − − + .
10 erg s−1 'Eνe ( 'Eν e ( 'Eνµ ( 'Eν µ (
(11.17)
A few seconds after bounce typical values might be 'Eνe ( = 11 MeV,
'Eν e ( = 16 MeV, and 'Eνµ ( = 'Eν µ ( = 25 MeV, the luminosities can
be taken to be the same at 3×1051 erg s−1 each, and R = 11 km. For
these conditions the estimated contributions to ∆V from electrons and
neutrinos are shown in Fig. 11.22 as a function of radius.
As long as no swap has occurred the neutrinos do not play a major
role because the contribution of νµ cancels exactly against ν µ . Further,
the difference between νe and ν e is smaller than the electron contribu-
tion, and it has the same sign whence it simply causes a slightly larger
effective matter density.
However, on resonance a relatively large number of νe ’s exchange
flavor with νµ ’s so that the neutrino contribution changes sign. More-
over, on resonance the vacuum ∆m2ν by definition cancels against the
medium-induced contribution. Therefore, “switching on” the neutrino
term shifts the resonance position for given vacuum mixing parameters.
Also, in a self-consistent treatment one needs to consider the full non-
Fig. 11.22. “Mass splitting” between νe and νµ (equivalently ντ ) of mo-

mentum pν = 10 MeV caused by the regular medium (the electrons), and
by different neutrino flavors according to Eq. (11.17). For the electrons the
density profile of Fig. 11.20 was used with Ye = 0.5; in a real SN core Ye is
much lower near the neutrino sphere so that the thick line would increase
less steeply toward the neutrino sphere than shown here. For the neutrinos,
a luminosity in each degree of freedom of 3×1051 erg s−1 was assumed, the
average energies were taken to be 'Eνe ( = 11 MeV, 'Eν e ( = 16 MeV, and
'Eνµ ( = 'Eν µ ( = 25 MeV, and the neutrino-sphere radius is 11 km. The
signs of the contributions relative to electrons are: + for νe and ν µ , − for
ν e and νµ . The contribution of νµ cancels exactly against ν µ unless a swap
νe ↔ νµ has taken place.
442 Chapter 11
linear equations of motion for the neutrino density matrix which causes
an “off-diagonal refractive index” as discussed in Sect. 9.3.2. Qian and
Fuller (1995) have performed an approximately self-consistent analysis
of this problem. All told, they found that the neutrino-neutrino interac-
tions have a relatively small impact on the parameter space where flavor
conversion disturbs the r-process. Within the overall precision of these
arguments and calculations, their final exclusion plot is nearly identi-
cal with the schematic picture shown in Fig. 11.21. Qian and Fuller’s
findings are corroborated by a study performed by Sigl (1995a).
c) Summary
In summary, there remains a large range of mixing angles where neu-
trino oscillations between νe and νµ or ντ with a cosmologically in-
teresting mass could help to explode supernovae, and yet not disturb
r-process nucleosynthesis. If one assumes a mass hierarchy with νe
dominated by the lightest, ντ by the heaviest mass eigenstate, the cos-
mologically relevant neutrino would be identified with ντ . It is inter-
esting that the relevant range of mixing angles, 3×10−4 < <
∼ θ ∼ 3×10 ,
−3
overlaps with the mixing angle among the first and third family quarks
which is in the range 0.002−0.005 (Eq. 7.6). Therefore, a scenario
where a massive ντ plays a cosmologically important role, helps to ex-
plode supernovae, and leaves r-process nucleosynthesis unscathed does
not appear to be entirely far-fetched. This scenario leaves the possi-
bility open that the MSW effect solves the solar neutrino problem by
νe -νµ oscillations.
11.4.6 A Caveat
All existing discussions of neutrino flavor oscillations in SNe were based
on spherically symmetric, smooth density profiles. However, there
can be significant density variations, convection, turbulence, and so
forth. Therefore, it is clear that many of my statements about medium-
induced oscillation effects are provisional. Further studies will be re-
quired to develop a more complete picture of SNe and their neutrino
oscillations as 3-dimensional events.
A first study of SN neutrino oscillations with an inhomogeneous den-
sity profile was recently performed by Loreti et al. (1995). They added
a random density field to the standard smooth profile and studied the
impact on neutrino oscillations. They found that the shock-revival sce-
nario involving MSW oscillations can be significantly affected in the
sense that less additional energy is transferred because the stochastic

density field can prevent a complete swap of the neutrino spectra. On
the other hand, for the r-process prevention a complete swap is not
necessary. Therefore, the mixing parameters for which r-process nucle-
osynthesis is prevented by oscillations is not significantly changed, even
if the amplitude of the stochastic density component is as large as 1%.
11.5 Neutrino Propulsion of Neutron Stars

Shortly after the discovery of pulsars (neutron stars) it became clear
that they tend to have the largest peculiar velocities of all stellar pop-
ulations. Recent determinations of pulsar proper motions by radio-
interferometric methods (Bailes et al. 1990; Fomalont et al. 1992; Har-
rison, Lyne, and Anderson 1993) and by interstellar scintillation obser-
vations (Cordes 1986) reveal typical speeds of a few 100 km s−1 . Tak-
ing into account selection effects against high-velocity pulsars, Lyne
and Lorimer (1994) argued that the mean pulsar velocity at birth was
450 ± 90 km s−1 . Associating certain pulsars and SN remnants would
indicate velocities of up to 2000 km s−1 (Frail and Kulkarni 1991; Car-
aveo 1993; Stewart et al. 1993) while the interaction of PSR 2224+65
with its local environment produces a nebula which reveals a transverse
velocity of at least 800 km s−1 (Cordes, Romani, and Lundgren 1993).
Therefore, the distribution of pulsar peculiar speeds appears to
have a mean of 400−500 km s−1 with the largest measured values of
1000−2000 km s−1 . Recall that the galactic rotation velocity is about
200 km s−1 , the escape velocity about 500 km s−1 . Therefore, the fastest
pulsars will eventually escape from the galaxy. In most cases the mi-
gration is away from the galactic disk in agreement with the picture
that pulsars are born in the disk where massive stars can form from
the interstellar gas. However, there seem to be a few puzzling cases of
pulsars which move toward the disk, apparently having formed in the
galactic halo.
Massive stars probably do not form with much larger velocities than
other stars and so pulsars are likely accelerated in conjunction with the
SN collapse that produced them or during their early evolution. One
suggestion for an acceleration mechanism holds that large neutron-star
“kick velocities” are related to the breakup of close binaries, notably
during the SN explosion of the second binary member (Gott, Gunn,
and Ostriker 1970; Dewey and Cordes 1987; Bailes 1989; see also the
review by Bhattacharya and van den Heuvel 1991).
444 Chapter 11
Another possibility is that the SN explosion itself is not spheri-

cally symmetric and thus imparts a kick velocity on the neutron star
(Shklovskiı̆ 1970). Indeed, as discussed in Sect. 11.1.3 SN explosions
likely involve large-scale convective overturns below and above the neu-
trino sphere which could lead to an explosion asymmetry of a few
percent, enough to accelerate the compact core to a speed of order
100 km s−1 , but not enough to account for the typically observed pulsar
velocities (Janka and Müller 1994).
An interesting acceleration mechanism was proposed by Harrison
and Tademaru (1975) who considered the rotation of an oblique mag-
netic dipole which is off-center with regard to the rotating neutron star.
The radiation of electromagnetic power is then asymmetric relative to
the rotation axis and so a substantial accelerating force obtains, enough
to cause velocities of several 100 km s−1 . The velocity reached should
not depend on the magnitude of the magnetic dipole moment while its
direction should correlate with the pulsar rotation axis. These predic-
tions do not seem to be borne out by the data sample of Anderson
and Lyne (1983) although the more recent observations may be less
disfavorable to the “electromagnetic rocket engine.” The correlation
between peculiar velocity and pulsar magnetic moment may now be less
convincing (Harrison, Lyne, and Anderson 1993; Itoh and Hiraki 1994).
Another intriguing mechanism first proposed by Chugaı̆ (1984) relies
on the asymmetric emission of neutrinos (“neutrino rocket engine”).
Recall that the total amount of binding energy released in neutrinos is
about 3×1053 erg; because neutrinos are relativistic they carry the same
amount of momentum. If the neutron-star mass is taken to be 1 M! ,
and if all neutrinos were emitted in one direction, a recoil velocity of
0.17 c = 5×104 km s−1 would obtain. Thus an asymmetric emission of
1.5% would be enough to impart a kick velocity of 800 km s−1 .
Neutrino emission deviates naturally from spherical symmetry if
large-scale convection obtains in the region of the neutrino sphere.
Janka and Müller (1994) believe that 500 km s−1 is a generous upper
limit on the kick velocity that can be achieved by this method. For a
reliable estimate one needs to know the typical size of the convective
cells as well the duration of the convective phase in the protoneutron
star. To this end one needs to perform a fully 3-dimensional calculation.
No such results are available at the present time.
An asymmetric neutrino emission would also obtain in strong mag-
netic fields because the opacity is directional for processes involving
initial- or final-state charged leptons, i.e. URCA processes of the type
e− + p → n + νe or e+ + n → p + ν e . The rates for such processes in the
presence of magnetic fields were discussed by a number of authors.68

Because of the left-handedness of the weak interaction both neutrinos
and antineutrinos would be emitted preferentially in the same direction
singled out by the magnetic field. Its effects become substantial only for
field strengths near and above the critical strength m2e /e = 4.4×1013 G.
In order to obtain neutron-star kick velocities of several 100 km s−1
it appears that magnetic fields several orders of magnitude larger are
required which, however, may possibly exist in some SN cores after
collapse. Also, the asymmetric emission of neutrinos may be aided by
the formation of a pion condensate and perhaps other processes (Par-
fenov 1988, 1989). Certain special field configurations seem to allow
for significantly anisotropic neutrino emission (Bisnovatyi-Kogan and
Janka 1995).
It remains to be seen if sufficiently anisotropic neutrino emission can
be established as a generic property of a protoneutron star’s Kelvin-
Helmholtz cooling phase. Meanwhile, the neutrino rocket engine re-
mains a fascinating speculation for accelerating neutron stars.
11.6 Future Supernovae

The neutrino observations from SN 1987A gave us a wealth of infor-
mation in the sense that they confirmed the broad picture of neutrino
cooling of the compact object formed after collapse. The data were
much too sparse, however, to distinguish between, say, different equa-
tions of state or different assumptions concerning neutrino transport, or
to detect or significantly constrain neutrino masses and mixing parame-
ters. Some worry is caused by the apparent anomalies of the SN 1987A
data, notably the angular distribution of the secondary charged par-
ticles. No doubt it would be extremely important to observe a SN
neutrino signal with greater statistical significance, or from a greater
distance. What is the prospect for such an observation?
SN neutrinos can be observed in a number of underground detec-
tors which are operational now or in the near future, or which have only
been proposed. An extensive overview was given by Burrows, Klein,
and Gandhi (1992). Of the experiments which will become operational
within the foreseeable future, the upcoming Superkamiokande water
Cherenkov detector (Sect. 10.9) would yield by far the largest num-
68
O’Connell and Matese (1969a,b); Matese and O’Connell (1969); Ivanov
and Shul’man (1980, 1981); Dorofeev, Rodionov, and Ternov (1984); Loskutov
(1984a,b); Cheng, Schramm, and Truran (1993).
446 Chapter 11
ber of events. Its fiducial mass for the detection of SN neutrinos is

about 32,000 t, to be compared with 2,140 t for Kamiokande, i.e. it has
a target mass about 15 times larger. Thus one may be able to recog-
nize a neutrino signal from a SN perhaps as much as 4 times farther
away than the Large Magellanic Cloud, which is out to about 200 kpc.
However, the closest large galaxy is M31 (Andromeda) at a distance
of about 700 kpc, allowing Superkamiokande (and all other near-future
detectors) to observe SNe only in our own galaxy and in the Large and
Small Magellanic Clouds.
The rate at which SNe occur in our Galaxy as well as in the LMC
is rather uncertain. From observations in other galaxies the rate of
core-collapse SNe in the Milky Way is estimated to be about 7.3 h2
per century with h the Hubble parameter in units of 100 km s−1 Mpc−1
(van den Bergh and Tammann 1991). Thus, for a low h of order 0.5
one may expect only about 2 such events per century. Roughly the
same number was found in a more recent study by Tammann, Löffler,
and Schröder (1994). The record of historical SNe, on the other hand,
suggests a significantly larger number. Thus it is optimistic, but not
entirely implausible, to hope for an observation within a decade of
Superkamiokande running time. The rate for the LMC is thought to
be about 0.5 per century (Tammann, Löffler, and Schröder 1994)—one
cannot reasonably expect another SN there within our lifetime.
To reach beyond the limits of our own galaxy and the LMC one
would need much more sensitive (much bigger) detectors. It would not
be enough to go as far as Andromeda because this galaxy appears to
have an anomalously low SN rate (van den Bergh and Tammann 1991).
In order to achieve a SN rate of at least 1 per year one may need to use
the Virgo cluster of galaxies at about 15 Mpc (300 times the distance to
the LMC) although it may be enough to reach to the nearby starburst
galaxies M82 and NGC 253 within about 4 Mpc which have a very high
SN rate because of their high rate of star formation (Becklin 1990).
However, a recent estimate of the SN rate for each of these galaxies is
only about 1 per 10 years (van Buren and Greenhouse 1994).
A novel detection scheme (Cline et al. 1990) that may allow one
to build big enough detectors is based on the neutral-current reaction
ν + (Z, N ) → (Z, N − 1) + n + ν which can have a much enhanced
cross section in some nuclei due to collective effects; one would detect
the final-state neutron. The necessary detector volume can be achieved
by using natural deposits of minerals which contain the relevant target
nuclei. Naturally, the main concern would be to reduce sources of
background in order to isolate the feeble signal from a distant SN.
An even more ambitious goal would be to measure the cosmic ν e

background flux from all past SNe in the universe. For energies below
around 10 MeV this flux is swamped by many orders of magnitude by
that from the nuclear power plants on Earth. Above a few 10 MeV the
atmospheric neutrino flux would dominate and so there is only a small
window where the cosmic SN flux might be detectable. However, even a
moderately sized (200 tons) scintillation detector located on the moon
would have a chance of measuring this flux because these backgrounds
do not exist there (Mann and Zhang 1990).
Returning to Earth one may speculate about what could be learned
if a galactic SN were indeed observed at Superkamiokande. For the pur-
pose of argument a distance of 10 kpc (5 times closer than the LMC)
is assumed. (Recall that the solar system is at a distance of about
8 kpc from the galactic center.) Then one expects at Superkamiokande
about 4000 events from the reaction ν e p → ne+ , compared with 270
at Kamiokande, and with 12 measured there from SN 1987A (Totsuka
1990). This would be enough to determine a statistically very signifi-
cant and very detailed “neutrino lightcurve.”
Interestingly, one expects about 13 events within the first few ms
from the prompt νe burst. Its presence would indicate that the νe ’s
have not oscillated, say, into ντ ’s as would be expected for a cosmolog-
ically interesting ντ mass (Sect. 11.4.2), thus excluding a large range
of masses and mixing angles (Fig. 11.17). Moreover, the prompt burst
could not have been dispersed by a neutrino mass and so a mνe bound of
order 1 eV could be derived. A number of other conclusions tentatively
reached for SN 1987A could be affirmed (Sect. 13.2). The nonobserva-
tion of the prompt burst, on the other hand, would be more difficult
to interpret as its absence could have a variety of causes ranging from
neutrino oscillations to some flaw in the standard picture of SN collapse.
The ν e lightcurve which would last, say, 10 s would not allow one to
extract interesting bounds on mνe relative to the ones already obtained
from SN 1987A and from laboratory experiments. From Eq. (11.9) one
concludes that for a distance D ≈ 10 kpc one is sensitive to masses in
the 100 eV range.
In Superkamiokande one would expect to see about 40 events for
each νµ +ν µ and ντ +ν τ from the elastic ν-e scattering process. Because
the final-state electrons are strongly forward peaked one can separate
them from the isotropic ν e p → ne+ signal. Therefore, a νµ or ντ mass
would manifest itself through late forward events. Seckel, Steigman,
and Walker (1991) found that from the signal in water Cherenkov de-
tectors one could be sensitive to a mass down to about 75 eV. Of
448 Chapter 11
course, masses so large and larger are excluded from cosmology—such

neutrinos would have to be unstable. In order to obey the cosmologi-
cal limits neutrinos emitted at a distance of 10 kpc would decay before
reaching Earth if their mass exceeds a few 10 keV. Thus, the obser-
vation of a galactic SN would allow one, at best, to probe the mass
window 100 eV < <
∼ mν ∼ 30 keV. A similar conclusion was reached by
Acker, Pakvasa, and Raghavan (1990) who considered the signature in
the proposed BOREX detector.
It may be possible, however, to probe a somewhat smaller mass
for, say, the ντ down to the cosmologically interesting range of 30 eV
if one takes advantage of all aspects of the observed neutrino signal
(Krauss et al. 1992). Because the late part of the neutrino lightcurve
is expected to be similar for ν e and the other flavors one can hope
to extract the behavior of the source from the ν e signal. Then one
would be more sensitive to modifications of the ντ lightcurve caused by
dispersion effects. However, in order to identify the ντ ’s one would have
to use energy cuts (νe ’s and ντ ’s have different spectra!) in addition to
angular cuts. In the analysis of Krauss et al. (1992) the possibility of
νe ↔ ντ oscillations was not included which would weaken the range of
accessible ντ masses because of the modified energy spectra. Of course,
the r-process nucleosynthesis argument of Sect. 11.4.5 would indicate
that MSW oscillations did not take place late even if they took place
early and rendered the prompt νe burst unobservable.
Still, a ντ mass relevant for the dark matter content of the universe
as well as for scenarios of structure formation may be much lower than
30 eV. Therefore, on the basis of current analyses the prospect of being
able to recognize a cosmologically relevant νµ or ντ in the neutrino
signal of a galactic SN appears relatively dim.
A far more positive view was taken by Cline et al. (1994) who ar-
gued that masses down to 15 eV may be accessible by the simultaneous
operation of Superkamiokande and their previously proposed (Cline
et al. 1990) Supernova Burst Observatory (SNBO) which is based on
neutral-current reactions alone. This would obviate the need to sep-
arate the charged-current ν e n → pe+ detection at Superkamiokande
from the neutral-current reaction νe → eν by angular and energy cuts.
One could use the Superkamiokande ν e signal to monitor the SN neu-
trino lightcurve, notably its sharp onset, and relate it to the onset of
the neutral-current events at SNBO which would be “washed out” for
a cosmologically interesting mass of, say, the ντ . One must hope that
SNBO will become a real project in the near future.
Chapter 12
Radiative Particle Decays

from Distant Sources
If neutrinos, axions, or other low-mass particles had radiative decay

channels, the decay photons would appear as x- or γ-ray fluxes from
stellar sources where these particles can be produced by nuclear or
plasma processes. This chapter is devoted to limits on such decays,
including decays into charged leptons, that are based on observational
limits on photon or positron fluxes from stellar sources, notably the Sun
and supernova 1987A. For comparison, laboratory and cosmological
limits are also reviewed.
12.1 Preliminaries
This book is largely about the properties of electrically neutral particles
whose electromagnetic interactions are correspondingly weak. However,
because they can virtually dissociate into charged states, they will still
interact with photons through higher-order amplitudes. The focus of
the present chapter is the possibility of radiative decays of the form
ν → ν ! γ (neutrinos) or a → γγ (axions), but also ν → ν ! e+ e− and
ν → ν ! e+ e− γ. Cowsik (1977) was the first to recognize that the huge
path lengths available in the astrophysical environment allow one to
obtain much more restrictive limits on such decays than from labora-
tory experiments. For example, the absence of single-photon counts in
a detector near a fission reactor indicates a bound69 τγ /mνe > 22 s/eV
69
In this chapter τγ will always denote the partial neutrino decay time into radia-
tion while τtot is the total decay time if hypothetical invisible channels are included.
−1
Then τγ−1 = Bγ τtot with the branching ratio Bγ .
449
450 Chapter 12
on the process νe → ν ! γ. The corresponding limit based on a compar-

ison between the measured solar neutrino flux and the measured limit
on x- or γ-rays from the quiet Sun is 7×109 s/eV, almost 9 orders of
magnitude more restrictive. Moreover, this method is fully analogous
to a laboratory experiment as it is based on a measured neutrino flux
and a measured upper limit photon flux. To a lesser degree this re-
mark also applies to the even better SN 1987A constraints which are
applicable to all neutrino flavors.
In addition, less directly established particle fluxes can be used such
as those from the stars in the galactic bulge or from all hydrogen-
burning stars or supernovae in the universe. Even more indirectly, one
may study the impact of the radiative decay of the cosmic background
sea of neutrinos or axions which are predicted to exist in the framework
of the big-bang theory of the early universe.
Usually, the bounds on ν → ν ! γ thus obtained are presented as lim-
its on the radiative decay time τγ . Even if one does not aim at an imme-
diate theoretical interpretation, however, the significance of τγ is limited
because it represents a combination of the final-state phase-space vol-
ume and the matrix element. The latter can be expressed in terms
of an effective transition moment µeff as in Eq. (7.12) of Sect. 7.2.2.
This moment characterizes the interaction strength independently of
phase-space effects, providing a much more direct link between the ex-
perimental results and an underlying theory.
A “heavy” neutrino νh with mh > 2me ≈ 1 MeV can decay into
νe e+ e− . This channel is often included in the notion of “radiative” de-
cays because relativistic charged leptons cause experimental signatures
similar to γ rays. If this decay proceeds by virtue of a mixing amplitude
Ueh between νh and νe the rate is given by Eq. (7.9). Therefore, it is
characterized by |Ueh | in a phase-space independent way.
If νe is a mixture of different mass eigenstates, any νe source such
as a power reactor or the Sun produces all components. If their mass
differences are small one needs to consider in detail the phenomenon of
neutrino oscillations as in Chapter 8. However, if the oscillation length
is much smaller than the distance between the detector and the source,
the neutrino flux can be considered an incoherent mixture of all mass
eigenstates; at a reactor this is the case for ∆m2 > 2
∼ 1eV . The flux of
“heavy” neutrinos from a νe source is then given by
Fνh (Eν ) = |Ueh |2 β(Eν ) Fνe (Eν ) . (12.1)
The velocity β(Eν ) = (1 − m2h /Eν2 )1/2 enters from the phase space of
nonrelativistic neutrinos in the production process. From νe sources
Radiative Particle Decays 451
one can then derive limits on Ueh for any νh , even a hypothetical sterile
one, if mh >
∼ 1 MeV because it is the same mixing amplitude that allows
for its production in the source and for its νh → νe e− e+ decay.
In the following I will discuss radiative lifetime limits from different
sources approximately in the order of available decay paths, from labo-
ratory experiments (a few meters) to the radius of the visible universe
(about 1010 light years).
12.2 Laboratory Experiments

12.2.1 Spectrum of Decay Photons
Perhaps the simplest neutrino source to use for laboratory experiments
is a nuclear power reactor which produces a strong ν e flux from the weak
decays of the uranium and plutonium fission products. If a γ detector
is placed at a certain distance from the reactor core one may assume
that the local neutrino flux Fν (Eν ) is known (units cm−2 s−1 MeV−1 ).
As a first step one then needs to compute the expected flux Fγ (Eγ ) of
photons from the decay ν → ν ! γ. The result will also apply to station-
ary stellar sources such as the Sun while for the short neutrino burst
from SN 1987A one needs to derive a separate expression (Sect. 12.4).
Because the decay is a dipole transition, the general form of the
photon angular distribution in the rest frame of the parent neutrino is
dNγ /d cos ϑ = 12 (1 − α cos ϑ), (12.2)
where ϑ is the angle between the ν spin polarization vector and the
photon momentum. For Majorana neutrinos the decay is isotropic,
independently of their polarization, and thus α = 0. Similarly for
axion decays, a → γγ, as these particles have no spin so that in their
rest frame no spatial direction is favored. For polarized Dirac neutrinos
the possible parameter range is −1 ≤ α ≤ 1.
The decay photon has an energy ω = δm mν /2 in the rest frame of
the parent neutrino—see Eq. (7.12). If it is emitted in a direction θ
with regard to the laboratory direction of motion, the energy in the
laboratory frame is Eγ = ω (Eν + pν cos θ)/mν . Hence,
Eγ = 12 δm Eν (1 + β cos θ), (12.3)
where β = pν /Eν = (1 − m2ν /Eν2 )1/2 is the neutrino velocity. For a left-
handed parent neutrino the spin is polarized opposite to its momentum
so that cos θ = − cos ϑ and dNγ /d cos θ = 12 (1 + α cos θ). For a very
452 Chapter 12
relativistic parent neutrino β = 1 and so the normalized photon energy

distribution corresponding to Eq. (12.2) is
! "
dNγ 1 Eγ
= 1 − α + 2α , 0 < E γ < δ m Eν . (12.4)
dEγ δ m Eν δm E ν
The decays of Majorana neutrinos or axions (α = 0) produce a box-

shaped spectrum while for α = ±1 it is triangle shaped (Fig. 12.1).
Fig. 12.1. Photon spectrum from the decay of a relativistic neutrino (en-
ergy Eν ) according to Eq. (12.4).
A fraction (mν /Eν )(dγ /τγ ) of neutrinos decay before reaching the
detector if the laboratory lifetime is large compared with the decay
path dγ . Here, τγ is the rest-frame radiative decay time and Eν /mν
the time dilation factor. Integrating over the neutrino source spectrum
then yields
! "
mν # ∞ dEν Eγ Fν (Eν )
Fγ (Eγ ) = dγ 1 − α + 2α . (12.5)
τγ Eγ /δm δm δm E ν Eν2
Most sources emit either antineutrinos (for example ν e ’s from a fission

reactor) or neutrinos (for example νe ’s from the Sun). In these cases
Eq. (12.5) gives us directly the expected γ flux as a function of the as-
sumed value for α. However, there are some examples for simultaneous
ν and ν sources. In those cases one needs to know which value of α to
use for ν if a certain value for ν has been assumed.
Under very general assumptions the laws of particle physics are
invariant under a simultaneous transformation which takes particles
into antiparticles (charge conjugation C), reflects all spatial coordinates
(parity transformation P), and inverts motions (time reversal T). In this
case the CPT theorem states that the masses and total decay times of
corresponding particles and antiparticles are the same. However, the

partial decay rates into specific channels need not be identical, and
indeed, K ◦ and K ◦ show such CP-violating decays: A transformation
under CP leads to a “mirror world” which is different from the one we
live in. Therefore, in the most general case one may only assume that
ν and ν have the same total lifetimes while their radiative decay rates
may be different.
However, it is common to analyze the available data under the as-
sumption of CP conservation for all neutrino interactions. For the
radiative decay of a polarized ν in its rest frame, the CP-mirrored de-
cay is one where a polarized ν with the same spin decays into ν ! and
γ of reversed momenta.70 Therefore, ν and ν of the same polarization
are characterized by opposite values for α in Eq. (12.2). In the source,
however, the ν’s and ν’s are produced by weak interactions which vio-
late parity maximally. If they are relativistic they both have negative
(left-handed) chiralities which means that the ν’s have negative and the
ν’s positive helicities. Thus, relative to their momentum left-handed
ν’s and ν’s show the same distribution of decay photons so that in
Eq. (12.5) one must use the same α for both.
12.2.2 Electron Neutrinos from Reactors

Fission reactors are superb neutrino sources. At a thermal power of
2800 MW, for example, one expects about 5×1020 ν e /s. At a distance
of 30 m this corresponds to a flux of about 4×1012 cm−2 s−1 , almost a
hundred times larger than the solar νe flux of about 6.6×1010 cm−2 s−1 .
Reactor ν e ’s emerge from many weak decays of the products of the
neutron induced fission of 235 U and 239 Pu while the fission of 238 U and
241
Pu contributes less than 10% to the total rate. The spectral distribu-
tion can be inferred from a measurement of the corresponding β spectra
together with the reasonably well-known distribution of the end point
energies of the fission products (von Feilitzsch et al. 1982; Schrecken-
bach et al. 1985). The distribution of ν e energies per fission of 235 U and
239
Pu is shown in Fig. 12.2 with a total of about 6 ν e ’s per fission.
An early, relatively crude, but often-quoted (Particle Data Group
1994) limit on the decay ν e → ν ! γ was derived by Reines, Sobel, and
Gurr (1974) on the basis of the upper limit γ flux in a scintillation detec-
tor near the Savannah River reactor (U.S.A.). A weaker but more reli-
70
A parity transformation P inverts all polar vectors, e.g. momenta, currents,
or electric fields, while it leaves axial vectors unchanged, e.g. angular momenta,
magnetic moments, or magnetic fields.
454 Chapter 12
Fig. 12.2. Spectrum of reactor ν e ’s per fission of 239 Pu and 235 U (von Feil-
itzsch et al. 1982; Schreckenbach et al. 1985). The width of the lines gives
the total error of the spectra.
able bound was inferred by Vogel (1984) from data taken at the Gösgen
reactor (Switzerland). The most recent analysis is, again, based on
data taken with a scintillation counter at Gösgen (Oberauer, von Feil-
itzsch, and Mössbauer 1987). From a comparison of the “reactor on”
with the “reactor off” photon counts for several energy channels in the
MeV range and using the experimentally established neutrino spectrum
(Fig. 12.2), these authors found the 68% CL lower limits on the ν e ra-
diative decay times of τγ /mνe > 22 s/eV for α = −1, 38 s/eV for α = 0,
and 59 s/eV for α = +1. It was assumed that ν ! is massless so that
δm = 1 in Eq. (12.5). With Eq. (7.12) the α = −1 constraint translates
into a bound on the effective electromagnetic transition moment of
µeff < 0.092 µB m−2

eV , (12.6)
not a very restrictive limit even if νe saturates its upper mass bound of
about 5 eV.
Even this weak limit would cease to apply if νe and ν ! became
nearly degenerate. Therefore, Bouchez et al. (1988) performed an ex-
periment where they searched for optical decay photons at the Bugey
reactor (France). They excluded a certain region in the parameter
plane spanned by τγ /mνe and δm . Their greatest sensitivity was ap-
proximately at δm = 2×10−5 where they found τγ /mνe > ∼ 0.04 s/eV.
< 7 2
With Eq. (7.12) this is µeff ∼ 10 µB /meV which, unfortunately, is ir-
relevant as a constraint. Hence, for nearly degenerate neutrinos there
is no meaningful limit on µeff from decay experiments. This example

highlights the importance of separating the intrinsic coupling strength,
or the magnitude of the matrix element, from phase-space effects in the
interpretation of such results.
12.2.3 Heavy Neutrinos from Reactors

Considering the decay of νe into a different neutrino species implies
entertaining the notion of the nonconservation of the electron lepton
number, forcing one to contemplate the possibility of neutrino flavor
mixing as well. Therefore, reactors will be sources for other flavors
and notably of “heavy neutrinos” according to Eq. (12.1). The photon
flux from radiative νh decays can now be calculated as before, except
that the dwelling time for nonrelativistic νh ’s in the decay volume is
increased by β −1 so that the velocity cancels between this factor and
Eq. (12.1). However, the photon spectrum shown in Fig. 12.1 must
be modified for the decays of nonrelativistic neutrinos because β < 1
in Eq. (12.3). Therefore, the overall expression for Fγ (Eγ ) becomes
somewhat more involved.
As long as mh < ∼ 1 MeV one may treat the νh flux as relativistic.
Then the radiative decay limits are the same as for ν e , except that
they are diminished by the reduced neutrino flux. Thus, the bound
Eq. (12.6) translates into
µeff < 0.92×10−13 µB |Ueh |−1 m−2

MeV . (12.7)
With mh up to an MeV this bound has a lot more teeth than the one
on electron neutrinos.
For mh > 2me ≈ 1 MeV the decays νh → νe e+ e− will become
kinematically possible and probably dominate. The scintillation coun-
ters that were used to search for decay photons near a power reactor
are equally sensitive to electrons and positrons—for many purposes
relativistic charged particles may be treated almost on the same foot-
ing as γ rays. Therefore, the same Gösgen data have been analyzed
to constrain the mixing amplitude Ueh (Oberauer, von Feilitzsch, and
Mössbauer 1987; Oberauer 1992). Even more restrictive limits were
obtained from data taken at the Rovno reactor (Fayons, Kopeykin,
and Mikaelyan 1991), and most recently at the Bugey reactor (Hagner
et al. 1995). Note that in this method the same mixing probability
|Ueh |2 appears in Eq. (12.1) to obtain the νh flux from a νe source, and
in Eq. (7.9) to obtain the decay probability. Hence, the expected e+ e−
flux is proportional to |Ueh |4 .
456 Chapter 12
In Fig. 12.3 the excluded range of masses and mixing angles71 is

shown together with similar constraints from other neutrino sources.
The reactor bounds are weaker than those from the Sun and SN 1987A,
but they remain important because of the short decay path involved!
If one accepts the big-bang nucleosynthesis bounds (Sect. 7.1.5) the ντ
total lifetime must be so short that it would not escape from the mantle
of a SN before decaying, and perhaps not even from the Sun.
12.2.4 Neutrinos from a Beam Stop

Another powerful laboratory source for both νe ’s and νµ ’s is a beam
stop where neutrinos are produced from the decay of stopped pions,
π + → µ+ νµ and the subsequent decay of stopped muons, µ+ → e+ νe ν µ .
In a recent experiment of this sort (Krakauer et al. 1991), the neu-
trino intensity was 4.3×1013 νµ /s with a total of 8.52×1019 νµ . In these
decays the νµ has a fixed energy of (m2π − m2µ )/2mπ = 29.8 MeV
while the other normalized spectra are (3/Y 4 )(3Y − 2Eν )Eν2 for ν µ
and (12/Y 4 )(Y − Eν )Eν2 for νe with Y ≡ 12 mµ = 52.8 MeV. For νe ,
the 90% CL radiative lifetime limit as a function of the “anisotropy
parameter” α is τγ /mνe > (15.9 + 9.8α + 0.3α2 ) s/eV, somewhat less
restrictive than the reactor results.
For νµ and ν µ one obtains slightly different limits because of the
different source spectra. Under the assumption of CP invariance the
radiative lifetimes for νµ and ν µ are the same. In this case the combined
limit is τγ /mνµ > (36.3 + 21.65α + 0.75α2 ) s/eV while the individual
limits are about half this value. In terms of an effective transition
moment the most conservative case (α = −1) yields
µeff < 0.11 µB m−2

eV . (12.8)
For the admixture of other mass eigenstates this result may be trans-
lated in a fashion analogous to the discussion of reactor neutrinos.
Beam-stop neutrinos from meson decays may also be used to con-
strain the e+ e− decays of heavy admixtures. Because of the larger
amount of available energy one may probe higher masses for νh while
the reactor bounds drop out above a few MeV because of the relatively
soft spectrum. In Fig. 12.3 the most restrictive such constraints are
summarized.
71
It is customary to display |Ueh |2 when constraining the mixing parameters of
heavy, decaying neutrinos while one shows sin2 2θeh for light, oscillating neutrinos
as in Chapter 8. For easier comparison I always use the mixing angle. Recall that
|Ueh | = sin θeh so that for small mixing angles sin2 2θeh = 4|Ueh |2 .
Fig. 12.3. Bounds on νe -νh mixing from the absence of νh → νe e+ e− decays.

Meson decays: (a,b) Leener-Rosier et al. (1986) and (c) Bryman et al. (1983).
Reactor neutrinos: (d) Hagner et al. (1995). Absence of reactor neutrino
oscillations: (e) Zacek et al. (1986). Absence of solar positrons: Toussaint
and Wilczek (1981); see also Sect. 12.3.2.
These results are based on νe sources which produce νh by their

mixing which also leads to the subsequent νh → νe e+ e− decay. Of the
known sequential neutrinos νh can be identified only with ν3 , the dom-
inant mass component of ντ with an allowed mass of up to 24 MeV.
If one makes this identification, stronger limits are obtained from di-
rect ντ ≈ ν3 sources. One example is the beam stop at the Big Euro-
pean Bubble Chamber (BEBC) where a strong flux of charmed strange
mesons72 Ds was produced which subsequently can decay as Ds → τ ντ
besides the dominant hadronic modes (WA66 Collaboration 1985). Ac-
cording to Babu, Gould, and Rothstein (1994) who quote a private
communication from the WA66 collaboration, a model-independent
constraint from the BEBC experiment is τγ /m3 > 0.15 s/MeV or
µeff < 1.1×103 µB m−2

eV . (12.9)
Another constraint is τe+ e− /m3 > 0.18 s/MeV. As a constraint on

the mixing amplitude, |Ue3 |2 < 1.6×105 m−6MeV , it is weaker than those
shown in Fig. 12.3. However, as it is based on a direct ντ flux it is valid
even if the decays are not induced by mixing but by exotic intermediate
states (Babu, Gould, and Rothstein 1993).
72
The Ds used to be called F as in the quoted reference.
458 Chapter 12
12.3 Particles from the Sun
12.3.1 Electron Neutrinos
Like a terrestrial power reactor, the Sun is a prolific neutrino source ex-
cept that it emits νe ’s rather than ν e ’s. The expected spectrum as well
as the relevant measurements were discussed in Chapter 10. Suffice
it to recall that the solar neutrino flux is now experimentally estab-
lished without a shred of doubt. There remain significant discrepancies
between the predicted and measured spectral shape of the spectrum
which may be explained by neutrino oscillations. However, the solar
neutrino problem is a fine point in the context of the present discus-
sion because the following results depend mostly on the low-energy
pp flux.
One may proceed exactly as in the previous section in order to trans-
late the solar neutrino spectrum shown in Fig. 10.1 into an expected
flux of x- and γ-rays from the Sun. In Fig. 12.4 I show this flux at Earth
for τγ /mνe = 10 s/eV (about the laboratory lifetime limit) and for the
values ±1 for the “anisotropy parameter” α. The spectrum as shown is
based on the calculated neutrino flux. The shoulders corresponding to
other than the pp neutrinos likely would have to be reduced somewhat.
The magnitude of the photon flux is enormous because the decay path
dγ is the entire distance to the Sun of 1.5×1013 cm = 500 s.
The quiet Sun is a significant source of soft x-rays from the quasi-
thermal emission of the hot corona at T ≈ 4.5×106 K. The flux mea-
surements of Chodil et al. (1965) are marked as open diamonds in
Fig. 12.4. The flux of decay photons in Fig. 12.4 would outshine the
solar corona by some 4 orders of magnitude! Moreover, the corona
spectrum falls off sharply at larger energies. In the hard x- and soft
γ-ray band very restrictive upper limits exist on the emission of the
quiet Sun that are shown in Fig. 12.4. They are based on balloon-
borne detectors flown many years ago (Frost et al. 1966; Peterson et al.
1966). These upper limits are still far above the estimated albedo (ra-
diation from cosmic rays hitting the surface of the Sun), leaving much
room for improvement. Alas, the quiet Sun is not an object of great
interest to γ-ray astronomers and so more recent measurements do not
seem to exist.
In order to respect these measured upper limit photon fluxes one
must shift the decay spectrum in Fig. 12.4 down by about 8 orders of
magnitude (thin solid line in Fig. 12.4). This yields a lower radiative
liftime limit for νe of τγ /mνe > 9
∼ 7×10 s/eV (Cowsik 1977; Raffelt 1985).
Fig. 12.4. Spectrum of photons from the solar neutrino decay νe → ν ! γ for
the indicated values of the anisotropy parameter α. Measurements of the
x-ray emission of the solar corona (open diamonds) according to Chodil et
al. (1965). Upper limit x- and γ-ray fluxes according to Frost et al. (1966)
and Peterson et al. (1966). Estimated albedo according to Peterson et al.
(1966). Thin solid line: Maximally allowed photon spectrum from neutrino
decay. (Figure adapted from Raffelt 1985.)
With Eq. (7.12) this translates into
µeff < 5×10−6 µB m−2

eV . (12.10)
It must be stressed, again, that the recent progress in solar neutrino

astronomy has placed this result on the same footing as a terrestrial
experiment—the magnitude of the solar neutrino flux and its main
spectral features are now experimentally established!
460 Chapter 12
This bound is diminished if νe and ν ! are nearly degenerate so that

in Eq. (7.12) δm ( 1. The structure of the photon flux as a function of
δm in Eq. (12.5) is such that for δm < 1 the spectrum can be obtained,
in a doubly logarithmic representation such as Fig. 12.4, by shifting
it “to the left” and “upward” by the amount | log δm | each, the shape
itself remaining unchanged (Raffelt 1985). The excluded regime in the
plane of τγ /mνe and δm is shown in Fig. 12.5 (left panel). Again, it
is more appropriate to express these limits in terms of µeff and δm by
virtue of Eq. (7.12), leading to Fig. 12.5 (right panel). As expected, for
fixed mνe the limits on µeff quickly degrade with small δm .
Fig. 12.5. Excluded parameters for νe → ν ! γ from the Sun for nearly degen-
erate neutrino masses (adapted from Raffelt 1985).
A limit on τγ /mνe similar to the solar one can be obtained from

the central bulge of the galaxy. Within 2.5 kpc it contains a luminosity
of about 2×1010 L% and thus a neutrino luminosity similarly enhanced.
With its distance of (8.7±0.6) kpc it is about 2×109 times farther away
from us, leading to a much smaller local neutrino flux than that from
the Sun. However, the neutrino decay path is also 2×109 times larger.
Even though the neutrino flux scales with the inverse of the distance
squared from the source, the flux of decay photons scales only with the
inverse distance! Therefore, the local flux of decay photons would be
larger than the solar one by perhaps a factor of ten. The measured hard
x- and soft γ-ray flux from the central region of the galaxy is similar
in magnitude to the upper limit solar flux so that one obtains a similar
constraint on radiative decays. Because no dramatic improvement is
expected a detailed analysis is not warranted. However, a substantial
improvement is achieved by considering all hydrogen-burning stars in
the universe as a source (Sect. 12.6).
12.3.2 Heavy Neutrino Admixtures
In full analogy to the case of reactor experiments the solar bound

Eq. (12.10) can be reinterpreted as a limit on radiative decays of heavy
νe admixtures. To this end one interprets meV = mh /eV and introduces
the factor |Ueh |−1 on the r.h.s. of Eq. (12.10). It must be stressed, how-
ever, that this simple procedure is only applicable to mh < ∼ 30 keV
because the main part of the solar neutrino spectrum is relatively soft.
In the remaining range of interest, 30 keV < <
∼ mh ∼ 1 MeV, the νh flux is
partly suppressed. Moreover, a large fraction of it will be nonrelativis-
tic or only moderately relativistic so that the flux of decay photons will
have a nonnegligible angular divergence. The detectors used to derive
the constraints shown in Fig. 12.4 had a limited forward aperture of
about 0.15 sr, relevant in the energy range 18−185 keV (Peterson et al.
1966), and 1 sr, relevant in the energy range 163−774 keV (Frost et al.
1966). Therefore, a certain part of the photon flux that would have
come from angles relatively far away from the Sun would have been cut
out, weakening the bounds on radiative decays of νh .
For mh > 2me one would, again, expect the decay νh → νe e+ e−
to dominate. For mh up to about 14 MeV one may use the 8 B neu-
trinos from the Sun as a source spectrum. The flux of interplanetary
positrons from cosmic ray secondaries is measured to be approximately
10−4 cm−2 s−1 sr−1 MeV−1 at kinetic energies of about 5 MeV. Inter-
preting this flux as an upper limit to possible decay positrons from
solar neutrinos, Toussaint and Wilczek (1981) derived

−4
 2×10
 (mh = 2 MeV),
|Ueh |2 <
∼  2×10
−5
(mh = 5 MeV), (12.11)

3×10−6 (mh = 10 MeV).
Because they used the theoretically expected rather than the experi-
mentally measured solar 8 B neutrino flux I have discounted their orig-
inal numbers by a factor of 3. These bounds are included in Fig. 12.3;
in the applicable mass range they are more restrictive than those from
laboratory experiments. They are valid only if the νh flux is not dimin-
ished by invisible decay channels on its way between Sun and Earth, i.e.
the total νh (laboratory) lifetime must exceed about 500 s. Because in
the mass range of a few MeV the time dilation factor for about 10 MeV
neutrinos is not large, the bounds apply for total νh lifetimes exceeding
about 100 s. Such “long-lived” MeV-mass neutrinos are in conflict with
the big-bang nucleosynthesis constraints shown in Fig. 7.2.
462 Chapter 12
12.3.3 New Particles
Besides neutrinos, stars can also produce other weakly interacting par-
ticles by both plasma and nuclear processes. With a temperature of
about 1.3 keV in the solar center the former reactions would produce a
relatively soft spectrum and so I focus on nuclear reactions where MeV
energies are available. If the new particle is a boson and if it couples
to nucleons, it will substitute for a photon with certain relative rates r
in reactions with final-state γ-rays. A short glance at the nuclear reac-
tion chains shown in Fig. 10.2 reveals that a particularly useful case is
p + d → 3 He + γ with Eγ = 5.5 MeV. This reaction occurs about 1.87
times for every 4 He nucleus produced by fusion in the Sun and so it
must occur about 1.7×1038 s−1 . A certain fraction of the particles pro-
duced will be reabsorbed or decay within the Sun. If their probability
for escaping is p the Sun emits r p 1.7×1038 s−1 of the new objects.
If the new particle is a scalar boson a it will have a decay channel
a → 2γ. Because this decay is isotropic in a’s rest frame the spectrum
of decay photons is box-shaped (Fig. 12.1) with an upper endpoint of
5.5 MeV. If a fraction q of the particles decays between the Sun and
Earth the local γ flux is r p q 2.2×1010 cm−2 s−1 MeV−1 . The upper limit
photon flux shown in Fig. 12.4 at 5.5 MeV is 0.8×10−3 cm−2 s−1 MeV−1
(Peterson et al. 1966). From there, Raffelt and Stodolsky (1982) found
the general upper bound r p q < 4×10−14 . They also calculated r, p,
and q for the specific case of “standard axions” and were able to derive
a strong limit on the properties of this hypothetical particle. Together
with many laboratory constraints (Particle Data Group 1994) standard
axions are now entirely excluded, the main motivation to consider “in-
visible axions” instead (Chapter 14).
12.4 Supernova 1987A
12.4.1 Decay Photons from Low-Mass Neutrinos
The most significant constraints on radiative particle decays from stellar

sources can be derived on the basis of the neutrino burst from supernova
(SN) 1987A. The neutrino observations and their interpretation were
discussed in Chapter 11. For the present purpose it is enough to know
that the neutrinos arrived in a short burst lasting a few seconds with
a total emitted energy per flavor of about 1×1053 erg = 6.2×1058 MeV.
For ν e the measured spectral distribution is consistent with a thermal
emission at Tνe ≈ 4 MeV so that the fluence73 of ν’s plus ν’s per flavor
was about
Fν = 1.4×1010 cm−2 (4 MeV/Tν ) , (12.12)
taking )Eν * = 3Tν . Approximately the same result is thought to apply

to νµ and ντ with about 1.3−1.7 times the temperature, although for
those flavors there is no direct measurement.
Because SN 1987A occurred in the Large Magellanic Cloud (LMC)
at an approximate distance of dLMC = 50 kpc = 1.5×1023 cm an enor-
mous decay path was available for the neutrinos from this measured
source. A very restrictive upper limit photon flux was provided by the
gamma ray spectrometer on the solar maximum mission (SMM) satel-
lite which was operational at the time of the neutrino signal and did
not register any excess γ counts above the normal background. In or-
der to use this result one needs to compute the expected γ signal from
neutrino decay. Because one is dealing with a short neutrino burst the
previous results for stationary sources do not apply directly: the time
structure of the expected photon burst must be taken into account.
This is easy when the mass of the parent neutrino is below about
40 eV; the pulse dispersion is then not much larger than the duration
of the observed ν e burst. Because for low-mass neutrinos the decay
photons have essentially the same time structure as the neutrino burst
one considers the γ fluence for a time interval of about 10 s around the
first neutrino arrival. The non-νe flavors could be heavier if they violate
the cosmological mass limit of a few 10 eV. Then the photon pulse will
be correspondingly stretched, a case to be studied in Sect. 12.4.4 below.
If the neutrino masses are not degenerate so that in Eq. (12.5)
δm = 1 the expected differential fluence is
# ∞ ! "
mν Eγ Φν (Eν )
Fγ! (Eγ ) = Fν dLMC dEν 1 − α + 2α ,
τγ Eγ Eν Eν2
(12.13)
where Φν (Eν ) ≡ Fν! (Eν )/Fν is a normalized spectrum (units MeV−1 ).

For this expression CP conservation was assumed so that ν’s and ν’s are
characterized by the same values of τγ and α as discussed in Sect. 12.2.1.
73
With fluence one means the time-integrated flux. In this book I use the sym-
bol F for a differential particle flux (cm−2 s−1 MeV−1 ), the symbol F for a fluence
(cm−2 ), and F " = dF/dE for a “differential fluence” which includes spectral infor-
mation (cm−2 MeV−1 ).
464 Chapter 12
Fig. 12.6. Expected fluence of photons from the decay ν → ν ! γ of low-mass

SN neutrinos according to Eq. (12.15) with mν /τγ = 10−15 eV/s and the
indicated neutrino temperatures. The shaded bands for each temperature
are for the range −1 ≤ α ≤ +1 with the harder edge corresponding to
α = +1. Also shown are the upper limits from the GRS channels taken from
the 10 s column of Tab. 12.1.
Because details of the spectral form are not known it is easiest to

use a Boltzmann distribution as a generic case,
Eν2 e−Eν /Tν

Φν (Eν ) = . (12.14)
2Tν3
Then one finds explicitly74
mν dLMC ( )
Fγ! (Eγ ) = Fν (1 − α) e −ε
+ 2α ε E 1 (ε) , (12.15)
τγ 2Tν2
where ε ≡ Eγ /Tν . In Fig. 12.6 this spectrum is shown for Tν = 4 and

8 MeV with mν /τγ = 10−15 eV/s. The envelopes of the shaded bands
in Fig. 12.6 correspond to α = ±1 where for each temperature the
“harder” edge corresponds to α = +1.
The observational constraints give a limiting γ fluence for certain
energy bands. Thus one needs the expected fluence for a given energy
*∞
74
The exponential integral function is defined as En (x) = 1
dt e−x t /tn . Note
that En (∞) = 0 while for n > 1 En (0) = (n − 1)−1 .
range (Eγ,1 , Eγ,2 ) for which one finds by integration of Eq. (12.15)
mν dLMC + , --ε1
−ε
Fγ,1,2 = Fν (1 − α) e + 2α [ε E2 (ε) + E3 (ε)] -- ,
τγ 2Tν ε2
(12.16)
where the expression in braces is meant to be taken as a difference

between the two limits for ε. For ε1 = 0 and ε2 = ∞ it is equal
to 1, independently of α, as it must because the angular distribution of
photon emission leaves the total number of decay photons unchanged.
12.4.2 SMM Observations
The gamma ray spectrometer (GRS) on the SMM satellite consists of

seven NaI detectors surrounded on the sides by a CsI annulus and at
the back by a CsI detector plate (Forrest et al. 1980). The three energy
bands shown in Tab. 12.1 have been analyzed for γ-ray emission from
SN 1987A (Chupp, Vestrand, and Reppin 1989; Oberauer et al. 1993).
At the detection time of the first neutrino event at IMB (7:35:41.37 UT)
the GRS was observing the Sun. A time interval of 223.232 s until it
went into calibration mode was used to search for a photon signal above
background in each energy band. The background was determined by
analyzing the rates measured during an interval of 151.6 s before the
first neutrino event. In Fig. 12.7 the recorded number of events per
2.048 s is shown in each band as a function of time. No excess counts
were found in any of them and the distributions of the rates are in good
agreement with a Gaussian shape.
Because the GRS was observing the Sun, γ-rays associated with
the neutrino burst would have hit the instrument almost exactly from
the side and so they had to traverse about 2.5 g cm2 of spacecraft alu-
minum before being recognized in one of the detectors. This effect
has been included to calculate the effective detector areas which allow
one to convert “counts” into a γ-ray fluence. In Tab. 12.1 the corre-
sponding 3σ limits are shown for the time until 223.232 s after the first
neutrino arrival (Oberauer et al. 1993). In order to constrain low-mass
neutrinos, only a time interval of 10 s around the burst is of interest;
the corresponding limits are also given (Chupp, Vestrand, and Rep-
pin 1989). The fluence limits (cm−2 ) can be expressed as limits on an
average differential fluence (cm−2 MeV−1 ) by dividing with the width
∆Eγ of a given channel. For the 10 s column they are shown as dotted
histograms in Fig. 12.6 while for 223.2 s they are shown in Fig. 12.13.
466 Chapter 12
Fig. 12.7. Event rates measured in the Gamma Ray Spectrometer (GRS)
of the Solar Maximum Mission (SMM) satellite encompassing the observed
neutrino burst of SN 1987A (the dashed line is for the first neutrinos observed
in the IMB detector). The time interval for each bin is 2.048 s. The rates to
the left of the dashed line are used to determine the background while the
ones to the right would include photons from neutrino decay. (Figure from
Oberauer et al. 1993 with permission.)
Table 12.1. GRS 3σ upper fluence limits.
Channel Energy Band γ Fluence Limit [cm−2 ]

[MeV] (10 s)a (223.2 s)b
1 4.1−6.4 0.9 6.11
2 10−25 0.4 1.48
3 25−100 0.6 1.84
a Chupp, Vestrand, and Reppin (1989)
b Oberauer et al. (1993)
For the 10 s and 223.2 s time intervals one can compute an average
flux limit (cm−2 s−1 ) for each channel. Then one expects that for the
longer time interval it is more restrictive by the ratio of (∆t)1/2 , i.e. by
(10 s/223.2 s)1/2 = 0.21. This expectation is approximately borne out
by the data in Tab. 12.1, confirming their consistency.
12.4.3 Radiative Decay Limit: Low-Mass Neutrinos

We are now armed to derive a radiative lifetime limit for low-mass
neutrinos (mν < ∼ 40 eV) by comparing the expected fluence according to
Eqs. (12.12) and (12.16) for channels 1, 2, and 3 with the observational
upper limits given in Tab. 12.1 for a 10 s time interval surrounding
the observed SN 1987A ν e burst. Depending on the assumed neutrino
spectral distribution which was parametrized by Tν and the anisotropy
parameter α, different channels give the most restrictive limits. These
are shown in Fig. 12.8 as a function of Tν for α = 0, ±1.
For normal neutrinos the relevant temperature range is between
4 and 8 MeV. In Fig. 12.8 a much larger range is shown because one
may also consider the emission of sterile neutrinos or axions from the
deep interior of a SN core where temperatures of several 10 MeV and
Fermi energies of several 100 MeV are available (Sect. 12.4.6).
For Dirac neutrinos the most conservative case is α = −1 where
for 4 MeV < <
∼ Tν ∼ 8 MeV the bound is approximately constant at
15
τγ /mν > 0.8×10 s/eV. Therefore, it applies equally to νe and low-
mass νµ and ντ . For Majorana neutrinos (α = 0) the limit is more
restrictive by a factor of 2−3, depending on the assumed Tν . With
Eq. (7.12) the most conservative overall limit75 (α = −1) translates into
µeff < 1.5×10−8 µB m−2

eV . (12.17)
It applies if the total laboratory lifetime exceeds the time of flight of

5.7×1012 s from the LMC to us. With a typical Eν = 20 MeV one must
require τtot /mν > 5
∼ 3×10 s/eV in the neutrino rest frame.
If the total lifetime is shorter than this limit all neutrinos decay
before they reach the Earth. Therefore, one can only derive a limit on
the branching ratio Bγ of the radiative channel. One easily finds that
Eq. (12.13) is to be replaced by
# ∞
Fγ! (Eγ ) = Fν B γ dEν (1 − α + 2α Eγ /Eν ) Φν (Eν )/Eν . (12.18)
Eγ
Therefore, using the same Boltzmann source spectrum the photon spec-
trum is slightly harder. Going through the same steps as before one
finds the upper limits on Bγ as a function of the assumed Tν and α
75
Somewhat stronger constraints found in the literature were based on a less de-
tailed analysis, notably with regard to the spectral dependence and the dependence
on α. Published results are τγ /mν > 0.83×1015 s/eV (von Feilitzsch and Oberauer
1988), 1.7×1015 (Kolb and Turner 1989), 6.3×1015 (Chupp, Vestrand, and Reppin
1989), and 2.8×1015 (Bludman 1992).
468 Chapter 12
Fig. 12.8. Lower limit on τγ /mν for mν < > 5

∼ 40 eV and τtot /mν ∼ 5×10 s/eV
(most neutrinos pass the Earth before decaying).
Fig. 12.9. Upper limit on the radiative branching ratio Bγ for mν <
∼ 40 eV
< 5
and τtot /mν ∼ 5×10 s/eV (most neutrinos decay between SN 1987A and
Earth).
shown in Fig. 12.9. The most conservative case yields approximately

Bγ < 3×10−10 . As a bound on an effective transition moment this is
µeff < 0.7×10−5 µB m−2

eV (meV /τs )
1/2
, (12.19)
where τs = τtot /s. Of course, if τtot became so short that the neutrinos
would decay while still within the envelope of the progenitor even this
weak limit would not apply. This is the case for τlab < ∼ Renv ≈ 100 s
(envelope radius Renv of the progenitor star) and so with Eν ≈ 20 MeV
one needs to require τtot /mν > −5
∼ 10 s/eV.
12.4.4 Decay Photons from High-Mass Neutrinos

For “high-mass” neutrinos with mν > ∼ 40 eV a calculation of the photon
fluence is more involved because of the dispersion of the neutrino burst
and the corresponding delay of the decay photons. The parent neutrino
travels with a velocity β = (1 − m2ν /Eν2 )1/2 ≈ 1 − m2ν /2Eν2 so that it
arrives at Earth with a time delay of about (m2ν /2Eν2 ) dLMC relative
to massless ones; here, dLMC = 50 kpc = 5.1×1012 s is our distance
to the LMC where SN 1987A had occurred. With Tν ≈ 6 MeV for
νµ or ντ an average neutrino energy is 3Tν ≈ 20 MeV, spreading out
the arrival times of massive neutrinos over an approximate interval of
10−2 s (mν /eV)2 . As the radiative decay may occur anywhere between
the LMC and here, the arrival times of the decay photons will be spread
out by a similar amount even though the photons themselves travel
with the speed of light. Thus for mν < ∼ 40 eV all decay photons fall
within about a 10 s time window around the arrival time of the first
ν e ; the results of Sect. 12.4.3 apply to this case. For mν < ∼ 200 eV
they fall within the 223.2 s interval for which GRS fluence limits exist.
Therefore, one may easily scale the previous limits to this case by using
the 223.2 s fluence limits in Tab. 12.1 instead of the 10 s ones. Of course,
for a given neutrino mass there would be an optimum time window for
which fluence limits could be derived on the basis of the original data.
For larger masses only a certain portion of the photon pulse falls
into the 223.2 s window. In order to calculate this fraction, I follow
Oberauer et al. (1993) and begin with the simple case where all neu-
trinos are emitted at the same time with a fixed energy Eν . The ra-
diative decay occurs at a time tD after emission and thus at a distance
dD = βtD from the source (neutrino velocity β), and the photon is emit-
ted at an angle θlab relative to the neutrino momentum (Fig. 12.10).
It has to travel a distance dγ until it arrives here; elementary geom-
etry yields dγ = [d2LMC − d2D (1 − cos2 θlab )]1/2 − dD cos θlab . Therefore,
relative to the first (massless) neutrinos the photons are delayed by
t = tD + dγ − dLMC . This delay has two sources: The parent moves
with a speed less than that of light, and the photon is emitted at an
angle so that a detour is taken from the LMC to us. For ultrarelativis-
tic parents both effects disappear as the relativistic transformations
squeeze all laboratory emission angles into the forward direction.
470 Chapter 12
Fig. 12.10. Geometry of the radiative neutrino decay.
The photons which arrive first are the ones from decays near the
source. Then the limit dD ( dLMC leads to dγ = dLMC − dD cos θlab and
t = tD (1 − β cos θlab ). With cos θlab = (β + cos θ)/(1 + β cos θ) where θ
is the angle of photon emission in the parent frame one finds
tD
t= , (12.20)
γ 2 (1 + βx)
where γ = Eν /mν is the neutrino Lorentz factor and x ≡ cos θ. There-
fore, photons detected between t and t + dt result from decays at
tD = γ 2 (1 + βx) t during an interval dtD = γ 2 (1 + βx) dt. (Recall
that t is measured after the first massless neutrinos arrived while tD
is measured after emission at the source.) The number of parent neu-
trinos diminishes in time as e−tD /γτtot with τtot the total decay time.
Therefore, the number of photons traversing a spherical shell of radius
dLMC per unit time is
γ(1 + βx) −γ(1+βx) t/τtot
Ṅγ (t) = e , (12.21)
τγ
where as before τγ is the radiative decay time.
Photons produced before the parent has left the envelope of the
progenitor star (radius Renv ) cannot be detected at Earth. If this
absorption effect is to be included, Eq. (12.21) will involve a step
function76 Θ(dD − Renv ) with dD = βtD = βγ 2 (1 + βx) t. Put an-
other way, photons emitted at an angle θ in the rest frame will first
arrive at a time t0 = Renv [βγ 2 (1 + βx)]−1 . The progenitor of SN 1987A
has been unambiguously identified as the blue supergiant Sanduleak
−69 202 (Schramm and Truran 1990). From its surface temperature
(15,000 K), its luminosity (5×1038 erg/s) and the distance to the LMC
one can infer its radius to be Renv ≈ 3×1012 cm = 100 s. Using this
76
The step function is defined by Θ(z) = 0 for z < 0 and Θ(z) = 1 for z > 0.
value, t0 is shown in Fig. 12.11 as a function of the neutrino velocity

β for several values of x = cos θ. The absorption effect is negligible
except for nonrelativistic neutrinos or for backward emission which, in
the laboratory frame, corresponds to very soft photon energies.
Fig. 12.11. Arrival time t0 of first decay photons from a parent neutrino
with velocity β, taking the envelope radius of the source to be Renv = 100 s.
The curves are marked with the respective values of x = cos θ, the direction
of photon emission in the neutrino rest frame.
In the parent frame the photons (energy ω) follow a normalized

distribution f (ω, x) which yields d3 Nγ (t, ω, x) = Ṅγ (t) f (ω, x) dt dω dx
with Ṅγ (t) from Eq. (12.21). The usual relativistic transformations
lead to a laboratory photon energy of Eγ = γ(1 + βx) ω. Transforming
from dωdx to dωdEγ and integrating over the unobserved rest-frame
energy ω yields
. /
d2 Nγ 1 Eγ # ω− f (ω, x) −Eγ t/ωτtot βγEγ t
= dω e Θ −ω ,
dEγ dt τγ βγ ω+ ω2 Renv
(12.22)
where ω± = Eγ [γ(1 ± β)]−1 and x = (Eγ /γω − 1)/β. The photon flux
at Earth is obtained by multiplication with the neutrino fluence Fν of
Eq. (12.12) and integration over a suitable spectrum Φν (Eν ) of neutrino
energies.
If the neutrinos are sufficiently long-lived (the exact meaning of
this is quantified below) the exponential can be ignored. If one also
ignores the absorption effect by the progenitor star (Renv = 0), the
472 Chapter 12
time structure of Eq. (12.22) reduces to Θ(t), i.e. its spectral form is
time independent except that it begins at t = 0. This is somewhat
surprising because the energy of a photon in the laboratory frame is
related to the angle of emission in the neutrino rest frame which in
turn determines the “detour” taken from the source to us (Fig. 12.10).
However, the first photons come from decays immediately at the source
and so any angle of emission leads to the same initial arrival time.
It must be stressed that the expression Eq. (12.22) depends on the
assumption of decays not too far from the source (dD ( dLMC ) and so
only the head of the photon pulse is correctly described while its tail
would require including decays even close to the Earth. Strictly speak-
ing, the photon burst never ends because even if the parent neutrinos
have passed the Earth, some photons will be received from backward
emission. However, because one is interested in neutrino masses so
large (mν >∼ 200 eV) that the photon burst is much longer than the
GRS measurement window, it is enough to account for the head of the
photon pulse.
12.4.5 Radiative Decay Limits: High-Mass Neutrinos

As a first explicit case for the distribution of photon energies and emis-
sion angles I take the two-body decay ν → ν ! γ with a massless daugh-
ter neutrino and with the dipole angular distribution of Eq. (12.2) with
x = cos θ = − cos ϑ for a left-handed parent. This amounts to
f (ω, x) = 12 (1 + αx) δ(ω − 12 mν ) (12.23)
in Eq. (12.22). Because of the δ function it is trivial to integrate,
0 1
d2 Nγ 1 2Eγ 2Eγ − Eν
= 1+α e−t/τ∗ Θ(t − tenv ), (12.24)
dEγ dt mν τγ pν pν
where
mν m2ν
τ∗ ≡ τtot , tenv ≡ Renv . (12.25)
2Eγ 2Eγ pν
Moreover, the flux vanishes if the chosen value for Eγ does not fall
between 12 (Eν ± pν ), or equivalently, unless
Eν > Eγ + m2ν /4Eγ , (12.26)
a condition on the minimum required neutrino energy.
For the simplest case when the neutrinos are relativistic (pν = Eν ),
long-lived (e−t/τ∗ = 1), and absorption effects by the progenitor can be
ignored (tenv = 0) the spectrum is shown in Fig. 12.12 for the anisotropy
parameters α = 0, ±1. Note the difference to the triangular shape of
Fig. 12.1 for a stationary source. High-energy photons are now en-
hanced because lower-energy ones correspond to larger emission angles
in the parent frame and so they take a larger “detour” from the source
to us (Fig. 12.10). Hence, their flux is spread out over a larger time
interval even though photons of all energies begin to arrive at the same
time if tenv = 0.
Fig. 12.12. Photon spectrum from the decay of a short burst of relativistic
neutrinos, energy Eν , according to Eq. (12.24) taking pν = Eν and e−t/τ∗ = 1
(relativistic and long-lived parent), and ignoring absorption effects by the
progenitor (tenv = 0).
In order to compare with the GRS fluence limits one needs to in-
tegrate the expected flux between t = 0 and t = tGRS = 223.2 s.
The Θ function in Eq. (12.24) is accounted for by using tenv as a lower
limit of integration. Integrating also over the neutrino source spectrum
Fν Φν (Eν ) yields
0 1
tGRS # ∞ 2Eγ 2Eγ − Eν
Fγ! = Fν dEν Φν 1+α I, (12.27)
mν τγ Emin pν pν
where
e−tenv /τ∗ − e−tGRS /τ∗
I≡ . (12.28)
tGRS /τ∗
For sufficiently long-lived parents (τ∗ , tGRS ) the exponentials can
be expanded and I = (tGRS − tenv )/tGRS . The lower limit of integra-
tion is set by the condition Eq. (12.26) and by the requirement that
tenv < tGRS , i.e. that I > 0. This condition may be expressed as
pν > (m2ν /2Eγ ) (Renv /tGRS ).
474 Chapter 12
The two-body decay is mostly interesting for mν <∼ 2me , a limit in

which one may safely ignore all nonrelativistic corrections, including
the progenitor absorption effect. In this case
1 − e−tGRS /τ∗
I= , (12.29)
tGRS /τ∗
which is unity for τ∗ >

∼ tGRS . With typical photon energies of 3Tν ≈
20 MeV and tGRS = 223.2 s this requirement translates into mν τtot >
∼
1010 eV s.
In the relativistic limit and with I = 1 one can easily integrate
Eq. (12.27) with the Boltzmann spectrum Eq. (12.14) and finds
tGRS ( )
Fγ! = Fν (1 − α) ε + (1 + α) ε2 e−ε , (12.30)
mν τγ
where ε = Eγ /Tν . This spectrum is shown in Fig. 12.13 for Tν = 4
and 8 MeV with mν τγ = 1018 eV s. The envelopes of the shaded bands
in Fig. 12.13 correspond to α = ±1 where for each temperature the
“harder” edge corresponds to α = +1. The expected fluence for each
GRS channel of Tab. 12.1 is found by integration. The GRS fluence
limits then yield the lower bounds on mν τγ shown in Fig. 12.14.
Fig. 12.13. Expected fluence of photons from the decay ν → ν ! γ of “high-

mass” but relativistic SN neutrinos, 200 eV < <
∼ mν ∼ 1 MeV, according to
18
Eq. (12.30) with mν τγ = 10 eV s and the indicated neutrino temperatures.
The shaded bands for each temperature are for the range −1 ≤ α ≤ +1 with
the harder edge corresponding to α = +1. Also shown are the upper limits
from the GRS channels taken from the 223.2 s column of Tab. 12.1.
Fig. 12.14. Lower limit on mν τγ for 200 eV < <

∼ mν ∼ 1 MeV, assuming
mν τtot > 10
∼ 10 eV s.
For Dirac neutrinos the most conservative case is α = −1 and the

temperature range relevant for νµ and ντ is between 6 and 8 MeV. This
yields an approximately temperature independent bound of mν τγ >
7×1018 eV s. For Majorana neutrinos (α = 0) the limit is about77
mν τγ > 12×1018 eV s. The most conservative case (α = −1) trans-
lates with Eq. (7.12) into
µeff < 1.6×10−10 µB m−1

eV , (12.31)
assuming mν τtot > 10

∼ 10 eV s. Note the different dependence on meV
relative to the low-mass result Eq. (12.17).
If mν τtot violates this condition because it is below 1010 eV s means
that the neutrinos decay so fast that the photon burst effectively ends
before the GRS integration time tGRS is over. Then the photon burst is
again “short” even though the neutrino mass is large. In this case one
can state a limit on Bγ as in Sect. 12.4.3. If τtot is only slightly shorter
so that mν τtot < 8
∼ 4×10 eV s, all photons arrive within tGRS = 10 s
and one may directly apply Eq. (12.19), originally derived for small
neutrino masses. For the narrow region where the photon burst ends
between 10 and 223.2 s this bound must be discounted by about a
factor of 2 because of the less restrictive fluence limits for the larger
integration time.
77
The result here is after Oberauer et al. (1993) which is similar to 6×1018 eV s
of Bludman (1992) but substantially more restrictive than 0.84×1018 eV s of Kolb
and Turner (1989). These works all refer to the isotropic case (α = 0).
476 Chapter 12
Fig. 12.15. Nonrelativistic correction of the expected photon fluence for

the GRS channels of Tab. 12.1 according to Eq. (12.32) with Renv = 100 s,
tGRS = 223.2 s, and Tν = 6 MeV.
So far the nonrelativistic corrections for a neutrino mass in the

10 MeV mass range have been ignored. As an example I consider the
photon fluence of Eq. (12.27) in the limit of a large τ∗ where the ex-
ponentials in Eq. (12.28) can be expanded. The Boltzmann spectrum
for nonrelativistic neutrinos should include an extra factor β = pν /Eν .
Then for α = 0 the fluence is78
. /
tGRS # ∞ Eν e−Eν /Tν m2ν Renv
Fγ! (Eγ ) = Fν dEν Eγ − , (12.32)
mν τγ Emin Tν3 2pν tGRS
with
  
 . /2 1/2 . /
 mν Renv m2ν 
Emin = max mν 1 +  , Eγ + . (12.33)
 2Eγ tGRS 4Eγ 

Relative to the massless case, the integral expression is suppressed if

mν >∼ Tν . A straightforward numerical integration then yields the sup-
pression of the expected fluence for each GRS channel as shown in
78
Strictly speaking, the fluence of massive neutrinos must be calculated by de-
termining their neutrino sphere which is different from the massless case. This
problem was recently tackled by Sigl and Turner (1995) by solving the Boltzmann
collision equation by means of an approximation method known from calculations
of particle freeze-out in the early universe. However, because only masses of up
to 24 MeV are presently considered, a precise treatment of the neutrino spectrum
changes the resulting limits only by a small amount.
Fig. 12.15. (The overall factor m−1 ν of Eq. 12.32 is not included, of
course.) Up to neutrino masses of about 10 MeV one may essentially
ignore the nonrelativistic corrections while for larger masses one has to
worry about them. However, because this discussion applies to stan-
dard neutrinos, the largest relevant mass is about 24 MeV and so the
nonrelativistic corrections never overwhelm the result.
12.4.6 Summary of ν → ν ! γ Limits

In order to summarize the decay limits I begin in Fig. 12.16 with the rel-
evant regimes of mν and τtot . Above the upper dotted line the neutrinos
live long enough so that most of them pass the Earth before decaying
while below the lower dotted line they decay within the envelope of
the progenitor star. In the areas 1 and 5 the photon burst is “short”
(∆tγ < < <
∼ 10 s), in 2 and 4 it is “intermediate” (10 s ∼ ∆tγ ∼ 223.2 s),
and in 3 it is “long” (223.2 s <∼ ∆tγ ). The exact boundaries as well as
the relevant constraints are summarized in Tab. 12.2. In Fig. 12.17 the
limits on µeff are summarized as a contour plot.
If one restricts possible neutrino decays to the radiative channel one
has τtot = τγ which depends only on µeff and mν . Then one may use
directly the upper limits on µeff given in Tab. 12.2 for the areas 1−3,
depending on the assumed mass. Put another way, the conditions on
τtot are then automatically satisfied.
These limits certainly apply to νe as the ν e burst from SN 1987A has
been measured. The fluxes of the other flavors were only theoretically
implied. If they have only standard weak interactions they must have
been emitted approximately with the same efficiency as νe . Large dipole
moments, however, imply large nonstandard interactions: The same
electromagnetic interaction vertex that allows for radiative decays also
allows for scattering on charged particles by photon exchange! For
MeV energies, for example, the scattering cross section on electrons by
regular weak interactions and that by photon exchange are the same
for µeff of order 10−10 µB . Hence in the lower left corner of Fig. 12.17
the neutrinos would interact much more strongly by photon exchange
than by ordinary weak interactions, causing them to emerge from higher
layers of the SN core than normally assumed. Their fluence and effective
temperature is then much smaller than standard. Put another way, for
µeff > −10
∼ 10 µB the above constraints are not self-consistent (Hatsuda,
Lim, and Yoshimura 1988). However, because large dipole moments can
be constrained by other methods (Sect. 7.5.1) a detailed investigation
of their impact on SN physics is not warranted.
478 Chapter 12
Fig. 12.16. Different regimes of neutrino masses and total lifetimes referred
to in the text. ∆tγ is the duration of the burst of decay photons. The
radiative lifetime limits in the areas 1−6 are summarized in Tab. 12.2.
Fig. 12.17. Upper limits on effective neutrino transition moments from

SN 1987A as given in Tab. 12.2. The contours are marked with log(µeff /µB ).
In the shaded lower right region the neutrinos decay within the progenitor.
Toward the lower left side the bounds are not self-consistent because large
dipole moments induce large nonstandard scattering cross sections which
enhance neutrino trapping.
The discussion of the previous sections focussed on standard neu-

trinos which are emitted approximately with the same efficiency and
similar energies as νe ’s and ν e ’s. It is possible, however, that these neu-
Table 12.2. Neutrino radiative lifetime limitsa from SN 1987A.
Areab Boundariesc Radiative lifetime Transition momentd

limitc µeff /µB
1 mν < 40 τγ m−1
ν > 0.8×10
15
1.5×10−8 m−2
ν
3×105 < τtot m−1
ν
2 40 < mν < 200 τγ m−1

ν > 0.2×10
15
0.8×10−8 m−2
ν
3×105 < τtot m−1
ν
3 200 < mν < 107 τγ mν > 7×1018 1.6×10−10 m−1

ν
9×109 < τtot mν
−1/2
4 mν < 107 Bγ < 1.2×10−9 1.4×10−5 m−3/2
ν τtot
4×108 < τtot mν < 9×109
10−5 < τtot m−1
ν < 3×10
5
−1/2
5 mν < 107 Bγ < 3×10−10 0.7×10−5 m−3/2
ν τtot
τtot mν < 4×108
10−5 < τtot m−1
ν < 3×10
5
6 τtot m−1
ν < 10
−5
Bγ < 0.01 —
a Forthe anisotropy parameter α = −1.
b Numbered as in Fig. 12.16.
c Neutrino masses in eV, lifetimes in s.
d Upper limit.
trinos have Dirac masses and thus right-handed partners which could
be emitted from the inner core of the SN by helicity-flipping processes
(Sect. 13.8). Moreover, entirely new particles could be produced and
escape from there.
The present bounds can be scaled to such cases if one calculates the
total energy Ex,tot = fx 1053 erg emitted in the new x particles, where
1×1053 erg is the total energy that was used for a standard ν plus ν.
Self-consistency requires fx < 1, of course. In addition, one needs the
average energy )Ex * of the new objects which allows one to define an
approximate equivalent temperature Tx = 13 )Ex *. Depending on the x
mass and total lifetime one can then read the radiative lifetime limits
directly from Figs. 12.8, 12.9, and 12.14, except that they must be
relaxed by a factor fx for the reduced fluence.
480 Chapter 12
12.4.7 Limit on ντ → νe e+ e−
Neutrinos with a mass exceeding 2me can decay into νe e+ e− , a channel

which probably dominates over ν ! γ. Among the standard neutrinos, the
role of the parent can be played only by ντ (or rather ν3 ) with its upper
experimental mass limit of about 24 MeV. Within the standard model
where the decay is due to flavor mixing the rate is given by Eq. (7.9).
In order to derive bounds on the e+ e− channel from the GRS obser-
vations, photons need to be produced. At first one may think that the
positrons would quickly annihilate so that a strong prompt γ flux can
be expected (Takahara and Sato 1987; Cowsik, Schramm, and Höflich
1989). Following Mohapatra, Nussinov, and Zhang (1994), however, the
gas density outside of the progenitor is too low, in spite of a substan-
tial stellar wind during the progenitor’s supergiant evolution. Also, the
annihilation of the charged leptons from the decay among each other is
moderately efficient only if the decays occur close to the source. Typ-
ical galactic magnetic fields have a strength of about 3 µG; they may
well be larger in the Large Magellanic Cloud, and the circumstellar field
of the SN 1987A progenitor may have been larger still. The gyromag-
netic radii for 5 MeV positrons is then less than 1010 cm ( Renv so that
one may think that the charged leptons were locally trapped (Cowsik,
Schramm, and Höflich 1989). However, the momentum carried by the
flux of the charged decay products is so large that those fields would
have been swept away (Mohapatra, Nussinov, and Zhang 1994).
Altogether it appears that the decay positrons will linger in inter-
stellar space for a long time before meeting annihilation partners unless
most decays occur immediately outside of the progenitor. Therefore,
prompt photons are mostly produced by bremsstrahlung ντ → νe e+ e− γ
which is suppressed relative to the decay rate only by a factor of about
α/π ≈ 10−3 (Dar and Dado 1987).
Neutrinos with masses in the MeV range which are emitted at MeV
temperatures are nearly nonrelativistic. Their rest frame is then ap-
proximately equal to the laboratory frame, but they still move essen-
tially with the speed of light. To escape from the progenitor before
decaying their rest-frame lifetime τtot must exceed a few 100 s. This
also guarantees that the pulse of decay photons will outlast the GRS
integration time: one is automatically in the region of a “long” photon
burst which was “case 3” in Fig. 12.16. In fact, if one assumes that ντ
decays are induced by mixing, the decay rate Eq. (7.9) together with
the laboratory bounds on Ueh shown in Fig. 12.3 easily guarantees that
they fulfill this requirement.
In order to estimate the expected photon flux from Eq. (12.22) one
needs to know the distribution of photon energies and emission angles
in the frame of the parent neutrino. In the absence of a detailed calcu-
lation I follow Oberauer et al. (1993) and assume approximate isotropy
for the photon emission. The soft part of the spectrum dNγ /dω from
a bremsstrahlung process is given by the rate of the primary process
times (α/π) ω −1 (Jackson 1975). Extending this behavior up to photon
energies of 12 mν I use
1 α/π 1
f (ω, x) = Θ( 12 mν − ω) , (12.34)
τγ τe+ e− 2ω
an expression which does not depend on x because of the assumed
isotropy. (α = 1/137 is the fine-structure constant, not the previous
anisotropy parameter.)
After dropping the exponential in Eq. (12.22) because the neutri-
nos are long-lived, one integrates over a Boltzmann source spectrum
for nonrelativistic neutrinos, integrates over the GRS energy channels,
and compares the expected fluence with the measured upper limits of
Tab. 12.1. The resulting bound on τe+ e− corresponds with Eq. (7.9)
directly to a limit on |Ue3 |2 . In Fig. 12.18 I show these bounds (trans-
formed into bounds on the mixing angle) as a function of the assumed
neutrino mass for Tν = 4 and 6 MeV. There remains a strong limit
even for masses far exceeding the temperature because the exponential
Fig. 12.18. SN 1987A limits on sin2 2θe3 = 4|Ue3 |2 with ν3 ≈ ντ for

Tν = 6 MeV (solid line) and 4 MeV (dashed line). Also shown are the corre-
sponding laboratory and solar limits from Fig. 12.3.
482 Chapter 12
suppression of the flux is compensated by the steep phase-space factor

m5ν in the expression for the decay rate.
These bounds are far more restrictive than those from laboratory
experiments (Fig. 12.3). The reason is that a SN explosion is a strong
ντ source; such a source is difficult to make in the laboratory. This is
the reason why the ντ has never been directly measured by its charged-
current conversion into τ .
The strong SN 1987A bounds on |Ue3 |2 imply that a heavy ντ with
only standard-model interactions must be rather long-lived, in fact too
long to be compatible with cosmological limits derived from big-bang
nucleonsynthesis (Sect. 7.1.5). Those bounds together with the present
results imply that a heavy ντ cannot exist unless it has fast, invis-
ible decays induced by interactions beyond the standard model. If
this were the case it could well decay before leaving the SN progeni-
tor. Therefore, the laboratory experiments with their short distance
between source and decay volume remain important for anomalously
short-lived neutrinos.
12.4.8 Heavy, Sterile Neutrinos
The bounds on Ue3 from reactors, beam stops, or the Sun were based on
a νe flux which partially converts into ν3 ’s which subsequently decay.
Because the SN emits about equal numbers of all ordinary neutrino
flavors, this approach is obsolete with regard to ν3 . However, one may
still consider hypothetical sterile neutrinos which interact only by virtue
of their mixing with νe . By assumption these states would not interact
through ordinary weak interactions and so they would not be trapped
in the SN core. Hence the expected νh flux would emerge from the deep
interior rather than the surface of the core.
In this case, however, the sterile neutrinos would carry away energy
much more efficiently than the ordinary ones and so the requirement
that enough energy was left for the observed ν e ’s from SN 1987A al-
ready gives one the approximate limit |Ueh |2 < ∼ 10
−10
(Sect. 9.6). If
this limit is approximately saturated one expects that about as much
energy is carried away by νh as by the ordinary flavors. Taking account
of the harder energies of neutrinos emitted from the SN core one still
obtains about the same limit on |Ueh |2 as on |Ue3 |2 before. Put an-
other way, the GRS observations do not dramatically improve on the
cooling argument of Sect. 9.6, although they range in the same general
magnitude.
12.4.9 Axions
Another hypothetical particle that could have been emitted abundantly
from SN 1987A is the axion. The impact of the axionic energy loss is
discussed in Sect. 13.5. If axions are more strongly interacting than a
certain limit, implying that their mass is larger than a few eV, they
are emitted from the surface of the SN core with a luminosity similar
to that of neutrinos. Kolb and Turner (1989) found that the axion
−12/11
fluence from SN 1987A would have been Fa ≈ 6×1010 cm−2 meV
−4/11
with meV ≡ ma /eV at a temperature of Ta ≈ 15 MeV meV . From
the GRS fluence limits, Kolb and Turner found that ma must be less
than a few 10 eV, a bound which is less restrictive than, for example,
the limit from globular cluster stars (Sect. 5.2.5).
12.4.10 Supernova Energetics

To derive the various GRS limits one had to assume that the radiative
decays occurred outside of the progenitor’s envelope and so neutrinos
falling into the shaded area in Fig. 12.16 were not accessible to these
arguments. However, in this case the stellar envelope itself serves as a
“detector” as discussed by Falk and Schramm (1978) many years ago;
see also Takahara and Sato (1986). Supernova observations in general,
and those of SN 1987A in particular, indicate that of the approximately
3×1053 erg of released gravitational binding energy only a small fraction
on the order of one percent becomes directly visible in the form of the
optical explosion as well as the kinetic energy of the ejecta. In contrast,
even if only one of the neutrino species decayed radiatively within the
progenitor, about 30% of the binding energy would light up!
If the lifetime were so short that the parent neutrinos would never
get far from the SN core one would not have to worry. Therefore, the
critical range of decay times is between the core dimensions of about
30 km = 10−4 s and the envelope radius of about 100 s. If the neutrinos
had nonradiative decay modes, and if their total laboratory lifetime fell
into this range, one could only conclude that Bγ < ∼ 10 .
−2
If they decayed only into radiation, and taking Eν to be 10 MeV,

the quantity τγ /mν cannot lie between about 10−11 and 10−5 s/eV (for
heavy neutrinos τγ includes the e+ e− channel). For an effective tran-
sition moment µeff , an interval between about 102 and 105 µB m−2 eV is
excluded, moderately interesting only for large masses. Then, however,
the cosmological limits strongly suggest the presence of nonradiative
decay channels.
484 Chapter 12
12.5 Galactic Supernovae and ντ → νe e+ e−

12.5.1 Bounds on the Positron Flux
A core-collapse SN produces about 3×1057 ντ ’s which may subsequently

decay into νe e+ e− . What is the long-term fate of all these positrons? If
the ντ ’s decay mostly outside of the galaxy it is well possible that the
positrons will linger in intergalactic space “forever.” Those positrons
produced within the galaxy, however, will be trapped by the magnetic
fields (typical strength a few µG) which render the galactic disk a mag-
netic “bottle” for charged particles. The interstellar electron density is
on the order 1 cm−3 , leading to a positron lifetime against annihilation
of order 105 years. Moreover, elastic e+ e− scattering (Bhabha scatter-
ing) is very efficient at slowing down relativistic positrons because of
the perfect mass match which allows for an efficient energy exchange
in collisions. Therefore, most annihilations occur at rest, producing a
sharp γ-ray feature at 511 keV.
The galactic SN rate is a few per century while the decay positrons
annihilate on a much longer time scale. Therefore, the galactic disk
should contain a stationary positron population with a density deter-
mined by the galactic SN rate and the ντ lifetime. A comparison with
the measured photon flux at Eγ = 511 keV of about 5×10−3 cm−2 s−1
then leads to a very restrictive limit on the e+ e− decay channel (Dar,
Goodman, and Nussinov 1987).
In detail these authors used a galactic rate of two core-collapse SN
per century, an e+ lifetime against annihilation of 105 yr, and a typical
distance of the decays from Earth of 10 kpc. If most neutrinos decay
within the galactic disk, these assumptions lead to an expected photon
flux of 400 cm−2 s−1 . Thus, it is enough that one in 105 neutrinos decays
within the galactic disk to outshine the measured flux.
This estimate is corroborated by the more recent work of Skibo, Ra-
maty, and Leventhal (1992) who devised detailed models of the positron
distribution in the galaxy in order to account for the 511 keV diffuse
galactic line feature measured in the direction away from the galac-
tic center. They found a total stationary positron annihilation rate in
the galaxy of 0.6−3 × 1043 s−1 , where the precise coefficient depends on
model assumptions. With two core-collapse SN per century the average
galactic ντ production rate is 2×1048 s−1 . Again, it is enough if one ντ
in 105 injects an e+ into the galaxy to account for the observations.
If all positrons produced within about 1 kpc = 3×1021 cm from the
source (the scale height of the galactic disk) were magnetically trapped,
while those produced further away escaped into intergalactic space, a

lifetime below 105 × 1 kpc ≈ 1016 s in the laboratory frame is excluded.
Because SN neutrinos with MeV masses are nearly nonrelativistic the
rest-frame lifetime is identical with the laboratory lifetime to within a
factor of a few, excluding τe+ e− < 15
∼ 10 s.
If the decays occur too close to the source the annihilation with
electrons from the neutrino decay is of some importance (Mohapa-
tra, Nussinov, and Zhang 1994). Therefore, decay times below about
104 s cannot be excluded by the present argument (Dar, Goodman, and
Nussinov 1987).
Actually, the galactic positron flux is thought to be associated with
supernovae, albeit not from ντ decay but rather from the β + decays
of certain nuclei which are synthesized in a SN explosion (Chan and
Lingenfelter 1993). These authors performed a detailed analysis of the
probability for positrons to escape without annihilation from the SN
environment into the galaxy.
12.5.2 Can the Tau Neutrino Be Heavy?

Armed with this result we can return to the question raised in Sect. 7.2.2
if a ντ with a mass exceeding 2me is compatible with the cosmologi-
cal requirement that such particles and their decay products do not
“overclose” the universe. It turns out that the SN constraints on
ντ → νe e+ e− presented in this chapter exclude this possibility so that
either ντ respects the cosmological mass limit of a few 10 eV or else it
must have fast invisible decay channels which inevitably require inter-
actions beyond the standard model.
In Fig. 12.19 the available constraints on τe+ e− are summarized. The
SN 1987A bound from the absence of a prompt γ burst, the cosmological
requirement, and the above limit from galactic positron annihilation
together exclude the entire range of possible masses and lifetimes. The
margins of overlap are so enormous that each of the arguments has
several orders of magnitude to spare for unaccounted uncertainties.
A heavy standard ντ is also excluded on the basis of arguments in-
volving big-bang nucleosynthesis (BBN). The usual limit on the num-
ber of effective neutrino degrees of freedom at nucleosynthesis alone
is enough to reach this conclusion (Sect. 7.1.5). Moreover, charged
leptons and secondary photons from the e+ e− decay channel would de-
stroy some of the synthesized nuclei (Lindley 1979, 1985; Krauss 1984;
Kawasaki, Terasawa, and Sato 1986). The main virtue of the SN limits
is, therefore, that no reference to BBN is required to exclude a heavy ντ .
486 Chapter 12
Fig. 12.19. Excluded areas of the ντ mass and lifetime if the standard-model
decay ντ → νe e+ e− is the only available channel. The laboratory results
refer to the bounds on sin2 2θe3 of Fig. 12.3, translated into a limit on τe+ e−
by virtue of Eq. (7.9). The SN 1987A bound is that from Fig. 12.18 while
the cosmological one is from Fig. 7.2. The excluded range indicated by the
vertical arrow refers to the argument of Sect. 12.5.1.
12.6 Neutrinos from All Stars

All stars in the universe contribute to a diffuse cosmic background flux
of MeV neutrinos. If they decayed radiatively they would produce a
cosmic x- and γ-ray background which must not exceed the measured
levels. Because the entire radius of the visible universe is available as a
decay path, one can derive rather restrictive limits on τγ (Cowsik 1977).
In order to derive such limits I assume that neutrinos of energy Eν
are produced with a constant rate Ṅν (cm−3 s−1 ). Assuming a zero-
curvature model of the universe, Kolb and Turner (1989) found for the
resulting isotropic flux of decay photons
d 2 Fγ mν 1 9 Ṅν t2U
= , (12.35)
dEγ dΩ τγ 4π 21/2 5 Eν3/2 Eγ1/2
where tU is the age of the universe. Moreover, it was assumed that
in ν → ν ! γ the daughter neutrino is massless, and that the decays
are isotropic in the parent frame (anisotropy parameter α = 0). In
a flat universe one has tU = 23 H0−1 = h−1 2.05×1017 s where H0 =
h 100 km s−1 Mpc−1 is the present-day Hubble expansion parameter and

where observationally 0.4 < <
∼ h ∼ 1.
There exist numerous measurements of the diffuse cosmic x- and
γ-radiation (for example Schönfelder, Graml, and Penningfeld 1980).
Between a few 100 keV and a few 10 MeV the isotropic flux is reasonably
well approximated by
. /2
d 2 Fγ MeV
= 2×10−2 cm−2 s−1 sr−1 MeV−1 . (12.36)
dEγ dΩ Eγ
Comparing this with Eq. (12.35), the most restrictive limit on τγ is

obtained for the highest possible photon energy, Eγ = Eν . The re-
quirement that the decay flux does not exceed the measurements at
this energy leads to the upper limit
−3 −1
mν < −40 eV cm s
∼ 1.6×10 h2 , (12.37)
τγ s Ṅν
which does not depend on the assumed value for Eν because Eq. (12.35)
and (12.36) both scale with E −2 . This limit applies if the total neutrino
lifetime τtot exceeds tU ; otherwise only a limit on the branching ratio
Bγ can be found.
The most prolific stellar neutrino source in the universe are hy-
drogen-burning stars which produce two νe ’s with MeV energies for
every synthesized 4 He nucleus. Because most of the binding energy
that can be liberated by nuclear fusion is set free when single nucleons
are combined to form 4 He, most of the energy emitted by stars can
be attributed to hydrogen burning. Therefore, it is easy to translate
the optical luminosity density of the universe into an average rate of
neutrino production.
The average luminosity density of the universe in the blue (B) spec-
tral band is about h 2.4×108 L%,B Mpc−3 where L%,B is the solar B lu-
minosity. The Sun produces about 1×1038 νe /s and so one arrives at
Ṅνe ≈ h 2.4×1046 Mpc−3 s−1 = 0.8×10−27 cm−3 s−1 . (Of course, this es-
timate is relatively crude in that the neutrino luminosity scales directly
with the average bolometric luminosity of a stellar population, but not
precisely with LB .) With Eq. (12.37) this leads to a constraint for νe
of τγ /mνe > 12 < −7
∼ 5×10 s/eV, or µeff ∼ 2×10 µB meV .
−2
The (core-collapse) supernovae in the universe are also very promi-

nent neutrino sources, and, more importantly, they are thought to pro-
duce MeV neutrinos of all flavors. Such SNe do not occur in ellipti-
cal galaxies, and their present-day rate in spirals depends sensitively
488 Chapter 12
on the Hubble type, varying from 0.2 h2 SNu for Sa spirals to about
5 h2 SNu for Sd (van den Bergh and Tammann 1991) where the su-
pernova unit is defined by 1 SNu ≡ 1 SN per century per 1010 L%,B .
Adopting 1 h2 SNu as a representative value and about 5×1057 neu-
trinos plus antineutrinos of a given flavor per SN yields for each fla-
vor Ṅν ≈ h3 1.3×10−27 cm−3 s−1 , a rate almost identical to that from
hydrogen-burning stars.79 The radiative lifetime limit is then also iden-
tical, except that it applies to neutrinos of all flavors.
All of these bounds are weaker than those from SN 1987A. There-
fore, decaying stellar neutrinos cannot actually contribute to the ob-
served x- and γ-ray background.
12.7 Cosmological Bounds

12.7.1 Neutrinos
Within the big-bang scenario a cosmic background sea of neutrinos is
an inevitable consequence of the hot early universe. Its contribution to
the cosmic energy density was already used in Sect. 7.1.5 to derive ex-
tremely restrictive neutrino mass limits. If neutrinos decay radiatively,
further constraints can be obtained. For one, the decay photons can
show up directly as a diffuse, isotropic cosmic background radiation.
If the decays occur before recombination, i.e. before the universe be-
came transparent to radiation, but so late that the photons could not
be thermalized entirely, they contribute to a spectral distortion of the
cosmic microwave background radiation (CMBR). The resulting limits
were discussed, for example, by Kolb and Turner (1990) who found
that those areas of masses and lifetimes are excluded that are hatched
in Fig. 12.20. It was assumed that neutrinos decay only radiatively.
The contribution of the neutrinos and their decay products to the
mass density of the universe leads to the constraints shown in Fig. 7.2
which are based on the present-day value of Ωh2 and on the expansion
79
Multiplying this rate with the age of the universe of about 3×1017 s and the
speed of light of 3×1010 cm/s, and using h = 0.5 one finds an estimated present-day
flux at Earth of about 1 cm−2 s−1 . In a recent detailed study, Totani and Sato
(1995) find a flux which is larger than this crude estimate by as much as a factor
of 30. The Kamiokande II detector has set an upper limit on the cosmic background
flux of ν e of about 103 cm−2 s−1 for effective temperatures in the 3−4 MeV range
(Zhang et al. 1988). It is conceivable that this background will be measured by
the Superkamiokande detector. Note that at the Kamiokande site the ν e flux from
power reactors is roughly 1000 times larger than the background flux, except that
it falls off sharply beyond about 10 MeV.
Fig. 12.20. Cosmological limits on neutrino radiative lifetimes according to

Kolb and Turner (1990). The radiative mode is assumed to be the only decay
channel. The shaded area is excluded according to Sect. 7.1.5 (Fig. 7.2).
rate at nucleosynthesis. These limits (shaded area in Fig. 12.20) are

more general because they do not depend on the nature of the final
states in the decay.
Kolb and Turner’s (1990) exclusion plot is somewhat schematic.
Ressell and Turner (1990) performed a much more detailed analysis on
the basis of the diffuse photon backgrounds in all wavebands. Probably
the most interesting region is that of small neutrino masses and large
lifetimes (the upper left corner of Fig. 12.20). The excluded range of
effective electromagnetic transition moments for mν < 30 eV is shown
in Fig. 12.21. It may be useful to approximate the excluded range
analytically by
µeff < −11
∼ 3×10 µB (eV/mν )
2.3
(12.38)
which is shown as a dashed line in Fig. 12.21.
Using favored cosmological parameters (Ωh2 ≈ 0.3) neutrinos with
mν ≈ 30 eV would be the dark matter of the universe. With smaller
masses there would have to be another component, but neutrinos could
490 Chapter 12
Fig. 12.21. Limits on µeff according to Ressell and Turner’s (1990) bounds
on the radiative lifetime of long-lived neutrinos from the diffuse cosmic back-
ground radiations. The dashed line corresponds to Eq. (12.38).
Fig. 12.22. Limits on decaying neutrinos in clusters of galaxies. (a) A1413,

A2218, and A2256 (Bershady, Ressell, and Turner 1991). (b) Coma
and Virgo (Henry and Feldman 1981). (c) A665 (Davidsen et al. 1991).
(d) Extragalactic background light (Overduin, Wesson, and Bowyer 1993).
(e) Background light from decay of unclustered neutrinos (Ressell and Turner
1990).
still play a significant dynamical role. Recently, such mixed dark matter
scenarios have received much attention where mν = 5 eV is a favored
value. Such low-mass particles cannot cluster on galactic scales, but
likely they would reside in clusters of galaxies. With radiative decays
ν → ν ! γ and a total lifetime exceeding the age of the universe one then
expects clusters of galaxies to be strong sources of optical or ultraviolet
photons. Several limits are summarized in Fig. 12.22.
A case has been made that radiatively decaying neutrino dark mat-
ter is actually required to solve certain problems, notably the ioniza-
tion of galactic hydrogen clouds (e.g. Melott and Sciama 1981; Melott,
McKay, and Ralston 1988; Sciama 1990a,b; Sciama 1993a,b, 1995).
The predictions are very specific: An energy of decay photons of Eγ =
(14.4 ± 0.5) eV and thus a neutrino mass of mν = (28.9 ± 1.1) eV
with a radiative lifetime of τγ = (2 ± 1) × 1023 s which translates into
µeff = (6.3 ± 2) × 10−15 µB . Such a large transition moment would
require particle physics beyond the standard model.
Nominally, this possibility is already excluded by the absence of a
uv line from the cluster A665 (Davidsen et al. 1991). However, a bound
from a single source is always subject to the uncertainty of unrecognized
absorbing material in the line of sight or internal absorption. Moreover,
the dark matter in the core of this cluster may be mainly baryonic
(Sciama, Persic, and Salucci 1993; Melott et al. 1994). Bounds from the
diffuse extragalactic background light are more reliable in this regard.
While they marginally exclude Sciama’s neutrino (Overduin, Wesson,
and Bowyer 1993) it is perhaps too early to pronounce it entirely dead.
A decisive test will be performed with a future satellite experiment
where the uv line from neutrinos decaying in the solar neighborhood
definitely would have to show up if neutrinos were the bulk of the
galactic dark matter (e.g. Sciama 1993b).
12.7.2 Axions
The axion lifetime from a → 2γ is τ = 6.3×1024 s (ma /eV)5 /ξ 2 where ξ
is a model-dependent number of order unity (Sect. 14.3.2). Therefore,
axions with eV masses have radiative lifetimes in the neighborhood of
the above neutrino limits whence they can be constrained by similar
methods (Kephart and Weiler 1987). Moreover, such axions would con-
tribute substantially to the mass density of the universe because they
would have been in thermal equilibrium until relatively late.80 Their
80
In the early universe, axions are also produced by the relaxation of the coherent
initial field configuration at the onset of the QCD phase transition. This process
492 Chapter 12
contribution to the cosmic mass density would be Ωa h2 = 0.082 ma /eV

(Turner 1987; Ressell 1991) and so one can expect that clusters of galax-
ies contain substantial amounts of axions even if they are not the main
dark matter component.
Observations of the diffuse extragalactic background radiation limit
the axion mass to values below about 8 eV unless ξ is very small (Ressell
1991). Overduin and Wesson (1993) found ξ < 0.43, 0.07, and 0.02 for
ma /eV = 5.3, 8.6, and 13, respectively (Fig. 12.23). Moreover, some
axions would reside in the halo of our own galaxy so that their decays
would light up the night sky. Its brightness yields a conservative bound
of ξ < (6 eV/ma )5 (Ressell 1991).
Fig. 12.23. Constraints on axion decays in galaxies and galaxy clusters; ξ

parametrizes the coupling to photons with ξ = 1 corresponding to common
axion models. (a) Line emission from clusters A2256 and A2218 (Bershady,
Ressell, and Turner 1991; Ressell 1991). (b) Diffuse extragalactic background
radiation according to Ressell (1991) and (c) Overduin and Wesson (1993).
(d) Our galaxy (Ressell 1991).
The most interesting limits arise from a search for axion decay lines
from the intergalactic space in the clusters of galaxies A2256 and A2218.
Bershady, Ressell, and Turner (1991) and Ressell (1991) found ξ < 0.16
0.078, 0.039, 0.032, 0.016, and 0.011 for ma /eV = 3.5, 4.0, 4.5, 5.0,
6.0, and 7.5 respectively (Fig. 12.23). These limits are placed into the
context of other constraints in Fig. 5.9.
yields Ωa h2 ≈ (10−5 eV/ma )1.175 —see Eq. (14.5). While the overall coefficient of
this expression is very uncertain it is clear that Ωa = 1 saturates for ma somewhere
between 1 µeV and 1 meV. In this range axions never achieved thermal equilibrium.
Chapter 13
What Have We Learned from

SN 1987A?
The lessons for particle physics from the SN 1987A neutrino burst are
studied. First, neutrinos could have decayed or oscillated into other
states on their way out of the SN core and to us. Second, propagation
effects could have caused a time delay between photons and neutrinos
or between ν e ’s of different energy. Third, nonstandard cooling agents
could have shortened the neutrino burst below its observed duration.
These arguments are applied to a variety of specific cases.
13.1 Introduction
In Chapter 11 the neutrino observations from SN 1987A were dis-
cussed and it was shown that they agree well with standard theoret-
ical expectations from the core collapse and subsequent explosion of
an evolved massive star. The signal displays several anomalies (time
gap at Kamiokande, anisotropy in both detectors) which render it a
less beautiful specimen of the expected signal characteristics than is
sometimes stated in the literature. Still, in the absence of plausible
alternatives one must accept that the Kamiokande II, IMB, and Bak-
san event clusters observed at 7:35 UT on 27 February 1987 represent
the ν e component of the neutrino burst from the core collapse of the
SN 1987A progenitor star rather than some other particle flux, or some
other reaction than the expected dominant ν e p → ne+ process.
Accepting this, there is a host of consequences concerning a variety
of fundamental physics issues. The first and simplest set of arguments
is based on the fact that the ν e pulse and perhaps the prompt νe burst
493
494 Chapter 13
were observed, constraining various mechanisms that could have re-

moved neutrinos from the beam such as decays. Equally important, the
nonobservation of a γ-ray burst in coincidence with the neutrino burst
constrains radiative decays of neutrinos and other particles (Sect. 12.4).
More intricate arguments involve signal dispersion, either between
photons and neutrinos, between ν e ’s and νe ’s, or the intrinsic dispersion
of the ν e burst, constraining various effects that could cause signal
dispersion such as a nonzero neutrino mass or charge.
Most importantly, the inferred cooling time scale of a few seconds
of the newborn neutron star precludes an efficient operation of a non-
standard cooling agent and thus yields constraints on the emission of
new particles from the SN core, notably of right-handed (r.h.) neutri-
nos or axions. This line of reasoning is analogous to the “energy-loss
argument” which for normal stars has been advanced in Chapter 2.
13.2 Basic Characteristics of the Neutrino Burst

13.2.1 Fluence
The neutrinos from a SN are expected to consist of two major compo-
nents: the prompt νe burst, and quasi-thermal emission of about equal
total amounts of energy in (anti)neutrinos of all flavors. The water
Cherenkov detectors would register the νe burst by virtue of the reac-
tion νe + e− → e− + νe where the scattered electron is strongly forward
peaked, while the cooling signal is registered by ν e + p → n + e+ with
an essentially isotropic e+ signal. Even though both detectors observed
a forward peaked overall signal, it cannot be associated with νe -e colli-
sions (Sect. 11.3.5). Most or all of the events are interpreted as ν e ’s.
The observation of a ν e fluence (time-integrated flux) roughly in
agreement with what is expected from a stellar collapse precludes that
these particles have decayed on their way from the SN to us, yielding
a constraint on their lifetime of (Frieman, Haber, and Freese 1988)
τνe /mνe > 5

∼ 6×10 s/eV. (13.1)
However, this simple result must be interpreted with care because mas-
sive neutrinos are expected to mix. The heavy νe admixtures could
decay and may violate this bound.
The ν e ’s were not removed by excessive scattering on cosmic back-
ground neutrinos, majorons, dark-matter particles etc., leading to con-
straints on “secret interactions” (Kolb and Turner 1987). Take the
What Have We Learned from SN 1987A? 495
scattering on cosmic background neutrinos as an example. The present-

day density of primordial neutrinos is about 100 cm−3 in each neu-
trino and antineutrino flavor. With a distance to the Large Magellanic
Cloud of about 50 kpc = 1.5×1023 cm one has a column density be-
tween SN 1987A and Earth of about 1025 cm−2 so that the ν e -ν cross
section must be less than about 10−25 cm2 . If the cosmic background
neutrinos are massless they have a temperature of about 1.8 K and
so $Eν % ≈ 3Tν ≈ 5×10−4 eV. Because the measured SN neutrinos
have
√ a characteristic energy of 30 MeV the center of mass energy is
s ≈ 200 eV.
The cross-section bound is not particularly impressive compared
with a standard weak cross section of order G2F s ≈ 10−51 cm2 . How-
ever, νν cross sections have never been directly measured and so the
SN 1987A limit provides nontrivial information. As an example, neu-
trinos could scatter by majoron exchange, or they could scatter di-
rectly on a background of primordial majorons. Kolb and Turner then
found a certain constraint on the neutrino-majoron Yukawa coupling
(Sect. 15.7.2). As another example, the proposition that the solar neu-
trino flux could be substantially depleted by scatterings on cosmic back-
ground particles (Slad’ 1983) is excluded.
Other particles besides neutrinos may have been emitted from the
SN and could have caused detectable events. Engel, Seckel, and Hayes
(1990) have discussed the case of axions; they can be absorbed in water
by oxygen nuclei, a16 O → 16 O∗ , which subsequently produce γ rays by
decays of the sort 16 O∗ → 16 O γ, 16 O∗ → 15 O n γ, and 16 O∗ → 15 N p γ.
The γ rays would cause electromagnetic cascades and so they are de-
tectable about as efficiently as e± . The axion emission was estimated by
identifying their unit optical depth for a given interaction strength in a
simplified model of the SN temperature and density profile. More than
10 extra events would be expected at Kamiokande for an axion-nucleon
Yukawa coupling in the range
1×10−6 < <

∼ gaN ∼ 1×10
−3
(13.2)
which is thus excluded. In the middle of this interval, up to 300 ad-

ditional events would have been expected. However, axions with cou-
plings in this interval are also excluded by other methods (Sect. 14.4).
13.2.2 Energy Distribution

The energy distribution of the events at the IMB and Kamiokande
detectors broadly confirms the expected quasi-thermal emission with
496 Chapter 13
a temperature of around 4 MeV. This precludes that a major swap

by oscillations with the higher-energetic ν µ or ν τ flux has taken place.
The impact of neutrino oscillations on the observable signal has been
discussed in Sect. 11.4.
Also, νµ,τ or ν µ,τ decays with final-state ν e ’s would produce addi-
tional higher-energy events. While the SN 1987A data are probably too
sparse to extract significant information on the presence or absence of
this effect, a future galactic SN would certainly allow one to exclude a
certain range of masses and decay times or to detect this effect (Soares
and Wolfenstein 1989).
The trapping of neutrinos in a SN core together with the condition
of β equilibrium inevitably implies that there is a large νe chemical
potential, leading to typical νe energies of order 200 MeV. Moreover,
the inner temperature during deleptonization reaches values of up to
40−70 MeV so that typical thermal (anti)neutrino energies of up to
100−200 MeV are available. Therefore, if neutrinos could escape di-
rectly from the inner core they would cause high-energy events in the
detectors which have not been observed.
A mechanism to tap the inner-core heat bath directly is the pro-
duction of r.h. neutrinos by a variety of possible effects such as spin-
flip scattering by a Dirac mass term or a magnetic dipole moment
(Sect. 13.8). R.h. states could not be detected directly because they are
sterile with regard to standard l.h. weak interactions, a property which
allows them to avoid the SN trapping. However, they could produce
detectable l.h. states by decays (Dodelson, Scott, and Turner 1992) or
by magnetic oscillations (Nötzold 1988; Barbieri and Mohapatra 1988).
13.2.3 Prompt νe Burst

The prompt νe burst can be seen in a water Cherenkov detector by the
reaction νe e → eνe where the final-state electron essentially preserves
the direction of the incident neutrino. At Kamiokande, the directional-
ity of the first event81 is consistent with the interpretation that it was
caused by this reaction. However, the expected fluence corresponds
only to a fraction of an event and so the first event may also be due
to the ν e p → ne+ reaction and point coincidentally in the forward di-
rection. A random direction has about a 5% chance of being forward
within 25◦ which is approximately the uncertainty of the Kamiokande
directional event reconstruction.
81
In the first publication of the Kamiokande group (Hirata et al. 1987) the second
event was also reported forward; its most probable direction was later revised.
Still, the observation of the prompt νe burst from a future galactic

SN would allow for a number of interesting conclusions. For example,
one could exclude or find evidence for neutrino oscillations (Sect. 11.4).
Signal dispersion caused by a neutrino mass or other effects which are
discussed below for the cooling-phase signal would be even more sig-
nificant for the prompt burst because of its short duration. For exam-
ple, the cooling signal with a duration of about 10 s is sensitive to νe
masses in the 10 eV regime. As the prompt burst is at least 100 times
shorter one is sensitive to a factor of 10 smaller masses, i.e. to mνe in
the eV range.
The (anti)neutrino signal during the prompt burst phase allows one
to decide if the SN consisted of antimatter rather than matter. In
that case one would expect a prompt ν e burst with a scattering cross
section on electrons which is about a factor of 2.4 smaller (Eq. 10.17).
Moreover, ν e ’s are dominantly absorbed by the isotropic ν e p → ne+
reaction and so the prompt burst would cause a substantial isotropic
signal within the first 50 ms. In a matter SN the cooling ν e ’s have
larger energies than the νe ’s; the reverse for antimatter. Therefore, the
observable ν e signal from the cooling phase would be reduced. In a
detector like Kamiokande one would then expect 6−20% of the total
ν e signal from the prompt burst, in contrast with at most 1% for a
regular matter SN (Barnes, Weiler, and Pakvasa 1987). Of course, in
the foreseeable future one can hope to acquire the relevant data only
from a galactic SN which, no doubt, consists of matter.
13.2.4 Nonobservation of a γ-Ray Burst

No γ rays in conjunction with the SN 1987A neutrino burst were ob-
served by the solar maximum mission (SMM) satellite which was oper-
ational at the relevant time. Therefore, one can derive some of the most
restrictive limits on neutrino radiative decays as detailed in Sect. 12.4.
13.3 Dispersion Effects

13.3.1 Photons vs. Antineutrinos
The optical sighting of SN 1987A followed the detection of the ν e burst
by only a few hours (Fig. 11.7), a delay which is expected on the basis
of the simple reasoning that some time must pass before the mantle of
a SN “notices” the collapse of the inner core. Hence the two signals
must have propagated through space with an almost identical velocity
498 Chapter 13
so that the speed of light and that of neutrinos are equal to within
(Longo 1987; Stodolsky 1988)
! !
!c − c !
! ν γ! <
! ! 2×10−9 , (13.3)
! cγ ! ∼
assuming an uncertainty of ±3 h in the relative duration of the transit

times from the LMC to us. This was interpreted as the most stringent
test of special relativity to date in the sense that it proves with high
precision the universality of a relativistic limiting velocity.82
This result can also be interpreted as testing the weak equivalence
principle of general relativity (Krauss and Tremaine 1988). In the post-
Newtonian approximation one predicts that a gravitational potential
V (r) delays a light signal (Shapiro time delay) by an amount
" A
∆t = −2 V [r(t)] dt, (13.4)
E
where the integral is taken along the trajectory r(t) of the beam between
the points of emission (E) and absorption (A). This delay is the same
for neutrinos and photons to within
! !
! ∆t − ∆t !
! ν γ!
! ! < 0.7−4×10−3 , (13.5)
! ∆tγ !
where the uncertainty reflects the uncertain modelling of the gravita-

tional potential between Earth and SN 1987A.83 This result has been
used to constrain the parameters of a specific model of C- and P-
violating gravitational forces (Almeida, Matsas, and Natale 1989), and
to constrain the parameters of a class of nonmetric theories of gravity
(Coley and Tremaine 1988).
13.3.2 Neutrinos vs. Antineutrinos

Assuming that the first event at Kamiokande represents the prompt νe
burst one may also constrain the difference in transit time between νe
and ν e and thus confirm the equivalence principle between matter and
antimatter (LoSecco 1988; Pakvasa, Simmons, and Weiler 1989). Of
course, in order to make such results reliable one would need to observe
the prompt burst from a future SN with greater statistical significance.
82
For a recent laboratory experiment which addresses the Lorentz limiting veloc-
ity, see Greene et al. (1991), and references there to earlier works. See also the book
by Will (1993).
83
See Will (1993) for a review of many other empirical tests of general relativity.
13.3.3 Intrinsic Dispersion of the ν e -Pulse

a) Neutrino Mass
So far the transit time of different particle species was compared un-
der the assumption of a fixed velocity each. However, the most likely
effect of signal propagation over large distances is dispersion due to an
energy-dependent speed of propagation. The most widely discussed84
case is that of a nonzero neutrino mass (Zatsepin 1968). The main
problem at extracting information about the signal dispersion is the
unknown behavior of the source which must be modelled according
to some theoretical assumptions. A particularly detailed discussion is
that of Loredo and Lamb (1989) who found a mass limit of mνe < 23 eV
(Sect. 11.3.4). In a similar analysis which included the 13% dead-time
effect at IMB, Kernan and Krauss (1995) found mνe < 20 eV.
b) Neutrino Charge
The absence of an energy-dependent dispersion of the neutrino pulse
can be used to constrain other neutrino properties. A small electric
charge eν would bend the neutrino path in the galactic magnetic field,
leading to a time delay of
∆t e2 (BT dB )2
= ν , (13.6)
t 6Eν2
where BT is the transverse magnetic field and dB the path length within
the field. This leads to a constraint of
# $# $
eν < 1 µG 1 kpc
∼ 3×10−17 (13.7)
e BT dB
(Barbiellini and Cocconi 1987; Bahcall 1989). Note that a typical field
strength for the ordered magnetic field in the galactic spiral arms is
2−3 µG and that the path length of the neutrinos within the galactic
disk is only of order 1 kpc because the LMC lies high above the disk
(galactic latitude about 33◦ ).
84
Limits on mνe from the SN 1987A data were derived, among others, by Abbott,
de Rújula, and Walker (1988), Adams (1988), Arnett and Rosner (1987), Bah-
call and Glashow (1987), Burrows and Lattimer (1987), Burrows (1988a), Chiu,
Chan, and Kondo (1988), Cowsik (1988), Kolb, Stebbins, and Turner (1987a,b),
Midorikawa, Terazawa, and Akama (1987), Sato and Suzuki (1987a,b), Spergel and
Bahcall (1988), Loredo and Lamb (1989), and Kernan and Krauss (1995).
500 Chapter 13
c) Long-Range Forces
Speculating further one may imagine some sort of neutrino “fifth-force
charge.” If electrons, protons, or dark-matter particles also carry such
a charge the bending of the neutrino trajectory in the fifth-force field of
the galaxy would lead to an energy-dependent time delay. This and re-
lated arguments were advanced by a number of authors (Pakvasa, Sim-
mons, and Weiler 1989; Grifols, Massó, and Peris 1988, 1994; Fiorentini
and Mezzorani 1989; Malaney, Starkman, and Tremaine 1995).
The most plausible form for such a long-range interaction is one me-
diated by a massless vector boson, i.e. a new gauge interaction, perhaps
related to a novel leptonic charge (Sect. 3.6.4). In this case neutrinos
and antineutrinos would carry opposite charges so that the cosmic neu-
trino background would be essentially a neutral plasma with regard to
the new interaction. The resulting screening effects then invalidate the
SN 1987A argument (Dolgov and Raffelt 1995).
Screening effects would not operate if the force were due to a spin-0
or spin-2 boson which always cause attractive forces. However, any
force mediated by a massless spin-2 boson must couple to the energy-
momentum tensor and thus is identical with gravity. The force medi-
ated by a scalar boson between a static source and a relativistic neutrino
is suppressed by a Lorentz factor. Therefore, even if scalar-mediated
forces existed between macroscopic bodies, their effect would be weak-
ened for relativistic neutrinos.
In summary, the SN 1987A signal does not seem to carry any simple
information concerning putative nongravitational long-range forces.
d) Fundamental Length Scale

Fujiwara has proposed a quantum field theory where the velocity of par-
ticles increases with energy, leading to an energy-dependent advance of
the arrival times by ∆t/t = − 12 (%0 Eν )2 . Here, %0 is a fundamental length
scale. Whatever the merits of this theory, a value %0 < −18
∼ 10 cm would
not be in conflict with the SN 1987A neutrino signal (Fujiwara 1989).
e) Lorentz Addition of Velocities

If relativistic particles (photons, massless neutrinos) are emitted by a
moving source (velocity vS ) their velocity c$ in the laboratory frame
should be equal to c (velocity in the frame of the source). The Galilean
addition of velocities, on the other hand, would give c$ = c + vS . In
general one may assume that velocities add according to c$ = c + KvS
with K = 0 representing the Lorentzian, K = 1 the Galilean law of

adding velocities. For photons, the most stringent laboratory bound is
Kγ < −4
∼ 10 derived from the time of flight of decay photons π → 2γ
◦
◦
from a pulsed π source (Alväger et al. 1964). A much more stringent
constraint (Kγ < −9
∼ 2×10 ) obtains from an analysis of the photon signal
from a pulsed x-ray source (Brecher 1977).
The absence of dispersion of the ν e pulse can be used to derive con-
straints on Kν (Atzmon and Nussinov 1994). The observed SN 1987A
neutrinos were produced by microscopic processes involving nearly rel-
ativistic nucleons, and subsequently scattered several times on such
nucleons before leaving the star. If the last nucleon on which they
scatter is considered the source with vS ≈ 0.2 c their laboratory speed
c$ν will be represented by a distribution of approximate width 0.2 Kν
around c because of the random orientation and distribution in magni-
tude of vS . Thus one derives a bound Kν < ∼ 10
−11
from the absence of
a spread in arrival times exceeding about 10 s.
Atzmon and Nussinov (1994) warn, however, that this simple argu-
ment may be too naive as the motion through the progenitor’s enve-
lope may cause the particles of the envelope to be the true source of
the “neutrino waves” as there is a substantial amount of refraction be-
tween the neutrino sphere and the stellar surface. If one follows Atzmon
and Nussinov’s reasoning, there remains only a much weaker bound of
Kν < −5
∼ 10 from the absence of an anomalous time delay between the
neutrino signal and the optical sighting of the SN.
13.4 Duration of Neutrino Emission

13.4.1 General Argument
The most intricate way to use SN 1987A as a laboratory arises from the
observed duration of neutrino cooling. While the neutrino luminosity
during the first few 100 ms until the shock has been revived is largely
powered by accretion and by the contraction and settling of the bloated
outer core, the long tail is associated with cooling, i.e. emission from
the neutrino sphere which is powered by energy originally stored deep
in the inner core. If a direct cooling channel existed for that region,
such as the emission of r.h. neutrinos or axions, the late cooling phase
would be deprived of energy. Put another way, a novel cooling channel
from the inner core would leave the schematic neutrino light curves
of Fig. 11.3 more or less unchanged before about 1 s while the long
Kelvin-Helmholtz cooling phase would be curtailed.
502 Chapter 13
Of course, this reasoning is identical with the energy-loss argument

previously studied for normal stars in Chapters 1 and 2. The main
difference is that neutrinos are trapped so that particles which interact
more weakly can dominate the thermal evolution by volume emission.
In normal stars, photons are trapped and neutrinos can dominate the
energy loss by volume emission as, for example, in the early cooling of
a white dwarf.
This general argument is best illustrated with axion emission. These
particles are pseudoscalars which for the purpose of this argument are
taken to interact with neutrons and protons with a common Yukawa
coupling strength ga which is the only free parameter in the problem.
For very small values of ga axions will play no role, but with an increas-
ing coupling strength their emission from the inner core by bremsstrah-
lung processes, N N → N N a, will begin to compete with neutrino cool-
ing. Of course, if ga exceeds some critical value axions will be trapped
and emitted from an “axion sphere” at about unit optical depth. Be-
yond some large coupling they will be trapped so effectively that their
contribution to the cooling of the SN core is, again, negligible and the
neutrino signal assumes its standard duration. This general behavior is
shown in Fig. 13.1 on the basis of the numerical cooling calculations85
of Burrows, Turner, and Brinkmann (1989) and Burrows, Ressell, and
Turner (1990). These authors used the quantity ∆t90% as a measure of
the cooling time; it represents the time at which 90% of the expected
number of events have arrived at a detector. ∆t90% was calculated sep-
arately for Kamiokande II and IMB; in Fig. 13.1 an average relative
signal duration is shown, normalized to the value when axions are not
important. It is apparent that a large range of ga values can be excluded
on the basis of the observed duration of the neutrino signal.
One is here considering the time scale of neutrino emission at the
source while the detectors register a pulse which conceivably could have
been lengthened by dispersion effects. However, in view of the recent
laboratory limits of mνe <∼ 5 eV this is not a serious concern.
If one contemplates nonstandard neutrinos, a relatively short emis-
sion time scale at the source is compatible with the observations if νµ ’s
85
In the free-streaming regime these calculations were based on axion emission
rates which do not take the high-density multiple-scattering effects into account
that were discussed in Sect. 4.6.7. Therefore, the free-streaming part of Fig. 13.1
probably overestimates the import of axion emission. For the present purpose of
discussing the general aspects of a novel cooling channel, however, this problem is
of no concern. The axion case is the only one where numerical cooling calculations
are available for both the volume-emission (free-streaming) and the trapping limit.
Fig. 13.1. Relative duration of neutrino cooling of a SN core as a function

of the axion-nucleon Yukawa coupling ga . In the free-streaming limit axions
are emitted from the entire volume of the protoneutron star, in the trapping
limit from the “axion sphere” at about unit optical depth. The solid line is
according to the numerical cooling calculations (case B) of Burrows, Turner,
and Brinkmann (1989) and Burrows, Ressell, and Turner (1990); the dotted
line is an arbitrary completion of the curve to guide the eye. The signal
duration is measured by the quantity ∆t90% discussed in the text; an average
for the IMB and Kamiokande detectors was taken.
or ντ ’s decay. The final states could include ν e ’s which are detectable

at IMB and Kamiokande so that one can obtain late-time events by
a suitable combination of mass and lifetime. For a 17 keV Majorana
neutrino a lifetime around 104 s would allow one to explain the signal
duration even with a short emission time scale at the source (Simpson
1991; see also Cline 1992). In the following it will be assumed that this
is not the explanation of the observed signal duration.
Another loophole is that some or all of the late-time events at Ka-
miokande, which are separated from the main bunch by a 7 s gap, were
caused by effects other than core cooling. Recall that a similar problem
exists with the x-ray observations of old neutrons stars (Sect. 2.3) where
it is not always clear that one is observing blackbody surface emission
from thermal cooling rather than magnetospherically produced x-rays.
In the present case, one possibility is the fall-back of material onto
the core, i.e. late-time accretion which could cause significant neutrino
emission. However, on the basis of an analytic estimate Janka (1995b)
has argued that even with extreme assumptions this is not a likely
explanation of the late events.
504 Chapter 13
While it is clear that a large range of ga values can be excluded,

the quantity ∆t90% is a relatively crude measure of the length of the
cooling phase. In principle, one should perform a maximum likelihood
analysis for a given range of particle properties. Moreover, one would
need to consider a variety of models for the protoneutron star where the
equation of state (EOS), mass, accretion rate, neutrino opacities, and
perhaps other parameters should be varied to optimize the agreement
with the observed signal when a novel cooling mechanism operates.
Another caveat applies to the “trapping regime” of the new par-
ticles. One may expect that they play a significant role during the
infall phase and shock formation of a SN collapse, an issue that was
addressed only by a small number of authors in the context of majoron
bounds (Fuller, Mayle, and Wilson 1988) and bounds on neutrino dipole
moments (Nötzold 1988). Hence, in general it is not obvious that pa-
rameters allowed by the cooling argument on the trapping side would
remain allowed if one took account of these effects. Moreover, on the
trapping side the novel particles interact about as strongly as neutrinos
and so they could also cause a signal in the detectors. For axions, this
argument rules out the values of ga given in Eq. (13.2).
13.4.2 Analytic Criterion in the Free-Streaming Limit

In order to estimate the impact of a novel cooling channel on the neu-
trino signal it is obviously useful to evolve a protoneutron star numer-
ically with the new physics included, and to calculate the expected
neutrino signal for a varying strength of the new effect. Considering
the many uncertainties involved in this procedure one may well ask if
it is not just as reliable to perform a simple analytic estimate.
At about 1 s after core bounce the neutrino luminosity in all six
(anti)neutrino degrees of freedom together is about 3×1052 erg s−1 . The
mass of the object is around 1.5 M% = 3×1033 g so that its average
energy-loss rate is Lν /M ≈ 1×1019 erg g−1 s−1 . A novel cooling agent
would have to compete with this energy-loss rate in order to affect the
total cooling time scale significantly. Therefore, the observed signal
duration indicates that a novel energy-loss rate is bounded by
'x < 19 −1 −1
∼ 10 erg g s . (13.8)
It is to be evaluated at typical core conditions, i.e. at a temperature

of around 30 MeV and a density of around 3×1014 g cm−3 . The nuclear
medium is then at the borderline between degeneracy and nondegener-
acy while the electrons are highly degenerate.
Fig. 13.2. Profile of various pa-

rameters for the protoneutron
star model S2BH 0 of Keil,
Janka, and Raffelt (1995), 1 s
after core bounce. The degen-
eracy parameters were approx-
imated by ηN = (EF − mN )/T
with EF2 = p2F + m2N and with
the effective nucleon mass.
506 Chapter 13
The profile of various parameters as a function of the mass coor-

dinate is shown in Fig. 13.2 for model S2BH 0 of the cooling calcula-
tions of Keil, Janka, and Raffelt (1995); it illustrates typical physical
conditions encountered in the core of a protoneutron star during the
Kelvin-Helmholtz phase. For this model, the average value of (ρ/ρ0 )n
with the nuclear density ρ0 = 3×1014 g cm−3 and of (T /30 MeV)n is
shown in Fig. 13.3 as a function of n.
Fig. 13.3. Average values for (ρ/ρ0 )n with the nuclear density ρ0 =
3×1014 g cm−3 and of (T /30 MeV)n for the protoneutron star model of
Fig. 13.2.
As an example one may apply this criterion to the bremsstrahlung

energy-loss rate N N → N N a for the emission of some pseudoscalar
boson a (axion) with a Yukawa coupling ga . The nondegenerate energy-
loss rate Eq. (4.8) is 'a = ga2 2×1039 erg g−1 s−1 ρ15 T30
3.5
where T30 =
15 −3
T /30 MeV and ρ15 = ρ/10 g cm . From Fig. 13.3 one finds that
3.5
$ρ15 % ≈ 0.4 and $T30 % ≈ 1.4. The criterion Eq. (13.8) then yields
ga <
∼ 10 −10
, similar to what one would conclude from Fig. 13.1. Using
the degenerate emission rate Eq. (4.10) yields an almost identical result.
Therefore, a simple criterion like Eq. (13.1) is not a bad first estimate
for the import of a novel energy-loss rate.
13.4.3 Trapping Limit

When the new particles (for example, axions) interact strongly enough,
they will be emitted from a spherical shell where their optical depth is
about unity rather than by volume emission. Again, one is concerned
mostly with a time later than 0.5−1 s where the outer core has set-
tled and the shock has begun to escape. The density of the protoneu-
tron star falls within a thin shell from supranuclear levels to nearly
zero, causing the “photosphere” radius rx of the new particles to be
essentially the radius R ≈ 10 km of the settled compact star. With a
“photosphere” temperature Tx of the new objects their luminosity is
4πr2 σTx4 with the Stefan-Boltzmann constant σ which is gπ 2 /120 in
natural units with g the effective number of degrees of freedom (2 for
photons). Therefore, one must demand that
Tx <
∼ 8 MeV g
−1/4
, (13.9)
in order to stay below the total neutrino luminosity of 3×1052 erg s−1 .
It is nontrivial, however, to determine the temperature Tx which
corresponds to about unit optical depth. Following the approach of
Turner (1988) who carried this analysis through for axions one may
assume a simple model for the run of temperature and density above
the settled inner core. A simple power-law ansatz is ρ(r) = ρR (R/r)n
with the density ρR = 1014 g cm−3 at a radius R ≈ 10 km. A plausible
ansatz for the temperature profile is T (r) = TR [ρ(r)/ρR ]1/3 with TR
(temperature at radius R) of around 10 MeV. From the opacity κ as a
function of density %and temperature one may then calculate the optical
depth as τ (rx ) = r∞ x
κρ dr. From the condition τ (rx ) ≈ 23 one can
determine the “photosphere” radius rx and thus its temperature Tx .
The opacity of axions for a medium of nondegenerate nucleons was
given in Eq. (4.28). One may define τR ≡ κR ρR R so that κρR =
τR (ρ/ρR )2 (TR /T )1/2 where Eq. (4.28) yields τR = ga2 3.4×1016 . Then
one finds for Turner’s model an optical depth τa at the axion-sphere
temperature Ta
τa = τR ( 11
6
n − 1) (Ta /TR )11/2−3/n . (13.10)
Because n is a relatively large number such as 3−7 the criterion τa <
∼3
2
yields τR > 6 <

∼ n (TR /Ta ) . With the requirement Ta ∼ 8 MeV and with
>
TR ≈ 20 MeV as taken by Turner one finds ga ∼ 2×10−7 , not in bad
agreement with what one would conclude from the numerical results
shown in Fig. 13.1.
Still, this argument is rather sensitive to the detailed model as-
sumptions concerning the protoneutron star structure. Also, as axions
contribute to the transfer of energy within the star, a self-consistent
model must take this effect into account. Moreover, for novel fermions
such as r.h. neutrinos one must distinguish carefully between their neu-
trino sphere (from where they can escape almost freely) and the deeper
508 Chapter 13
region where their energy flux is set. Put another way, for fermions the
concept of blackbody emission from a neutrino sphere is not adequate,
making it impossible to apply the Stefan-Boltzmann in a simplistic way.
The transport of r.h. neutrinos in the trapping limit is an equally com-
plicated problem as that of l.h. ones! Therefore, a proper treatment of
the trapping limit is generally a tricky subject; axions are the only case
where it has been studied in some detail.
Occasionally one may wish to construct a particle-physics model
that avoids the SN limit. It would be incorrect to believe that this
is achieved when the interaction strength has been tuned such that
the mean free path is of order the neutron star radius. On the con-
trary, when this condition obtains the impact on the cooling rate is
maximized. This is analogous to the impact of novel particles on the
structure and evolution of the Sun as depicted in Fig. 1.2; the cooling
rate is maximized when the mfp corresponds to a typical geometric di-
mension of the object. In the trapping regime a new particle is harmless
only if it interacts about as strongly as the particles which provide the
standard mode of energy transfer.
13.5 Axions
13.5.1 Numerical Studies
The most-studied application of the SN cooling-time argument is that
of invisible axions as these particles are well motivated (Chapter 14).
Moreover, they have attracted much interest because they are one of
the few particle-physics motivated candidates for the cosmic dark mat-
ter. Early analytic studies in the free-streaming limit are Ellis and Olive
(1987), Raffelt and Seckel (1988), and Turner (1988) who also discussed
the trapping regime; his line of reasoning was presented in Sect. 13.4.3
above. Numerical studies in the free-streaming limit were performed
by Mayle et al. (1988, 1989) and by Burrows, Turner, and Brinkmann
(1989) while the trapping regime was numerically studied by Burrows,
Ressell, and Turner (1990). The numerical studies by different work-
ers in the free-streaming limit used different assumptions concerning
the axion couplings, emission rates, and other aspects. In my pre-
vious review (Raffelt 1990d) I have attempted to reduce the results of
these works to a common and consistent set of assumptions; apart from
relatively minor differences which could be blamed on different input
physics (e.g. softer equation of state and thus higher temperatures in
the Mayle et al. papers) the results seemed reasonably consistent. A
recent numerical study by Keil (1994) who used the same axion emis-
sion rates as Burrows, Turner, and Brinkmann (1988) confirmed their
results.
Here, I present the numerical studies of Burrows and his collabora-
tors where axions were assumed to couple with equal strength to pro-
tons and neutrons. The axial-vector coupling to nucleons is written in
the form (C/2fa ) ψγµ γ5 ψ ∂ µ a with a model-dependent numerical factor
C, the Peccei-Quinn energy scale fa , the nucleon Dirac field ψ, and the
axion field a. Under certain assumptions detailed in Sect. 14.2.3 it can
be written in the pseudoscalar form −i ga ψγ5 ψ where ga = CmN /fa
is a dimensionless Yukawa coupling (nucleon mass mN ); Burrows et al.
used C = 12 . All results will be discussed in terms of ga and as such they
apply to any pseudoscalar particle which couples to nucleons accord-
ingly. In Sect. 14.4 the available constraints on axions will be expressed
in terms of the axion mass ma .
In the free-streaming limit the energy loss by axions was imple-
mented according to the numerical rates of Brinkmann and Turner
(1988); limiting cases of these rates were discussed in Sect. 4.2. In the
trapping regime, the transfer of energy by axions as well as axion cool-
ing from an “axion sphere” was implemented by means of an effective
radiative opacity as discussed in Sect. 4.4. The protoneutron star mod-
els are those of Burrows and Lattimer (1986) and of Burrows (1988b).
In the latter study, cooling sequences were presented for different equa-
tions of state (EOS), and different assumptions concerning the mass
and early accretion rate of the stars. A fiducial case in these studies is
model 55 with a “stiff EOS,” an initial baryon mass of 1.3 M% , and an
initial accretion of 0.2 M% .
The compatibility of a given model with the SN 1987A observations
should be tested by a maximum-likelihood analysis of the time and
energy distributions of the events in both the IMB and Kamiokande II
detectors. In practice, it is easier to consider a few simple observables.
Burrows and his collaborators chose the total number of events NKII
and NIMB in the two detectors as well as the signal duration defined by
the expected times tKII and tIMB it takes to accrue 90% of the expected
total number of events. As both detectors measured approximately
10 events each, the time of the last event probably is a reasonable
estimate of tKII and tIMB . Finally, Burrows et al. calculated the total
energy carried away by neutrinos and axions.
The run of these quantities with ga is shown in Fig. 13.4. Re-
call from Sect. 11.3.2 that the observed SN 1987A numbers of events
are NIMB = 8 and NKII = 10−12, depending on whether event No. 6
510 Chapter 13
Fig. 13.4. Results from protoneutron star cooling sequences with axions.
The free-streaming regime (small ga ) is according to Burrows, Turner, and
Brinkmann (1989), the trapping regime (large ga ) according to Burrows,
Ressell, and Turner (1990). For models A, B, and C (corresponding to
models 57, 55, and 62 of Burrows 1988b) the amount of early accretion and
the type of EOS (“stiff” or “soft”) is indicated. The models were calculated
until 20 s after collapse.
was actually due to background (quite possible) and whether event

No. 1 was due to the prompt νe burst (possible but not necessary).
The last events were registered at 5.6 s after the first (IMB) and 12.4 s
(Kamiokande II). Recall also that the absolute timing between the two
detectors is uncertain to within a minute although it seems plausible
that in both cases the first event essentially marks the arrival of the
first neutrinos. Finally, recall that the signal at Kamiokande II exhibits
a peculiar 7.3 s time gap before the last three events; event No. 9 was
registered at 1.9 s after the first. Naturally, it is worrisome that the
large Kamiokande time scale rests on the last three events, i.e. in order
to take the Kamiokande pulse duration seriously one needs to appeal
to a rare statistical fluctuation.
The total number of events observed is not very sensitive to the
amount of axion cooling which has an impact mostly on the late-time
neutrino signal. Interestingly, in the trapping regime (large ga ) the
number of events at IMB actually increases because the axionic energy
transfer heats the neutrino sphere to higher temperatures. NIMB re-
sponds sensitively to the neutrino spectrum because of the high thresh-
old at IMB. However, for this reason it is a bad indicator for the actual
neutrino flux because the high-energy tail of the spectrum is relatively
uncertain. For example, if it is described by a Fermi-Dirac function
with a degeneracy parameter η = 2−3 rather than η = 0 reduces NIMB
by about a factor of two (Fig. 11.11).
No numerical results are available in the intermediate regime be-
tween free streaming and trapping where the axion mean free path is
of order the neutron star radius. In this range of coupling constants
the impact of axions on the star is maximized. Moreover, a substantial
modification of the initial collapse phase obtains.
As emphasized before, the most sensitive observable is the duration
of the neutrino signal at the detectors. Therefore, nominally a range
of coupling constants 1×10−10 < < −7
∼ ga ∼ 3×10 is excluded. Within this
range the observed neutrino signal likely would be shortened too much
to be compatible with the observations.
13.5.2 Impact of Multiple-Scattering Effects

The results presented in the previous section were based on a naive
perturbative calculation of the axion emission rate without taking the
modification of the spin-density structure function into account that
must occur at high density as outlined in Sect. 4.6.7. The density de-
pendence of the axion emission rate is encapsuled in the spin-fluctuation
512 Chapter 13
rate Γσ which, in the nondegenerate limit, was given in Eq. (4.7) on the
basis of a perturbative one-pion exchange (OPE) calculation. For the
protoneutron star model displayed in Fig. 13.2 the profile of this Γσ /T
is shown in Fig. 13.5. In Fig. 4.8 the axion emission rate was shown as a
function of Γσ , revealing that for the conditions of interest one is in the
neighborhood of the maximum of the solid curve. In a realistic nuclear
medium, the true spin fluctuation rate may be smaller than the OPE
calculated value, taking one perhaps somewhat to the left of the maxi-
mum. Therefore, the true axion emission rate corresponds to the naive
one (dashed line in Fig. 4.8) at Γσ /T ≈ 3−5 which at temperatures
around 30 MeV corresponds to around 20% nuclear density.
Fig. 13.5. Profile for the nondegenerate spin-fluctuation rate Γσ of Eq. (4.7)
in the protoneutron star model S2BH 0 of Keil, Janka, and Raffelt (1995)
shown in Fig. 13.2.
Given the overall uncertainties involved in this discussion it is best

to derive a plausible limit on ga by the analytic criterion Eq. (13.8). The
relevant average temperature is about 30 MeV for which the maximum
emission rate corresponds to the naive one at about 5×1013 g cm−3 .
From Eq. (4.8) one finds an approximate axion energy-loss rate of
ga2 1×1038 erg g−1 s−1 . Then Eq. (13.8) indicates that one needs to re-
quire ga < −10
∼ 3×10 , about a factor of 3 less restrictive than the nomi-
nal bound from the numerical calculations above. Altogether one may
adopt
3×10−10 < <
∼ ga ∼ 3×10
−7
(13.11)
as a range excluded by the SN 1987A cooling-time argument.
13.6 How Many Neutrino Flavors?

One may ask how the neutrino signal from a SN would be modified
if there existed additional light sequential neutrino flavors beyond νe ,
νµ , and ντ . Of course, the Z ◦ decay width measured at CERN already
reveals that there are exactly three sequential neutrino flavors (Particle
Data Group 1994); the same conclusion is reached from studies of big
bang nucleosynthesis (e.g. Kolb and Turner 1990).
Burrows, Ressell, and Turner (1990) calculated several protoneu-
tron star cooling sequences, varying the number of flavors from 3, the
standard value, to 11. This increases the efficiency of energy transfer
within the SN core and also allows for a more efficient radiation from
the neutrino sphere as there are more degrees of freedom. Thus one
expects a shortened signal in the Kamiokande II and IMB detectors,
as well as a reduced number of events because the available energy is
shared between more neutrino degrees of freedom of which mostly the
Fig. 13.6. Number of events NKII and NIMB in the Kamiokande and IMB
detectors as well as the signal duration tKII and tIMB (in sec) as a function
of the assumed number of neutrino flavors (Burrows, Ressell, and Turner
1990). The signal duration is defined as the time it takes to accrue 90% of
the total expected number of events.
514 Chapter 13
ν e ’s are detected. These expectations are borne out by the numeri-

cal results shown in Fig. 13.6. A doubling of the number of flavors is
probably excluded by the observed signal duration.
13.7 Neutrino Opacity

The cooling time scale of a young SN core is determined by the neutrino
opacities which in turn are dominated by the neutral-current scattering
ν + N → N + ν of neutrinos on nucleons. Apart from final-state Pauli
blocking effects these opacities are given in terms of the scattering cross
section σ = (G2F /π) (CV2 + 3CA2 ) Eν2 where the neutral-current nucleon
weak-coupling constants CV,A were given in Appendix B. Therefore, the
neutrino opacities are dominated by the axial-vector, i.e. the nucleon
spin-dependent interaction. In Sect. 4.6.7 it was discussed that a naive
application of perturbation theory in a nuclear medium likely is not
appropriate because of the large spin fluctuation rate implied by this
method. It would indicate that the spin of a given nucleon fluctuates
so fast in a SN core that a neutrino would “see” on average a nearly
vanishing contribution. This would lead to a decrease of a typical axial-
vector scattering rate as estimated in Fig. 4.9.
To test if such a suppression effect is compatible with the SN 1987A
neutrino signal, Keil, Janka, and Raffelt (1995) calculated a series of
protoneutron star cooling sequences with modified neutrino opacities.
To this end they substituted CA2 → F CA2 in the numerical subroutine
which evaluates the opacities where
a
F = (1 − a) + (13.12)
1+b
with
# $2 & '2
1 Γσ ρ 10 MeV
b= ≈ . (13.13)
12 T 3×10 g cm−3
13 T
Here, a = 1 represents full suppression while smaller values of a allow
one to dial a lesser reduction of the opacities.
The predicted neutrino signal at IMB and Kamiokande II was,
again, characterized by the total number of expected events NIMB and
NKII as well as the signal durations tIMB and tKII which represent the
time at which 90% of the total number of expected events have been ac-
crued. In Fig. 13.7 these quantities are shown as a function of a where
the suppression effect was implemented for both, neutral- and charged-
current axial-vector interactions. The results marked with open circles
Fig. 13.7. Number of events in the Kamiokande and IMB detectors as well
as the signal durations as a function of the assumed “opacity suppression
parameter” defined by Eq. (13.12). Filled circles refer to a suppression of
both neutral- and charged-current axial-vector interactions while open circles
refer to a suppression of neutral-current interactions only. (Adapted from
Keil, Janka, and Raffelt 1995.)
refer to a suppression of the neutral-current reactions alone. The modi-

fication of the results between those cases is relatively minor, indicating
that the neutral-current interactions represent the dominant opacity
source for the overall cooling time scale.
The increase of the counting rates at the two detectors with decreas-
ing opacities is explained by the neutrino sphere moving to deeper and
hotter layers, yielding larger neutrino energies. The detectors register
mostly ν e ’s so that the number of events is relatively sensitive to the
charged-current opacity which affects only the electron flavor. NIMB is
particularly sensitive to the ν e spectrum because of its high threshold.
516 Chapter 13
By the same token, NIMB is quite sensitive to spectral pinching, an ef-

fect not included in these calculations where an equilibrium neutrino
transport scheme was used. Therefore, the total number of events,
notably at IMB, is a poor measure to characterize the neutrino signal.
In the calculations of Keil, Janka, and Raffelt (1995) the mass of
the initial neutron-star model as well as its temperature profile and the
equation of state were varied. While such modifications cause changes
in the predicted signal durations and event counts, none of these param-
eters appears likely to be able to compensate for an extreme suppres-
sion of the neutrino opacities. The SN 1987A neutrino signal excludes
a suppression effect stronger than, say, a >∼ 0.5.
These findings are in agreement with those of the previous section
where the number of neutrino flavors had been increased. Essentially
that procedure amounted to increasing the efficiency of neutral-current
energy transfer and so it is not very different from the decreased opac-
ities used here.
Therefore, it appears that the “standard” opacities which ignore
fast spin fluctuations provide a reasonable representation of what is
observed. This result appears to imply that the spin-fluctuation rate
Γσ does not exceed O(T ) in a SN core—see Sect. 4.6.7 for a discussion
of these matters. However, it is surprising that the best fit is achieved
by the naive opacities because the spin-fluctuation effect is only one
reason to expect reduced axial-vector opacities. Other reasons include
reduced effective values for CA in a nuclear medium, and spin-spin
correlations which tend to “pair” the spins and thus tend to reduce the
opacities. The question of the appropriate neutrino opacities in a SN
medium remains worrisome.
However, with regard to particle bounds the effect of reduced opaci-
ties goes in the direction of making those constraints more conservative
as a reduction of the opacities, like an anomalous energy loss, shortens
the neutrino signal.
13.8 Right-Handed Neutrinos

13.8.1 Dirac Mass
Right-handed neutrinos (helicity-minus neutrinos, helicity-plus anti-

neutrinos) do not interact by the standard weak interactions and so
they would not be trapped in the interior of a SN core. Therefore, the
SN 1987A neutrino signal allows one to constrain any mechanism that
could produce these “wrong-helicity” states.86 The main possibilities

are the existence of novel r.h. interactions which couple directly to r.h.
neutrinos, the existence of neutrino magnetic or electric dipole moments
which allow for left-right scatterings or magnetic oscillations, and the
existence of neutrino Dirac masses. Of course, these possibilities are
not necessarily distinct as the existence of r.h. currents or a Dirac mass
would usually also induce magnetic dipole moments.
Beginning with the assumption that neutrinos have a Dirac mass,
the mismatch between chirality and helicity for massive fermions im-
plies that in purely l.h. interactions a final-state neutrino or antineu-
trino sometimes has the “wrong” helicity and thus nearly r.h. chirality.
Then it is essentially noninteracting and thus may escape almost freely.
In principle, there are two production channels, the spin-flip scattering
of trapped l.h. states, and the production of pairs νL ν R or νR ν L by
the medium. In Sect. 4.10 it was shown that the neutrino phase space
favors the spin-flip process by a large margin. Moreover, in a nonrel-
ativistic medium the spin-flip scattering rate was found to be simply
the nonflip scattering rate times the factor (mν /2Eν )2 . The energy-
loss rate was then given by Qscat in Eq. (4.94) in terms of a dynamic
structure function of the medium.87
In a dilute medium consisting of only one species of nucleons one
has S(ω) = (CV2 + 3CA2 ) 2πδ(ω), leading to an energy-loss rate of
3(CV2 + 3CA2 ) G2F m2ν T 4

'R =
2π 3 mN
# $2 # $4
mν T
≈ 0.7×1019 erg g−1 s−1 , (13.14)
30 keV 30 MeV
where CV2 +3CA2 ≈ 1 was used (see Appendix B). Even though the axial-
vector structure function is not a δ function, Eq. (13.14) is probably a
reasonable estimate because the results of the previous section indicate
that the neutrino scattering rate in a dense medium is probably not
86
Such bounds naturally can be avoided if one assumes that the r.h. neutrinos
have other novel interactions which are strong enough to trap them efficiently in a
SN core. Explicit models were constructed, for example, by Babu, Mohapatra, and
Rothstein (1992) or Rajpoot (1993). These authors aimed at avoiding the SN 1987A
bound on Dirac neutrino masses.
87
Besides the neutrino spin-flip rate due to the neutrino weak interactions with
nucleons or other particles, there is also a spin-flip scattering term in the gravita-
tional field of the entire neutron star (Choudhury, Hari Dass, and Murthy 1989).
However, the resulting energy loss was found to be small except for low-energy
neutrinos.
518 Chapter 13
too different from that found in the dilute-medium limit. If one applies
the analytic criterion Eq. (13.8) one finds
mν <
∼ 30 keV (13.15)
as a limit on a possible Dirac neutrino mass. This agrees with the

bounds originally estimated by Raffelt and Seckel (1988) and Gaemers,
Gandhi, and Lattimer (1989) while Grifols and Massó (1990a) esti-
mated a slightly more restrictive limit (14 keV). This sort of bound
only applies if the mass is not so large that the “wrong-helicity” states
interact strongly enough to be trapped themselves. This would occur
for a mass beyond a few MeV. Therefore, a Dirac-mass ντ with, say,
mντ = O(10 MeV) is not excluded by this argument.
In a numerical study Gandhi and Burrows (1990) implemented the
spin-flip energy-loss rate and calculated the expected event counts and
signal durations at the IMB and Kamiokande II detectors. They found
a bound almost identical with Eq. (13.15). A similar numerical study
by Burrows, Gandhi, and Turner (1992) corroborated this result. An-
other numerical study was performed by Mayle et al. (1993) who found
a somewhat more restrictive limit of about 10 keV, essentially because
their equation of state allows the core to heat up to much higher tem-
peratures than are found in Burrows’ implementation with a stiff EOS.
In the Mayle et al. (1993) study, a more restrictive bound of around
3 keV was claimed if the pion-induced pair emission process π + N →
N +νR +ν L was included. This result is incorrect if one accepts the pre-
dominance of the spin-flip scattering over the pair-emission processes
that was discussed in Sect. 4.10. It does not seem believable that the
presence of pions would enhance the scattering cross section on nucle-
ons. In the form implemented by Mayle et al. (1993), the pair emission
rate and their neutrino opacities were not based on a common and
consistent axial-vector dynamical structure function.
A massive νµ or ντ likely would mix with νe . In this case the degen-
erate νe sea initially present in a SN core would partially convert into a
degenerate νµ or ντ sea (Maalampi and Peltoniemi 1991; Turner 1992;
Pantaleone 1992a). In Sect. 9.5 it was shown that the flavor conversion
would be very fast even for rather small mixing angles. In this case the
spin-flip scattering energy-loss rate involves initial-state neutrinos with
much larger average energies than those of a nondegenerate distribu-
tion that was used above. However, even though the initial energy-loss
rate in r.h. neutrino is much larger than before, the degeneracy effect
disappears after the core has been deleptonized and so the late-time
neutrino signal is not affected as significantly as one might have ex-

pected. The Dirac mass limit becomes only slightly more restrictive if
mixing is assumed (Burrows, Gandhi, and Turner 1992).
Neutrinos with masses in the keV range must decay sufficiently fast
in order to avoid “overclosing” the universe. If r.h. Dirac-mass neu-
trinos escape directly from the inner core of a SN they have energies
far in excess of l.h. neutrinos emitted from the neutrino sphere. If
their decay products involve sequential l.h. neutrinos or antineutrinos,
these daughter states would have caused high-energy events at IMB or
Kamiokande II, contrary to the observations. Dodelson, Frieman, and
Turner (1992) found that this argument excludes the lifetime range
10−9 s/keV < < 7

∼ τ /mν ∼ 5×10 s/keV, (13.16)
for Dirac masses in the range 1 keV < <

∼ mν ∼ 300 keV, assuming that
the “visible” channel dominates.
13.8.2 Right-Handed Currents

On some level r.h. weak gauge interactions may exist as, e.g. in left-
right symmetric models where the gauge bosons which couple to r.h.
currents would differ from the standard ones only in their mass. In the
low-energy limit relevant for processes in stars one may account for the
novel couplings by a “r.h. Fermi constant” which is given as 'GF with '
some small dimensionless number which may be different for charged-
and neutral-current processes. In left-right symmetric models one finds
explicitly for charged-current reactions (Barbieri and Mohapatra 1989)
'2CC = ζ 2 + (mWL /mWR )4 , (13.17)
where mWR,L are the r.h. and l.h. charged gauge boson masses while ζ
is the left-right mixing parameter.
In order to constrain 'CC one assumes the existence of r.h. νe ’s so
that the dominant energy-loss mechanism of a SN core is e + p →
n + νe,R where the final-state r.h. neutrino escapes freely. Initially, a
substantial fraction of the thermal energy of a SN core is stored in
the degenerate electron sea. Therefore, the time scale of cooling is
estimated by the inverse scattering rate for e + p → n + νe,R . The usual
charged-current weak scattering cross section involving nonrelativistic
nucleons is G2F (CV2 +3CA2 )Ee2 /π with CV2 +3CA2 ≈ 4 in a nuclear medium
(Appendix B). Using a proton density corresponding to nuclear matter
at 1015 g cm−3 and using 100 MeV for a typical electron energy one finds
520 Chapter 13
a charged-current scattering rate of 0.6×1010 s−1 or an approximate

cooling time-scale by r.h. neutrinos of '−2
CC 2×10
−10
s. The requirement
that this timescale exceeds a few seconds leads to the constraint
'CC < −5
∼ 10 , (13.18)
in agreement with a result of Barbieri and Mohapatra (1989) while Raf-
felt and Seckel (1988) found a somewhat less restrictive limit of 'CC < ∼
3×10−5 . Laboratory experiments yield a limit of order 'CC < ∼ 3×10 −2
(e.g. Jodidio et al. 1986) which is much weaker but does not depend
on the assumed existence of r.h. neutrinos. Mohapatra and Nussinov
(1989) extended the SN 1987A bound to the case of r.h. Majorana
neutrinos which mix with νe .
In order to constrain r.h. neutral currents, equivalent to constraining
the mass of putative r.h. Z ◦ gauge bosons, one considers the emission
of r.h. neutrino pairs νR ν R . The dominant emission process is by the
nucleons of the medium; in a dilute medium it can be represented as the
bremsstrahlung process N N → N N νR ν R . Apart from a global scaling
factor '2NC , the bremsstrahlung energy-loss rate for a nondegenerate
medium was given in Eq. (4.23). However, in a dense medium this
rate probably saturates at around 10% nuclear density as in the case
of axion emission (Sect. 4.6.7). Evaluating Eq. (4.23) at 10% nuclear
density (ρ15 = 0.03) and at T = 30 MeV, and applying the analytic
criterion Eq. (13.8) one finds
'NC < −3
∼ 3×10 . (13.19)
This is less restrictive from what was found by Raffelt and Seckel (1988)
or Barbieri and Mohapatra (1989).
The translation of a limit on 'NC into one on a r.h. gauge boson mass
depends on details of the couplings to quarks and leptons, and notably
on the mixing angle between the new and the standard Z bosons. De-
tailed analyses were presented by Grifols and Massó (1990b), Grifols,
Massó, and Rizzo (1990), and Rizzo (1991). Because these authors did
not consider multiple-scattering effects and the resulting saturation of
the bremsstrahlung process, their bounds on the Z $ mass are somewhat
too restrictive, perhaps by a factor of 2 or 3. Still, mZ ! has to exceed
at least 1 TeV, except for special choices of the mixing angle.
The SN 1987A limits on r.h. neutral currents are weaker than those
from big bang nucleosynthesis ('NC < −3
∼ 10 ) which are based on the
requirement that r.h. neutrinos must not have come to thermal equi-
librium after the QCD phase transition (at T < ∼ 200 MeV) in the early
universe (e.g. Olive, Schramm, and Steigman 1981; Ellis et al. 1986).
13.8.3 Magnetic Dipole Moments
Turning to neutrino magnetic dipole moments, the main production

process of r.h. states would be spin-flip scattering on charged parti-
cles, notably on protons. The scattering cross section was discussed in
Sect. 7.4. It involves the usual Coulomb divergence which in a medium
is cut off by screening effects. In a SN core, the electrons are initially
very degenerate while the protons are essentially nondegenerate and
so the main contribution to screening is from the protons. Accord-
ing to Eq. (D.17) they are electrically weakly coupled at the prevailing
temperatures and densities so that Debye screening should be an ap-
proximately adequate prescription; the Debye scale for the protons is
found to be around 30 MeV. As typical neutrino energies are somewhat
larger but of the same order, the Coulomb logarithm is approximately
unity. Thus the cross section is approximately αµ2ν with µν the mag-
netic dipole or transition moment.
This is to be compared with the spin-flip cross section from a Dirac
mass which is approximately G2F m2ν /4π. In Sect. 13.8.1 a Dirac mass
bound of about 30 keV was derived which translates into
µν < −12
∼ 4×10 µB (13.20)
with µB = e/2me the Bohr magneton. This bound is similar to that

derived by Barbieri and Mohapatra (1988), but less restrictive by about
an order of magnitude than that claimed by Lattimer and Cooperstein
(1988). See also Nussinov and Rephaeli (1987), Goldman et al. (1988),
and Goyal, Dutta, and Choudhury (1995).
Eq. (13.20) applies to all magnetic, electric, and transition moments
of Dirac neutrinos. Numerically, it is similar to the bound Eq. (6.97)
derived from the absence of excessive plasmon decay in globular cluster
stars. However, because this latter result applies also to Majorana
transition moments it is more general than the SN limit. The SN limit,
on the other hand, applies to masses up to a few MeV while the globular
cluster bound only for mν < ∼ 5 keV.
The simple cooling argument that led to Eq. (13.20) is not necessar-
ily the end of the story of neutrino magnetic dipole moments in SNe. If
r.h. neutrinos were indeed produced in the inner core and escaped freely,
they could rotate back into l.h. ones in the magnetic field around the
SN and in the galaxy. For a galactic magnetic field of order 10−6 Gauss,
extended over, say, 1 kpc (the field is confined to the disk, the LMC lies
far above the disk) this effect would be important for µν > −12
∼ 10 µB .
522 Chapter 13
As the r.h. neutrinos escape from the inner core with much larger ener-
gies than those from the neutrino sphere one would expect high-energy
events in the Kamiokande and IMB detectors, contrary to the observa-
tions. Therefore, one probably needs to require µν < −12
∼ 10 µB for the
diagonal dipole moments; spin-flavor oscillations could be suppressed
by the neutrino mass differences. As the spin precession is the same for
all neutrino energies, this limit would not apply if the Earth happened
to be in a node of the oscillation pattern between SN 1987A and us.
Neutrino magnetic moments of order 10−12 µB could also affect the
infall phase of SNe. The spin-flip scattering on nuclei would be coher-
ently enhanced relative to protons. Therefore, neutrinos could escape
in the r.h. channel for much longer so that effectively trapping would
set in much later than in the standard picture (Nötzold 1988).
In and near the SN core there probably exist strong magnetic fields
of order 1012 Gauss or more which would induce spin-precessions be-
tween r.h. and l.h. neutrinos. Therefore, the sterile states produced in
the deep interior by spin-flip scattering could back-convert into active
ones near the neutrino sphere. Depending on details of the matter-
induced neutrino energy shifts, the vacuum mass differences, and the
magnetic field strengths and configurations this conversion could take
place inside or outside of the neutrino sphere. The observable neutrino
signal could be affected, but also the energy transfer within the SN core
and outside of the neutrino sphere. Perhaps, a more efficient transfer of
energy to the stalled shock wave could help to explode SNe in the de-
layed explosion scenario. Various aspects of these scenarios have been
studied by Dar (1987), Nussinov and Rephaeli (1987), Goldman et al.
(1988), Voloshin (1988), Okun (1988), Blinnikov and Okun (1988), and
Athar, Peltoniemi, and Smirnov (1995).
Clearly, Dirac magnetic or transition moments in the 10−12 µB range
and below would affect SN dynamics and the observable neutrino signal
in interesting ways. However, because there are so many parameters
and possible field configurations, it is hard to develop a clear view of
the excluded or desired neutrino properties. If compelling evidence for
nonstandard neutrino electromagnetic properties in this range were to
emerge, SN dynamics likely would have to be rethought from scratch.
13.8.4 Millicharges
Within the particle physics standard model it is not entirely impos-
sible that neutrinos have small electric charges (Sect. 15.8). In this
case neutrinos would have to be Dirac fermions and so the r.h. states
can be produced in pairs by their coupling to the electromagnetic field.

Obvious production processes are the plasmon decay γpl → ν R νR and
pair annihilation e+ e− → ν R νR with an intermediate photon. An in-
termediate photon can also be coupled to hadronic components of the
medium. Mohapatra and Rothstein (1990) considered nucleon-nucleon
bremsstrahlung, and notably an amplitude where the electromagnetic
field is coupled to an intermediate charged pion. One may well wonder,
however, if this sort of naive perturbative bremsstrahlung calculation
is adequate in a nuclear medium.
A simple estimate of the plasmon decay process begins with the
energy-loss rate Eq. (6.94). Because the electrons in a SN core are
highly relativistic the plasma frequency is given by Eq. (6.43) as ωP2 =
(4α/3π) (µ2e + 13 π 2 T 2 ) where µe is the electron chemical potential. For
a SN core ωP ≈ 10 MeV is a reasonable estimate while the relevant
temperature is about T = 30 MeV. Taking approximately Q1 = 1 in
Eq. (6.94) and applying the approximate criterion Eq. (13.8) one finds
eν < −9
∼ 10 e. (13.21)
This is similar to Mohapatra and Rothstein’s (1990) result. Including

the e+ e− annihilation process would slightly improve this limit and
extend it to somewhat larger masses.
The bound Eq. (13.21) would apply to any millicharged particle
which is not trapped in the SN core. Mohapatra and Rothstein (1990)
estimated that for a charge in excess of about 10−7 e the particles would
be sufficiently trapped by scatterings off electrons to leave the SN cool-
ing time scale essentially unaffected. If they are Dirac neutrinos with a
mass in excess of a few MeV trapping by spin-flip scattering would be-
come important. Again, it is not obvious how strong the impact of such
trapped particles would be during the infall phase of SN collapse, i.e.
one should not infer that millicharged particles in the trapping regime
would not have a strong impact on SN physics just because their impact
on the Kelvin-Helmholtz cooling phase is small.
13.8.5 Charge Radius

If r.h. neutrinos existed and had an effective electromagnetic interaction
by virtue of a charge radius, they would be produced in a SN core by
the same processes as above where they were assumed to have a charge.
On the basis of the e+ e− annihilation process Grifols and Massó (1989)
found a limit of about 3×10−17 cm on a r.h. charge radius.
Chapter 14
Axions
The idea of axions is introduced and their phenomenological properties

are reviewed. The constraints on pseudoscalars that have been derived
throughout this book are systematically applied to axions.
14.1 The Strong CP-Problem

All fermions, with the possible exception of neutrinos, have magnetic
dipole moments—see Tab. 14.1 for several important examples. The
minimal electromagnetic coupling of charged spin- 12 fermions (charge q,
mass m) automatically yields a “Dirac moment” q/2m. Higher-order
amplitudes lead to an additional “anomalous” contribution, which is
the only one for neutral particles. For example, the magnetic moments
of massive neutrinos calculated in the standard model were given in
Eq. (7.16). The QED prediction of the electron anomalous magnetic
moment is perhaps the most stunning quantitative success of theoretical
physics.
On the other hand, no particle electric dipole moment has ever
been detected—see Tab. 14.1 for some upper limits. At first this is
quite satisfying because an electric dipole moment would allow one to
distinguish between matter and antimatter in an absolute sense.88 The
electroweak and strong gauge interactions are CP-conserving, so the
observed broad symmetry between particles and antiparticles appears
natural.
88
In the nonrelativistic limit the electric dipole operator d must be proportional
to the spin operator s which is an axial vector. Because the electric field is a polar
vector the energy d·E reverses sign under P and thus under CP whence its absolute
sign yields an absolute distinction between fermions and antifermions.
524
Axions 525
Table 14.1. Particle magnetic and electric dipole moments.
Fermion Magnetic Momentc Electric Momentd

[10−26 e cm]
Protona 2.792, 847, 3(86) µN 4000 ± 6000
Neutrona −1.913, 042, (75) µN < 11e
Electrona 1.001, 159, 652, 1(93) µB −0.3 ± 0.8
Neutrinob < 3×10−12 µB < 6000
∼ ∼
a Particle Data Group (1994).
b All flavors with mν <
∼ 5 keV (Sect. 6.5.6).
c Bohr magneton µ = e/2m ; nuclear magneton µ = e/2m .
B e N p
d 1 e cm = 5.18×1010 µ = 0.951×1014 µ (Appendix A).
B N
e 95% CL.
This symmetry is not respected, however, by the generally complex

Yukawa couplings to the Higgs field which are thought to induce the
fermion masses (Sects. 7.1 and 7.2). The resulting complex quark mass
matrix Mq can be made real and diagonal by suitable transformations
of the quark fields. This involves a global chiral phase transformation
(angle Θ = arg det Mq ), leading to a term in the QCD Lagrangian
αs !
LΘ = Θ GG . (14.1)
8π
Here, αs is the fine-structure constant of strong interactions and GG ! ≡

µν ! µν ! 1 ρσ
Gb Gbµν where Gb is the color field strength tensor, G bµν = 2 εµνρσ Gb
its dual, and the implied summation over b refers to the color degrees
of freedom. Of course, det Mq and thus Θ would vanish if one of the
quarks were exactly massless, but this does not seem to be the case.
Under the combined action of charge conjugation (C) and a parity
transformation (P) the Lagrangian Eq. (14.1) changes sign,89 violating
the CP invariance of QCD. It leads to a neutron electric dipole moment
|dn | ≈ |Θ| (0.04 − 2.0)×10−15 e cm (Baluni 1979; Crewther et al. 1979;
see also Cheng 1988). This is |dn | ≈ |Θ| (0.004 − 0.2) µN in units of
nuclear magnetons. Hence, for |Θ| of order unity one expects a neu-
tron electric dipole moment almost as large as its magnetic one. The
89
The structure of GG ! is Ecolor · Bcolor , i.e. the scalar product of a polar with an
axial vector and so it is CP-odd.
526 Chapter 14
experimental limit (Tab. 14.1), however, indicates |Θ| < −9

∼ 10 , surpris-
−3
ingly small in view of the phase δ = 3.3×10 which appears in the
Cabbibo-Kobayashi-Maskawa matrix Eq. (7.6) and which explains the
observed CP-violating effects in the K ◦ -K ◦ system.
Even worse, QCD alone produces a term like Eq. (14.1) because of
the nontrivial topological structure of its ground state (Callan, Dashen,
and Gross 1976; Jackiw and Rebbi 1976; t’Hooft 1976a,b). The coeffi-
cient ΘQCD is a parameter characterizing the “Θ-vacuum.” It is mapped
onto itself by a transformation ΘQCD → ΘQCD + 2π so that different
ground states are characterized by values in the range 0 ≤ ΘQCD < 2π.
The phase of the quark mass matrix and the QCD-vacuum together
yield Θ ≡ ΘQCD + arg det Mq as a compound coefficient for Eq. (14.1).
The experimental bounds then translate into
" "
" "
"ΘQCD + arg det Mq " < −9
∼ 10 . (14.2)
The CP-Problem of strong interactions consists of the smallness of Θ

which implies that the numbers ΘQCD and arg det Mq are either sepa-
rately very small, or cancel each other with very high accuracy. How-
ever, both are expected to be of order unity, or perhaps of order δ in
the case of arg det Mq , and completely unrelated to each other.
14.2 The Peccei-Quinn Mechanism

14.2.1 Generic Features
An attempt to explain the smallness of Θ may be overambitious as
long as we do not have an understanding of the origin of the Yukawa
couplings that went into Mq and of the other seemingly arbitrary pa-
rameters of the standard model. Still, whatever determines these “con-
stants of nature,” the strong CP-problem can be elegantly explained
by the existence of a new physical field, the axion field, which allows
Θ to vanish dynamically (Peccei and Quinn 1977a,b; Weinberg 1978;
Wilczek 1978). In this scheme, the CP-violating Lagrangian Eq. (14.1)
is literally switched off by its own force.
To this end the new field a(x) must be a pseudoscalar which couples
to gluons according to Eq. (14.1) with Θ replaced by −a/fa . The
constant90 fa with the dimension of an energy is the Peccei-Quinn scale
90
In the literature one often finds fa /N , Fa /N , vPQ /N etc. for what I call fa . It
was stressed, e.g. by Georgi, Kaplan, and Randall (1986) that a discussion of the
generic properties of all axion models does not require a specification of the model-
dependent integer N which can be conveniently absorbed in the definition of fa .
Axions 527
or axion decay constant. Axions must be fundamentally massless so that

all observable effects remain unchanged under a global shift a(x) →
a(x) + a0 where a0 is a constant. (A mass term 12 m2a a2 would spoil this
possibility.) This invariance allows one to absorb Θ in the definition of
the axion field. Including a kinetic term, Eq. (14.1) is replaced by
αs !.
LΘ → La = 12 (∂µ a)2 − a GG (14.3)
8π fa
It conserves CP because axions were assumed to be pseudoscalar (odd
under CP), similar to neutral pions.
Even though axions were constructed to be massless they acquire an
effective mass by their interaction with gluons. It induces transitions to
qq states and thus to neutral pions (Fig. 14.1) which means physically
that a and π ◦ mix with each other. Axions thereby pick up a small mass
which is approximately given by (Bardeen and Tye 1978; Kandaswamy,
Salomonson, and Schechter 1978)
ma fa ≈ mπ fπ , (14.4)
where mπ = 135 MeV is the pion mass and fπ ≈ 93 MeV its decay
constant. This mass term implies that at low energies the axion La-
grangian contains a potential V (a) which expands to lowest order as
1 2 2
m a . Because of the invariance of LΘ with respect to Θ → Θ + 2π
2 a
the gluon-induced potential V (a) is a function periodic with 2πfa .
Fig. 14.1. Axion mixing with qq states and thus π ◦ . The curly lines represent
gluons, the solid lines quarks.
The ground state of the axion field is at the minimum of its po-
tential at a = 0, explaining the absence of a neutron electric dipole
moment. If one could produce a static nonvanishing axion field a0 in
some region of space, neutrons there would exhibit an electric dipole
moment corresponding to Θ = −a0 /fa .
The Lagrangian Eq. (14.3) is the minimal ingredient for any ax-
ion model: the aGG ! coupling is their defining feature as opposed to
other pseudoscalar particles. Then axions inevitably acquire an effec-
tive mass at low energies. Thus the concept of a “massless axion” for
some arbitrary pseudoscalar is a contradiction in terms.
528 Chapter 14
Because of the mixing with π ◦ , axions share not only their mass, but
also their couplings to photons and nucleons with a strength reduced
by about fπ /fa . Therefore, they generically couple to photons so that
the general discussion of Chapter 5 applies directly except for those
aspects which required massless pseudoscalars.
The effective axion mass is a low-energy phenomenon below ΛQCD ≈
200 MeV. Above this energy pions and other hadrons dissociate in
favor of a quark-gluon plasma. Then a = 0 is no longer singled out so
that any value in the interval 0 ≤ a < 2πfa is physically equivalent.
Because the universe is believed to begin with a hot and dense “big
bang,” any initial value for a is equally plausible, or different initial
conditions in different regions of space. As the universe expands and
cools below ΛQCD , however, the axion field must relax to its newly
singled-out ground state at a = 0. This relaxation process produces a
population of cosmic background axions which is, in units of the cosmic
critical density (e.g. Kolb and Turner 1990),
Ωa h2 ≈ (fa /1012 GeV)1.175 . (14.5)
The exact value depends on details of the cosmic scenario and of the
relaxation process. Modulo this uncertainty, values exceeding fa ≈
1012 GeV are excluded as axions would overdominate the dynamics of
the universe. With Eq. (14.4) this corresponds to ma < −5
∼ 10 eV; axions
near this bound would be the cosmic dark matter. A search strategy
for galactic axions in this mass range was discussed in Sect. 5.3.
14.2.2 Axions as Nambu-Goldstone Bosons

The invariance of LΘ in Eq. (14.1) against transformations of the form
Θ → Θ + 2π, and the corresponding invariance of the axion Lagrangian
against transformations a → a + 2πfa , calls for a very simple interpre-
tation of the axion field as the phase of a new scalar field.
A transparent illustration is provided by the KSVZ axion model
(Kim 1979; Shifman, Vainshtein, and Zakharov 1980) where one in-
troduces a new complex scalar field Φ which does not participate in
the weak interactions, i.e. an SU(2)×U(1) singlet. There is also a new
massless fermion field Ψ and one considers a Lagrangian with the usual
kinetic terms, a potential V for the scalar field, and an interaction term,
# $
L= i
2
Ψ∂µ γ µ Ψ + h.c. + ∂µ Φ† ∂ µ Φ − V (|Φ|)
# $
− h ΨL ΨR Φ + h.c. . (14.6)
Axions 529
The Yukawa coupling h is chosen to be positive, and ΨL ≡ 12 (1 − γ5 )Ψ

and ΨR ≡ 12 (1 + γ5 )Ψ are the usual left- and right-handed projections.
This Lagrangian is invariant under a chiral phase transformation of
the form
Φ → eiα Φ , ΨL → eiα/2 ΨL , ΨR → e−iα/2 ΨR , (14.7)
where the left- and right-handed fields pick up opposite phases. This
chiral symmetry is usually referred to as the Peccei-Quinn (PQ) sym-
metry UPQ (1).
The potential V (|Φ|) is chosen
√ to be a “Mexican hat” with an abso-
lute minimum at |Φ| = fPQ / 2 where fPQ is some large energy scale.
The ground state is characterized
√ iϕ by a nonvanishing vacuum expecta-
tion value )Φ* = (fPQ / 2) e where ϕ is an arbitrary phase. It spon-
taneously breaks the PQ symmetry because it is not invariant under a
transformation of the type Eq. (14.7). One may then write
fPQ + ρ ia/fPQ
Φ= √ e (14.8)
2
in terms of two real fields ρ and a which represent the “radial” and
“angular” excitations.
The potential V provides a large mass for ρ, a field which will be of
no further interest for these low-energy considerations. Neglecting all
terms involving ρ the Lagrangian Eq. (14.6) is
# $
L= Ψ∂µ γ µ Ψ + h.c. + 12 (∂µ a)2 − m Ψeiγ5 a/fPQ Ψ ,
i
2
(14.9)
√
where m ≡ hfPQ / 2. The variation of the fermion fields under a
PQ transformation is given by Eq. (14.7) while a → a + αfPQ . The
invariance of Eq. (14.9) against such shifts is a manifestation of the
UPQ (1) symmetry. It implies that a represents a massless particle, the
Nambu-Goldstone boson of the PQ symmetry.
Expanding the last term in Eq. (14.9) in powers of a/fPQ , the zeroth-
order term mΨΨ plays the role of an effective fermion mass. Higher
orders describe the interaction of a with Ψ,
m m
Lint = −i a Ψγ5 Ψ + 2 a2 ΨΨ + . . . . (14.10)
fPQ 2fPQ
The dimensionless Yukawa coupling ga ≡ m/fPQ is proportional to the
fermion mass.
The fermion Ψ is taken to be some exotic heavy quark with the usual
strong interactions, i.e. an SUC (3) triplet. The lowest-order interaction
530 Chapter 14
of a with gluons is then given by the triangle graph of Fig. 14.2. With
the first term of Eq. (14.10) it yields an effective a-gluon interaction of
ga α s !,
LaG = − a GG (14.11)
m 8π
where αs ≡ gs2 /4π. All external momenta were taken to be small relative
to the mass m of the loop fermion.
Fig. 14.2. Triangle loop diagram for the interaction of axions with gluons
(strong coupling constant gs , axion-fermion Yukawa coupling ga ). An anal-
ogous graph pertains to the coupling of axions with photons if the fermion
carries an electric charge which replaces gs .
In more general models, several conventional or exotic quark fields

Ψj may participate in this scheme. The transformation of each field
under a UPQ (1) transformation is characterized by its PQ charge Xj ,
ΨjL → eiXj α/2 ΨjL . (14.12)
The total aGG ! interaction is obtained as a sum over Eq. (14.11) for all
Ψj . Because gaj = Xj mj /fPQ the fermion masses drop out. With
%
N≡ j
Xj and fa ≡ fPQ /N (14.13)
one has then found the required coupling Eq. (14.3) which allows one
to interpret a as the axion field.
The potential V (a) is periodic with 2πfa = 2πfPQ /N . The interpre-
tation of a as the phase of Φ, on the other hand, implies a periodicity
with 2πfPQ so that N must be a nonzero integer. This requirement
restricts the possible assignment of PQ charges to the quark fields. It
also implies that there remain N different equivalent ground states for
the axion field, each of which satisfies Θ = 0 and thus solves the CP
problem.
Axions 531
14.2.3 Pseudoscalar vs. Derivative Interaction

There has been considerable confusion in the literature concerning the
proper structure for the coupling of axions to fermions. The lowest-
order term in the expansion Eq. (14.10) is a pseudoscalar interaction
which is frequently used because of its simplicity. However, there is an
infinite series of terms which sometimes must be taken into account.
For example, the axion scattering on fermions is second order in the
first term of Eq. (14.10), but first order in the second whence both must
be included.
Such complications can be avoided if one redefines the fermion field
in Eq. (14.9) by a local transformation,
ψL ≡ e−ia/2fPQ ΨL , ψR ≡ eia/2fPQ ΨR . (14.14)
The last term in Eq. (14.10) is then a simple mass term mψψ. The
interaction between ψ and a now arises from the kinetic Ψ term in
Eq. (14.9),
1
Lint = ψγµ γ5 ψ ∂ µ a . (14.15)
2fPQ
This interaction is of derivative nature, and it is linear in a with no
higher-order terms. The fermion part has the form of an axial-vector
current, bringing out a useful similarity between neutrino and axion
interactions.
Pions play the role of Nambu-Goldstone bosons of a spontaneously
broken U(2)L−R symmetry of QCD and so their interactions with nucle-
ons should involve similar higher-order terms. The difference between
derivative and “naive” pseudoscalar couplings should become apparent
in bremsstrahlung processes of the type shown in Fig. 14.3 where two
Nambu-Goldstone bosons are attached to one fermion line. A useful
template is provided by p + p → p + p + π ◦ where existing data, in-
deed, favor the derivative case (Choi, K. Kang, and Kim 1989; Turner,
H.-S. Kang, and Steigman 1989).
It should be noted in this context that the pion-nucleon interaction
is described by the Lagrangian (Carena and Peccei 1989),
Lint = gπN N γ µ γ5 τ N · ∂µ π + fπN
2
N γ µ τ N · ∂µ π×π , (14.16)
where gπN ≡ f /mπ ≈ 1/mπ and fπN ≡ 1/2fπ are the relevant coupling
constants, τ is a vector of Pauli isospin matrices, π is the isovector of
the neutral and charged pion fields, and N is the isodoublet of neutron
and proton. Thus there appears an extra dimension-6 term compared
532 Chapter 14
Fig. 14.3. Nucleon-nucleon bremsstrahlung emission of axions or pions.
with the axion example Eq. (14.15) where only one Nambu-Goldstone
boson was present as opposed to the pion isotriplet. However, this ad-
ditional term does not contribute to the bremsstrahlung process in the
limit of nonrelativistic nucleons so that the above conclusion regarding
the derivative coupling remains valid.
In the process N N → N N a (Fig. 14.3) axions and pions appear so
that again two Nambu-Goldstone bosons are attached to one fermion
line. It is then necessary to use a derivative coupling for at least one
of them (Raffelt and Seckel 1988). For other bremsstrahlung processes
such as e− p → pe− a, where the particles interact through a virtual
photon (a gauge boson) the pseudoscalar coupling causes no trouble.
Also for the Compton process γe− → e− a one may use either the pseu-
doscalar or the derivative axion coupling: both yield the same result.
Because it is not always a priori obvious whether the pseudoscalar and
derivative couplings yield the same result it is a safe strategy to use the
derivative coupling in all calculations.
14.2.4 The Onslaught of Quantum Gravity
A heavy critique was levied against the PQ mechanism by quantum-

gravity inspired phenomenological considerations. The main idea is
that generally the PQ symmetry, like any other global symmetry, will
not be respected by gravity (Georgi, Hall, and Wise 1981). For exam-
ple, a black hole can “swallow” any amount of PQ charge without a
trace, while a swallowed electric charge remains visible by its Coulomb
force. At energy scales exceeding the Planck mass mPl = 1.2×1019 GeV
quantum gravitational effects are expected and so mPl is a phenomeno-
logical cutoff for any quantum theory which does not fundamentally
include gravitation. In the “low-energy” world it should manifest itself
by all sorts of effective interactions which are not forbidden by a sym-
metry and which likely involve inverse powers of the cutoff scale mPl .
Axions 533
Notably, the Higgs field Φ which gives rise to the axion probably
exhibits effective interactions of dimension 2m + n
(ΦΦ† )m Φn
Vgrav (Φ) = g eiδ , (14.17)
m2m+n−4
Pl
where g and δ are real numbers. Because such interactions violate the
PQ symmetry for n += 0 they induce an effective potential for the axion
after spontaneous symmetry breaking. The full potential is then of the
form (Kamionkowski and March-Russell 1992)
V (a) & ' & '
2 2
= m QCD 1 − cos(a/f a ) + m grav 1 − cos(δ + na/f a ) ,(14.18)
fa2
where mQCD is the usual QCD axion mass while gravity induces
√
m2grav = g m2Pl (fa / 2 mPl )2m+n−2 . (14.19)
For g of order unity and for low values of m and n one needs a very
small fa for the QCD effect to dominate. Therefore, unless gravity for
some reason favors a minimum at the CP-conserving position for a the
PQ scheme will be ruined entirely.91
Whatever the ultimate quantum theory of gravitation, no doubt it
will be very special. Therefore, it is by no means obvious that the
above arguments, which do not go far beyond a dimensional analysis,
correctly represent the low-energy effects of Planck-scale physics. Even
then the PQ mechanism still works if the PQ global symmetry is an
“automatic symmetry” of a gauge theory; in this case it is protected
from the assault of quantum gravity. Such models can be constructed
(Holman et al. 1992) and in fact may be quite generic (Barr 1994).
Either way, in order for axions to solve the strong CP problem one
must assume that the PQ scheme is not ruined by quantum gravity.
This discussion illustrates an important feature of axion models,
or any model involving a broken global symmetry and its Nambu-
Goldstone boson. These particles are interlopers in the low-energy
world—axions really belong to the high-energy world at the PQ scale.
These roots make them susceptible to physics at large energy scales,
at the Planck mass, for example. By the same token, if axions were
ever detected, for example by the galactic axion search (Sect. 5.3), they
would be one of the few messengers that we can ever hope to receive
from a high-energy world which is otherwise inaccessible to experimen-
tal enquiry.
91
These issues were studied by Barr and Seckel (1992) and by Kamionkowski and
March-Russell (1992). See also the earlier papers by Georgi, Hall, and Wise (1981),
Lazarides, Panagiotakopoulos, and Shafi (1986), and Dine and Seiberg (1986).
534 Chapter 14
14.3 Fine Points of Axion Properties
14.3.1 The Most Common Axion Models
Axions generically mix with pions so that their mass and their couplings
to photons and nucleons are crudely fπ /fa times those of π ◦ . In detail,
however, these properties depend on the specific implementation of
the PQ mechanism. Therefore, it is useful to review briefly the most
common axion models which may serve as generic examples for an
interpretation of the astrophysical evidence.
In the standard model, the would-be Nambu-Goldstone boson from
the spontaneous breakdown of SU(2)×U(1) is interpreted as the third
component of the neutral gauge boson Z ◦ , making it impossible for the
scalar field Φ of which axions are the phase to be the standard Higgs
field. Therefore, one needs to introduce two independent
√ √Higgs fields Φ1
and Φ2 with vacuum expectation √ values f 1 / 2 and f 2 / 2 which must
2 2 1/2 −1/2
obey (f1 + f2 ) = fweak ≡ ( 2 GF ) ≈ 250 GeV. In this standard
axion model (Peccei and Quinn 1977a,b; Weinberg 1978; Wilczek 1978)
Φ1 gives masses to the up- and Φ2 to the down-quarks and charged
leptons. With x ≡ f1 /f2 and 3 families the axion decay constant is
fa = fweak [3 (x + 1/x)]−1 <
∼ 42 GeV. This and related “variant” models
(Peccei, Wu, and Yanagida 1986; Krauss and Wilczek 1986), however,
are ruled out by overwhelming experimental and astrophysical evidence;
for reviews see Kim (1987), Cheng (1988), and Peccei (1989).
Therefore, one is led to introduce an√electroweak singlet Higgs field
with a vacuum expectation value fPQ / 2 which is not related to the
weak scale. Taking fPQ , fweak , the mass of the axion becomes
very small, its interactions very weak. Such models are generically re-
ferred to as invisible axion models. The first of its kind was the KSVZ
model (Kim 1979; Shifman, Vainshtein, and Zakharov 1980) discussed
in Sect. 14.2.2. It is very simple because the PQ mechanism entirely
decouples from the ordinary particles: at low energies, axions interact
with matter and radiation only by virtue of their two-gluon coupling
which is generic for the PQ scheme. The KSVZ model in its simplest
form is determined by only one free parameter, fa = fPQ , although one
may introduce N > 1 exotic quarks whence fa = fPQ /N .
Also widely discussed is the DFSZ model introduced by Zhitnitskiı̆
(1980) and by Dine, Fischler, and Srednicki (1981). It is a hybrid be-
tween the standard and KSVZ models in that it uses an √ electroweak
singlet scalar field Φ with a vacuum expectation value fPQ / 2 and two
electroweak doublet fields Φ1 and Φ2 . There is no need, however, for ex-
Axions 535
otic heavy quarks: only the known fermions carry Peccei-Quinn charges.
Therefore, N is the number of standard families. Probably N = 3 so
that the remaining free parameters of this model are fa = fPQ /N and
x = f1 /f2 which is often parametrized by x = cot β or equivalently by
cos2 β = x2 /(x2 + 1).
From a practical perspective, the main difference between the KSVZ
and DFSZ models is that in the latter axions couple to charged leptons
in addition to nucleons and photons. The former is an example for the
category of hadronic axion models.
Because fPQ , fweak in these models, one may attempt to identify
fPQ with the grand unification scale fGUT ≈ 1016 GeV (Wise, Georgi,
and Glashow 1981; Nilles and Raby 1982). However, the cosmologi-
cal bound fa < 12
∼ 10 GeV disfavors the GUT assignment. There exist
numerous other axion models, and many attempts to connect the PQ
scale with other scales—for a review see Kim (1987). In the absence
of a compelling model fa should be viewed as a free phenomenological
parameter.
14.3.2 Axion Mass and Coupling to Photons
The axion mass which arises from its mixing with π ◦ can be obtained
with the methods of current algebra to be (Bardeen and Tye 1978;
Kandaswamy, Salomonson, and Schechter 1978; Srednicki 1985; Georgi,
Kaplan, and Randall 1986; Peccei, Bardeen, and Yanagida 1987)
( )1/2
fπ mπ z
ma =
fa (1 + z + w)(1 + z)
107 GeV
= 0.60 eV , (14.20)
fa
where the quark mass ratios are (Gasser and Leutwyler 1982)
z ≡ mu /md = 0.568 ± 0.042,

w ≡ mu /ms = 0.0290 ± 0.0043. (14.21)
Aside from these uncertainties there are higher-order corrections to

the current-algebra axion mass which have not been estimated in the
literature.
536 Chapter 14
By their generic coupling to gluons, axions necessarily mix with

pions and hence couple to photons according to
Lint = − 14 gaγ Fµν F! µν a = gaγ E · B a, (14.22)
where F is the electromagnetic field strength tensor and F! its dual. In
models where the quarks and leptons which carry PQ charges also carry
electric charges, there is a contribution from a triangle loop diagram as
in Fig. 14.2, replacing gs with the electric charge Qj e of the lepton. It
yields an axion-photon coupling proportional to
%
E≡2 j
Xj Q2j Dj , (14.23)
where Dj = 3 for color triplets (quarks) and 1 for color singlets (charged
leptons). The total axion-photon coupling strength is then (Kaplan
1985; Srednicki 1985)
α 3 meV
gaγ = − ξ= ξ, (14.24)
2πfa 4 0.69×1010 GeV
where
* + * +
4 E 2 4+z+w 4 E
ξ≡ − = − 1.92 ± 0.08 (14.25)
3 N 3 1+z+w 3 N
and meV ≡ ma /eV.
In the DFSZ or grand unified models one has for a given fam-
ily of quarks and leptons E/N = 8/3. Neglecting w this yields ξ ≈
(8/3) z/(1 + z) ≈ 1. However, one may equally consider models where
E/N = 2 so that ξ = 0.1±0.1, i.e. the axion-photon coupling is strongly
suppressed and may actually vanish (Kaplan 1985).
14.3.3 Model-Dependent Axion-Fermion Coupling

The discussion in Sect. 14.2.3 implies that axions interact with a given
fermion j (mass mj ) according to a pseudoscalar or a derivative axial-
vector interaction,
Cj m j Cj
Lint = −i Ψj γ5 Ψj a or Ψ j γ µ γ 5 Ψj ∂ µ a , (14.26)
fa 2fa
where Cj is an effective PQ charge of order unity to be defined below.
Evidently gaj ≡ Cj mj /fa plays the role of a Yukawa coupling and
2
αaj = gaj /4π that of an “axionic fine structure constant.” Numerically,
gae = Ce me /fa = Ce 0.85×10−10 meV ,
gaN = CN mN /fa = CN 1.56×10−7 meV (14.27)
for electrons and nucleons.
Axions 537
Various axion models differ in their assignment of PQ charges. How-

,
ever, in all models N = quarks Xj is a nonzero integer. The assignment
at high energies is not maintained in the low-energy sector because
the spontaneous breakdown of the weak SUL (2)×UY (1) symmetry at
fweak ≈ 250 GeV mixes the axion with the would-be Nambu-Goldstone
boson which becomes the longitudinal component of the Z ◦ gauge bo-
son. Hence the PQ charges must be shifted such that the physical axion
does not mix with the Z ◦ ; these shifted values are denoted as Xj# . Also,
below the QCD scale ΛQCD ≈ 200 MeV free quarks do not exist, so one
needs to consider the effective coupling to nucleons which arises from
the direct axion coupling to quarks and from the mixing with π ◦ and η,
leading to PQ charges Xp# and Xn# for protons and neutrons. The above
effective PQ charges are obtained by Cj ≡ Xj# /N in order to absorb N
in their definition just as it was absorbed in fa = fPQ /N .
In the KSVZ model Ce = 0 at tree level (“hadronic axions”) al-
though there are small radiatively induced couplings (Srednicki 1985).
In the DFSZ model
Ce = cos2 β/Nf , (14.28)
where Nf is the number of families, probably 3.

The nucleon interactions in general axion models were investigated
by Kaplan (1985) and Srednicki (1985). They were revisited by Mayle
et al. (1988, 1989),
Cp = (Cu − η)∆u + (Cd − ηz)∆d + (Cs − ηw)∆s ,

Cn = (Cu − η)∆d + (Cd − ηz)∆u + (Cs − ηw)∆s , (14.29)
where η ≡ (1 + z + w)−1 with z and w were given in Eq. (14.21).

For a given quark flavor, q = u, d, or s, the interaction strength
with protons depends on the proton spin content carried by this par-
ticular quark flavor, Sµ ∆q ≡ )p|qγµ γ5 q|p* where Sµ is the proton spin.
Similar expressions pertain to the coupling with neutrons; the two sets
of expressions are related by isospin invariance. Neutron and hyperon
β-decays as well as polarized lepton scattering experiments on nucleons
yield a consistent set of ∆q’s (Ellis and Karliner 1995)
∆u = +0.85, ∆d = −0.41, ∆s = −0.08 (14.30)
with an approximate uncertainty of ±0.03 each.

In the DFSZ model, Cs = Cd = Ce , Cu + Cd = 1/Nf , and Cu − Cd =
− cos2 β/Nf , leading to Cu = sin2 β/Nf and Cd = Cs = Ce = cos2 β/Nf .
538 Chapter 14
With Nf = 3 one finds

Cp = −0.10 − 0.45 cos2 β,
Cn = −0.18 + 0.39 cos2 β . (14.31)
In the KSVZ model and in other hadronic axion models Cu = Cd =
Cs = 0 which yields
Cp = −0.39,
Cn = −0.04. (14.32)
In Fig. 14.4 these couplings are shown, for DFSZ axions as a function
of cos2 β. They are all uncertain to within about ±0.05, but even then
Cp and Cn never seem to vanish simultaneously.
Fig. 14.4. Axion couplings to fermions according to Eqs. (14.28), (14.31),

and (14.32).
14.4 Astrophysical Axion Bounds

Because axions couple to nucleons, photons, and electrons it is easy
to translate the bounds on such couplings derived in previous chapters
of this book into bounds on the Peccei-Quinn scale or equivalently, on
the axion mass. A possible modification of the usual axion models by
the quantum gravity effects discussed in Sect. 14.2.4 is ignored for the
present discussion.92
92
Barr and Seckel (1992) studied astrophysical axion bounds when quantum grav-
ity effects are taken seriously.
Axions 539
Bounds on the Yukawa coupling to electrons of various novel parti-

cles were derived in Chapter 3; for pseudoscalars a summary was given
in Tab. 3.1. The most restrictive limit was obtained from the delay of
helium ignition in low-mass red giants that would be caused by exces-
sive axion emission; in terms of the axion-electron Yukawa coupling it
is gae < −13
∼ 2.5×10 . With Eq. (14.27) this translates into
ma C e < > 9
∼ 0.003 eV and fa /Ce ∼ 2×10 GeV. (14.33)
Axions which interact too strongly to escape freely from the interior
of stars would still contribute to the transfer of energy. For the Sun,
this issue was studied in Sect. 1.3.5. One easily finds that for Ce = 1
Fig. 1.2 excludes axion masses below about 50 keV.
In hadronic axion models Ce = 0 at tree level and so no interesting
bounds on ma and fa obtain. In the DFSZ model, Ce was given in
Eq. (14.28). Taking the number of families to be Nf = 3 one finds
ma cos2 β < 2 > 9

∼ 0.01 eV and fa / cos β ∼ 0.7×10 GeV. (14.34)
These limits depend on the parameter cos2 β which, in principle, can

be equal to 0.
The axion-photon coupling is best constrained by the lifetime of
horizontal-branch (HB) stars as outlined in Sect. 5.2.5. The limit
Eq. (5.23) translates into
ma ξ < > 7
∼ 0.4 eV and fa /ξ ∼ 1.5×10 GeV, (14.35)
where ξ was defined in Eq. (14.25). In addition, approximately the

mass range 4−14 eV is excluded by the “telescope search” for a line
from the radiative decay of cosmic axions (Fig. 12.23).
The most restrictive limit on the axion-nucleon coupling arises from
the duration of the neutrino signal of SN 1987A. The formally ex-
cluded range for the axion-nucleon Yukawa coupling was specified in
Eq. (13.11); as discussed in Sect. 13.5 it is fraught with uncertainties
because no reliable calculation of the axion emission rate from a nuclear
medium is available at the present time. In terms of the axion mass
and axion decay constant the nominally excluded range is
0.002 eV < <

∼ CN ma ∼ 2 eV,
3×106 GeV < < 9
∼ fa /CN ∼ 3×10 GeV. (14.36)
The case of large ma (small fa ) is the trapping regime where axions

contribute to the energy transfer in a SN core, and where they are
540 Chapter 14
emitted from an “axion sphere” rather than the entire volume of the
protoneutron star.
These bounds were derived assuming equal couplings to protons and
neutrons. However, a glance at Fig. 14.4 reveals that KSVZ axions es-
sentially do not couple to neutrons while Cp ≈ −0.36. For DFSZ axions
the couplings vary with cos2 β, although for cos2 β ≈ 0.5 about the same
values as for KSVZ axions apply which are thus taken as generic. As-
suming a proton fraction of about 0.3 for the relevant regions of the SN
core I estimate an effective nucleon coupling of CN ≈ 0.31/2 0.36 ≈ 0.2.
Therefore,
0.01 eV < <
∼ ma ∼ 10 eV,
0.6×106 GeV < < 9
∼ fa ∼ 0.6×10 GeV (14.37)
are formally adopted as the SN 1987A excluded axion parameters.
Axions on the “trapping side” of the SN argument can still be ex-
cluded because they would have caused additional events in the IMB
and Kamiokande water Cherenkov detectors. The excluded range of
Eq. (13.2) translates into the approximate mass exclusion range of
20 eV−20 keV.
These limits on the axion mass and decay constant are summarized
in Fig. 14.5. The slanted end of the SN 1987A exclusion bar is a
reminder of the potentially large uncertainty of this limit. Except for
very special choices of model-dependent parameters axions with a mass
above 0.01 eV are excluded.
The high-mass end of the stellar exclusion bars in Fig. 14.5 has
not been worked out in detail because of their overlap with laboratory
limits. The globular cluster limits apply without modification up to a
mass of, say, 30 keV because the temperature in the cores of HB stars
and red giants are about 10 keV. However, axions with masses in this
range interact much more strongly than those at the low-mass end of the
exclusion bar so that the Boltzmann suppression of the emission rate
for a large mass is partly balanced by the increased coupling strength.
14.5 Cosmological Limits

The stellar-evolution bounds on axions have received much attention
because they push the Peccei-Quinn scale to such large values that
axions appear to play a significant cosmological role if they exist at
all. Therefore, to place the stellar constraints into context it may be
useful to close this chapter with a brief summary of the cosmological
Axions 541
Fig. 14.5. Astrophysical and cosmological bounds on “invisible axions.” One

globular-cluster limit (white exclusion bar) is based on the axion-electron
coupling and thus applies only if axions are of the DFSZ type (cos2 β = 1
was used). For the coupling to photons ξ = 1 was assumed. Slanted ends of
exclusion bars indicate an estimated uncertainty of the bounds. The antic-
ipated range of sensitivity of the Livermore and Kyoto search experiments
are also indicated (Sect. 5.3).
limits which I quote from my contribution to the axion session of the

XVth Moriond Workshop Dark Matter in Cosmology, Clocks, and Tests
of Fundamental Laws, (Villars-sur-Ollon, Switzerland, January 21–28,
1995). The proceedings of this meeting will provide many up-to-date
accounts of different aspects of the axion saga.
If axions were sufficiently strongly interacting (fa < 8
∼ 10 GeV) they
would have come into thermal equilibrium before the QCD phase tran-
sition and so we would have a background sea of invisible axions in
analogy to the one expected for neutrinos (Turner 1987). This pa-
rameter range is excluded by the astrophysical arguments summarized
in Fig. 14.5. Hence axions must be so weakly interacting that they
have never come into thermal equilibrium. Still, the well-known mis-
542 Chapter 14
alignment mechanism will excite coherent oscillations of the axion field

(Abbott and Sikivie 1983; Dine and Fischler 1983; Preskill, Wise, and
Wilczek 1983; Turner 1986). When the temperature of the universe
falls below fa the axion field settles somewhere in the brim of its Mex-
ican hat potential. When the hat tilts at the QCD phase transition,
corresponding to the appearance of a mass term for the axion, the field
begins to move and finally oscillates when the expansion rate of the uni-
verse has become smaller than the axion mass. In units of the cosmic
critical density one finds for the axionic mass density
Ωa h2 ≈ 0.23×10±0.6 (fa /1012 GeV)1.175 Θ2i F (Θi ), (14.38)
where h is the present-day Hubble expansion parameter in units of
100 km s−1 Mpc−1 . The stated range reflects recognized uncertainties of
the cosmic conditions at the QCD phase transition and uncertainties in
the calculations of the temperature-dependent axion mass. The cosmic
axion density thus depends on the initial misalignment angle Θi which
could have any value between 0 and π. The function F (Θi ) encapsules
anharmonic corrections to the axion potential for Θ , 0. (For a recent
analytic determination of F see Strobl and Weiler 1994.)
The age of the universe indicates that Ωh2 ≈ 0.3, causing a problem
with Ω = 1 models if h is around 0.8 as indicated by recent measure-
ments. For the present purpose I take Ωa h2 = 0.3 × 2±1 for axions
which constitute the dark matter where the adopted uncertainty of a
factor of 2 likely covers the whole range of plausible cosmological mod-
els. Then, axions with ma = O(1 µeV) are the cosmic dark matter if
Θi is of order 1.
Because the corresponding Peccei-Quinn scale of fa = O(1012 GeV)
is far below the GUT scale one may speculate that cosmic inflation, if
it occurred at all, did not occur after the PQ phase transition. If it
did not occur at all, or if it did occur before the PQ transition with
Treheat > fa , the axion field will start with a different Θi in each region
which is causally connected at T ≈ fa and so one has to average over all
possibilities to obtain the present-day axion density. More importantly,
because axions are the Nambu-Goldstone mode of a complex Higgs field
after the spontaneous breaking of a global U(1) symmetry, cosmic axion
strings will form by the Kibble mechanism (Davis 1986). The motion of
these global strings is damped primarily by the emission of axions rather
than by gravitational waves. At the QCD phase transition, the U(1)
symmetry is explicitly broken (axions acquire a mass) and so domain
walls bounded by strings will form, get sliced up by the interaction
with strings, and the entire string and domain wall system will quickly
Axions 543
decay into axions. This complicated sequence of events leads to the

production of the dominant contribution of cosmic axions. Most of
them are produced near the QCD transition at T ≈ ΛQCD ≈ 200 MeV.
After they acquire a mass they are nonrelativistic or mildly relativistic
so that they are quickly redshifted to nonrelativistic velocities. Thus,
even the string and domain-wall produced axions form a cold dark
matter component.
In their recent treatment of axion radiation from global strings,
Battye and Shellard (1994a,b) found that the dominant source of axion
radiation are string loops rather than long strings, contrary to what
was assumed in the previous works by Davis (1986) and Davis and
Shellard (1989). At a given cosmic time t the average loop creation
size is parametrized as )-* = αt while the radiation power from loops
is P = κµ with µ the renormalized string tension. The exact values
of the parameters α and κ are not known; the cosmic axion density is
a function of the combination α/κ. For α/κ < 1 the dependence of
Ωa h2 on α/κ is found to be rather weak. Battye and Shellard favor
α/κ ≈ 0.1 for which Ωa h2 = 18×10±0.6 (fa /1012 GeV)1.175 , about an
order of magnitude smaller than originally found by Davis (1986) and
Davis and Shellard (1989). The overall uncertainty has the same source
as in Eq. (14.38) above. With Ωa h2 = 0.3×2±1 the mass of dark-matter
axions is found to be ma = 30−1000 µeV; the cosmologically excluded
range of axion masses is indicated in Fig. 14.5.
These results are plagued with systematic uncertainties. Battye and
Shellard (1994a,b) argue that the largest uncertainty was the impact
of the backreaction of axion emission on the string network. (They
believe that a current numerical study will allow them to pin down
the parameter α/κ to within, say, a factor of 2.) Further, Sikivie and
his collaborators (Harari and Sikivie 1987; Hagmann and Sikivie 1991)
have consistently argued that the motion of global strings was over-
damped, leading to an axion spectrum emitted from strings or loops
with a flat frequency spectrum. In Battye and Shellard’s treatment,
wavelengths corresponding to the loop size are strongly peaked, the
motion is not overdamped. In Sikivie et al.’s picture, much more of
the string-radiated energy goes into kinetic axion energy which is red-
shifted so that ultimately there are fewer axions; it was argued that
the cosmic axion density was then of order the misalignment contribu-
tion. Therefore, following Sikivie et al. one would estimate the mass of
dark-matter axions at about ma = 4−150 µeV where the range reflects
the same overall uncertainties that bedevil the Battye and Shellard
estimate, or the misalignment contribution.
544 Chapter 14
While the cosmic axion bounds claimed by both groups of authors

still differ significantly, the overall uncertainty within either scenario
is larger than the mutual disagreement, i.e. the range of masses where
axions could be the dark matter overlaps significantly between the pre-
dictions of the two groups (Fig. 14.5). Moreover, there remain difficult
to control uncertainties, for example, with the “dilute instanton gas”
calculation of the temperature-dependent axion mass near the QCD
phase transition. There may be other unaccounted systematic prob-
lems which may increase the adopted uncertainty of the cosmic mass
prediction which is represented in Fig. 14.5 by the slanted end of the
cosmic exclusion bar.
The astrophysical and cosmological limits on axions leave a narrow
window of parameters where axions could still exist (Fig. 14.5); they
would then be some or all of the dark matter of the universe. While the
SN 1987A as well as the cosmological bound are each very uncertain,
there is a recent trend toward an allowed range near ma = O(1 meV)
where no current or proposed experimental effort appears to be sensi-
tive. Of course, the overall quantitative uncertainty of the predicted
cosmic axion density is large, especially if one includes the possibility
of late-time inflation after the Peccei-Quinn phase transition or late-
time entropy production after the QCD phase transition. Therefore,
the microwave cavity experiments (Sect. 5.3) in Livermore and Kyoto
no doubt have a fair chance of detecting galactic axions if they are the
dark matter. In September 1995, the Livermore search experiment has
taken up operation (K. van Bibber, private communication).
Chapter 15
Miscellaneous Exotica
Stellar-evolution constraints on a variety of hypotheses are discussed

and compared with limits from other sources. Specifically, a possible
time variation of Fermi’s and Newton’s constant, the validity of the
equivalence principle, a photon mass and charge, the existence of free
quarks and supersymmetric particles, and the role of majorons and
millicharged particles are considered.
15.1 Constancy of Fermi’s Constant

One of the basic physical assumptions commonly made in astrophysi-
cal research is that the laws of nature are the same at different places
in the universe, and at earlier times here and elsewhere. Apparently
this assumption has never been challenged seriously by any experiment
or observation that would have indicated a spatial or temporal varia-
tion of parameters such as particle masses or coupling constants. At
the present time it is not known what fixes the values of such “funda-
mental numbers.” Therefore, the possibility that they vary in time or
space cannot be a priori rejected. Notably, Dirac (1937, 1938) is often
quoted for his speculation that the large value of some dimensionless
numbers occurring in physics are related to variations of some physical
constants on cosmological time scales. Whatever the merit of Dirac’s
large numbers hypothesis, it remains an interesting task to isolate sim-
ple observables that are sensitive to variations of certain “constants.”
One instructive stellar-evolution example was discussed by Scherrer
and Spergel (1993) who considered the constancy of Fermi’s constant
GF which governs weak-interaction
√ 2 physics. The particle-physics stan-
−1
dard model gives GF = 2 Φ0 in terms of the Higgs-field vacuum
expectation value Φ0 . Scherrer and Spergel noted that a nonconstant
545
546 Chapter 15
vacuum expectation value of a physical field was more plausibly subject

to variations than dimensionless constants such as gauge and Yukawa
couplings.
Type Ia supernovae are thought to represent the nuclear deflagration
of a white dwarf pushed beyond its Chandrasekhar limit by accretion
(e.g. Woosley and Weaver 1986b). Therefore, their light curves are
very reproducible and indeed serve as standard candles in an attempt
to improve measurements of the cosmic expansion rate. The shape of
the lightcurve is determined by radioactive heating of the SN remnant
by the decay of 56 Co; its lifetime is proportional to G−2
F . A change of
order 10% in the slope of type Ia SN lightcurves in neighboring galaxies
(Leibundgut et al. 1991) would be readily observable so that GF must
be constant within at least 5% over 30 Mpc distances.
In addition, Scherrer and Spergel showed that the agreement be-
tween the observed primordial light element abundances and big-bang
nucleosynthesis calculations imply that GF lay within −1% and +9%
of its standard value at this early epoch.
A stringent constraint on the local93 time variation of GF obtains
from the analysis of ores from the Oklo uranium mine where a natu-
ral fission reactor is thought to have operated about 2 Gyr ago (e.g.
Maurette 1976). A shift of the ground state difference between 150 Sm
and 149 Sm by more than 0.02 eV is inconsistent with the isotope ra-
tios at Oklo (Shlyakhter 1976, 1983). According to this author the
contribution of the weak interaction to the nuclear binding energy is
about 200 eV, excluding a change of GF by more than 0.1% over the
past 2 Gyr.
15.2 Constancy of Newton’s Constant

15.2.1 Present-Day Constraints from Celestial Mechanics
Another “constant of nature” that might vary in time is Newton’s con-
stant GN . Indeed, there exist self-consistent alternative theories to
general relativity which actually predict a temporal variation of GN on
cosmological time scales—see Will (1993) for a summary of such the-
ories and detailed references. A typical scale for the rate of change
is the cosmic expansion parameter H so that it is natural to write
ĠN /GN = σH with σ a dimensionless model-dependent number. Some
93
The Earth moves with the galaxy and the local group relative to the cosmic
microwave background (CMB). Taking 600 km s−1 for this peculiar velocity the
Earth moves by about 1.2 Mpc in 2 Gyr relative to a frame defined by the CMB.
Miscellaneous Exotica 547
Table 15.1. Bounds on the present-day ĠN /GN . (Adapted from Will 1993.)
Method ĠN /GN References

[10−12 yr−1 ]
Laser ranging (Moon) 0 ± 10 Müller et al. (1991)
Radar ranging (Mars) −2 ± 10 Shapiro (1990)
Binary pulsar 1913+16 11 ± 11 Damour and Taylor (1991)
Spin-down PSR 0655+64 < 55 Goldman (1990)
recent discussions have also addressed the possibility of an oscillating

GN (Hill, Steinhardt, and Turner 1990; Accetta and Steinhardt 1991)
with a rate of change much faster than H. The following discussion
does not generically address such extreme model assumptions.
The GN rate of change can be tested by a detailed study of the
orbits of celestial bodies. Particularly precise data exist in the so-
lar system from laser ranging of the moon and radar ranging of plan-
ets, notably by the Viking landers on Mars. Very precise orbital data
also exist beginning 1974 for the binary pulsar PSR 1913+16; among
other things its orbital decay reveals the emission of gravitational ra-
diation. A weaker but also less model-dependent bound can be de-
rived from the spin-down rate of the pulsar PSR 0655+64. These
present-day limits on ĠN /GN are summarized in Tab. 15.1; altogether
|ĠN /GN | < −12 −1
∼ 20×10 yr is probably a safe limit. The Hubble expan-
sion parameter today is H0 = h 100 km s−1 Mpc−1 = h 1.02×10−10 yr−1
(observationally 0.4 < < < −1
∼ h ∼ 1). Thus today |σ| ∼ 0.2 h .
15.2.2 Big-Bang Nucleosynthesis

An interesting constraint on the value of GN in the early universe arises
from the observed primordial light element abundances as first dis-
cussed by Barrow (1978).94 In a Friedman-Robertson-Walker model of
the universe the expansion rate is given by H 2 = (Ṙ/R)2 = 8π 3
GN ρ
in terms of the energy density ρ which, during the epoch of nucleosyn-
thesis, is dominated by radiation (photons, neutrinos). It is a standard
94
Apparently there is an earlier discussion of this limit by G. Steigman in an
unpublished essay for the 1976 Gravity Research Foundation Awards. Subse-
quent refinements include Rothman and Matzner (1982), Accetta, Krauss, and
Romanelli (1990), Damour and Gundlach (1991), and Casas, Garcı́a-Bellido, and
Quirós (1992).
548 Chapter 15
argument to constrain ρ from the yield of 4 He and other light elements,

and thus to constrain the effective number of neutrino degrees of free-
dom at nucleosynthesis (Yang et al. 1979, 1984; Olive et al. 1990; Walker
et al. 1991). Because the number of low-mass sequential neutrino fam-
ilies is now known to be 3, such constraints can be translated into
constraints on the value of GN pertaining to the nucleosynthesis epoch.
An extra neutrino species would add around 15% to ρ. It appears
reasonably conservative to assume that big-bang nucleosynthesis does
not allow for a deviation of the standard number of effective neutrino
degrees of freedom by more than 1 so that GN at that time must
have been within about ±15% of its present-day value. It is possible,
however, that GN was considerably smaller if this reduction was com-
pensated by additional exotic degrees of freedom such as right-handed
neutrinos which increase ρ. Moreover, it was assumed that Fermi’s
constant had its present-day value at nucleosynthesis, contrary to the
speculations discussed in Sect. 15.1. Still, barring fortuitous compen-
sating effects, nucleosynthesis excludes an O(1) deviation of GN from
its standard value at nucleosynthesis.
This sort of result can be compared with the present-day bounds
of Sect. 15.2.1 only by assuming a specific functional form for GN (t)
which is often taken to be
GN (t) = GN (t0 ) (t0 /t)β , (15.1)
where t0 refers to the present epoch. Assuming that GN at nucleosyn-

thesis was within ±50% of its standard value one finds |β| < ∼ 0.01 which
would imply |ĠN /GN |today <
∼ 10 −12
yr −1
, at least a factor of ten below
the present-day limits. One should keep in mind, however, that a power-
law variation of GN is a relatively arbitrary assumption. For example,
in scalar-tensor extensions of general relativity such as the Brans-Dicke
theory GN varies as a power law during the matter-dominated epoch
while it remains constant when radiation dominates. Either way, while
the nucleosynthesis bounds are probably somewhat more restrictive
than the celestial-mechanics ones it is interesting that the resulting
bounds |ĠN /GN |today < −12 −1
∼ 1−10×10 yr are of the same general order
of magnitude. They leave room for a considerable variation of GN over
cosmic time scales.
15.2.3 Properties of the Sun

If GN did vary in time one would expect a modification of the standard
course of stellar evolution as first stressed by Teller (1948). By means
of a beautifully simple homology argument, Teller showed that the lu-
minosity of the Sun is approximately proportional to G7N M5 with M
the solar mass which was also allowed to vary in the spirit of Dirac’s
(1937, 1938) large numbers hypothesis. Teller then proceeded to esti-
mate the temperature on Earth in the past, taking a modification of its
orbit from the GN variation into account. Depending on whether GN
was larger or smaller in the past the average terrestrial surface tem-
perature would have been larger or smaller. Teller estimated that if
GN fell outside ±10% of its standard value it would be unlikely that
life on Earth could be sustained. Therefore, GN should have remained
constant to within this accuracy at least during the past 500 million
years or more where life has been known to exist on Earth.95
Later, a similar homology argument based on the solar age was pre-
sented by Gamow (1967). Detailed models of the Sun with a varying
GN were constructed by Pochoda and Schwarzschild (1964), Ezer and
Cameron (1966), Roeder and Demarque (1966), Shaviv and Bahcall
(1969), Chin and Stothers (1975, 1976), Demarque et al. (1994), and
Guenther et al. (1995). The crux with constraining ĠN from the Sun
is that the presolar helium abundance Yinitial and the mixing-length pa-
rameter α can and must be tuned to reproduce the Sun’s present-day
luminosity and radius. In Sect. 1.3.2 it became clear that even extreme
anomalous energy-loss rates could be compensated by an adjustment
of Yinitial ; a similar effect pertains to variable-GN solar models. Even
though the present-day central temperature, density, and helium abun-
dance could differ vastly from standard predictions, their main impact
would be on the neutrino flux which, however, is not a reliable probe
of the solar central conditions as it may get modified by neutrino oscil-
lations.
At the present time the most sensitive probe of a variant internal
solar structure is afforded by the measured p-mode frequencies which
agree well with standard predictions, especially when the gravitational
settling of helium is taken into account. Demarque et al. (1994) have
constructed solar models with a varying GN and then analyzed their
p-mode spectra in comparison with the observations. They assumed a
GN time variation of the form Eq. (15.1) with t0 = 15 Gyr for the age of
95
Apparently there is more recent evidence for primitive lifeforms on Earth as
early as 3.5 Gyr ago (e.g. Gould 1994).
550 Chapter 15
Table 15.2. Characteristics of the Demarque et al. (1994) solar models with
a varying GN according to Eq. (15.1).
37
β α Xinitial Xc Tc ρc Renv Cl 71 Ga
[%] [%] [106 K] [g/cm3 ] [R" ] [SNU] [SNU]
−0.4 1.832 69.42 45.6 15.14 125.5 0.739
−0.2 1.895 70.10 42.1 15.28 134.0 0.731 5.5 117
−0.1 1.936 70.51 40.0 15.37 139.5 0.724
0.0 1.983 70.98 37.6 15.47 146.2 0.721 6.8 124
0.1 2.036 71.51 34.8 15.58 154.6 0.716
0.2 2.104 72.11 31.7 15.72 165.1 0.710 8.7 134
0.4 2.291 73.56 23.9 16.07 197.3 0.695
the universe. Some characterisitics of their solar models as a function of

β are summarized in Tab. 15.2 where α is the mixing-length parameter,
Xinitial the presolar hydrogen abundance, the quantitites with index c
refer to central conditions of the present-day Sun, and Renv to the radius
of its convective envelope. The last two columns are the predicted
counting rates in the chlorine and gallium solar neutrino experiments
under the assumption that there are no neutrino oscillations.
The most important effect of the GN variation is a shift of the base
of the convective envelope which is caused by the required change of α
and the initial helium abundance Yinitial . It is this modification of the
convection zone which has the largest impact on the observable p-mode
frequencies. However, an identical shift can be produced by other effects
such as modified opacities, equation of state, surface boundary condi-
tions, degree of gravitational helium settling, and perhaps by magnetic
fields. Therefore, only relatively crude limits can be extracted at the
present time. Demarque et al. (1994) believe that |β| < ∼ 0.4 is a rea-
sonably conservative limit which probably can be improved by a factor
of four within the next decade by more precise p-mode observations.
With Eq. (15.1) the current limit corresponds to
|ĠN /GN | < −12 −1

∼ 30×10 yr , (15.2)
similar to the celestial-mechanics bounds of Tab. 15.1. The precise func-

tional form Eq. (15.1) is not crucial for the solar bound as it probes GN
only for the last 4.5 Gyr of the assumed 15 Gyr cosmic age. Therefore,
one could have equally assumed a linear form for GN (t).
Most recently, a similar study was completed by Guenther et al.

(1995) who focussed on the predicted g-mode spectrum. At the present
time there is no generally accepted observation of solar g-modes (re-
call that they are evanescent in the convection zone). If one were to
take the claimed observations by Hill and Gu (1990) seriously, a bound
|β| <
∼ 0.05 would obtain. Therefore, if an unambiguous identification
of g-modes would emerge from a number of forthcoming observational
projects, the Sun may yet provide one of the most restrictive limits on
the constancy of Newton’s constant.
15.2.4 White Dwarfs

A large impact of a time-varying gravitational constant can be expected
on the oldest stars which “integrate” GN (t) into the more distant past
than does the evolution of the Sun. One well understood case are
white dwarfs, the faintest of which likely formed shortly after the birth
of the galactic disk. Therefore, the age of the galactic disk implied by
the fast drop of the white-dwarf luminosity function at the faint end
(Sect. 2.2.1) depends on the GN evolution in the past.
In an early study Vila (1976) concluded on the basis of the observa-
tions then available that ĠN /GN as large as 75×10−12 yr−1 was not ex-
cluded. Garcı́a-Berro et al. (1995) constructed detailed luminosity func-
tions under the assumption of a decreasing GN . For an assumed age of
the galactic disk of 7 Gyr, which probably is a lower plausible limit, the
best fit for the faintest data point requires ĠN /GN = −10×10−12 yr−1
while the curves for 0 and −30×10−12 yr−1 lie somewhat outside of the
1σ error bar of this all-important data point. Still, the white-dwarf
luminosity function does not seem to yield significant limits relative to
the celestial-mechanics ones.
15.2.5 Globular Clusters

The oldest stellar objects in the galaxy are globular-cluster stars which
are thus expected to yield the most restrictive stellar-evolution limits
on ĠN /GN . A color-magnitude diagram for an intermediate-aged galac-
tic cluster was constructed by Roeder (1967) while detailed studies of
globular clusters were performed by Prather (1976) and Degl’Innocenti
et al. (1995). Roeder (1967) and Prather (1976) used a time variation
for a specific Brans-Dicke cosmology where GN decreases approximately
as in Eq. (15.1) with β ≈ 0.03 while Degl’Innocenti et al. (1995) con-
sidered more generic cases of GN (t).
552 Chapter 15
The main impact of the assumed GN variation is a change in the

time it takes for a star to burn out hydrogen at its center and thus to
leave the main sequence (MS). The subsequent fast evolution (ascend-
ing the RGB, HB evolution, etc.) is determined by the present-day
value of GN . The stellar evolutionary tracks in the color-magnitude di-
agram can look significantly different from the standard ones if the
GN variation was sufficiently severe. However, within the range of
possibilities left open by the above ĠN /GN bounds, the present-day
isochrone cannot be observationally distinguished from the standard
case (Degl’Innocenti et al. 1995). Apparently, then, the only signifi-
cant consequence of a time-varying gravitational constant is that the
true age τ of a globular cluster is different from its apparent age τ∗
which is inferred from its color-magnitude diagram in the framework of
a constant-gravity scenario.
The change of the MS lifetime can be estimated by Teller’s (1948)
homology relation L ∝ GγN . Based on a specific assumption for the
opacity variation with temperature and density Teller found γ = 7
while a more appropriate value for low-metallicity globular-cluster stars
is γ = 5.6 (Degl’Innocenti et al. 1995). Either way, one can easily
show that the true (τ ) and apparent age (τ∗ ) at the MS turnoff are
approximately related by
! t0 " #γ
τ∗ = dt GN (t)/GN (t0 ) (15.3)
t0 −τ
(Prather 1976; Degl’Innocenti et al. 1995). For all practical purposes

this analytic result can be considered to be exact because it agrees with
numerical calculations surprisingly well.
Unless one wishes to probe the very early universe it is fairly generic
to assume a linear GN variation of the form
" #
GN (t) = 1 + Γ0 (t − t0 ) GN (t0 ), (15.4)
where Γ0 = ĠN (t0 )/GN (t0 ) is the present-day rate of change of Newton’s
constant. Then one finds explicitly
τ γ 1 Γ0 τ 1 − (1 − γ1 Γ0 τ∗ )1/γ1
= = , (15.5)
τ∗ 1 − (1 − Γ0 τ )γ1 Γ0 τ∗
where γ1 ≡ γ + 1. Given a present-day rate of change Γ0 one can
thus determine the modification of the globular-cluster age if a certain
apparent age or a certain true age is assumed.
The observed color-magnitude diagrams of globular clusters yield
apparent ages τ∗ in the range 14 to 18 Gyr. With these values one
Fig. 15.1. Required present-day ĠN /GN in order to achieve a true globular-
cluster age τ , given that the apparent age is τ∗ . A linear GN (t) variation as
in Eq. (15.4) was assumed.
can relate a desired true age τ to a required value for Γ0 (Fig. 15.1).
Conversely, it is probably safe to assume that the true ages of globular
clusters do not exceed 20 Gyr. Then Fig. 15.1 implies that today
ĠN /GN < −12 −1

∼ 7×10 yr . (15.6)
A lower age limit is less certain. Taking 8 Gyr one finds ĠN /GN > ∼
−35×10−12 yr−1 which is less certain, and also less interesting relative
to the limits discussed in the previous sections.
Fig. 15.2. Summary of limits on the present-day ĠN /GN . The big-bang
nucleosynthesis limit is not shown as it depends sensitively on the assumed
GN (t) variation at early cosmic times.
554 Chapter 15
Actually, a certain reduction of the true globular-cluster ages rela-

tive to their apparent ones would be a welcome cosmological effect as
they are, at best, marginally compatible with other cosmic age indica-
tors. In view of the current limits on ĠN /GN summarized in Fig. 15.2
this possibility cannot be excluded at present.
15.3 Test of the Equivalence Principle
The equivalence principle of Einstein’s general theory of relativity im-

plies that the space-time trajectories of relativistic particles should be
independent of internal degrees of freedom such as spin or flavor, and
independent of the type of particle under consideration (photons, neu-
trinos). A number of astronomical observations allow one to test this
prediction. Laboratory tests of various consequences of the equivalence
principle are discussed in Will’s (1993) book.
Nonsymmetric extensions of general relativity (e.g. Moffat 1991)
predict that different polarization components of electromagnetic waves
propagate with different phase velocities in gravitational fields. This
birefringence effect would lead to the depolarization of the Zeeman
components of spectral lines emitted in magnetically active regions of
the Sun. The absence of this depolarization effect leads to significant
constraints on Moffat’s theory and others (Gabriel et al. 1991).
In a similar approach one uses the difference of the Shapiro time
delay between different particles or between different polarization states
of a given particle which propagate through the same gravitational
field. In Sect. 13.3 the absence of an anomalous shift between the
SN 1987A photon and neutrino arrival times gave limits on violations
of the equivalence principle because both pulses moved through the
same galactic gravitational potential.
Also, one may search for differences in the arrival times of left-
and right-handed polarized electromagnetic signals from distant pulsars
(LoSecco et al. 1989). The best bound was obtained from an analysis of
the pulse arrival times from PSR 1937+21 which is about 2.5 kpc away
from Earth. One may write the effective gravitational potential in the
form V (r) = V0 (r) [1 + A1 σ · r̂ + A2 σ · v + A3 r̂ · (v × σ)] where r, v, and
σ represent the location, velocity, and spin of the particles (photons,
neutrinos). The PSR 1937+21 data then yield a constraint |A1 | <
4×10−12 and |A2 | < 1×10−12 (Klein and Thorsett 1990), apparently
the most restrictive limits of their kind.
A violation of the equivalence principle could also manifest itself by

a relative shift of the energies of different neutrino flavors in a gravita-
tional field. For a given momentum p the matrix of energies in flavor
space (relativistic limit) is E = p + M 2 /2p + 2pφ(r)(1 + F ) where M 2
is the squared matrix of neutrino masses, φ(r) is the Newtonian gravi-
tational potential, and F is a matrix of dimensionless constants which
parametrize the violation of the equivalence principle; in general rela-
tivity F = 0. A nontrivial matrix F can lead to neutrino oscillations in
analogy to the standard vacuum oscillations which are caused by the
matrix M 2 (Gasperini 1988, 1989; Halprin and Leung 1991; Pantaleone,
Halprin, and Leung 1993; Iida, Minakata, and Yasuda 1993; Minakata
and Nunokawa 1995; Bahcall, Krastev, and Leung 1995). Values for
Fij in the general 10−14 −10−17 range could account for the solar neu-
trino problem and perhaps could be probed with future long-baseline
oscillation experiments.
15.4 Photon Mass and Charge

Even though in classical electrodynamics gauge invariance implies that
photons must be massless, quantum electrodynamics (QED) can be
formulated consistently with the inclusion of a photon mass, and the
limit mγ → 0 takes the modified theory smoothly over to massless QED
(Stückelberg 1941). Therefore, the possibility of a small photon mass
cannot be excluded theoretically; limits must be set by laboratory and
astrophysical methods. A still up-to-date review of the laboratory lim-
its was given by Goldhaber and Nieto (1971); the best is mγ < −14
∼ 10 eV
from a test of Coulomb’s law (a photon mass would modify the inverse-
square behavior). More recent experiments worked at low temperature
(Ryan, Accetta, and Austin 1985; Chernikov et al. 1992); the resulting
limits on mγ are relatively weak, however.
An astrophysical limit may be set by the absence of an anomalous
dispersion of photon signals from distant sources, notably the pulsed
signal from radiopulsars. This method is limited by the presence of the
ionized interstellar medium. It causes a dispersion relation for photons
which mimics the effect of a photon mass mγ = ωP where the plasma
frequency is given by ωP2 = 4παne /me . With a typical electron density
ne of order 0.1 cm−3 the photon plasma mass is of order 10−11 eV so
that a vacuum mass much smaller than this value cannot be probed.
In Sect. 13.3.3 a limit on a hypothetical νe charge was derived from
the absence of a dispersion of the SN 1987A neutrino pulse. The path of
556 Chapter 15
charged particles in the galactic magnetic field would be curved, lead-

ing to an energy-dependent time-delay (Barbiellini and Cocconi 1987).
Because the same argument can be applied to photons, the signals from
radio pulsars also allow one to set a limit on a putative photon electric
charge (Cocconi 1988). However, the resulting dispersion effect scales
with photon frequency in the same way as the effect caused by a photon
mass or by the plasma effect so that this method, again, is limited by
the standard dispersion effect (Raffelt 1994). One finds a bound on the
photon charge of Qγ < −29
∼ 10 e.
Returning to a hypothetical photon mass, its value can be extracted,
in principle, from the spatial distribution of static magnetic fields of
celestial bodies. The measured fields can be fitted by an appropriate
multipole expansion in which mγ is kept as a free parameter. The
most restricitve limit of this sort was derived from Jupiter’s magnetic
field on the basis of the Pioneer-10 observations; Davis, Goldhaber,
and Nieto (1975) found a limit mγ < −15
∼ 0.6×10 eV. The same method
applied to the Earth’s magnetic field yields an almost equivalent bound
of mγ < −15
∼ 0.8×10 eV (Fischbach et al. 1994).
As detailed in a review by Barrows and Burman (1984) more restric-
tive limits obtain from detailed considerations of astrophysical objects
in which magnetic fields, and hence the Maxwellian form of electrody-
namics, play a key role in maintaining equilibrium or creating long-lived
stable structures. The most restrictive such limit of mγ < −27
∼ 10 eV is
based on an argument by Chibisov (1976) concerning the magneto-
gravitational equilibrium of the gas in the Small Magellanic Cloud
which requires that the range of the interaction exceeds the charac-
teristic field scale of about 3 kpc. This limit, if correct, is surprisingly
close to 10−33 eV where the photon Compton wavelength would exceed
the radius of the observable universe and thus would cease to have any
observable consequences.
15.5 Free Quarks

It is thought that quarks cannot exist as free particles; they always
occur bound in hadrons which are neutral (“white”) with regard to the
“color charge” of the strong interaction. In order to test this hypoth-
esis of confinement it remains an important task to search for single
quarks. The observation of fractional charges in the experiment of
LaRue, Phillips, and Fairbanks (1981) has never been confirmed. How-
ever, if their observations were caused by unconfined quarks bound to
nuclei it would correspond to an abundance of one quarked nucleus

(Q-nucleus) in 6×1017 normal ones.
A small abundance of Q-nuclei could significantly alter the stel-
lar thermonuclear reaction chains and among other effects change the
solar neutrino predictions (Boyd et al. 1983). Detailed nuclear reac-
tion chains involving Q-nuclei were studied by Boyd et al. (1985). For
strangelets (lumps of strange quark matter) trapped in stars the nuclear
networks were investigated by Takahashi and Boyd (1988); the effect of
strangelets is similar to that of Q-nuclei. The Q-nuclear reactions were
implemented in a stellar evolution code by Joseph (1984). Predictions
for the solar neutrino flux were worked out by Sur and Boyd (1985).
With a Q-nuclear abundance of order 10−15 the modified reaction
chains would compete with the standard ones. In the Sun one could
achieve a reduction of the high-energy solar neutrino flux and thus solve
the “old solar neutrino problem” (missing boron neutrinos). However,
Sur and Boyd (1985) predict an increase of the low-energy flux, corre-
sponding to a substantially increased counting rate at the gallium solar
neutrino experiments. As this contradicts the findings of SAGE and
GALLEX (Sect. 10.3) one concludes that Q-nuclear burning is not the
answer to the solar neutrino problem. Turning the SAGE/GALLEX
observations around one concludes that in the Sun the abundance of
Q-nuclei is below about 10−15 .
15.6 Supersymmetric Particles

Supersymmetric extensions of the particle-physics standard model are
very popular, among other reasons because the lightest supersymmet-
ric particle (LSP) could play the role of the cosmic dark matter. In
these models, there is a fermionic partner to all standard bosons, and
a bosonic partner to all standard fermions. The supersymmetric part-
ners of the photon, the Z ◦ gauge boson, and the neutral Higgs bo-
son (photino, Zino and Higgsino) would be Majorana fermions. They
would be very much like Majorana neutrinos except that their interac-
tion strength is not fixed by the Fermi constant but rather depends on
details of the supersymmetric models.
If these “neutralinos” had low enough masses they would be pro-
duced in the interior of stars by the same processes that create neu-
trinos, except that the coupling strength has to be adjusted according
to the particular model that one has in mind. Limits to an anomalous
energy loss of stars yielded early constraints on supersymmetric models
558 Chapter 15
(Bouquet and Vayonakis 1982; Fukugita and Sakai 1982; Anand et al.
1984). The neutrino burst of SN 1987A yielded more interesting limits
for the case of low-mass photinos (Ellis et al. 1988; Grifols, Massó, and
Peris 1989; Grifols and Massó 1990b). To avoid that too much energy
is carried away by photinos they inferred that squark masses in the
approximate range 60 GeV to 2.5 TeV were excluded.
Low-mass neutralinos are disfavored by laboratory limits. More-
over, if the LSP plays the role of cold dark matter its mass likely is
above several 10 GeV. In this case the stellar energy-loss arguments
would not yield any constraints as all supersymmetric particles would
be too heavy to be emitted. Stars would still play an interesting role as
they could trap the dark-matter particles. Their annihilation in the Sun
or Earth would lead to a high-energy neutrino signal which has been
constrained by the Kamiokande detector (Mori et al. 1992). It may
well be found at the Cherenkov detectors Superkamiokande, NESTOR,
DUMAND, or AMANDA and thus lead to the indirect discovery of par-
ticle dark matter in the galaxy. These important issues are discussed
at length in the forthcoming review Supersymmetric Dark Matter by
Jungman, Kamionkowski, and Griest (1995).
15.7 Majorons
15.7.1 Particle-Physics and Cosmological Motivations
Axions (Chapter 14) are one representative of a variety of Nambu-
Goldstone bosons of spontaneously broken global symmetries that have
appeared in the literature over the years. Another widely discussed ex-
ample are the majorons first introduced by Chicashige, Mohapatra,
and Peccei (1981) as a scheme to generate small neutrino Majorana
masses. An important variation by Gelmini and Roncadelli (1981) and
Georgi, Glashow, and Nussinov (1981) led to a model where neutri-
nos had small Majorana masses and coupled to the massless majoron
(a pseudoscalar boson like the axion) with a relatively large Yukawa
strength. The main phenomenological interest in this sort of conjec-
ture lies in the intriguing possibility that neutrinos could have relatively
strong interactions with the majorons and with each other by virtue of
majoron exchange. As majorons would not necessarily show up in in-
teractions with ordinary matter one could well speculate that neutrinos
might have “secret interactions” which would be of relevance only in
a neutrino-dominated environment such as the early universe, perhaps
the present-day universe if neutrinos have a cosmologically significant
mass, and in supernovae. Another motivation to consider majoron

models is the possibility to account for fast neutrino decays in order to
avoid cosmological neutrino mass bounds. In the laboratory, majorons
could show up in experiments searching for neutrinoless 2β decays.
The main motivation for the introduction of majorons is the puz-
zling smallness of neutrino masses (if they have nonvanishing masses at
all) relative to other fermions. As outlined in Sect. 7.1, in the particle-
physics standard model it is thought that all Dirac fermions acquire a
mass by their interaction with a background Higgs field which takes on
a classical value (vacuum expectation value) Φ0 everywhere; neutrino
masses could well arise in the same fashion, except that the Yukawa
couplings to the Higgs field would have to be extremely small. Alter-
natively, one may speculate that neutrino masses are so small because
they arise in a different fashion. Notably, the known sequential neu-
trinos νe , νµ , and ντ could well be Majorana fermions, i.e. their own
antiparticles so that the ν e is really equivalent to a helicity-plus νe . As
long as neutrinos are massless this picture is equivalent to an inter-
pretation where the standard left-handed neutrinos are the two active
components of a four-component Dirac spinor while the two remain-
ing sterile components would never have been observed because they
do not interact. With a nonvanishing mass these interpretations are
vastly different because helicity flips in collisions would allow one to
produce the (almost) sterile “wrong-helicity” Dirac components. This
possibility was exploited in Sect. 13.8.1 to set bounds on a neutrino
Dirac mass from the SN 187A neutrino signal. For Majorana neu-
trinos, a helicity-flipping collision takes an active νe into an active ν e ,
thus violating lepton number by two units. Therefore, Majorana masses
could not arise from the coupling to the standard Higgs field which is
lepton-number conserving.
Majorana masses could arise, however, by interacting with a dif-
ferent Higgs field which would develop a vacuum expectation value by
virtue of the usual spontaneous breakdown of a global symmetry. The
resulting Nambu-Goldstone boson is the majoron. (Recall that the
Nambu-Goldstone boson of the standard Higgs field shows up as the
third polarization degree of the massive Z ◦ gauge boson so that there
is no massless Nambu-Goldstone degree of freedom in the standard
model.) In the original model of Chicashige, Mohapatra, and Peccei
(1981), the “singlet majoron model,” the new vacuum expectation value
was considered to be much larger than the standard Φ0 ≈ 250 GeV.
Large masses would be given primarily to sterile neutrinos postulated
to exist; the standard sequential neutrinos would obtain their small
560 Chapter 15
masses by a see-saw type mixing effect with the heavy states. The ma-
joron coupling to standard neutrinos would be extremely small in this
model, leading to no interesting consequences besides small Majorana
masses for νe , νµ , and ντ .
Gelmini and Roncadelli (1981) and Georgi, Glashow, and Nussinov
(1981) suggested instead to do away with the unobserved heavy sterile
neutrinos and give a small Majorana mass directly to the sequential
neutrinos by the interaction with the new Higgs field. The intriguing
feature of this model is that it requires a very small vacuum expectation
value v, perhaps in the keV regime. As all couplings of the new Higgs
field to fermions scale with the inverse of v, the majoron would have
a rather strong coupling to neutrinos. Among many fascinating phe-
nomenological and astrophysical consequences (e.g. Georgi, Glashow,
and Nussinov 1981; Gelmini, Nussinov, and Roncadelli 1982) this model
predicted, however, that the new Higgs field should contribute precisely
the equivalent of two massless neutrino species to the Z ◦ decay width.
The measurements of this width at CERN and SLAC in 1989−1990,
however, correspond exactly to the known three neutrino flavors (Parti-
cle Data Group 1994), leaving no room for this “triplet majoron model.”
Other “doublet majoron models” which would contribute one-half of an
effective neutrino species to the Z ◦ decay width are also excluded (for
references see, e.g. Berezhiani, Smirnov, and Valle 1992). It is possible,
however, to construct majoron models for Majorana neutrino masses
which retain the original idea of Chicashige, Mohapatra, and Peccei
(1981) and yet provide large majoron-neutrino couplings (e.g. Berezhi-
ani, Smirnov, and Valle 1992 and references therein; see also Burgess
and Cline 1994a; Kikuchi and Ma 1994, 1995).
The main motivation for going out of one’s way to construct such
models does not arise from particle theory but rather from experi-
ments and astrophysics. In Sect. 7.1.4 it was outlined that those nuclei
which decay predominantly by a double beta channel (emission of 2e−
and 2ν e ) can also decay in a neutrinoless mode if νe has a Majorana
mass, allowing an emitted ν e to be effectively reabsorbed as a νe . In
majoron models of Majorana neutrino masses there is a third decay
channel where the intermediate νe in the 0ν mode radiates a majoron
so that effectively 2e− plus one majoron χ are emitted. The expected
sum spectrum of the electron energies would be continuous as in the
2ν mode, but with a different spectral shape. Once in a while, ex-
periments which search for the 0ν mode (a sharp endpoint peak of
the 2e− sum spectrum) have reported a continuous spectral signature
which allegedly could not be ascribed to the dominant 2ν mode or other
backgrounds, although such claims have tended to disappear with the

collection of more significant data. If interpreted in terms of an up-
per limit to the majoron-νe Yukawa coupling, current experiments give
about g < −4
∼ 2×10 (Beck et al. 1993 and references therein). At the
present time there does not appear to exist a compelling signature for
majorons in any of the experiments, although several of them seem
to find certain spectral anomalies—see, e.g. Burgess and Cline (1993,
1994b) for an overview and references.
If the spectral anomalies were to represent the first evidence of ma-
joron emission in 2β decays, the Yukawa coupling would have to be near
the 10−4 level. The neutrino-majoron coupling is given as g = mν /v in
terms of the neutrino mass and the symmetry breaking scale v. With
mνe <∼ 1 eV for a Majorana mass from measured limits on the 0ν de-
cay mode one finds the requirement v < ∼ 10 keV which is an extremely
small scale of symmetry breaking. The majoron is the “angular degree
of freedom” of a complex scalar field; the “radial degree of freedom,”
often referred to as the ρ field, has a mass typically of order v. There-
fore, a low-mass scalar particle beyond the majoron would appear in
the low-energy sector of the theory.
One severe limitation on such models is provided by big-bang nu-
cleosynthesis which has been widely used to set limits on additional
low-mass degrees of freedom which are thermally excited during the
epoch of nucleosynthesis; an upper limit of about 0.3 is often quoted as
the maximum allowed extra contribution in units of effective neutrino
degrees of freedom (e.g. Walker et al. 1991). On the face of it, this
limit excludes majoron models where the majoron (and possibly the ρ)
interact sufficiently strongly with neutrinos to reach thermal equilib-
rium. In one recent study Chang and Choi (1994) found that one must
require g < −5
∼ 10 for the largest majoron Yukawa coupling to any neu-
trino species in order to avoid thermalization of the majorons before
nucleosynthesis.
Such constraints rely on the assumption of 3 standard light neu-
trino species being in thermal equilibrium at the epoch of nucleosyn-
thesis. However, in majoron models the usual cosmological neutrino
mass bound does not apply because of the possibility of fast decays of
the type ν → ν % χ (majoron χ) which are induced by flavor off-diagonal
Yukawa couplings. Therefore, ντ ’s may have a mass of, say, a few
MeV and may have disappeared by decays and annihilations before nu-
cleosynthesis, thus making room for majorons; for detailed numerical
studies see Kawasaki et al. (1994) and papers quoted there. Such heavy,
short-lived ντ ’s may provide interesting effects on scenarios of galaxy
562 Chapter 15
formation and thus could be a novel ingredient for cold dark matter
cosmological models (Dodelson, Gyuk, and Turner 1994).
Even if in the long run the 2β experiments do not yield any com-
pelling evidence for majoron emission, one may consider a decaying-
neutrino cosmology as a motivation in its own right for majoron mod-
els. Such cosmologies may explain the discrepancy between the cosmic
density fluctuation spectrum inferred from the cosmic microwave back-
ground and from galaxy correlations which persists in a purely cold
dark matter cosmology (Bond and Efstathiou 1991; Dodelson, Gyuk,
and Turner 1994; White, Gelmini, and Silk 1995). Another neutrino-
related explanation of this discrepancy is a hot plus cold dark matter
cosmology which involves neutrinos with a mass of a few eV.
In summary, the simplest majoron model which implied large cou-
plings to neutrinos (the Gelmini-Roncadelli model) is experimentally
excluded although one can construct more complicated ones which re-
tain sizeable neutrino-majoron couplings and yet are compatible with
the Z ◦ decay width. The possibility of such models is entertained be-
cause 2β decay experiments may yet turn up compelling evidence for
majoron decays, and because certain cosmological models of structure
formation may be taken to suggest massive, decaying neutrinos.
15.7.2 Majorons and Stars

Majorons could also have an impact on stellar evolution. Besides in-
teracting with neutrinos, they typically also couple to other fermions,
allowing one to apply the astrophysical bounds on pseudoscalars de-
rived throughout this book to majoron models.
Because of the possibility of fast decays ν → ν % χ the neutrino signal
from distant sources, notably from the Sun or from SN 1987A, would
be affected. Such decays are not likely to be able to explain the so-
lar neutrino problem as discussed in Sect. 10.8. It remains interesting,
however, that the matter-induced νe -ν e energy splitting allows for de-
cays νe → ν e χ (Sect. 6.8). Should a solar ν e flux show up in future
measurements, it could be an indication for such decays.
As for SN neutrinos, the decay of massive νµ ’s or ντ ’s with final-
state ν e ’s could modify the ν e signal observed in a detector, notably
the energy distribution and duration of the observed pulse (Soares and
Wolfenstein 1989; Simpson 1991). The SN 1987A data have not been
analyzed in detail with regard to this possibility, although the observed
signal can be accounted for without invoking such effects. It is not
clear if one could derive significant constraints on majoron models from
SN 1987A on the basis of this decay argument. Aharonov, Avignone,

and Nussinov (1988a) predicted for certain parameters a dramatic in-
crease of the number of observable events from the prompt νe burst.
One interesting SN 1987A limit is based on the interaction of the
pulse of observed ν e ’s with the cosmic majoron background that would
be expected to exist for majorons which thermalized in the early uni-
verse. In order not to deplete the pulse too much by collisions, Kolb
and Turner (1987) found a certain upper limit on the majoron-neutrino
Yukawa coupling. Unfortunately, their bound was based on an in-
correct cross section for the process νχ → νχ. They used a pseu-
doscalar coupling of the form igψ ν γ5 ψν χ rather than a derivative cou-
pling (1/2v) ψ ν γµ γ5 ψν ∂ µ χ where g = mν /v and v is the majoron sym-
metry breaking scale. As discussed in Sect. 14.2.3, in processes which
involve two Nambu-Goldstone bosons attached to one fermion it is
mandatory to use the derivative coupling in order to obtain the correct
interaction rate. The pseudoscalar coupling yields a scattering cross
section (g 4 /64π) s−1 times an expression of order unity which depends
on the neutrino mass and s, the squared CM energy. Put another way,
the cross section is (mν /v)4 s−1 times numerical factors. The derivative
coupling, on the other hand, leads to a cross section m2ν /v 4 times numer-
ical factors (Choi and Santamaria 1990). Therefore, Kolb and Turner’s
(1987) bound translates approximately into g (eV/mν )1/2 < −4
∼ 3×10 . It
is not a bound on g alone.
Typical energies of neutrinos and other particles which prevail in the
interior of a SN core are in the range of tens to hundreds of MeV. In
majoron models with a symmetry breaking scale below this range, the
symmetry may be restored in the interior of the star, and both com-
ponents of the complex Higgs field will be thermally excited. Quick
deleptonization may be achieved by reactions involving the majoron
field, leading to a high-entropy collapse. Substantial majoron emission
may shorten the SN 1987A ν e signal too much, although decays of heav-
ier neutrinos in flight could, perhaps, provide the late events observed
in the detectors. SN scenarios involving majorons were discussed by a
number of authors.96 There is little doubt that majoron models will
have an important impact on SN physics for Yukawa couplings some-
where in the range 10−6 −10−3 . From the available literature, however,
96
Kolb, Tubbs, and Dicus (1982); Dicus, Kolb, and Tubbs (1983); Manohar
(1987); Fuller, Mayle, and Wilson (1988); Aharonov, Avignone, and Nussinov
(1988b, 1989); Choi et al. (1988); Grifols, Massó, and Peris (1988); Konoplich
and Khlopov (1988); Dicus et al. (1989); Berezhiani and Smirnov (1989); Choi and
Santamaria (1990).
564 Chapter 15
the present author has not been able to develop a clear view of the
precise range of parameters that can be ruled out or ruled in by the
SN 1987A neutrino signal.97
15.8 Millicharged Particles

It is commonly assumed that all particles have charges in multiples of
1
3
e (electron charge); notably neutrinos are thought to be electrically
neutral. While gauge invariance and anomaly cancellation constraints
pose limits on the possible charge assignments in the standard model,
electric charge quantization is not entirely assured. Two of the three
neutrino species may carry small charges if one gives up the assumption
that the three fermion families differ only in the mass of their mem-
bers (Takasugi and Tanaka 1992; Babu and Volkas 1992; Foot, Lew,
and Volkas 1993). Moreover, if one allows for charge nonconservation,
all neutrinos could have small charges (Babu and Mohapatra 1990;
Maruno, Takasugi, and Tanaka 1991). Finally, the existence of novel
particles with small electric charges is possible and actually motivated
by certain models involving a “mirror sector” where the mirror symme-
try is slightly broken (Holdom 1986; see also Davidson, Campbell, and
Bailey 1991). Therefore, it is interesting to study the experimental, as-
trophysical, and cosmological bounds on the existence of particles with
small electric charge.98
The most severe constraints obtain for nonstandard charge assign-
ments in the first family of quarks and leptons. The most model-
independent charge limit on νe was derived from the absence of an
anomalous dispersion of the SN 1987A neutrino signal (Sect. 13.3.3)
97
An incomplete list of issues that ought to be considered in a study of majorons
in SNe are the following. If the symmetry is broken within the SN core, a derivative
majoron coupling to neutrinos should be used instead of a pseudoscalar one (Choi
and Santamaria 1990). For light neutrinos, the medium-induced dispersion rela-
tion may dominate the cross section result. Besides the medium-induced processes
ν → νχ, the process χ → νν should be included (Sect. 6.8). The decay of heavy
neutrinos outside of the SN could contribute to the measurable signal. The trapping
of neutrinos and majorons due to reactions with each other should be properly un-
derstood along the lines discussed by Dicus et al. (1989); it is dubious, for example,
that a process like χχ → χχ really contributes to the majoron opacity. The effect
of majorons during the infall phase must be understood, especially the possibility
of early deleptonization.
98
Such studies were performed by Dobroliubov and Ignatiev (1990), Davidson,
Campbell, and Bailey (1991), Babu and Volkas (1992), Mohapatra and Nussinov
(1992), and Davidson and Peskin (1994).
which led to (Barbiellini and Cocconi 1987; Bahcall 1989)
eνe < −17

∼ 3×10 e. (15.7)
If electric charge conservation is assumed to hold in β processes such

as neutron decay, one finds a more restrictive limit of
eνe < −21

∼ 3×10 e. (15.8)
It is based on a limit for the neutron charge of en = (−0.4±1.1)×10−21 e

(Baumann et al. 1988) and on the neutrality of matter which was found
to be ep +ee = (0.8±0.8)×10−21 e assuming a vanishing neutron charge
(Marinelli and Morpurgo 1984).
The deflection of charged neutrinos in the toroidal magnetic field
in the solar convection zone would modify the observable flux at Earth
(Ignatiev and Joshi 1994, 1995). However, in view of the above limits
an unrealistically large field gradient is required to obtain significant
flux modifications.
Babu and Volkas (1992) derived a limit on the νµ electric charge
from the measured νµ e cross section which would receive a contribution
from photon exchange,
eνµ < −9
∼ 10 e. (15.9)
Similar limits could be derived for νe .

The following arguments apply to millicharged neutrinos or novel
particles alike. They would appear as virtual states in higher-order am-
plitudes. For example, they would contribute to the anomalous mag-
netic moment of electrons and muons, and to the Lamb shift between
the 2P1/2 and 2S1/2 states of the hydrogen atom. Of these quantitities,
the Lamb shift gives the most restrictive limit (Davidson, Campbell,
and Bailey 1991),
ex < 0.11e mx /MeV, (15.10)
where ex and mx are the charge and mass of the millicharged particle,
respectively. This result applies to mx >∼ 1 keV.
Davidson, Campbell, and Bailey (1991) have reviewed more restric-
tive bounds from a host of accelerator experiments (Fig. 15.3).
A simple astrophysical constraint is based on avoiding excessive en-
ergy losses of stars which can produce millicharged particles by various
reactions, most notably the plasma decay process. To avoid an unac-
ceptable delay of helium ignition in low-mass red giants, and to avoid an
566 Chapter 15
Fig. 15.3. Summary on limits on the electric charge ex and mass mx of

generic millicharged particles which may be sequential neutrinos or novel
particles. (Adapted from Davidson, Bailey, and Campbell 1991.) In order
to avoid overclosing the universe, additional model-dependent parameter
regions are excluded. The big-bang nucleosynthesis (BBN) excluded region
is larger in some models.
undue shortening of the lifetime of horizontal-branch stars, one needs

to require (Sect. 6.5.6)
ex < −14
∼ 2×10 e. (15.11)
It applies for mx <∼ ωP /2 with ωP the plasma frequency. It is larger for
red giants before helium ignition than for HB stars because of the larger
average density of 2×105 g cm−3 which corresponds to ωP ≈ 8.6 keV.
The white-dwarf luminosity function yields about the same limit.
If ex exceeds around 10−8 e the mean free path of the millicharged
particles will be less than the physical size of a white dwarf or red-
giant core. For larger ex the particles will be trapped and contribute
to the transfer of energy. Their impact on stellar evolution will become
negligible when ex is so large that other forms of energy transfer (photon
radiation, convection) are more important. Because the new particles
act essentially as radiation their mean free path must be less than that
of photons, or very crudely, their charge must be of order an electron
charge. (The main opacity source is probably Coulomb scattering on
charged particles.) Therefore, if their mass is below a few keV, even the
properties of the Sun would imply that there is not an allowed range of
large ex on the trapping side of the red-giant argument, except perhaps
for ex so large that it is excluded by experimental arguments.
For a narrow range of charges, these limits can be extended to larger

masses by the SN 1987A cooling argument (Sect. 13.8.4). Accordingly,
10−9 e < < −7
∼ ex ∼ 10 e (15.12)
is excluded for mx up to several MeV, perhaps up to 10 MeV. On
the trapping side of this range (large ex ) these particles surely would
have an important impact on SN physics even though they cannot be
excluded on the basis of the simple cooling argument.
If millicharged particles reach thermal equilibrium in the early uni-
verse before nucleosynthesis they contribute to the energy density and
thus to the expansion rate. If they are one of the sequential neutrinos,
this means that the right-handed degrees of freedom of that species
are excited (they must be Dirac particles!), adding an effective neu-
trino degree of freedom. If they are nonneutrinos, even more energy
is contributed, depending on their spin degrees of freedom. Even one
additional effective neutrino degree of freedom is excluded and so for
any millicharged particle Davidson, Campbell, and Bailey (1991) found
ex <
∼ 3×10 e
−9
(15.13)
if mx <∼ 1 MeV. In certain models where the millicharged particles are
associated with a shadow sector, more stringent limits apply (Davidson
and Peskin 1994).
Additional regions in the mass-charge plane can be excluded by
the requirement that the novel objects do not overclose the universe.
However, these arguments depend on the annihilation cross section in
the early universe so that one needs to know all of their interactions
apart from the millicharge. It is hard to imagine novel particles which
interact only by their small electric charge! For certain specific cases
the excluded regime was derived by Davidson, Campbell, and Bailey
(1991) and Davidson and Peskin (1994).
All of these constraints leave the possibility open that ντ ’s have a
mass in the 1−24 MeV range and a charge in the 10−5 −10−3 e range.
Then they would annihilate sufficiently fast before nucleosynthesis to
actually reduce their effective contribution to the expansion rate (Foot
and Lew 1993). In the standard model with small neutrino charges,
however, two sequential neutrino species must carry a millicharge of
equal but opposite magnitude (Babu and Volkas 1992; Takasugi and
Tanaka 1992). Because the large charges required for Foot and Lew’s
scenario are excluded for νe and νµ one would need to require that only
ντ carries a relatively large charge, forcing one to espouse even more
exotic particle-physics models.
Chapter 16
Neutrinos: The Bottom Line
Most of the particle-physics arguments discussed in this book are closely

related to neutrino physics because these particles play an important
role in stellar evolution whether or not they have nonstandard prop-
erties. Besides a summary of some recent developments of standard-
neutrino astrophysics, a synthesis is attempted of what stars as neutrino
laboratories have taught us about these elusive objects, and what one
might reasonable hope to learn in the foreseeable future.
16.1 Standard Neutrinos

The main theme of this book has been an attempt to extract infor-
mation about the properties of neutrinos and other weakly interact-
ing particles from the established properties of stars. However, even
standard-model neutrinos (massless, no mixing, no exotic properties)
play a significant role in astrophysics. There have been some recent de-
velopments in “standard-neutrino astrophysics” which deserve mention
in a summary.
It is now thought that neutrinos play an active role in supernovae
besides carrying away the binding energy of the newborn neutron star
(Chapter 11). In the delayed-explosion scenario they are crucial to
revive the stalled shock wave which is supposed to expel the stellar
mantle and envelope. Moreover, they have a strong impact on r-process
nucleosynthesis which is thought to occur in the high-entropy region
above the neutron star a few seconds after collapse. For both purposes it
is crucial to calculate the SN “neutrino lightcurve” for the first seconds
after collapse. Convection below the neutrino sphere and large-scale
convective turnovers in the region between the neutron star and the
shock wave are both important and need to be understood better on
568
Neutrinos: The Bottom Line 569
the basis of 2- and 3-dimensional hydrodynamic calculations which are

only beginning to appear in the literature.
In addition, however, the neutrino opacities must be calculated with
greater reliability. In Chapter 4 the weaknesses of a naive calculation
of the axial-vector opacities have been amply demonstrated. In my
opinion, a far better understanding of the interaction rates of neutrinos
with a hot nuclear medium is required before one can calculate a SN
neutrino lightcurve (notably the duration of Kelvin-Helmholtz cooling)
with a reasonable precision.
Neutrinos presumably play a key role for the self-acceleration of neu-
tron stars which are observed to have huge “kick velocities” (Sect. 11.5).
The required anisotropic neutrino emission of 1−2% may be caused
by temperature fluctuations on the neutrino sphere (convection!), or
by magnetic-field induced anisotropies of the neutrino opacities. Of
course, other phenomena may be responsible for anisotropic neutrino
emission and for the kick velocities.
For the first time ever neutrinos have been observed from a collaps-
ing star (SN 1987A), confirming the expected signal behavior within
the large uncertainties caused by the small number of observed events
(Chapter 11).
Neutrinos are routinely observed from the Sun—currently in four
different detectors with three different spectral response characteristics
(Chapter 10). At least three further detectors will soon take up opera-
tions. The measured solar neutrino spectrum differs significantly from
theoretical predictions. This “solar neutrino problem” has no obvious
“astrophysical solution” that would involve plausible variations of in-
put parameters such as nuclear cross sections or photon opacities. Still,
the gallium detectors SAGE and GALLEX have for the first time mea-
sured the dominant low-energy flux of pp neutrinos, confirming that the
Sun cannot be completely different at its center from what had been
thought.
There is also some more benign news. The interest of some authors
to apply the methods of finite temperature field theory to astrophysical
problems has led to a new formulation of the photon dispersion relation
in a plasma (Sect. 6.3). It helped to correct some “fossilized errors” in
the literature on plasma neutrino emission (see also Appendix C.1). It
may well be worthwhile to scrutinize other standard aspects of stellar-
evolution input physics that involve subtle dispersion or screening ef-
fects.
570 Chapter 16
16.2 Minimally Extended Standard Model
16.2.1 Cosmological Mass Limit for All Flavors
Neutrinos with nonstandard properties would have more radical impli-

cations in astrophysics. A minimal and most plausible extension of the
standard model is the possibility that they have masses and mixings
like the other fermions. Taking this hypothesis in a literal sense means
that neutrinos would need to have Dirac masses like the charged fer-
mions. Therefore, one needs to postulate the existence of right-handed
neutrinos and Yukawa couplings to the standard Higgs field to generate
masses and mixings. What do we know about neutrinos in the context
of this Minimally Extended Standard Model?
Perhaps the most dramatic lesson is that the mass of all sequen-
tial neutrinos (νe , νµ , ντ ) must be less than the cosmological limit of
approximately 30 eV (Sect. 7.1.5). This conclusion is not to be taken
for granted as massive neutrinos with mixings can decay by virtue of
ν → ν ! γ, and by ν → νe e+ e− if mν exceeds about 2me . While the
standard-model radiative decays are too slow to avoid the cosmological
limit, the e+ e− channel can be fast on cosmological time scales if the
mixing angle is not too small. This decay channel is an option only
for ντ which may have a mass of up to 24 MeV while the experimental
mass limits on the other flavors are below 0.16 MeV.
Such a “heavy” ντ , however, can be excluded by several arguments.
The stellar evolution one based on SN arguments was presented in
Sect. 12.5.2. The bremsstrahlung emission of photons in ντ → νe e+ e− γ
would produce a γ-ray flux in excess of the SMM limits for SN 1987A
unless sin2 2θe3 < −9
∼ 10 . (For a detailed dependence of this limit on the
assumed mass see Fig. 12.18.) In addition, the integrated positron flux
from all galactic supernovae over the past, say, 100,000 years would
exceed the observed value unless the decays are very fast (near the SN)
or very slow (outside of the galactic disk). Together, these limits leave
no room for a heavy ντ (Fig. 12.19). In addition, the mass range
0.5 MeV < <
∼ mν ∼ 35 MeV can be excluded on the basis of big bang
nucleosynthesis arguments (Fig. 7.2) unless the neutrinos are shorter
lived than permitted by the Minimally Extended Standard Model.
Neutrino masses near the cosmological limit would be important for
cosmology as they could contribute some or all of the dark matter of
the universe. The latter option is disfavored by theories of structure
formation. However, a subdominant “hot dark matter” contribution in
the form of, say, 5 eV neutrinos might be cosmologically quite welcome.
Neutrino masses in this general range are also accessible to time-

of-flight measurements. The neutrino burst of SN 1987A has already
provided a limit of mνe <
∼ 20 eV (Sect. 11.3.4), a result which remains
interesting in view of the confusing situation with the tritium β de-
cay endpoint experiments which seem to be plagued by systematic ef-
fects which cause the appearance of a negative neutrino mass-square
(Sect. 7.1.3). The observation of the prompt νe burst from a future
galactic SN would allow one to reduce this limit to a few eV. More-
over, one may well be able to detect or constrain a νµ or ντ mass in
the 10−20 eV range, assuming the simultaneous operation of a water
Cherenkov detector such as Superkamiokande and a neutral-current de-
tector such as the proposed Supernova Burst Observatory (Sect. 11.6).
16.2.2 Oscillations of Solar Neutrinos
Small neutrino masses or rather, small neutrino mass differences can

have dramatic consequences if neutrinos also mix; this is expected in the
present scenario. Neutrino oscillations then lead to the possibility that
a different neutrino flavor is measured in a detector than was produced
in the source. The observed characteristics of the solar neutrino flux
strongly suggest that neutrino oscillations may in fact be occurring. In
terms of the mass difference and mixing angle between νe and νµ or ντ
there remain three solutions which account for the presently available
data from the chlorine (Homestake), gallium (GALLEX and SAGE),
and Cherenkov (Kamiokande) experiments (Tab. 16.1). In view of pos-
sible systematic uncertainties concerning such quantities as the solar
opacities and the p 7 Be cross section the values of the favored mixing
angles can be somewhat different from those shown in Tab. 16.1. How-
ever, the required mass differences remain rather stable against large
nonstandard modifications of the solar model.
Table 16.1. Approximate neutrino parameters which explain all current solar
neutrino observations in the framework of standard solar model assumptions.
Solution ∆m2 [eV2 ] sin2 2θ

Large-angle MSW 2×10−5 0.6
Nonadiabatic MSW 0.6×10−5 0.006
Vacuum oscillations 0.8×10−10 0.8−1
572 Chapter 16
The main aspect of the current situation is that there does not seem
to be a simple “astrophysical solution” to reconcile the solar source
spectrum with the measured fluxes in experiments with three different
spectral response characteristics. Even allowing for large modifications
of the p 7 Be cross section or the solar central temperature does not yield
consistency unless one stretches the experimental uncertainties of the
flux measurements beyond reasonable limits. Still, a final verdict on
the question of solar neutrino oscillations can be expected only from
the near-future experiments Superkamiokande, SNO, and BOREXINO
as discussed in Chapter 10.
16.2.3 Oscillation of Supernova Neutrinos

Naturally, neutrino oscillations would also affect the characteristics of
SN neutrinos which have been observed only from SN 1987A, although
it is not unrealistic to hope for the observation of a galactic supernova
at Superkamiokande or SNO within, say, a decade of operation. In wa-
ter Cherenkov detectors, the main signal of SN neutrinos is thought to
be from the ν e p → ne+ reaction, although the angular characteristics of
the SN 1987A observations do not square well with this assumption ex-
cept that there is no convincing alternate interpretation (Sect. 11.3.5).
The MSW solution in the Sun requires a “normal” mass hierarchy with
νe being dominated by the smaller mass eigenstate. In this case reso-
nant oscillations do not occur among the ν’s.
Still, for large-angle vacuum oscillations such as those corresponding
to the solar vacuum solution, the signal in the IMB and Kamiokande
detectors would have been “hardened” by the partial swap of, say, the
ν e with the ν µ spectrum. It is not entirely obvious from the current
literature if this effect is ruled in or ruled out by the SN 1987A obser-
vations (Sect. 11.4.3).
If the mixing angle of νe with both νµ and ντ is small (sin2 2θ <
∼ 0.1)
there is no impact on the ν e SN signal. Still, the prompt νe burst could
be affected even for small mixing angles because of the possibility of
resonant oscillations (Sect. 11.4.2) and so the observation of a future
galactic supernova could serve to measure this effect.
If one were to contemplate more general neutrino mass matrices
with an “inversion” so that νe is not dominated by the lowest-mass
eigenstate, there could be resonant oscillations in the ν sector which
would then lead to dramatic modifications of the SN signal in a large
range of masses and mixing angles. This possibility has not been ex-
plored much in the literature.
Apart from the detector signal, the oscillation of SN neutrinos can

have important implications for SN physics itself because of the swap
of, say, the νe with the ντ spectrum which is much harder. For a mass
difference corresponding to the cosmologically interesting range of a few
to a few tens of eV, resonant oscillations could occur so close to the neu-
trino sphere that the “crossover” point is within the stalling shock wave
in the delayed-explosion scenario. The effective hardening of the νe
spectrum would then enhance the neutrino energy transfer to the shock
wave, thus helping to explode supernovae (Sect. 11.4.4). Conversely, a
few seconds after collapse the same effect would drive the hot wind
proton rich which is driven from the surface of the compact remnant.
This effect would prevent the occurrence of r-process nucleosynthesis
which requires a neutron-rich environment (Sect. 11.4.5). Interestingly,
because the remnant is more compact at “late” times (few seconds after
collapse), the adiabaticity condition can be met only for relatively large
mixing angles. Thus, there is a plausible range of neutrino parameters
where oscillations may help to explode supernovae, and still r-process
nucleosynthesis may proceed undisturbed (Fig. 11.20). At any rate, it
is impossible to ignore neutrino oscillations for SN physics if neutrino
masses happen to lie in the cosmologically interesting range.
16.2.4 Electromagnetic Properties

In the Minimally Extended Standard Model neutrinos have magnetic
and electric diagonal and transition moments which are proportional to
their assumed masses (Sect. 7.2.2). Because of the cosmological mass
limit these quantities are so small that they do not seem to be important
anywhere. However, one may toy with the idea that neutrinos actually
carry small electric charges, a possibility that is not entirely excluded
by the structure of the Standard Model if one gives up the notion that
the second and third particle families are exact replicas of the first
except for the masses (Sect. 15.8).
The possible magnitude of a νe charge is limited by eνe < ∼ 3×10 e
−17
from the absence of an anomalous dispersion of the SN 1987A neu-

trino burst (Sect. 13.3.3). All neutrino charges are limited by eν < ∼
2×10−14 e from the absence of anomalous cooling of globular-cluster
stars (Sect. 6.5.6). For νµ and ντ the cosmological mass limit is crucial
for this bound because their emission would be suppressed by threshold
effects if their mass exceeded the relevant plasma frequency of a few
keV. The assumption of charge conservation in β decay yields a more
restrictive limit eνe <
∼ 3×10
−21
(Sect. 15.8).
574 Chapter 16
Either way, possible charges of all neutrinos must be so small that

the quantization of charge is very accurately realized among the fermi-
ons of the Standard Model. Therefore, charge quantization is probably
exact so that the neutrino electric charges vanish exactly, those of the
quarks are exactly 13 e and 23 e, respectively.
16.3 New Interactions

16.3.1 Majorana Masses
Because of the cosmological limit all neutrino masses are found to be so
small relative to those of the corresponding charged fermions (Fig. 7.1)
that it is hard to maintain the pretense that neutrinos are essentially
like the other fermions. In this sense the assumption of a Minimally
Extended Standard Model is self-defeating. One reaction may be to
return to the assumption of massless two-component neutrinos. It re-
mains to be seen for how much longer this option remains viable in view
of the expected progress in solar neutrino astronomy and, perhaps, in
laboratory experiments (including atmospheric neutrino observations).
They may soon yield unrefutable evidence for neutrino oscillations.
Another reaction is to embrace the notion of neutrinos being very
different from the charged leptons with a possible wealth of novel and
unexpected properties. The most benign assumption is to maintain the
notion of only two neutrino components per family which are charac-
terized by a Majorana mass term which may arise by new physics at
some large energy scale. In this case all that was said about neutrino
masses and oscillations in the previous section remains applicable.
16.3.2 “Heavy” Neutrinos and Fast Decays

If one postulates novel neutrino interactions one may speculate about
the possibility of neutrino decays which proceed faster than in the Min-
imally Extended Standard Model. Besides accelerated radiative decays
one may speculate about “fast invisible decays” of the form ν → ν ! ν !! ν !!!
or ν → ν ! χ where χ is some new boson such as the majoron (Sect. 15.7).
In this case one can escape the cosmological mass limit. In fact, such
fast-decaying neutrinos can be a welcome feature of theories for the
formation of structure in the universe (Sect. 7.1.5).
“Heavy” νµ ’s or ντ ’s are not in obvious conflict with normal stellar
evolution even though their emission could now be suppressed by the
mass threshold. Thermal production of neutrinos from a stellar plasma
is dominated by the plasmon decay process γ → νν for a large range

of stellar conditions (Appendix C). This process occurs mostly by the
vector-current coupling to electrons of the medium. Because the weak
mixing angle has the special value sin2 ΘW = 0.23 ≈ 14 the vector-
current coupling of νµ and ντ to electrons nearly vanishes (Appendix B)
and so γ → ν µ νµ and ν τ ντ plays near to no role anyway—in this sense
stars are “blind” to the issue of heavy νµ or ντ masses.
However, assuming that the heavy neutrinos are Dirac particles one
can limit their mass by a SN argument: The energy loss by right-handed
states produced in spin-flip collisions must not be too large, yielding a
limit of mν < ∼ 30 keV (Sect. 13.8.1). In addition, these neutrinos have
very high energies, typical of SN core temperatures, and so their decays
might produce high-energy daughter ν e ’s (100−200 MeV) which were
not observed from SN 1987A (Sect. 13.8.1).
Even Majorana neutrinos would not be harmless for SN physics if
they would mix sufficiently strongly with νe . In this case they would
effectively participate in β equilibrium (Sect. 9.5) so that the spin-
flip scattering of, say, ντ would effectively lead to ν τ ’s and thus to
deleptonization without the need of transporting lepton number to the
stellar surface.
In addition, even though heavy Majorana neutrinos would be emit-
ted from the neutrino sphere so that their decays would not produce
high-energy daughter products, the decays could still add to the de-
tectable signal. It is not obvious from the existing literature which
(if any) range of masses and decay times is ruled out or ruled in by the
SN 1987A observations (Sect. 13.2.2).
If the fast decays were due to some sort of majoron model, large
“secret” neutrino-neutrino interactions are conceivable, possibly in con-
junction with a small vacuum expectation value of a new Higgs field so
that the symmetry may be restored in a SN core. While a substantial
body of literature exists on this sort of scenario (Sect. 15.7.2) I believe
that in this context the story of SN physics would have to be rewritten
more systematically than has been done so far. However, there appears
to be little doubt that for neutrino-majoron Yukawa couplings in excess
of about 10−5 one would expect dramatic modifications of the transport
of energy and lepton number.
In summary, “heavy” neutrinos with fast invisible decays are a way
to circumvent the cosmological mass limit, and may indeed be desirable
in certain scenarios of cosmic structure formation. Depending on their
detailed properties they could have a substantial impact on SN physics
and the signal observable in a detector. It is difficult, however, to
576 Chapter 16
state general constraints as one may easily postulate, for example, that
the decays do not involve final-state ν e ’s, that even “wrong-helicity”
Dirac neutrinos are trapped in a SN core by novel interactions, or that
heavy Majorana ντ ’s have only negligible mixings with νe . Surely other
loopholes could be found.
16.3.3 Electromagnetic Properties
a) Spin and Spin-Flavor Oscillations
If one contemplates neutrino interactions beyond the Minimally Ex-

tended Standard Model, neutrino dipole and transition moments no
longer need to be small. In particular, they do not need to be pro-
portional to the neutrino masses; one example are left-right symmetric
models where even massless neutrinos would have large dipole moments
(Sect. 7.3.1). Dipole and transition moments can lead to spin or spin-
flavor oscillations in external magnetic fields, they allow for spin-flip
scattering on charged particles, for the plasmon decay γ → νν in stars,
and for radiative decays ν → ν ! γ.
One motivation for studying neutrino dipole moments is the ap-
parent flux variability of the solar neutrino signal in the Homestake
detector. It anticorrelates with solar magnetic activity too closely to
blame it comfortably on a statistical fluke (Sect. 10.4.3). The only
physical explanation put forth to date is that of Voloshin, Vysotskiı̆,
and Okun of a partial depletion of left-handed (measurable) neutrinos
by spin or spin-flavor oscillations. Unfortunately, the required value for
µν B (neutrino dipole moment µν , magnetic field B in the solar convec-
tion zone) exceeds by about two orders of magnitude what is allowed by
typical models of the solar magnetic field and by limits on µν . There-
fore, this scenario appears to be in big trouble. Still, if one ignores the
µν limits or speculates about large convection-zone magnetic fields one
may fit all currently available solar neutrino data (Sect. 10.7).
Spin or spin-flavor oscillations can be very important in and near
the cores of supernovae where fields of order 1012 G exist, and perhaps
pockets with much larger fields. If neutrinos are Dirac particles so that
their spin-flipped (right-handed) states are sterile, the combination of
spin-flip scattering on charged particles and the magnetic spin oscilla-
tion in large-scale magnetic fields can lead to nonlocal modes of energy
transfer where energy can be deposited in one region that was depleted
from a distant other region. Thus, energy transfer could no longer be
treated with simple differential equations which involve local gradients
of temperature and lepton number. (Of course, a nonlocal energy trans-

fer mechanism is already thought to be important for reviving the shock
wave in the delayed-explosion scenario.) As far as I know, nothing more
quantitative than back-of-the-envelope estimates of this scenario exist
in the literature. Therefore, alleged bounds of order 10−12 µB (Bohr
magneton µB = e/2me ) on Dirac-neutrino dipole moments probably
have to be used with some reservation (Sect. 13.8.3). Conversely, such
dipole moments may actually help to explode supernovae.
With regard to the observable neutrino signal, spin and spin-flavor
oscillations both in the SN and in the galactic magnetic field may cause
vast modifications of the ν e fluxes and spectra observable in water Che-
renkov detectors if neutrinos have dipole moments in the ballpark of
10−12 −10−14 µB .
Spin and spin-flavor oscillations can be very important in the early
universe where strong magnetic fields may exist, and where a popu-
lation of the r.h. degrees of freedom would accelerate the expansion
rate of the universe. These issues are being investigated in the current
literature; final conclusions do not seem to be available at the present
time. Still, it appears that this effect may well be the most significant
impact of small Dirac neutrino dipole or transition moments anywhere
in nature.
b) Laboratory Limits
Less problematic bounds on neutrino dipole moments arise from labo-
ratory experiments where one studies the recoil spectrum of electrons
in the reaction ν + e → e + ν ! where ν ! can be the same or a different
flavor (Sect. 7.5.1). A sensitivity down to, perhaps, as low as 10−11 µB
can be expected from a current effort involving reactor neutrinos as
a source (MUNU experiment). Current limits on dipole or transition
moments are about 2×10−10 µB if νe is involved, and about 7×10−10 µB
if νµ is involved. For transition moments, these limits are subject to
the assumption that there is no cancellation between a magnetic and
an electric dipole scattering amplitude.
c) Huge ντ Dipole Moments or Millicharges

In principle, the possibility of a large diagonal moment for ντ remains
open as it has not been possible to produce a strong ντ source in the lab-
oratory so that only extremely crude limits exist on the ντ -e-scattering
cross section. The globular-cluster µν bounds discussed below do not
578 Chapter 16
apply to a “heavy” ντ so that one is confronted with a nontrivial al-

lowed region in µντ -mντ space where even MeV masses become cos-
mologically allowed because of the dipole-induced annihilation process
ν τ ντ → e+ e− (µντ in the ballpark of 10−7 µB ). Of course, such large
dipole moments must be caused by a fairly nontrivial arrangement of
intermediate charged states and so one may wonder if a large ντ mag-
netic moment could be realistically the only manifestation of these new
particles and/or interactions.
Still, a large ντ dipole moment and the correspondingly large an-
nihilation cross section in the early universe is one possibility to toler-
ate a large ντ mass without the need for fast decays—such a particle
could be entirely stable. Another similar possibility is that ντ has a
“huge millicharge” in the neighborhood of 10−5 −10−3 e. Such a scheme
would require the violation of charge conservation as the possibility of
neutrino charges within a simple extension of the Standard Model dis-
cussed above always gives charges to two neutrino flavors; for νe or νµ
the required value is not tolerable (Sect. 15.8).
For the issues of stellar evolution, the only conceivable consequence
of such large ντ electromagnetic interaction cross sections would be
a reduced contribution to the energy transfer in SNe because of the
reduced mean free path. In the study discussed in Sect. 13.6 one should
have included the possibility of only two effective flavors! Still, there
is little doubt that large ντ cross sections could be accommodated in
what one knows about SNe today.
d) Astrophysical Bounds on Dipole and Transition Moments

For all neutrinos with a mass below a few keV a very restrictive limit
on dipole or transition magnetic or electric moments arises from the ab-
sence of anomalous neutrino emission from the cores of evolved globular-
cluster stars, notably of red-giant cores just before helium ignition
(Sect. 6.5.6). One finds a limit µν < −12
∼ 3×10 µB which applies to Dirac
and Majorana neutrinos, and which does not allow for a destructive
interference between electric and magnetic amplitudes.
Neutrino transition moments would reveal themselves by radiative
decays. Because the decay rate involves a phase-space factor m3ν this
method is suitable only for large masses. In the cosmologically allowed
range with mν < ∼ 30 eV, the only radiative limit which can compete
with the globular-cluster bound is from the cosmic diffuse background
radiations (Fig. 12.21). In fact, Sciama has proposed a scheme where a
28.9 eV neutrino with a radiative decay time corresponding to a tran-
sition moment of 0.6×10−14 µB plays a significant cosmological role

(Sect. 12.7.1). While this scenario is probably excluded it highlights
the possibility of interesting cosmological effects for radiatively decay-
ing neutrinos in the range allowed by the globular-cluster bound.
Radiative decay limits which are based on astronomical decay paths
suffer from the uncertainty of other invisible decay channels which may
compete with the radiative mode. Limits based on the cosmic back-
ground radiations (Fig. 12.20) imply that in a large range of cosmolog-
ically allowed neutrino masses and lifetimes the dominant decay chan-
nel must be nonradiative. Therefore, one should use the cosmic back-
ground radiations to derive limits on the branching ratio. One could
then construct a contour plot of the excluded transition moments in
the mν -τν plane.
From the SN 1987A radiative lifetime limits I have constructed
such a plot in Fig. 12.17. For large dipole moments in excess of,
say, 10−10 µB these bounds are not self-consistent because neutrinos
would be trapped too strongly by electromagnetic scatterings. How-
ever, the laboratory limits exclude large transition moments. More-
over, the globular-cluster bound yields more restrictive limits if mν is
less than a few keV. However, for relatively large neutrino masses, and
for lifetimes not so short that the decays would have occurred within
the progenitor star, SN 1987A yields the most restrictive limits on tran-
sition moments. Naturally, one must assume that νµ or ντ were actually
emitted with about the standard fluxes. Trapping effects by additional
new interactions could circumvent this assumption.
16.3.4 Summary
Neutrinos with nonstandard interactions may well saturate the experi-
mental mass limits, and may have a variety of novel properties. How-
ever, it is nearly impossible to derive generic constraints on quanti-
ties like magnetic transition moments without specifying an underlying
particle physics model. Many constraints, notably those related to
SN 1987A or to cosmology, can be circumvented by postulating suf-
ficiently bizarre neutrino properties. Therefore, it is probably more
important to know the arguments that can serve to learn something
about neutrinos in astrophysics than it is to know a list of alleged lim-
its. If a concrete conjecture turns up, or a specific theoretical model
needs to be constrained, one can easily go through the list of arguments
and check if they apply or not. Perhaps this book can be of help at
this task.
Appendix A
Units and Dimensions
In the astrophysical context, frequently occurring units of length are

centimeters, (light) seconds, light years, and parsecs. Conversion fac-
tors are given in Tab. A.1. For example, 1 pc = 3.26 ly.
Using both centimeters and (light) seconds as units of length implies
a system of units where the speed of light c is dimensionless and equal to
unity. In this book I always use natural units where Planck’s constant h̄
and Boltzmann’s constant kB are also dimensionless and equal to unity.
This implies that (length)−1 , (time)−1 , mass, energy, and temperature
can all be measured in the same unit by virtue of x = ct, E = mc2 ,
E = h̄ω, ω = 2π/t, and E = kB T . In Tab. A.2 conversion factors are
given. For example, 1 K = 0.862×10−4 eV or 1 erg = 0.948×1027 s−1 .
The most confusing aspect of natural units is that of an electromag-
netic field strength. The square of a field strength is an energy density
(erg/cm3 ) which, in natural units, is (energy)4 or (length)−4 . Thus,
an electric or magnetic field may be measured, for example, in eV2 or
cm−2 . In natural units, electric charges are dimensionless numbers.
However, there is a general ambiguity in the definition of charges
and field strengths because only their product (a force on a charged
particle) is operationally defined. All physical quantities stay the same
if the charges are multiplied with an arbitrary number and the field
strengths are divided by it. However, the fine-structure constant α ≈
1/137 is dimensionless in all systems of units, and its value does not
depend on this arbitrary choice. If e is the charge of the electron one has
α = e2 /4π in the rationalized system of (natural) units which is always
used in modern works on field theory, and is used throughout this book.
The energy density of an electromagnetic field is then 12 (E 2 + B 2 ). In
the older literature and some texts on electromagnetism, unrationalized
units are used where α = e2 and the energy density is (E 2 + B 2 )/8π.
580
Units and Dimensions 581
In the astrophysical literature the cgs system of units is very popular

where magnetic fields are measured in Gauss (G). Confusingly, this
system happens to be an unrationalized one. Field strengths given in
Gauss can be translated into our rationalized natural units by virtue of
!
1 erg/cm3
1G → = 1.953×10−2 eV2 = 0.502×108 cm−2 , (A.1)
4π
where I have converted erg and cm−1 into eV according to Tab. A.2.
The energy density of a magnetic field of strength 1 G is, therefore,
1
2
(1.953×10−2 eV2 )2 = 1.908×10−4 eV4 = 3.979×10−2 erg cm−3 =
(1/8π) erg cm−3 . For a further discussion of electromagnetic units see
Jackson (1975).
It is sometimes useful to measure very strong magnetic fields in
terms of a critical field strength Bcrit which is defined by the condition
that the quantum energy corresponding to the classical cyclotron fre-
quency h̄ (eB/me c) of an electron equals its rest energy me c2 so that in
natural units
Bcrit = m2e /e. (A.2)
Note that the Lorentz force on an electron in this field is proportional

to eBcrit so that the electron charge cancels. Hence, Eq. (A.2) is the
same in a rationalized
√ or unrationalized system of units. In 2our ratio-
nalized units e = 4πα = 0.303 so that Bcrit = (0.511 MeV) /0.303 =
0.862×1012 eV2 which, with Eq. (A.1), corresponds to 4.413×1013 G, in
accordance to what is found in the literature (Mészáros 1992).
Magnetic dipole moments of electrons and neutrinos are usually
discussed in terms of Bohr magnetons µB ≡ e/2me . For particle elec-
tric dipole moments, on the other hand, one commonly uses 1 e cm as
a unit. The conversion is achieved by 1 e cm = (2me cm) (e/2me ) =
5.18×1010 µB .
Tab. A.1. Conversion factors between different units of length.
582
cm s ly pc
cm 1 0.334×10−10 1.06×10−18 0.325×10−18
s 2.998×1010 1 0.317×10−7 0.973×10−8
ly 0.946×1018 3.156×107 1 0.307
pc 3.08×1018 1.028×108 3.26 1
Tab. A.2. Conversion factors in the system of natural units.
s−1 cm−1 K eV amua erg g

s−1 1 0.334×10−10 0.764×10−11 0.658×10−15 0.707×10−24 1.055×10−27 1.173×10−48
cm−1 2.998×1010 1 0.2289 1.973×10−5 2.118×10−14 3.161×10−17 0.352×10−37
K 1.310×1011 4.369 1 0.862×10−4 0.926×10−13 1.381×10−16 1.537×10−37
eV 1.519×1015 0.507×105 1.160×104 1 1.074×10−9 1.602×10−12 1.783×10−33
amu 1.415×1024 0.472×1014 1.081×1013 0.931×109 1 1.492×10−3 1.661×10−24
erg 0.948×1027 0.316×1017 0.724×1016 0.624×1012 0.670×103 1 1.113×10−21
g 0.852×1048 2.843×1037 0.651×1037 0.561×1033 0.602×1024 0.899×1021 1
Appendix A
a
Atomic mass unit.
Appendix B
Neutrino Coupling Constants
Neutrinos can interact with other fermions and with each other by
the exchange of W or Z bosons. Because the astrophysical phenomena
relevant for this book take place at very low energies compared with the
W or Z mass, one may always use an effective four-fermion coupling
which is parametrized in terms of the Fermi constant and the weak
mixing angle
GF = 1.166×10−5 GeV−2 ,
sin2 ΘW = 0.2325 ± 0.0008. (B.1)
The tree-level relationship of these quantities with the gauge-boson

masses is
√ πα πα
2 GF = 2 2 = 2 2 , (B.2)
mW sin ΘW mZ sin ΘW cos2 ΘW
where mZ = 91.2 GeV and mW = 80.2 GeV.

The effective charged-current interaction between nucleons and lep-
tons is written in the form
GF
Hint = √ ψ p γµ (CV − CA γ5 )ψn ψ ! γ µ (1 − γ5 )ψν! , (B.3)
2
where the ψj are the proton, neutron, charged-lepton, and the corre-
sponding neutrino field. The vector-current coupling constant is CV = 1
while the axial-vector coupling for free nucleons is CA = 1.26. However,
in large nuclei this value is suppressed somewhat, and the commonly
used value for nuclear matter is CA = 1.0 (e.g. Castle and Towner 1990).
This quantity would be relevant, for example, for β reactions in super-
nova cores and neutron stars.
583
584 Appendix B
The effective charged-current interaction between charged leptons

and their own neutrinos, e.g. between e− and νe , is written in the same
form with CV = CA = 1. By virtue of a Fierz transformation it is
brought into the form of a neutral current (see below).
Neutral-current interactions between a neutrino ν and a fermion f
are written in the form
GF
Hint = √ ψ f γµ (CV − CA γ5 )ψf ψ ν γ µ (1 − γ5 )ψν . (B.4)
2
If f is the charged lepton corresponding to ν, there is a contribution
with CV = CA = 1 from a Fierz-transformed charged current. The
compound effective CV ’s and CA ’s for various combinations of f and
ν are given in Tab. B.1. (Note that the |CV,A | for neutral currents are
typically 12 , a factor which is sometimes pulled out front so that the
√
global coefficient is GF /2 2 while the couplings are then twice those of
Tab. B.1.) For neutrinos interacting with neutrinos of the same flavor a
factor 2 for an exchange amplitude for identical fermions was included.
The CA ’s for nucleons were thought to be given by isospin invariance
to be ±1.26/2. However, because of the strange-quark contribution to
the nucleon spin there is an isoscalar piece as well giving rise to the
values shown in Tab. B.1—for a discussion and references to the original
literature see Raffelt and Seckel (1995). Moreover, in a nuclear medium
a certain suppression is expected to occur. In analogy to the charged-
current couplings Raffelt and Seckel (1995) suggested the values CAp ≈
1.09/2 and CAn ≈ −0.91/2.
Table B.1. Neutral-current couplings for the effective Hamiltonian Eq. (B.4)
in vacuum.
Fermion f Neutrino CV CA CV2 CA2

Electron νe + 2 + 2 sin2 ΘW
1
+ 12 0.9312 0.25
νµ,τ − 12 + 2 sin2 ΘW − 12 0.0012 0.25
Proton νe,µ,τ + 12 − 2 sin2 ΘW +1.37/2 0.0012 0.47
Neutron νe,µ,τ − 12 −1.15/2 0.25 0.33
Neutrino (νa ) νa +1 +1 1 1
νb"=a + 12 +2 1
0.25 0.25
Appendix C
Numerical Neutrino
Energy-Loss Rates
In normal stars with densities below nuclear there are four main reac-
tions that contribute to the energy loss by neutrino emission:
γpl → νν Plasma process,
γ e− → e− νν Photoneutrino process,
(C.1)
e+ e− → νν Pair annihilation,
e− (Z, A) → (Z, A) e− νν Bremsstrahlung.
Individual processes dominate in the regions of density and temperature
indicated in Fig. C.1. In the following, various analytic fit formulae for
these neutrino emission rates are reviewed.
C.1 Plasma Process

Widely used formulae for the plasma process are those of Beaudet, Pet-
rosian, and Salpeter (1967), Munakata, Kohyama, and Itoh (1985), and
Schinder et al. (1987), which all agree with each other to better than 1%
if the same effective coupling constants are used. All of these rates are
poor approximations for T < 8
∼ 10 K which is relevant for low-mass stars
because they were optimized for higher temperatures. Itoh et al. (1989)
have attempted to improve the accuracy at low temperatures, and Blin-
nikov and Dunina-Barkovskaya (1994) gave rates which were optimized
for low-mass stars but fail for temperatures above about 108 K. At high
temperatures and densities, a poor approximation to the photon dis-
persion relation was used in all of these works (Braaten 1991) whence
none of these rates are satisfactory. A new fit by Itoh et al. (1992) still
contains islands in the ρ-T -plane with errors of several 10%.
585
586 Appendix C
Fig. C.1. Regions of density and temperature where the indicated neutrino
emission processes contribute more than 90% of the total. µe is the electron
“mean molecular weight,” i.e. roughly the number of baryons per electron.
The bremsstrahlung contribution depends on the chemical composition. The
solid lines are for helium, the dotted ones for iron which yields a larger
bremsstrahlung rate.
A detailed comparison of these formulae with the exact rates was

performed by Haft, Raffelt, and Weiss (1994). They provided a new
fitting formula which approximates the analytic emission rate to within
5% in the entire regime where the plasma process dominates.
C.2 Photoneutrino and Pair-Annihilation

Process
Beaudet, Petrosian, and Salpeter (1967) provided analytic approxi-
mations for the photoneutrino and pair-annihilation processes. Dicus
(1972) gave global correction factors to these rates to include neutral-
current effects. Schinder et al. (1987) numerically recalculated the emis-
sion rates in the standard model and found good agreement with the
BPS formulae together with the Dicus correction factors. They supple-
mented the BPS rates for the temperature range 1010 −1011 K.
An alternate set of approximation formulae was provided by Itoh
et al. (1989) who improved on their previous work (Munakata, Koh-
yama, and Itoh 1985). In Fig. C.2 I show the relative deviation between
the Itoh et al. (1989) with the Schinder et al. (1987) rates. The total
Numerical Neutrino Energy-Loss Rates 587
Fig. C.2. Deviation between the Schinder et al. (1987) and the Itoh et al.
(1989) rates for the photoneutrino and pair-annihilation processes. Com-
pared are the total emission rates where the plasma rate of Haft, Raffelt and
Weiss (1994) and the bremsstrahlung rate (helium) of Itoh and Kohyama
(1983) were used. The contours indicate were the individual processes dom-
inate (Fig. C.1).
energy-loss rates for the photo and pair process was calculated accord-
ing to these authors, while in each case the plasma rate of Haft, Raffelt,
and Weiss (1994) and the bremsstrahlung rate for helium of Itoh and
Kohyama (1983) were taken. Therefore, deviations occur only in the
range of temperatures and densities where the photo or pair process
dominates. The largest deviations in the lower left corner of Fig. C.2
are around 25%. However, there the absolute magnitude of neutrino
emission is very small (see below) so that the difference between the
rates in this regime does not appear to be of much practical significance.
Another analytic approximation formula for the pair process was
derived by Blinnikov and Rudzskiı̆ (1989).
588 Appendix C
C.3 Bremsstrahlung
Bremsstrahlung dominates for low temperatures and high densities
where electrons are degenerate and the nuclei are strongly correlated.
In a series of papers the emission rate was calculated by Itoh and Ko-
hyama (1983), Itoh et al. (1984a,b), and Munakata, Kohyama, and Itoh
(1987).
For simple estimates one may use the approximate rate given in
Eq. (11.40). In Fig. C.3 I display the error of this approximation for
iron relative to the results of Itoh and Kohyama (1983). For orientation
the contours of Fig. C.1 for iron are also shown, but only the brems-
strahlung rates are compared. The simple approximation is not a bad
fit in the regions where bremsstrahlung could be of interest. For carbon
the fit is almost as good, but it is substantially worse for helium.
Fig. C.3. Relative deviation between the bremsstrahlung rates of Itoh and
Kohyama (1983) for iron and the simple approximation formula Eq. (11.40).
The contours where different processes dominate are for iron.
Numerical Neutrino Energy-Loss Rates 589
C.4 Total Emission Rate

The total neutrino energy-loss rate for helium is shown in Figs. C.4–C.6
where the photoneutrino and pair-annihilation rates are from Schinder
et al. (1987), the plasma process from Haft, Raffelt, and Weiss (1994),
and bremsstrahlung (helium) from Itoh and Kohyama (1983)
Fig. C.4. Contour plot for the total neutrino energy-loss rate !ν per unit
mass for helium. The thin contours are at intervals of a factor of 10 for !ν .
The regions where the individual processes dominate are also indicated.
590 Appendix C
Fig. C.5. Neutrino energy-loss rate as a function of density. The thin lines are
for temperatures 2, 3, 4, etc. times the value indicated on the corresponding
thick line.
Fig. C.6. Neutrino energy-loss rate as a function of temperature for the

indicated values of 2ρ/µe .
Appendix D
Characteristics of Stellar
Plasmas
D.1 Normal Matter

D.1.1 Temperatures and Densities
The material encountered in stars is usually in a state of thermal equi-
librium. In the absence of strong magnetic fields, the plasma is entirely
characterized by its temperature T , mass density ρ, and a set of chem-
ical composition parameters X, Y , X12 , etc. which determine the mass
fractions of the elements 1 H, 4 He, 12 C, and so forth. The mass fraction
of all elements heavier than helium (“metals”) is denoted by Z.
The number density of a species with mass fraction Xj , atomic
weight Aj , and charge Zj e is given by
nj = (ρ/mu ) Xj /Aj , (D.1)
where mu = 1.66×10−24 g = 0.932 GeV is the atomic mass unit.99 The
number density of electrons is
! ρ ! X j Zj ρ
ne = Zj n j = = , (D.2)
j mu j Aj µe mu
where µe is the “mean molecular weight” per electron, not to be con-

fused with the electron chemical potential. (Strictly speaking ne =
ne− − ne+ , the number density of electrons minus that of positrons.)
99
The proton and neutron mass are 0.9383 and 0.9396 GeV, respectively. An exact
translation between mass and number density thus requires taking nuclear binding
energies into account whence the Aj are not exact integers. For the purposes of this
book these differences are negligible.
591
592 Appendix D
For all elements except hydrogen Zj /Aj ≈ 12 . Notably, this applies

to helium and compounds of α-particles such as 12 C and 16 O. Therefore,
Ye ≈ µ−1 1
e ≈ X + 2 (Y + Z), (D.3)
where Ye is the mean number of electrons per baryon. Here, the mass
fraction Z of metals must not be confused with a nuclear charge.
Some examples for typical conditions encountered in stars are shown
in Fig. D.1, ignoring neutron stars (density around nuclear). Aside
from the example of an evolved massive star, all conditions refer to the
centers of stars. Except for the hydrogen main squence, the abscissa is
essentially the physical density because for most chemical compositions
Ye ≈ 12 . Horizontal-branch (HB) stars correspond essentially to the
helium main sequence at 0.5 M" .
Fig. D.1. Typical temperatures and densities encountered in stars, ignoring

neutron stars. The hydrogen, helium, and carbon main sequence (H-MS,
He-MS and C-MS) represent the conditions at the center of zero-age models;
for selected cases their M/M" is indicated (adapted from Kippenhahn and
Weigert 1990). The highly evolved 25 M" star is according to Woosely and
Weaver (1986b) where the open circles are marked with the energy source of
the different burning shells (He-burning etc.). There is no nuclear burning
at the center (Fe). Also shown is the evolution of the central conditions of
a 0.8 M" star from the hydrogen main sequence to the helium flash (Haft,
Raffelt, and Weiss 1994). The rear ends of the arrows mark the indicated
values of the absolute surface brightness in magnitudes.
Characteristics of Stellar Plasmas 593
D.1.2 Relativistic Conditions for Electrons

The nuclei in normal stellar matter are always nonrelativistic; relativis-
tic corrections begin to be important only in neutron stars. The elec-
trons, on the other hand, tend to be at least partially relativistic. Even
at the center of the Sun at a temperature of 1.3 keV, a typical thermal
electron velocity is about 9% of the speed of light. In Fig. D.2 contours
for the thermal average $v 2 %1/2 are shown in the ρ-T -plane. The loci of
the stellar models of Fig. D.1 are also indicated. For low-mass stars, the
electrons are mildly relativistic, although a nonrelativistic treatment is
often enough as a first approximation.
Fig. D.2. Contours for $v 2 %1/2 , the average thermal velocity of electrons.
The loci of the stellar models of Fig. D.1 are also indicated.
D.1.3 Electron Degeneracy

The phase-space occupation numbers of fermions in thermal equilib-
rium are characterized by a Fermi-Dirac distribution
1
fp = , (D.4)
e(Ep −µ)/T +1
where Ep is the energy of the momentum mode p. If dispersion effects
can be ignored, Ep2 = m2 + p2 with the fermion vacuum mass m. The
(relativistic) chemical potential is denoted by µ; for electrons it should
not be confused with the mean molecular weight µe . The distribution
594 Appendix D
of antifermions is given by the same expression with µ → −µ. Then µ

is implicitly given by the phase-space integral
" # $
2 d3 p 1 1
nf = − , (D.5)
(2π)3 e(Ep −µ)/T + 1 e(Ep +µ)/T + 1
where the second term represents antifermions and the factor 2 is for the
two spin degrees of freedom. Again, the fermion density is understood
to mean the density of fermions minus that of antifermions.
At vanishing temperature, fp becomes a step function Θ(µ − Ep ).
If µ > 0 so that nf > 0, i.e. an excess of fermions over antifermions,
there are no antifermions at all at T = 0. The fermion integral yields
nf = p3F /3π 2 , (D.6)
where the Fermi momentum is defined by µ20 = p2F + m2 with µ0 the
zero-temperature chemical potential. The Fermi energy is defined by
EF2 = p2F + m2e , i.e. EF = µ0 .
Equation D.6 is taken as the definition of the Fermi momentum even
at T > 0; it is a useful parameter to characterize the fermion density,
whether or not they are degenerate. Numerically it is
pF = 5.15 keV (Ye ρ)1/3 (D.7)
for electrons with the mass density ρ in units of g cm−3 .
In general, Eq. (D.5) cannot be made explicit for µ; it has to be
solved numerically or by an approximation method. In the ρ-T -plane,
contours for the electron chemical potential are shown in Fig. D.3.
Above the main plot, the electron density is characterized by pF . On
the right side, the temperature is shown in units of keV. Recall that
107 K = 0.8621 keV (Appendix A).
For nonrelativistic electrons the contours in Fig. D.3 are very sensi-
tive to the exact value of µ. Therefore, in this regime the nonrelativistic
chemical potential
µ̂ ≡ µ − m (D.8)
is a more appropriate parameter. Often µ̂ is referred to as the chem-
ical potential. This can be very confusing when relativistic effects are
important. In terms of µ̂, the relativistic Fermi-Dirac distribution is
1
fp = , (D.9)
e(Ekin −µ̂)/T +1
with the kinetic energy Ekin = Ep −m → p2 /2m (nonrelativistic limit).
Fig. D.3. Contours for the electron chemical potential µ. The solid lines are
marked with the relevant value for µ, the dotted lines with µ − me .
Fig. D.4. Contours for the electron degeneracy parameter η = (µ − me )/T .

Also shown are the loci of the stellar models of Fig. D.1.
596 Appendix D
A Fermi gas becomes degenerate when a typical thermal (kinetic)

energy is on the order of µ̂. Therefore, the degeneracy parameter
η ≡ µ̂/T = (µ − m)/T (D.10)
is frequently used to characterize the fermions. They are degenerate
for η larger than a few, and nondegenerate for η < 0. For electrons,
contours of η in the ρ-T -plane are shown in Fig. D.4. A comparison
with the loci of the stellar models of Fig. D.1 reveals that electrons in
stellar plasmas are often at least partially degenerate.
D.1.4 Plasma Frequency

Other characteristic properties of a plasma refer to the behavior of
electromagnetic waves. The photon dispersion relation (Sect. 7.4) is
characterized by the plasma frequency which at T = 0 is
ω02 = 4πα ne /EF = (4α/3π) p3F /EF . (D.11)
Nonrelativistically, it takes on the familiar form ω02 = 4πα ne /me . Nu-
merically it is
(Ye ρ)1/2
ω0 = 28.7 eV , (D.12)
[1 + (1.019×10−6 Ye ρ)2/3 ]1/4
where the mass density ρ is in units of g cm−3 .
D.1.5 Screening Scale

An electric test charge will be screened by the polarization of the
plasma. If the plasma is weakly coupled (see below) the screened
Coulomb potential takes the form of a Yukawa potential r−1 e−kS r where
kS is the screening scale; its inverse is the screening radius. The plasma
is polarized by the test charge because the positive constituents of the
plasma are repelled while the negative ones are attracted, or the re-
verse. Therefore, both electrons and ions contribute to screening. If
both are nondegenerate the total contribution is kS2 = kD 2
+ ki2 where
the electron contribution is known as the Debye scale,
2
kD = 4παne /T = (4α/3π) p3F /T. (D.13)
The ions (charge Zj e, atomic weight Aj ) contribute
4πα ! 2 4πα ρ ! Xj Zj2
ki2 = Z nj = . (D.14)
T j j T mu j Aj
For only one species of ions with charge Ze one has ki2 = Z kD
2
. Nu-
merically,
kD = 222 eV (Ye ρ/T8 )1/2 , (D.15)
where ρ is in units of g cm−3 and T8 = T /108 K. Contours in the ρ-T -

plane are shown in Fig. D.5.
Fig. D.5. Contours for the Debye scale kD in keV.
When the electrons are degenerate they cannot form a Debye-Hückel

cloud around a test charge. Rather, their distribution is characterized
2
by a Thomas-Fermi model which results in the screening scale kTF =
4αpF EF /π. Because kTF ( kD the electron screening can be neglected
relative to the ions whence kS ≈ ki .
Degenerate electrons form an essentially inert background of nega-
tive charge in which the ions move, subject to their mutual Coulomb
interaction. They can be treated as a weakly coupled Boltzmann gas as
long as a typical thermal energy exceeds a typical Coulomb interaction
energy. As a quantitative measure one uses the plasma parameter
Γ = Z 2 α/ai T, (D.16)
where Ze is the nuclear charge and ai the ion-sphere radius defined by

n−1
i = 4πa3i /3 with the ion density ni . Numerically this is
Γ = 1.806×10−3 T8−1 (Z 5 2Ye ρ)1/3 , (D.17)

598 Appendix D
with the mass density in units of g cm−3 and T8 = T /108 K. Recall that
for a single nuclear species Ye = Z/A (atomic weight A).
The plasma is weakly coupled for Γ < ∼ 1, it is in the liquid metal
<
phase for 1 ∼ Γ < 178, and forms a body-centered cubic lattice for
Γ > 178 (Slattery, Doolen, and DeWitt 1980, 1982). Debye screening
by the ions is appropriate for a weakly coupled plasma; otherwise the
Debye approximation for the ion-ion correlations is misleading. Con-
tours for Γ in the ρ-T -plane are shown in Fig. D.6. For the purposes
of this book, a strongly coupled plasma occurs only in the interior of
white dwarfs.
Fig. D.6. Contours for the plasma coupling parameter Γ. Also shown are
the loci of the stellar models of Fig. D.1 which had to be shifted relative
to each other according to the nuclear charge Z relevant for each chemical
composition.
D.1.6 Summary
The characteristic plasma properties for a number of typical astrophys-
ical sites that are important in the main body of the book are summa-
rized in Tab. D.1.
Table D.1. Plasma characteristics for some typical astrophysical sites.
Center of Red-giant core

standard Core of just before
solar model HB stars helium ignition White dwarf
Characteristic nondegenerate nondegenerate degenerate degenerate
nonrelativistic nonrelativistic weakly coupled strongly coupled
Temperature 1.55×107 K ≈ 108 K ≈ 108 K 3×106 −2×107 K
= 1.3 keV = 8.6 keV = 8.6 keV = 0.3−1.7 keV
Characteristics of Stellar Plasmas
Density 156 g cm−3 ≈ 104 g cm−3 ≈ 106 g cm−3 1.8×106 g cm−3 (a )

4 4 12
Composition X = 0.35 He, 12 C, 16 O He C, 16 O
Electron density 6.3×1025 cm−3 3.0×1027 cm−3 3.0×1029 cm−3 5.3×1029 cm−3
Fermi momentum 24.3 keV 88 keV 409 keV 495 keV
Fermi energyb 0.58 keV 7.6 keV 144 keV 200 keV
Plasma frequency 0.3 keV 2.0 keV 18 keV 23 keV
Plasma coupling Γ = 0.07 0.12 0.57 144−22
Debye screening electrons + ions electrons + ions ions —
2 2
kS = (kD + ki2 )1/2 kS = (kD + ki2 )1/2 kS = ki (strong
= 9.1 keV = 27 keV = 222 keV screening)
a b
Center of WD with M = 0.66 M" . Nonrelativstic Fermi energy EF − me .
599
600 Appendix D
D.2 Nuclear Matter

D.2.1 The Ideal p n e νe Gas
For the topics discussed in this book, the properties of hot nuclear
matter in a young supernova core are of great interest. The relevant
range of densities and temperatures is about 3×1012 −3×1015 g cm−3
and 3−100 MeV, respectively. The properties of matter at such con-
ditions is determined by its equation of state which takes the nuclear
interaction fully into account. However, in order to gain a rough un-
derstanding of the behavior of the main constituents of the medium
(protons, neutrons, electrons, and electron neutrinos) it is worthwhile
to study a simple toy model where these particles are treated as ideal
Fermi gases.
To this end, neutrinos and electrons are treated as massless. Their
dispersion relation is dominated by the interaction with the medium.
Because they interact only by electroweak forces their “effective mass”
is always much smaller than their energies.
D.2.2 Kinetic and Chemical Equilibrium

The reaction e p ↔ n νe which establishes β equilibrium is fast com-
pared to other relevant time scales. Therefore, the relative abundances
of n, p, e, and νe are determined by the conditions of kinetic and chem-
ical equilibrium. The physical condition of the medium is then deter-
mined by the baryon density nB , the temperature T , and the condition
of electric charge neutrality
nB = nn + np Baryon density,
np = ne Charge neutrality, (D.18)
µe + µp = µn + µνe β equilibrium.
Here, the µj are the relativistic chemical potentials of the fermions

which determine their number densities nj according to the Fermi-Dirac
distribution Eq. (D.5). Recall that nj is the difference between fermions
and antifermions of a given species.
In addition, one of two extreme assumptions is made. In a young
SN core the neutrinos are trapped so that the local lepton number is
conserved. In this case the lepton fraction YL is the fourth required
input parameter,
YL nB = ne + nνe Lepton conservation. (D.19)

As a neutron star cools it becomes transparent to neutrinos. In this

case their chemical potential vanishes which yields
µνe = 0 Free neutrino escape (D.20)
as the other extreme additional condition.
D.2.3 Cold Nuclear Matter

The limit T → 0 relevant for old neutron stars is particularly simple
because it allows one to express the Fermi-Dirac distributions as step-
functions. One may express all Fermi momenta in units of the effective
nucleon mass, i.e., xj ≡ pjF /m∗N . Then baryon conservation is
x3B = x3p + x3n , (D.21)
where nj = (pjF )3 /3π 3 was used, and

1/3
xB ≡ (3π 2 nB )1/3 /m∗N = 0.255 ρ14 (mN /m∗N ), (D.22)
with ρ14 the baryonic mass density in units of 1014 g cm−3 .

Because the star is transparent to neutrinos one may use µνe = 0.
Then the equation of β-equilibrium becomes µe + µp − µn = 0 or
xp + (1 + x2p )1/2 − (1 + x2n )1/2 = 0, (D.23)
where xe = xp was used from the condition of charge neutrality. This

is easily solved to yield (Shapiro and Teukolsky 1983)
x4n
x2p = . (D.24)
4 (1 + x2n )
This result may be expressed in terms of the usual composition param-
eters Yp and Yn which give the number of protons and neutrons per
baryon; Yp + Yn = 1. Then Yn,p = (xn,p /xB )3 so that
# $3
xB (1 − Yp )2
Yp = . (D.25)
2 [1 + x2B (1 − Yp )2/3 ]3/2
When xB ( 1 this is Yp = (xB /2)3 = 2.1×10−3 ρ14 (mN /m∗N )3 so that

the proton fraction is small—hence the term “neutron star”—although
the exact Yp for a given density depends sensitively on the nucleon
dispersion relation. For infinite density (xB → ∞) a maximum of
Yp = 19 is reached.
602 Appendix D
D.2.4 Hot Nuclear Matter
In a supernova core right after collapse the temperature is so high (sev-

eral tens of MeV) that the nucleons are nearly nondegenerate. More-
over, the neutrinos are trapped so that locally a fixed value for YL
determined by initial conditions is assumed. Most of the lepton num-
ber will reside in electrons, causing the proton concentration Yp to be
approximately equal to YL . A more accurate determination requires a
numerical solution of Eqs. (D.18) and (D.19).
In Figs. D.7 (a)−(c) the results of such an exercise are presented
for YL = 0.3, which is a typical value for the material in a SN core just
after collapse. The proton concentration Yp , the difference between the
neutron and proton chemical potentials, and the degeneracy parameters
for neutrons and protons are shown. In each case, a solid line refers
to the assumption of an effective nucleon mass as in Fig. 4.10 while
a dotted line refers to the vacuum mass. As expected, the reduced
effective mass increases somewhat the medium’s degeneracy.
Because µn − µp = µe − µνe , the contours of Fig. D.7 (b) also give
the difference between the surfaces of the electron and neutrino Fermi
seas. Note that the leptons are much more degenerate than the nucleons
because they are essentially massless. This remark does not apply to the
upper-left corner of the plots where actually an excess of antineutrinos
is enforced—there are more protons than leptons (Yp > YL )!
Fig. D.7. (a) Contours for Yp in hot neutron-star matter with YL = 0.3.
Solid lines for the effective nucleon mass as in Fig. D.7, dotted lines for the
vacuum mass.
Fig. D.7. (b) µn − µp = µe − µνe (in MeV).
Fig. D.7. (c) Degeneracy parameter for protons and neutrons.

604 Appendix D
Fig. D.8. Degeneracy suppression of neutrino scattering on “heavy” nucleons

for YL = 0.3 and an effective nucleon mass as in Fig. D.7.
Fig. D.9. Ratio of the suppression factor of Fig. D.8 between the case with
an effective nucleon mass and with the vacuum one.
When neutrinos scatter on a “heavy” nucleon recoil effects can be

neglected. The nucleon does not change its momentum so that the
degeneracy suppression is given by a factor
2 " d3 p
ξ= fp (1 − fp ), (D.26)
nN (2π)3
where fp is a Fermi-Dirac occupation number. In Fig. D.8 contours
for Yn ξn + Yp ξp are shown for the same parameters as in Fig. D.7, i.e.
YL = 0.3 and the effective nucleon mass of Fig. 4.10. In Fig. D.9 the
ratio of this factor between the case of an effective nucleon mass and
the vacuum one are shown, i.e. the additional Pauli suppression of the
neutrino scattering rate from using an effective nucleon mass.
References
Prefixes to authors’ names have been used as a full part of the name so
that, for example, van den Bergh, van Bibber, von Feilitzsch and others
are found under the letter V.
Abbott, L. F., de Rújula, A., and Walker, T. P. 1988, Nucl. Phys. B,

299, 734.
Abbott, L. F., and Sikivie, P. 1983, Phys. Lett. B, 120, 133.
Abdurashitov, J. N., et al. 1994, Phys. Lett. B, 328, 234.
Accetta, F. S., Krauss, L. M., and Romanelli, P. 1990, Phys. Lett. B,
248, 146.
Accetta, F. S., and Steinhardt, P. J. 1991, Phys. Rev. Lett., 67, 298.
Achkar, B., et al. 1995, Nucl. Phys. B, 434, 503.
Acker, A., and Pakvasa, S. 1994, Phys. Lett. B, 320, 320.
Acker, A., Pakvasa, S., and Raghavan, R. S. 1990, Phys. Lett. B, 238,
117.
Adams, E. N. 1988, Phys. Rev. D, 37, 2047.
Adams, J. B., Ruderman, M. A., and Woo, C.-H. 1963, Phys. Rev., 129,
1383.
Adler, S. L. 1971, Ann. Phys. (N.Y.), 67, 599.
Aharonov, Y., Avignone III, F. T., and Nussinov, S. 1988a, Phys. Lett.
B, 200, 122.
Aharonov, Y., Avignone III, F. T., and Nussinov, S. 1988b, Phys. Rev.
D, 37, 1360.
Aharonov, Y., Avignone III, F. T., and Nussinov, S. 1989, Phys. Rev.
D, 39, 985.
Ahrens, L. A., et al. 1985, Phys. Rev. D, 31, 2732.
Ahrens, L. A., et al. 1990, Phys. Rev. D, 41, 3301.
Akhmedov, E. Kh. 1988a, Yad. Fiz., 48, 599 (Sov. J. Nucl. Phys., 48,
382).
Akhmedov, E. Kh. 1988b, Phys. Lett. B, 213, 64.
Akhmedov, E. Kh., and Berezin, V. V. 1992, Z. Phys. C, 54, 661.
606
References 607
Akhmedov, E. Kh., and Berezhiani, Z. G. 1992, Nucl. Phys. B, 373,

479.
Akhmedov, E. Kh., Lanza, A., and Petcov, S. T. 1993, Phys. Lett. B,
303, 85.
Akhmedov, E. Kh., Lanza, A., and Petcov, S. T. 1995, Phys. Lett. B,
348, 124.
Akhmedov, E. Kh., Lipari, P., and Lusignoli, M. 1993, Phys. Lett. B,
300, 128.
Akhmedov, E. Kh., Petcov, S. T., and Smirnov, A. Yu. 1993a, Phys.
Lett. B, 309, 95.
Akhmedov, E. Kh., Petcov, S. T., and Smirnov, A. Yu. 1993b, Phys.
Rev. D, 48, 2167.
Alcock, C., and Olinto, A. V. 1988, Ann. Rev. Nucl. Part. Sci., 38, 161.
Alcock, C., et al. 1993, Nature, 365, 621.
Alcock, C., et al. 1995, Phys. Rev. Lett., 74, 2867.
ALEPH Collaboration 1995, Phys. Lett. B, 349, 585.
Alexeyev, E. N., et al. 1987, Pis’ma Zh. Eksp. Teor. Fiz., 45, 461 (JETP
Lett., 45, 589).
Alexeyev, E. N., et al. 1988, Phys. Lett. B, 205, 209.
Allen, C. W. 1963, Astrophysical Quantities (University of London
Press, London).
Almeida, L. D., Matsas, G. E. A., and Natale, A. A. 1989, Phys. Rev.
D, 39, 677.
Altherr, T. 1990, Z. Phys. C, 47, 559.
Altherr, T. 1991, Ann. Phys. (N.Y.), 207, 374.
Altherr, T., and Kraemmer, U. 1992, Astropart. Phys., 1, 133.
Altherr, T., Petitgirard, E., and del Rı́o Gaztelurrutia, T. 1993, As-
tropart. Phys., 1, 289.
Altherr, T., Petitgirard, E., and del Rı́o Gaztelurrutia, T. 1994, As-
tropart. Phys., 2, 175.
Altherr, T., and Salati, P. 1994, Nucl. Phys. B, 421, 662.
Alväger, T., et al. 1964, Phys. Lett., 12, 260.
Alvarez, L. 1949, University of California Radiation Laboratory Report
UCRL-328 (quoted after Davis, Mann, and Wolfenstein 1989).
Anand, J. D., Goyal, A., and Iha, R. N. 1990, Phys. Rev. D, 42, 996.
Anand, J. D., et al. 1984, Phys. Rev. D, 29, 1270.
Anderhub, H. B., et al. 1982, Phys. Lett. B, 114, 76.
Anderson, B., and Lyne, A. G. 1983, Nature, 303, 597.
Anderson, S. B., et al. 1993, Ap. J., 414, 867.
Aneziris, C., and Schechter, J. 1991, Int. J. Mod. Phys. A, 6, 2375.
608 References
Anselm, A. A. 1982, Pis’ma Zh. Eksp. Teor. Fiz., 36, 46 (JETP Lett.,
36, 55).
Anselm, A. A. 1985, Yad. Fiz., 42, 1480 (Sov. J. Nucl. Phys., 42, 936).
Anselm, A. A. 1988, Phys. Rev. D, 37, 2001.
Anselm, A. A., and Uraltsev, N. G. 1982a, Phys. Lett. B, 114, 39.
Anselm, A. A., and Uraltsev, N. G. 1982b, Phys. Lett. B, 116, 161.
Arafune, J., and Fukugita, M. 1987, Phys. Rev. Lett., 59, 367.
Arafune, J., et al. 1987a, Phys. Rev. Lett., 59, 1864.
Arafune, J., et al. 1987b, Phys. Lett. B, 194, 477.
ARGUS Collaboration 1988, Phys. Lett. B, 202, 149.
ARGUS Collaboration 1992, Phys. Lett. B, 292, 221.
Arnett, W. D., and Rosner, J. L. 1987, Phys. Rev. Lett., 58, 1906.
Arnett, W. D., et al. 1989, Ann. Rev. Astron. Astrophys., 27, 629.
Ashkin, A. and Dziedzic, J. M. 1973, Phys. Rev. Lett., 30, 139.
Assamagan, K., et al. 1994, Phys. Lett. B, 335, 231.
Athanassopoulos, C., et al. 1995, Phys. Rev. Lett., 75, 2650.
Athar, H., Peltoniemi, J. T., and Smirnov, A. Yu. 1995, Phys. Rev. D,
51, 6647.
Atzmon, E., and Nussinov, S. 1994, Phys. Lett. B, 328, 103.
Aubourg, E., et al. 1993, Nature, 365, 623.
Aubourg, E., et al. 1995, Astron. Astrophys., 301, 1.
Auriemma, G., Srivastava, Y., and Widom, A. 1987, Phys. Lett. B, 195,
254.
Avignone, F. T., et al. 1987, Phys. Rev. D, 35, 2752.
Babu, K. S., Gould, T. M., and Rothstein, I. Z. 1994, Phys. Lett. B,
321, 140.
Babu, K. S., and Mohapatra, R. N. 1990, Phys. Rev. D, 42, 3866.
Babu, K. S., Mohapatra, R. N., and Rothstein, I. Z. 1992, Phys. Rev.
D, 45, R3312.
Babu, K. S., and Volkas, R. R. 1992, Phys. Rev. D, 46, R2764.
Bahcall, J. N. 1989, Neutrino Astrophysics (Cambridge University
Press).
Bahcall, J. N. 1990, Phys. Rev. D, 41, 2964.
Bahcall, J. N. 1991, Phys. Rev. D, 44, 1644.
Bahcall, J. N. 1994a, Phys. Rev. D, 49, 3923.
Bahcall, J. N. 1994b, Phys. Lett. B, 338, 276.
Bahcall, J. N. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 98.
Bahcall, J. N., and Bethe, H. A. 1993, Phys. Rev. D, 47, 1298.
Bahcall, J. N., and Frautschi, S. C. 1969, Phys. Lett. B, 29, 623.
Bahcall, J. N., and Glashow, S. L. 1987, Nature, 326, 476.
Bahcall, J. N., and Holstein, B. R. 1986, Phys. Rev. C, 33, 2121.
References 609
Bahcall, J. N., Krastev, P. I., and Leung, C. N. 1995, Phys. Rev. D, 52,
1770.
Bahcall, J. N., and Pinsonneault, M. H. 1992, Rev. Mod. Phys., 64, 885.
Bahcall, J. N., and Pinsonneault, M. H. 1995, to be published in Rev.
Mod. Phys.
Bahcall, J. N., and Press, W. H. 1991, Ap. J., 370, 730.
Bahcall, J. N., and Ulrich, R. K. 1988, Rev. Mod. Phys., 60, 297.
Bahcall, J. N., and Wolf, R. A. 1965a, Phys. Rev. Lett., 14, 343.
Bahcall, J. N., and Wolf, R. A. 1965b, Phys. Rev., 5B, 1452.
Bahcall, J. N., et al. 1995, Nature, 375, 29.
Bailes, M. 1989, Ap. J., 342, 917.
Bailes, M., et al. 1990, Mon. Not. R. astr. Soc., 247, 322.
Bakalov, D., et al. 1994, Nucl. Phys. B (Proc. Suppl.), 35, 180.
Balantekin, A. B., and Loreti, F. 1992, Phys. Rev. D, 45, 1059.
Baluni, V. 1979, Phys. Rev. D, 19, 2227.
Balysh, A., et al. 1995, Phys. Lett. B, 356, 450.
Barbiellini, G., and Cocconi, G. 1987, Nature, 329, 21.
Barbieri, R., and Dolgov, A. 1991, Nucl. Phys. B, 349, 743.
Barbieri, R., and Fiorentini, G. 1988, Nucl. Phys. B, 304, 909.
Barbieri, R., and Mohapatra, R. N. 1988, Phys. Rev. Lett., 61, 27.
Barbieri, R., and Mohapatra, R. N. 1989, Phys. Rev. D, 39, 1229.
Barbieri, R., et al. 1991, Phys. Lett. B, 259, 119.
Bardeen, J., Bond, J., and Efstathiou, G. 1987, Ap. J., 321, 28.
Bardeen, W. A., Peccei, R. D., and Yanagida, T. 1987, Nucl. Phys. B,
279, 401.
Bardeen, W. A., and Tye, S. H. H. 1978, Phys. Lett. B, 74, 580.
Barger, V., Phillips, R. J. N., and Sarkar, S. 1995, Phys. Lett. B, 352,
365; (E) ibid., 356, 617.
Barger, V., Phillips, R. J. N., and Whisnant, K. 1991, Phys. Rev. D,
44, 1629.
Barger, V., Phillips, R. J. N., and Whisnant, K. 1992, Phys. Rev. Lett.,
69, 3135.
Baring, M. G. 1991, Astron. Astrophys., 249, 581.
Barnes, A. V., Weiler, T. J., and Pakvasa, S. 1987, Ap. J., 323, L31.
Barr, S. M., and Seckel, D. 1992, Phys. Rev. D, 46, 539.
Barroso, A., and Branco, G. C. 1982, Phys. Lett. B, 116, 247.
Barrow, J. D. 1978, Mon. Not. R. astr. Soc., 184, 677.
Barrow, J. D., and Burman, R. R. 1984, Nature, 307, 14.
Barstow, M. A. 1993, ed., White Dwarfs: Advances in Observation and
Theory (Kluwer, Dordrecht).
Battye, R. A., and Shellard, E. P. S. 1994a, Phys. Rev. Lett., 73, 2954.
610 References
Battye, R. A., and Shellard, E. P. S. 1994b, Nucl. Phys. B, 423, 260.

Baumann, J., et al. 1988, Phys. Rev. D, 37, 3107.
Baym, G. 1973, Phys. Rev. Lett., 30, 1340.
Bazilevskaya, G. A., Stozhkov, Yu. I., and Charakhch’yan, T. N. 1982,
Pis’ma Zh. Eksp. Teor. Fiz., 35, 273 (JETP Lett., 35, 341).
Beaudet, G., Petrosian, V., and Salpeter, E. E. 1967, Ap. J., 150, 979.
Beck, M., et al. 1993, Phys. Rev. Lett., 70, 2853.
Becker, W., and Aschenbach, B. 1995, in: M. A. Alpar et al. (eds.), The
Lives of the Neutron Stars (Kluwer Academic Publishers), pg. 47.
Becker, W., et al. 1992, Poster presented at the conference Physics of
Isolated Pulsars (Taos, New Mexico).
Becker-Szendy, R., et al. 1992, Phys. Rev. Lett., 69, 1010.
Becklin, E. 1990, Talk presented at the Supernova Watch Workshop
(Santa Monica).
Bég, M. A. B., Marciano, W. J., and Ruderman, M. 1978, Phys. Rev.
D, 17, 1395.
Belesev, A. I., et al. 1994, Report INR-862/94 (Institute for Nuclear
Research, Russia).
Berezhiani, Z. G., Moretti, M., and Rossi, A. 1993, Z. Phys. C, 58, 423.
Berezhiani, Z. G., and Rossi, A. 1994, Phys. Lett. B, 336, 439.
Berezhiani, Z. G., and Rossi, A. 1995, Phys. Rev. D, 51, 5229.
Berezhiani, Z. G., and Smirnov, A. Yu. 1989, Phys. Lett. B, 220, 279.
Berezhiani, Z. G., Smirnov, A. Yu., and Valle, J. W. F. 1992, Phys.
Lett. B, 291, 99.
Berezhiani, Z. G., and Vysotsky, M. I. 1987, Phys. Lett. B, 199, 281.
Berezhiani, Z. G., et al. 1992, Z. Phys. C, 54, 581.
Berezinsky, V. 1994, Comm. Nucl. Part. Phys., 21, 249.
Berezinsky, V., Fiorentini, G., and Lissia, M. 1994, Phys. Lett. B, 341,
38.
Bernabéu, J., et al. 1994, Nucl. Phys. B, 426, 434.
Bernstein, J., Ruderman, M. A., and Feinberg, G. 1963, Phys. Rev.,
132, 1227.
Bershady, M. A., Ressel, M. T., Turner, M. S. 1991, Phys. Rev. Lett.,
66, 1398.
Bethe, H. A. 1935, Proc. Camb. Phil. Soc., 31, 108.
Bethe, H. A. 1939, Phys. Rev., 55, 434.
Bethe, H. A. 1986, Phys. Rev. Lett., 56, 1305.
Bethe, H. A., and Wilson, J. R. 1985, Ap. J., 295, 14.
Bhattacharya, D., and van den Heuvel, E. 1991, Phys. Rep., 203, 1
Bica, E., et al. 1991, Ap. J., 381, L51.
Bieber, J. W., et al. 1990, Nature, 348, 407.
References 611
Bilenky, S. M., and Giunti, C. 1993, Phys. Lett. B, 311, 179.

Bilenky, S. M., and Giunti, C. 1994, Phys. Lett. B, 320, 323.
Bionta, R. M., et al. 1987, Phys. Rev. Lett., 58, 1494.
Bionta, R. M., et al. 1988, Phys. Rev. D, 38, 768.
Bisnovatyi-Kogan, G. V., and Janka, H.-T. 1995, work in progress.
Bjorken, J. D., and Drell, S. D. 1964, Relativistic Quantum Mechanics
(McGraw-Hill, New York).
Blinnikov, S. I., and Dunina-Barkovskaya, N. V. 1994, Mon. Not. R.
astr. Soc., 266, 289.
Blinnikov, S. I., and Okun, L. B. 1988, Pis’ma Astron. Zh., 14, 867
(Sov. Astron. Lett., 14, 368).
Blinnikov, S. I., and Rudzskiı̆, M. A. 1989, Astron. Zh., 66, 730 (Sov.
Astron., 33, 377).
Blinnikov, S. I., et al. 1995, Report ITEP-31-95 and hep-ph/9505444.
Bludman, S. A. 1992, Phys. Rev. D, 45, 4720.
Bludman, S., et al. 1993, Phys. Rev. D, 47, 2220.
Bludman, S., Kennedy, D., and Langacker, P. 1992a, Phys. Rev. D, 45,
1810.
Bludman, S., Kennedy, D., and Langacker, P. 1992b, Nucl. Phys. B,
374, 373.
Boehm, F., and Vogel, P. 1987, Physics of Massive Neutrinos (Cam-
bridge University Press).
Boguta, J. 1981, Phys. Lett. B, 106, 255.
Bond, J. R., and Efstathiou, G. 1991, Phys. Lett. B, 265, 245.
Bouquet, A., and Vayonakis, C. E. 1982, Phys. Lett. B, 116, 219.
Boris, S., et al. 1987, Phys. Rev. Lett., 58, 2019.
Born, M., and Wolf, E. 1959, Principles of Optics (Pergamon Press,
London).
Börner, G. 1992, The Early Universe, 2nd edition (Springer, Berlin).
Borodovsky, L., et al. 1992, Phys. Rev. Lett., 68, 274.
Botella, F. J., Lim, C.-S., and Marciano, W. J. 1987, Phys. Rev. D, 35,
896.
Bouchez, J., et al. 1988, Phys. Lett. B, 207, 217.
Boyd, R. N., et al. 1983, Phys. Rev. Lett., 51, 609.
Boyd, R. N., et al. 1985, Ap. J., 289, 155.
Braaten, E. 1991, Phys. Rev. Lett., 66, 1655.
Braaten, E. 1992, Ap. J., 392, 70.
Braaten, E., and Segel, D. 1993, Phys. Rev. D, 48, 1478.
Bratton, C. B., et al. 1988, Phys. Rev. D, 37, 3361.
Brecher, K. 1977, Phys. Rev. Lett., 39, 1051.
Brinkmann, R. P., and Turner, M. S. 1988, Phys. Rev. D, 38, 2338.
612 References
Brinkmann, W., and Ögelman, H. 1987, Astron. Astrophys., 182, 71.

Brodsky, S. J., et al. 1986, Phys. Rev. Lett., 56, 1763.
Broggini, C., et al. 1990, Experimental Proposal (Univ. Neuchâtel).
Brown, B. A., Csótó, A., and Sherr, R. 1995, Report nucl-th/9506004.
Brown, G. E. 1988, ed., Phys. Rep., 163, 1.
Brown, G. E., Bethe, H. A., and Baym, G. 1982, Nucl. Phys. A, 375,
481.
Bruenn, S. W. 1987, Phys. Rev. Lett., 59, 938.
Bruenn, S. W., and Mezzacappa, A. 1994, Ap. J., 433, L45.
Bryman, D. A., et al. 1983, Phys. Rev. Lett., 50, 1546.
Buonanno, R., et al. 1986, Mem. Soc. Astron. Ital., 57, 391.
Buonanno, R., Corsi, C. E., and Fusi Pecci, F. 1989, Astron. Astrophys.,
216, 80.
Burbidge, E. M., et al. 1957, Rev. Mod. Phys., 29, 547.
Burgess, C. P., and Cline, J. M. 1993, Phys. Lett. B, 298, 141.
Burgess, C. P., and Cline, J. M. 1994a, Phys. Rev. D, 5925.
Burgess, C. P., and Cline, J. M. 1994b, Report hep-ph/9401334, Int.
Conf. Nonaccelerator Particle Physics, Bangalore, India, Jan. 1994.
Burrows, A. 1979, Phys. Rev. D, 20, 1816.
Burrows, A. 1988a, Ap. J., 328, L51.
Burrows, A. 1988b, Ap. J., 334, 891.
Burrows, A. 1990a, Ann. Rev. Nucl. Part. Sci., 40, 181.
Burrows, A. 1990b, in: Petschek 1990.
Burrows, A., and Fryxell, B. A. 1992, Science, 258, 430.
Burrows, A., and Fryxell, B. A. 1993, Ap. J., 418, L33.
Burrows, A., Gandhi, R., and Turner, M. S. 1992, Phys. Rev. Lett., 68,
3834.
Burrows, A., Hayes, J., and Fryxell, B. A. 1995, Ap. J., 450, 830.
Burrows, A., Klein, D., and Gandhi, R. 1992, Phys. Rev. D, 45, 3361.
Burrows, A., and Lattimer, J. M. 1986, Ap. J., 307, 178.
Burrows, A., and Lattimer, J. M. 1987, Ap. J., 318, L63.
Burrows, A., and Lattimer, J. M. 1988, Phys. Rep., 163, 51.
Burrows, A., Ressell, T., and Turner, M. S. 1990, Phys. Rev. D, 42,
3297.
Burrows, A., Turner, M. S., and Brinkmann, R. P. 1989, Phys. Rev. D,
39, 1020.
Buzzoni, A., et al. 1983, Astron. Astrophys., 128, 94.
Callan, C. G., Dashen, R. F., and Gross, D. J. 1976, Phys. Lett. B, 63,
334.
Cameron, A. G. W. 1957, Publ. Astr. Soc. Pacific, 69, 201.
Cameron, R., et al. 1993, Phys. Rev. D, 47, 3707.
References 613
Cannon, R. D. 1970, Mon. Not. R. astr. Soc., 150, 111.

Cantatore, G., et al. 1991, Phys. Lett. B, 265, 418.
Caraveo, P. A. 1993, Ap. J., 415, L111.
Carena, M., and Peccei, R. D. 1989, Phys. Rev. D, 40, 652.
Carlson, E. D. 1995, Phys. Lett. B, 344, 245.
Carlson, E. D., and Garretson, W. D. 1994, Phys. Lett. B, 336, 431.
Carlson, E. D., and Salati, P. 1989, Phys. Lett. B, 218, 79.
Carlson, E. D., and Tseng, L.-S. 1995, Report HUTP-95/A025 and
hep-ph/9507345, to be published in Phys. Lett. B.
Casas, J. A., Garcı́a-Bellido, J., and Quirós, M. 1992, Phys. Lett. B,
278, 94.
Castellani, M., and Castellani, V. 1993, Ap. J., 407, 649.
Castellani, M., and Degl’Innocenti, S. 1993, Ap. J., 402, 574.
Castellani, V., Degl’Innocenti, S., and Fiorentini, G. 1993, Phys. Lett.
B, 303, 68.
Castellani, V., Degl’Innocenti, S., and Romaniello, M. 1994, Ap. J.,
423, 266.
Castellani, V., et al. 1994a, Phys. Lett. B, 324, 425.
Castellani, V., et al. 1994b, Phys. Rev. D, 50, 4749.
Castle, B., and Towner, I. 1990, Modern Theories of Nuclear Moments
(Clarendon Press, Oxford).
Catelan, M., de Freitas Pacheco, J. A., and Horvath, J. E. 1995, Report
astro-ph/9509062, to be published in Ap. J.
Cazzola, P., de Zotti, G., and Saggion, A. 1971, Phys. Rev. D, 3, 1722.
CDF Collaboration 1995, Phys. Rev. Lett., 74, 2627.
Chaboyer, B., et al. 1992, Ap. J., 388, 372.
Chan, K.-W., and Lingenfelter, R. E. 1993, Ap. J., 405, 614.
Chanda, R., Nieves, J. F., and Pal, P. B. 1988, Phys. Rev. D, 37, 2714.
Chandrasekhar, S. 1939, An Introduction to the Study of Stellar Struc-
ture (University of Chicago Press, Chicago).
Chang, L. N., and Zia, R. K. P. 1988, Phys. Rev. D, 38, 1669.
Chang, S., and Choi, K. 1994, Phys. Rev. D, 49, R12.
CHARM II Collaboration 1993, Phys. Lett. B, 309, 463.
Chen, P. 1995, Phys. Rev. Lett., 74, 634; (E) ibid., 3091.
Cheng, B., Schramm, D. N., and Truran, J. W. 1993, Phys. Lett. B,
316, 521.
Cheng, H.-Y. 1988, Phys. Rep., 158, 1.
Chernikov, M. A., et al. 1992, Phys. Rev. Lett., 68, 3383; (E) ibid., 69,
2999.
Chibisov, G. V. 1976, Sov. Phys. Usp., 19, 624.
614 References
Chicashige, Y., Mohapatra, R. N., and Peccei, R. D. 1981, Phys. Lett.

B, 98, 265.
Chieffi, A., Straniero, O., and Salaris, M. 1991, in: K. Janes (ed.), The
Formation and Evolution of Star Clusters (ASP Conference Series,
13), pg. 219.
Chin, C.-W., and Stothers, R. 1975, Nature, 254, 206.
Chin, C.-W., and Stothers, R. 1976, Phys. Rev. Lett., 36, 833.
Chiu, H.-Y., Chan, K. L., and Kondo, Y. 1988, Ap. J., 329, 326.
Chiu, H.-Y., and Salpeter, E. E. 1964, Phys. Rev. Lett., 12, 413.
Chiu, H.-Y., and Stabler, R. C. 1961, Phys. Rev., 122, 1317.
Chodil, G., et al. 1965, Phys. Rev. Lett., 15, 605.
Choi, K., Kang, K., and Kim, J. E. 1989, Phys. Rev. Lett., 62, 849.
Choi, K., and Santamaria, A. 1990, Phys. Rev. D, 42, 293.
Choi, K., et al. 1988, Phys. Rev. D, 37, 3225.
Choudhury, D., Hari Dass, N. D., and Murthy, M. V. N. 1989, Class.
Quantum Grav., 6, L167.
Christensen-Dalsgaard, J. 1992, Ap. J., 385, 354.
Christensen-Dalsgaard, J., Proffitt, C. R., and Thompson, M. J. 1993,
Ap. J., 403, L75.
Chugaı̆, N. N. 1984, Pis’ma Astron. Zh., 10, 210 (Sov. Astron. Lett.,
10, 87).
Chupp, E. L., Vestrand, W. T., and Reppin, C. 1989, Phys. Rev. Lett.,
62, 505.
Çiftçi, A. K., Sultansoi, S., and Türköz, S. 1994, Ankara University
Preprint AU/94-03/HEP (quoted after Blinnikov et al. 1995).
Cisneros, A. 1971, Astrophys. Space Sci., 10, 87.
Clayton, D. D. 1968, Principles of Stellar Evolution and Nucleosynthe-
sis (University of Chicago Press, Chicago).
CLEO Collaboration 1993, Phys. Rev. Lett., 70, 3700.
Cline, D., et al. 1990, Astro. Lett. and Communications, 27, 403.
Cline, D., et al. 1994, Phys. Rev. D, 50, 720.
Cline, J. M. 1992, Phys. Rev. D, 45, 1628.
Cocconi, G. 1988, Phys. Lett. B, 206, 705.
Cohen, J. G., and Frogel, J. A. 1982, Ap. J., 255, L39.
Cohen, J. G., Frogel, J. A., and Persson, S. E. 1978, Ap. J., 222, 165.
Coley, A. A., and Tremaine, S. 1988, Phys. Rev. D, 38, 2927.
Colgate, S. A., and Johnson, M. H. 1960, Phys. Rev. Lett., 5, 235.
Colgate, S. A., and White, R. H. 1966, Ap. J., 143, 626.
Commins, E. D., and Bucksbaum, P. 1983, Weak Interactions of Lep-
tons and Quarks (Cambridge University Press, Cambridge).
Cooper, L., and Stedman, G. E. 1995, Phys. Lett. B, 357, 464.
References 615
Cooper-Sarkar, A. M., et al. 1992, Phys. Lett. B, 280, 153.

Cooperstein, J. 1988, Phys. Rep., 163, 95.
Cooperstein, J., van den Horn, L. J., and Baron, E. A. 1987, Ap. J.,
321, L129.
Cordes, J. M. 1986, Ap. J., 311, 183.
Cordes, J. M., Romani, R. W., and Lundgren, S. C. 1993, Nature, 362,
133.
Cowsik, R. 1977, Phys. Rev. Lett., 39, 784.
Cowsik, R. 1988, Phys. Rev. D, 37, 1685.
Cowsik, R., Schramm, D., and Höflich, P. 1989, Phys. Lett. B, 218, 91.
Cox, J. P. 1980, Theory of Stellar Pulsation (Princeton University
Press, Princeton).
Crewther, R., et al. 1979, Phys. Lett. B, 88, 123; (E) 1980, ibid., 91,
487.
Cung, V. K., and Yoshimura, M. 1975, Nuovo Cim., 29 A, 557.
D0 Collaboration 1995, Phys. Rev. Lett., 74, 2633.
Da Costa, G. S., and Armandroff, T. E. 1990, Astron. J., 100, 162.
Da Costa, G. S., Frogel, J. A., and Cohen, J. G. 1981, Ap. J., 248, 612.
Damour, T., and Gundlach, C. 1991, Phys. Rev. D, 43, 3873.
Damour, T., and Taylor, J. H. 1991, Ap. J., 366, 501.
D’Antona, F., and Mazzitelli, I. 1990, Ann. Rev. Astron. Astrophys.,
28, 139.
Dar, A. 1987, Report, Institute for Advanced Study (unpublished).
Dar, A., and Dado, S. 1987, Phys. Rev. Lett., 59, 2368.
Dar, A., Goodman, J., and Nussinov, S. 1987, Phys. Rev. Lett., 58,
2146.
Daum, K. 1994, Report WUB 94-9, Contributed paper to Neutrino 94
(unpublished).
Daum, M., et al. 1991, Phys. Lett. B, 265, 425.
Davidsen, A. F., et al. 1991, Nature, 351, 128.
Davidson, S., Campbell, B., and Bailey, D. 1991, Phys. Rev. D, 43,
2314.
Davidson, S., and Peskin, M. 1994, Phys. Rev. D, 49, 2114.
Davis, R. L. 1986, Phys. Lett. B, 180, 225.
Davis, R. L., and Shellard, E. P. S. 1989, Nucl. Phys. B, 324, 167.
Davis Jr., L., Goldhaber, A. S., and Nieto, M. M. 1975, Phys. Rev.
Lett., 35, 1402.
Davis Jr., R. 1964, Phys. Rev. Lett., 12, 303.
Davis Jr., R., Harmer, D. S., and Hoffman, K. C. 1968, Phys. Rev. Lett.,
20, 1205.
616 References
Davis Jr., R., Mann, A. K., and Wolfenstein, L. 1989, Ann. Rev. Nucl.
Part. Sci., 39, 467.
Dearborn, D. S. P., Schramm, D. N., and Steigman, G. 1986, Phys.
Rev. Lett., 56, 26.
Dearborn, D. S. P., et al. 1990, Ap. J., 354, 568.
Debye, P., and Hückel, E. 1923, Phys. Z., 24, 185.
Degl’Innocenti, S., Fiorentini, G., and Lissia, M. 1995, Nucl. Phys. B
(Proc. Suppl.), 43, 66.
Degl’Innocenti, S., et al. 1995, Report MPI-PTh/95-78 and astro-ph/
9509090, submitted to Astron. Astrophys.
Degrassi, G., Sirlin, A., and Marciano, W. J. 1989, Phys. Rev. D, 39,
287.
del Campo, S., and Ford, L. H. 1988, Phys. Rev. D, 38, 3657.
Demarque, P., et al. 1994, Ap. J., 437, 870.
Dewey, R. J., and Cordes, J. M. 1987, Ap. J., 321, 780.
Dicus, D. A. 1972, Phys. Rev. D, 6, 961.
Dicus, D. A., Kolb, E. W., and Teplitz, V. L. 1977, Phys. Rev. Lett.,
39, 169.
Dicus, D. A., Kolb, E. W., and Tubbs, D. L. 1983, Nucl. Phys. B, 223,
532.
Dicus, D. A., and Repko, W. W. 1993, Phys. Rev. D, 48, 5106.
Dicus, D. A., et al. 1976, Ap. J., 210, 481.
Dicus, D. A., et al. 1978, Phys. Rev. D, 18, 1829.
Dicus, D. A., et al. 1980, Phys. Rev. D, 22, 839.
Dicus, D. A., et al. 1989, Phys. Lett. B, 218, 84.
DiLella, L. 1993, Neutrino 92, Nucl. Phys. B (Proc. Suppl.), 31, 319.
Dimopoulos, S., Starkman, G. D., and Lynn, B. W. 1986a, Phys. Lett.
B, 167, 145.
Dimopoulos, S., Starkman, G. D., and Lynn, B. W. 1986b, Mod. Phys.
Lett. A, 8, 491.
Dimopoulos, S., et al. 1986, Phys. Lett. B, 179, 223.
Dine, M., and Fischler, W. 1983, Phys. Lett. B, 120, 137.
Dine, M., Fischler, W., and Srednicki, M. 1981, Phys. Lett. B, 104, 199.
Dine, M., and Seiberg, N. 1986, Nucl. Phys. B, 273, 109.
Dirac, P. A. M. 1937, Nature, 139, 323.
Dirac, P. A. M. 1938, Proc. R. Soc., A165, 199.
Dobroliubov, M. I., and Ignatiev, A. Yu. 1990, Phys. Rev. Lett., 65,
679.
Dodd, A. C., Papageorgiu, E., and Ranfone, S. 1991, Phys. Lett. B,
266, 434.
Dodelson, S., and Feinberg, G. 1991, Phys. Rev. D, 43, 913.
References 617
Dodelson, S., Frieman, J. A., and Turner, M. S. 1992, Phys. Rev. Lett.,
68, 2572.
Dodelson, S., Gyuk, G., and Turner, M. S. 1994, Phys. Rev. Lett., 72,
3754.
Dolgov, A. D. 1981, Yad. Fiz., 33, 1309 (Sov. J. Nucl. Phys., 33, 700).
Dolgov, A. D., Kainulainen, K., and Rothstein, I. Z. 1995, Phys. Rev.
D, 51, 4129.
Dolgov, A. D., and Rothstein, I. Z. 1993, Phys. Rev. Lett., 71, 476.
Dolgov, A. D., and Raffelt, G. G. 1995, Phys. Rev. D, 52, 2581.
D’Olivo, J. C., Nieves, J. F., and Pal, P. B. 1989, Phys. Rev. D, 40,
3679.
D’Olivo, J. C., Nieves, J. F., and Pal, P. B. 1990, Phys. Rev. Lett., 64,
1088.
Domokos, G., and Kovesi-Domokos, S. 1995, Phys. Lett. B, 346, 317.
Donelly, T. W., et al. 1978, Phys. Rev. D, 18, 1607.
Dorofeev, O. F., Rodionov, V. N., and Ternov, I. M. 1985, Pis’ma As-
tron. Zh., 11, 302 (Sov. Astron. Lett., 11, 123).
Dydak, F., et al. 1984, Phys. Lett. B, 134, 281.
Dziembowski, W. A., et al. 1994, Ap. J., 432, 417.
Eggleton, P. P., and Cannon, R. C. 1991, Ap. J., 383, 757.
Eggleton, P. P., and Faulkner, J. 1981, in: Iben and Renzini (1981).
Ellis, J., and Karliner, M. 1995, Phys. Lett. B, 341, 397.
Ellis, J., and Olive, K. A. 1983, Nucl. Phys. B, 233, 252.
Ellis, J., and Olive, K. A. 1987, Phys. Lett. B, 193, 525.
Ellis, J., and Salati, P. 1990, Nucl. Phys. B, 342, 317.
Ellis, J., et al. 1986, Phys. Lett. B, 167, 457.
Engel, J., Krastev, P. I., and Lande, K. 1995, Report hep-ph/9501219,
submitted to Phys. Rev. D.
Engel, J., Seckel, D., and Hayes, A. C. 1990, Phys. Rev. Lett., 65, 960.
Enqvist, K., Rez, A. I., and Semikoz, V. B. 1995, Nucl. Phys. B, 436,
49.
Enqvist, K., and Uibo, H. 1993, Phys. Lett. B, 301, 376.
Ezer, D., and Cameron, A. G. W. 1966, Can. J. Phys., 44, 593.
Falk, S. W., and Schramm, D. N. 1978, Phys. Lett. B, 79, 511.
Faulkner, J., and Swenson, F. J. 1988, Ap. J., 329, L47.
Fayons, S. A., Kopeykin, V. I., and Mikaelyan, L. A. 1991, quoted after
L. Moscoso, Neutrino 90, Nucl. Phys. B (Proc. Suppl.), 19, 147
(1991).
Feinberg, E. L., and Pomeranchuk, I. 1956, Nuovo Cim. Suppl., 3, 652.
618 References
Festa G. G., and Ruderman, M. A. 1969, Phys. Rev., 180, 1227.

Filippone, B. W., and Vogel, P. 1990, Phys. Lett. B, 246, 546.
Filippone, B. W., et al. 1983, Phys. Rev. C, 28, 2222.
Finley, J. P., Ögelman, H., Kiziloğlu, Ü. 1992, Ap. J., 394, L21.
Fiorentini, G., and Mezzorani, G. 1989, Phys. Lett. B, 221, 353.
Fiorentini, G., et al. 1994, Phys. Rev. D, 49, 6298.
Fischbach, E., and Talmadge, C. 1992, Nature, 356, 207.
Fischbach, E., et al. 1976, Phys. Rev. D, 13, 1523.
Fischbach, E., et al. 1977, Phys. Rev. D, 16, 2377.
Fischbach, E., et al. 1986, Phys. Rev. Lett., 56, 3.
Fischbach, E., et al. 1994, Phys. Rev. Lett., 73, 514.
Fleming, T. A., Liebert, J., and Green, R. F. 1986, Ap. J., 308, 176.
Flowers, E. 1973, Ap. J., 180, 911.
Flowers, E. 1974, Ap. J., 190, 381.
Flynn, J. M., and Randall, L. 1988, Report LBL-25115, unpublished.
Fogli, G. L., and Lisi, E. 1995, Phys. Rev. D, 52, 2775.
Foldy, L. L. 1945, Phys. Rev., 67, 107.
Fomalont, E. B., et al. 1992, Mon. Not. R. astr. Soc., 258, 497.
Foot, R. 1994, Phys. Rev. D, 49, 3617.
Foot, R., and Lew, H. 1993, Mod. Phys. Lett. A, 8, 3767.
Foot, R., Lew, H., and Volkas, R. R. 1993, J. Phys. G: Nucl. Part.
Phys., 19, 361; (E) ibid., 1067.
Forrest, D. J., et al. 1980, Sol. Phys, 65, 15.
Frail, D. A., and Kulkarni, S. R. 1991, Nature, 352, 785.
Freedman, D. Z. 1974, Phys. Rev. D, 9, 1389.
Fréjus Collaboration 1990, Phys. Lett. B, 245, 305.
Fréjus Collaboration 1995, Z. Phys. C, 66, 417.
Frieman, J. A., Dimopoulos, S., and Turner, M. S. 1987, Phys. Rev. D,
36, 2201.
Frieman, J. A., Haber, H. E., and Freese, K. 1988, Phys. Lett. B, 200,
115.
Friman, B. L., and Maxwell, O. V. 1979, Ap. J., 232, 541.
Fritzsch, H., and Plankl, J. 1987, Phys. Rev. D, 35, 1732.
Frogel, J. A., Cohen, J. G., and Persson, S. E. 1983, Ap. J., 275, 773.
Frogel, J. A., Persson, S. E., and Cohen, J. G. 1981, Ap. J., 246, 842.
Frogel, J. A., Persson, S. E., and Cohen, J. G. 1983, Ap. J. Suppl., 53,
713.
Frost, K. J., Rothe, E. D., and Peterson, L. E. 1966, J. Geophys. Res.,
71, 4079.
Fujikawa, K., and Shrock, E. 1980, Phys. Rev. Lett., 45, 963.
Fujiwara, K. 1989, Phys. Rev. D, 39, 1764.
References 619
Fukuda, Y., et al. 1994, Phys. Lett. B, 335, 237.

Fukugita, M., and Sakai, N. 1982, Phys. Lett. B, 114, 23.
Fukugita, M., Watamura, S., and Yoshimura, M. 1982a, Phys. Rev.
Lett., 48, 1522.
Fukugita, M., Watamura, S., and Yoshimura, M. 1982b, Phys. Rev. D,
26, 1840.
Fukugita, M., and Yanagida, T. 1988, Phys. Lett. B, 206, 93.
Fukugita, M., and Yazaki, S. 1987, Phys. Rev. D, 36, 3817.
Fuller, G. M., and Malaney, R. A. 1991, Phys. Rev. D, 43, 3136.
Fuller, G. M., Mayle, R., and Wilson, J. R. 1988, Ap. J., 332, 826.
Fuller, G. M., et al. 1992, Ap. J., 389, 517.
Fusi Pecci, F., et al. 1990, Astron. Astrophys., 238, 95.
Gabriel, M. D., et al. 1991, Phys. Rev. Lett., 67, 2123.
Gaemers, K. J. F., Gandhi, R., and Lattimer, J. M. 1989, Phys. Rev.
D, 40, 309.
Gai, M. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 77.
Gai, M., and Bertulani, C. A. 1995, Phys. Rev. C, 52, 1706.
Galam, S., and Hansen, J.-P. 1976, Phys. Rev. A, 14, 816.
GALLEX Collaboration 1994, Phys. Lett. B, 327, 377.
GALLEX Collaboration 1995a, Phys. Lett. B, 342, 440.
GALLEX Collaboration 1995b, Report GX 75-1995, submitted to Phys.
Lett. B.
Gamow, G. 1967, Proc. Natl. Acad. Sci., 57, 187.
Gamow, G., and Schoenberg, M. 1940, Phys. Rev., 58, 1117.
Gamow, G., and Schoenberg, M. 1941, Phys. Rev., 59, 539.
Gandel’man, G. M., and Pinaev, V. S. 1959, Zh. Eksp. Teor. Fiz., 37,
1072 (Sov. Phys. JETP, 10, 764).
Gandhi, R., and Burrows, A. 1990, Phys. Lett. B, 246, 149; (E) 1991,
ibid., 261, 519.
Garcia A., et al. 1991, Phys. Rev. Lett., 67, 3654.
Garcı́a-Berro, E., et al. 1995, in: D. Koester and K. Werner (eds.),
White Dwarfs, Proc. 9th European Workshop on White Dwarfs,
Kiel, Germany, 29 August–1 Sept. 1994 (Springer, Berlin).
Gasperini, M. 1987, Phys. Rev. Lett., 59, 396.
Gasperini, M. 1988, Phys. Rev. D, 38, 2635.
Gasperini, M. 1989, Phys. Rev. D, 39, 3606.
Gasser, J., and Leutwyler, H. 1982, Phys. Rep., 87, 77.
Gates, E., Krauss, L. M., and White, M. 1995, Phys. Rev. D, 51, 2631.
Gavrin, V. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 60.
Gell-Mann, M. 1961, Phys. Rev. Lett., 6, 70.
620 References
Gelmini, G. B., Nussinov, S., and Roncadelli, M. 1982, Nucl. Phys. B,

209, 157.
Gelmini, G. B., and Roncadelli, M. 1981, Phys. Lett. B, 99, 411.
Georgi, H., Glashow, S. L., and Nussinov, S. 1981, Nucl. Phys. B, 193,
297.
Georgi, H., Hall, L., and Wise, M. 1981, Nucl. Phys. B, 192, 409.
Georgi, H., Kaplan, D. B., and Randall, L. 1986, Phys. Lett. B, 169,
73.
Ghosh, R. K. 1984, Phys. Rev. D, 29, 493.
Ghosh, S. K., Phatak, S. C., and Sahu, P. K. 1994, Mod. Phys. Lett.
A, 9, 1717.
Gilliland, R. L., and Däppen, W. 1987, Ap. J., 313, 429.
Giudice, G. F. 1990, Phys. Lett. B, 251, 460.
Giunti, C., Kim, C. W., and Lam, W. P. 1991, Phys. Rev. D, 43, 164.
Giunti, C., Kim, C. W., and Lee, U. W. 1991, Phys. Rev. D, 44, 3635.
Giunti, C., Kim, C. W., and Lee, U. W. 1992, Phys. Rev. D, 45, 2414.
Giunti, C., et al. 1992, Phys. Rev. D, 45, 1557.
Giunti, C., et al. 1993, Phys. Rev. D, 48, 4310.
Glashow, S. L., Iliopoulos, J., and Maiani, L. 1970, Phys. Rev. D, 2,
1285.
Glass, E. N., and Szamosi, G. 1987, Phys. Rev. D, 35, 1205.
Glass, E. N., and Szamosi, G. 1989, Phys. Rev. D, 39, 1054.
Gnedin, Yu. N., and Krasnikov, S. V. 1992, Zh. Eksp. Teor. Fiz., 102,
1729 (Sov. Phys. JETP, 75, 933).
Goldhaber, A. S., and Nieto, M. M. 1971, Rev. Mod. Phys., 43, 277.
Goldman, I. 1990, Mon. Not. R. astr. Soc., 244, 184.
Goldman, I., et al. 1988, Phys. Rev. Lett., 60, 1789.
Goldman, V. M., Zisman, G. A., and Shaulov, R. Ya. 1972, Tematichen-
skiı̆ Sbornik LGPI (quoted after Blinnikov et al. 1995).
Goldreich, P., and Julian, W. H. 1969, Ap. J., 157, 869.
Goldreich, P., and Weber, S. 1980, Ap. J., 238, 991.
Góngora-T., A., and Stuart, R. G. 1992, Z. Phys. C, 55, 101.
Goodman, M. C. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38,
337.
Goodman, J., Dar, A., and Nussinov, S. 1987, Ap. J., 314, L7.
Gordon, J. P. 1973, Phys. Rev. A, 8, 14.
Gott, J. R., Gunn, J. E., and Ostriker, J. P. 1970, Ap. J., 160, L91.
Gould, R. J. 1985, Ap. J., 288, 789.
Gould, S. J. 1994, Scientific American, 271:4, 85.
Goyal, A., and Anand, J. D. 1990, Phys. Rev. D, 42, 992.
References 621
Goyal, A., Dutta, S., and Choudhury, S. R. 1995, Phys. Lett. B, 346,
312.
Greene, G. L., et al. 1991, Phys. Rev. D, 44, R2216.
Gregores, E. M., et al. 1995, Phys. Rev. D, 51, 4587.
Gribov, N. V., and Pontecorvo, B. M. 1969, Phys. Lett. B, 28, 493.
Grifols, J. A., and Massó, E. 1986, Phys. Lett. B, 173, 237.
Grifols, J. A., and Massó, E. 1989, Phys. Rev. D, 40, 3819.
Grifols, J. A., and Massó, E. 1990a, Phys. Lett. B, 242, 77.
Grifols, J. A., and Massó, E. 1990b, Nucl. Phys. B, 331, 244.
Grifols, J. A., Massó, E., and Peris, S. 1988a, Phys. Lett. B, 207, 493.
Grifols, J. A., Massó, E., and Peris, S. 1988b, Phys. Lett. B, 215, 593.
Grifols, J. A., Massó, E., and Peris, S. 1989a, Phys. Lett. B, 220, 591.
Grifols, J. A., Massó, E., and Peris, S. 1989b, Mod. Phys. Lett. A, 4,
311.
Grifols, J. A., Massó, E., and Peris, S. 1994, Astropart. Phys., 2, 161.
Grifols, J. A., Massó, E., and Rizzo, T. G. 1990, Phys. Rev. D, 42, 3293.
Grifols, J. A., and Tortosa, S. 1994, Phys. Lett. B, 328, 98.
Grimus, W., and Neufeld, H. 1993, Phys. Lett. B, 315, 129.
Guenther, D. B., et al. 1995, Ap. J., 445, 148.
Guzzo, M. M., Bellandi, J., and Aquino, V. M. 1994, Phys. Rev. D, 49,
1404.
Guzzo, M. M., Masiero, A., and Petcov, S. T. 1991, Phys. Lett. B, 260,
154.
Guzzo, M. M., and Petcov, S. T. 1991, Phys. Lett. B, 271, 172.
Guzzo, M. M., and Pulido, J. 1993, Phys. Lett. B, 317, 125.
Gvozdev, A. A., Mikheev, N. V., and Vassilevskaya, L. A. 1992a, Phys.
Lett. B, 289, 103.
Gvozdev, A. A., Mikheev, N. V., and Vassilevskaya, L. A. 1992b, Phys.
Lett. B, 292, 176.
Gvozdev, A. A., Mikheev, N. V., and Vassilevskaya, L. A. 1993, Phys.
Lett. B, 313, 161.
Gvozdev, A. A., Mikheev, N. V., and Vassilevskaya, L. A. 1994a, Phys.
Lett. B, 321, 108.
Gvozdev, A. A., Mikheev, N. V., and Vassilevskaya, L. A. 1994b, Phys.
Lett. B, 323, 179.
Haensel, P., and Jerzak, A. J. 1987, Astron. Astrophys., 179, 127.
Haft, M. 1993, Master’s Thesis, University of Munich, unpublished.
Haft, M., Raffelt, G., and Weiss, A. 1994, Ap. J., 425, 222; (E) 1995,
ibid., 438, 1017.
Hagmann, C., et al. 1990, Phys. Rev. D, 42, 1297.
Hagmann, C., and Sikivie, P. 1991, Nucl. Phys. B, 363, 247.
622 References
Hagner, C., et al. 1995, Phys. Rev. D, 52, 1343.

Halpern, J. P., and Ruderman, M. 1993, Ap. J., 415, 286.
Halprin, A. 1975, Phys. Rev. D, 11, 147.
Halprin, A., and Leung, C. N. 1991, Phys. Rev. Lett., 67, 1833.
Hansen, J.-P. 1973, Phys. Rev. A, 8, 3096.
Harari, D., and Sikivie, P. 1987, Phys. Lett. B, 195, 361.
Harari, D., and Sikivie, P. 1992, Phys. Lett. B, 289, 67.
Harrison, P. A., Lyne, A. G., and Anderson, B. 1993, Mon. Not. R.
astr. Soc., 261, 113.
Harrison, E. R., and Tademaru, E. 1975, Ap. J., 201, 447.
Hata, N., Bludman, S., and Langacker, P. 1994, Phys. Rev. D, 49, 3622.
Hata, N., and Haxton, W. 1995, Phys. Lett. B, 353, 422.
Hata, N., and Langacker, P. 1994, Phys. Rev. D, 50, 632.
Hatsuda, T., Lim, C. S., and Yoshimura, M. 1988, Phys. Lett. B, 203,
462.
Haxton, W. C. 1987, Phys. Rev. D, 36, 2283.
Haxton, W. C. 1995, Report, to appear in Ann. Rev. Astron. Astrophys.
Haxton, W. C., and Zhang, W.-M. 1991, Phys. Rev. D, 43, 2484.
Heisenberg, W., and Euler, H. 1936, Z. Phys., 98, 714.
Hellings, R. W., et al. 1983, Phys. Rev. Lett., 51, 1609.
Henry, R. C., and Feldmann, P. D. 1981, Phys. Rev. Lett., 47, 618.
Herant, M., Benz, W., and Colgate, S. 1992, Ap. J., 395, 642.
Herant, M., et al. 1994, Ap. J., 435, 339.
Hernanz, M., et al. 1994, Ap. J., 434, 652.
Hill, C. T., Steinhardt, P. J., and Turner, M. S. 1990, Phys. Lett. B,
252, 343.
Hill, H. A., and Gu, Ye-ming 1990, Science in China (Series A), 37:7,
854.
Hill, J. E. 1995, Phys. Rev. Lett., 75, 2654.
Hirata, K. S. 1991, Ph. D. Thesis (Univ. of Tokyo); Report ICRR-239-
91-8.
Hirata, K. S., et al. 1987, Phys. Rev. Lett., 58, 1490.
Hirata, K. S., et al. 1988, Phys. Rev. D, 38, 448.
Hirata, K. S., et al. 1991a, Phys. Rev. Lett., 66, 9.
Hirata, K. S., et al. 1991b, Phys. Rev. D, 44, 2241.
Hirata, K. S., et al. 1992, Phys. Lett. B, 280, 146.
Hoffmann, S. 1987, Phys. Lett. B, 193, 117.
Holdom, B. 1986, Phys. Lett. B, 166, 196.
Holman, R., et al. 1992, Phys. Lett. B, 132.
Holzschuh, E., et al. 1992, Phys. Lett. B, 287, 381.
Hoogeveen, F. 1990, Phys. Lett. B, 243, 455.
References 623
Hoogeveen, F., and Stuart, R. G. 1992, Phys. Lett. B, 286, 165.

Hoogeveen, F., and Ziegenhagen, T. 1991, Nucl. Phys. B, 358, 3.
Horowitz, C. J., and Serot, B. D. 1987, Nucl. Phys. A, 464, 613.
Horowitz, C. J., and Wehrberger, K. 1991a, Phys. Rev. Lett., 66, 272.
Horowitz, C. J., and Wehrberger, K. 1991b, Phys. Lett. B, 266, 236.
Horvat, R. 1993, Phys. Rev. D, 48, 2345.
Hubbard, W. B. 1978, Fund. Cosmic Phys., 3, 167.
Hulse, R. A., and Taylor, J. H. 1975, Ap. J., 195, L51.
Iacopini, E., and Zavattini, E. 1979, Phys. Lett. B, 85, 151.
Iben Jr., I., and Laughlin, G. 1989, Ap. J., 341, 312.
Iben Jr., I., and Renzini, A. 1981, eds., Physical Processes in Red Giants
(Reidel, Dordrecht).
Iben Jr., I., and Renzini, A. 1984, Phys. Rep., 105, 329.
Iben Jr., I., and Tutukov, A. V. 1984, Ap. J., 282, 615.
Iglesias, C. A., and Rogers, F. J. 1991a, Ap. J., 371, 408.
Iglesias, C. A., and Rogers, F. J. 1991b, Ap. J., 371, L73.
Iglesias, C. A., Rogers, F. J., and Wilson, B. G. 1990, Ap. J., 360, 221.
Ignatiev, A. Yu., and Joshi, G. C. 1994, Mod. Phys. Lett. A, 9, 1479.
Ignatiev, A. Yu., and Joshi, G. C. 1995, Phys. Rev. D, 51, 2411.
Iida, K., Minakata, H., and Yasuda, O. 1993, Mod. Phys. Lett. A, 8,
1037.
Isern, J., Hernanz, M., and Garcı́a-Berro, E. 1992, Ap. J., 392, L23.
Ishizuka, N., and Yoshimura, M. 1990, Prog. Theor. Phys., 84, 233.
Itoh, N., and Hiraki, K. 1994, Ap. J., 435, 784.
Itoh, N., and Kohyama, Y. 1983, Ap. J., 275, 858.
Itoh, N., et al. 1984a, Ap. J., 279, 413.
Itoh, N., et al. 1984b, Ap. J., 280, 787; (E) 1987, ibid., 322, 584.
Itoh, N., et al. 1989, Ap. J., 339, 354; (E) 1990, ibid., 360, 741.
Itoh, N., et al. 1992, Ap. J., 395, 622; (E) 1993, ibid., 404, 418.
Itzykson, C., and Zuber, J.-B. 1980, Quantum Field Theory (McGraw-
Hill, New York).
Ivanov, M. A., and Shul’man, G. A. 1980, Astron. Zh., 57, 537 (Sov.
Astron., 24, 311).
Ivanov, M. A., and Shul’man, G. A. 1981, Astron. Zh., 58, 138 (Sov.
Astron., 25, 76).
Iwamoto, N. 1980, Phys. Rev. Lett., 44, 1537.
Iwamoto, N. 1982, Ann. Phys. (Leipzig), 141, 1.
Iwamoto, N. 1984, Phys. Rev. Lett., 53, 1198.
Iwamoto, N., and Pethick, C. J. 1982, Phys. Rev. D, 25, 313.
Iwamoto, N., et al. 1995, Phys. Rev. D, 51, 348.
Jackiw, R., and Rebbi, C. 1976, Phys. Rev. Lett., 37, 177.
624 References
Jackson, J. D. 1975, Classical Electrodynamics, 2nd ed. (John Wiley,

New York).
Jancovici, B. 1962, Nuovo Cim., 25, 428.
Janka, H.-T. 1991, Astron. Astrophys., 244, 378.
Janka, H.-T. 1993, in: F. Giovannelli and G. Mannocchi (eds.), Proc.
Vulcano Workshop 1992 Frontier Objects in Astrophysics and Par-
ticle Physics, Conf. Proc. Vol. 40 (Soc. Ital. Fis.).
Janka, H.-T. 1995a, Astropart. Phys., 3, 377.
Janka, H.-T. 1995b, Report astro-ph/9505034, Proc. Ringberg Work-
shop Dark Matter.
Janka, H.-T., and Hillebrandt, W. 1989a, Astron. Astrophys. Suppl.,
78, 375.
Janka, H.-T., and Hillebrandt, W. 1989b, Astron. Astrophys., 224, 49.
Janka, H.-T., and Müller, E. 1993a, in: R. McCray and W. Zhenru
(eds.), Proc. of the IAU Coll. No. 145, Xian, China, May 24–29
(Cambridge University Press, Cambridge).
Janka, H.-T., and Müller, E. 1993b, in: Y. Suzuki and K. Nakamura
(eds.), Frontiers of Neutrino Astrophysics (Universal Academy
Press, Tokyo).
Janka, H.-T., and Müller, E. 1994, Astron. Astrophys., 290, 496.
Janka, H.-T., and Müller, E. 1995a, Phys. Rep., 256, 135.
Janka, H.-T., and Müller, E. 1995b, MPA-Report 863, Astron. Astro-
phys., in press.
Jeckelmann, B., Goudsmit, P. F. A., and Leisi, H. J. 1994, Phys. Lett.
B, 335, 326.
Jodidio, A., et al. 1986, Phys. Rev. D, 34, 1967; (E) 1988, ibid., 37, 237.
Johnson, C. W., et al. 1992, Ap. J., 392, 320.
Joseph, C. L. 1984, Nature, 312, 254.
Jungman, G., Kamionkowski, M., and Griest, K. 1995, Phys. Rep., in
press.
Kainulainen, K., Maalampi, J., and Peltoniemi, J. T. 1991, Nucl. Phys.
B, 358, 435.
Kamionkowski, M., and March-Russell, J. 1992, Phys. Lett. B, 282, 137.
Kandaswamy, J., Salomonson, P., and Schechter, J. 1978, Phys. Rev.
D, 17, 3051.
Kapetanakis, D., Mayr, P., and Nilles, H. P. 1992, Phys. Lett. B, 282,
95.
Kaplan, D. B. 1985, Nucl. Phys. B, 260, 215.
KARMEN Collaboration 1995, Phys. Lett. B, 348, 19.
Kavanagh, R. W. 1960, Nucl. Phys., 15, 411.
Kavanagh, R. W., et al. 1969, Bull. Am. Phys. Soc., 14, 1209.
References 625
Kawakami, H., et al. 1991, Phys. Lett. B, 256, 105.

Kawano, L. H. 1992, Phys. Lett. B, 275, 487.
Kawasaki, M., Terasawa, N., and Sato, K. 1986, Phys. Lett. B, 178, 71.
Kawasaki, M., et al. 1994, Nucl. Phys. B, 419, 105.
Kayser, B. 1982, Phys. Rev. D, 26, 1662.
Keil, W. 1994, Master’s Thesis, University of Munich (unpublished).
Keil, W., and Janka, H.-T. 1995, Astron. Astrophys., 296, 145.
Keil, W., Janka, H.-T., and Raffelt, G. 1995, Phys. Rev. D, 51, 6635.
Keil, W., et al. 1995, Report astro-ph/9507023, submitted to Phys. Rev.
Lett.
Kephart, T. W., and Weiler, T. J. 1987, Phys. Rev. Lett., 58, 171.
Kepler, S. O., et al. 1991, Ap. J., 378, L45.
Kernan, P. J., and Krauss, L. M. 1995, Nucl. Phys. B, 437, 243.
Kie#lczewska, D. 1990, Phys. Rev. D, 41, 2967.
Kikuchi, H., and Ma, E. 1994, Phys. Lett. B, 335, 444.
Kikuchi, H., and Ma, E. 1995, Phys. Rev. D, 51, R296.
Kim, C. W., Kim, J., and Sze, W. K. 1988, Phys. Rev. D, 37, 1072.
Kim, J. E. 1976, Phys. Rev. D, 14, 3000.
Kim, J. E. 1979, Phys. Rev. Lett., 43, 103.
Kim, J. E. 1987, Phys. Rep., 150, 1.
Kippenhahn, R., and Weigert, A. 1990, Stellar Structure and Evolution
(Springer, Berlin).
Kirsten, T. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 68.
Kirzhnits, D. A. 1987, Usp. Fiz. Nauk, 152, 399 (Sov. Phys. Usp., 30,
575).
Kirzhnits, D. A., Losyakov, V. V., and Chechin, V. A. 1990, Zh. Eksp.
Teor. Fiz., 97, 1089 (Sov. Phys. JETP, 70, 609).
Klein, J. R., and Thorsett, S. E. 1990, Phys. Lett. A, 145, 79.
Klimov, V. V. 1981, Sov. J. Nucl. Phys., 33, 934.
Klimov, V. V. 1982, Zh. Eksp. Teor. Fiz., 82, 336 (Sov. Phys. JETP,
55, 199).
Knoll, J., and Voskresensky, D. N. 1995, Phys. Lett. B, 351, 43.
Koester, D., and Schönberner, D. 1986, Astron. Astrophys., 154, 125.
Kolb, E. W., Stebbins, A. J., and Turner, M. S. 1987a, Phys. Rev. D,
35, 3598.
Kolb, E. W., Stebbins, A. J., and Turner, M. S. 1987b, Phys. Rev. D,
36, 3820.
Kolb, E. W., Tubbs, D. L., and Dicus, D. A. 1982, Ap. J., 255, L57.
Kolb, E. W., and Turner, M. S. 1987, Phys. Rev. D, 36, 2895.
Kolb, E. W., and Turner, M. S. 1989, Phys. Rev. Lett., 62, 509.
626 References
Kolb, E. W., and Turner, M. S. 1990, The Early Universe (Addison-

Wesley, Reading, Mass.).
Kolb, E. W., et al. 1991, Phys. Rev. Lett., 67, 533.
Kopysov, Yu. S., and Kuzmin, V. A. 1968, Can. J. Phys., 46, S488.
Kohyama, Y., Itoh, N., and Munakata, H. 1986, Ap. J., 310, 815.
Kohyama, Y., et al. 1993, Ap. J., 415, 267.
Konoplich, R. V., and Khlopov, M. Yu. 1988, Yad. Fiz., 47, 891 (Sov.
J. Nucl. Phys., 47, 565).
Koshiba, M. 1992, Phys. Rep., 220, 229.
Kostelecký, V. A., Pantaleone, J., and Samuel, S. 1993, Phys. Lett. B,
315, 46.
Kovetz, A., and Shaviv, G. 1994, Ap. J., 426, 787.
Krakauer, D. A., et al. 1990, Phys. Lett. B, 252, 177.
Krakauer, D. A., et al. 1991, Phys. Rev. D, 44, R6.
Krastev, P. I. 1993, Phys. Lett. B, 303, 75.
Krastev, P. I., and Petcov, S. T. 1994, Phys. Rev. Lett., 72, 1960.
Krastev, P. I., and Smirnov, A. Yu. 1994, Phys. Lett. B, 338, 282.
Krauss, L. M. 1984, Phys. Rev. Lett., 53, 1976.
Krauss, L. M. 1991, Phys. Lett. B, 263, 441.
Krauss, L. M., Moody, J. E., and Wilczek, F. 1984, Phys. Lett. B, 144,
391.
Krauss, L. M., and Tremaine, S. 1988, Phys. Rev. Lett., 60, 176.
Krauss, L. M., and Wilczek, F. 1986, Phys. Lett. B, 173, 189.
Krauss, L. M., et al. 1985, Phys. Rev. Lett., 55, 1797.
Krauss, L. M., et al. 1992, Nucl. Phys. B, 380, 507.
Kuhn, J. R. 1988, in: J. Christensen-Dalsgaard and S. Frandsen (eds.),
Advances in Helio- and Asteroseismology (Reidel, Dordrecht).
Kunihiro, T., et al. 1993, Prog. Theor. Phys. Suppl., No. 112.
Kuo, T. K., and Pantaleone, J. 1988, Phys. Rev. D, 37, 298.
Kuo, T. K., and Pantaleone, J. 1989, Rev. Mod. Phys., 61, 937.
Kuo, T. K., and Pantaleone, J. 1990, Phys. Lett. B, 246, 144.
Kuznetsov, A. V., and Mikheev, N. V. 1993, Phys. Lett. B, 299, 367.
Kwong, W., and Rosen, S. P. 1994, Phys. Rev. Lett., 73, 369.
Kyuldjiev, A. V. 1984, Nucl. Phys. B, 243.
Lagage, P. O., et al. 1987, Phys. Lett. B, 193, 127.
Lam, W. P., and Ng, K.-W. 1991, Phys. Rev. D, 44, 3345.
Lam, W. P., and Ng, K.-W. 1992, Phys. Lett. B, 284, 331.
Lamb, D. Q., and van Horn, H. M. 1975, Ap. J., 200 306.
Landau, L. D. 1946, Sov. Phys. JETP, 16, 574.
Landau, L. D., and Lifshitz, E. M. 1958, Statistical Physics (Addison-
Wesley, Reading, Mass.).
References 627
Landau, L. D., and Pomeranchuk, I. 1953a, Dokl. Akad. Nauk. SSSR,

92, 535.
Landau, L. D., and Pomeranchuk, I. 1953b, Dokl. Akad. Nauk. SSSR,
92, 735.
Lande, K. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 47, and
unpublished viewgraphs at the conference.
Langacker, P., Leveille, J. P., and Sheiman, J. 1983, Phys. Rev. D, 27,
1228.
Langacker, P., and Liu, J. 1992, Phys. Rev. D, 46, 4140.
Langanke, K., and Shoppa, T. D. 1994, Phys. Rev. C, 49, R1771; (E)
1995, ibid., 51, 2844.
Langanke, K., and Shoppa, T. D. 1995, Phys. Rev. C, 52, 1709.
Langmuir, I. 1926, Proc. Natl. Acad. Sci., 14, 627.
LaRue, G. S., Phillips, J. D., and Fairbank, W. M. 1981, Phys. Rev.
Lett., 46, 967.
Lattimer, J. M., and Cooperstein, J. 1988, Phys. Rev. Lett., 61, 23.
Lattimer, J. M., and Yahil, A. 1989, Ap. J., 340, 426.
Lattimer, J. M., et al. 1991, Phys. Rev. Lett., 66, 2701.
Lattimer, J. M., et al. 1994, Ap. J., 425, 802.
Lazarides, G., Panagiotakopoulos, C., and Shafi, Q. 1986, Phys. Rev.
Lett., 56, 432.
Lazarus, D. M., et al. 1992, Phys. Rev. Lett., 69, 2333.
Learned, J. G., and Pakvasa, S. 1995, Astropart. Phys., 3, 267.
Lee, T. D., and Yang, C. N. 1955, Phys. Rev., 98, 1501.
Lee, Y.-W. 1990, Ap. J., 363, 159.
Lee, Y.-W., Demarque, P., and Zinn, R. 1990, Ap. J., 350, 155.
Lee, Y.-W., Demarque, P., and Zinn, R. 1994, Ap. J., 423, 248.
Leener-Rosier, N. de, et al. 1986, Phys. Lett. B, 177, 228.
Leibundgut, B., et al. 1991, Astron. Astrophys. Suppl., 89, 537.
Leinson, L. B. 1993, Ap. J., 415, 759.
Lenard, A. 1953, Phys. Rev., 90, 968.
Lesko, K. T., et al. 1993, Rev. Mex. Fı́s., 39, Supl. 2, 162.
Levine, M. J. 1967, Nuovo Cim., 48, 67.
Li, L. F., and Wilczek, F. 1982, Phys. Rev. D, 25, 143.
Liebert, J. 1980, Ann. Rev. Astron. Astrophys., 18, 363.
Liebert, J., Dahn, C. C., and Monet, D. G. 1988, Ap. J., 332, 891.
Lim, C.-S., and Marciano, W. J. 1988, Phys. Rev. D, 37, 1368.
Lindley, D. 1979, Mon. Not. R. astr. Soc., 188, 15P.
Lindley, D. 1985, Ap. J., 294, 1.
Lipunov, V. M. 1992, Astrophysics of Neutron Stars (Springer, Berlin).
Liu, J. 1991, Phys. Rev. D, 44, 2879.
628 References
Longo, M. J. 1987, Phys. Rev. D, 36, 3276.

Loredo, T. J., and Lamb, D. Q. 1989, in: E. J. Fenyves (ed.), Fourteenth
Texas Symposium on Relativistic Astrophysics, Ann. N.Y. Acad.
Sci., 571, 601.
Loredo, T. J., and Lamb, D. Q. 1995, Report, submitted to Phys. Rev.
D.
Loreti, F. N., et al. 1995, Report astro-ph/9508106, submitted to Phys.
Rev. D.
LoSecco, J. M. 1988, Phys. Rev. D, 38, 3313.
LoSecco, J. M. 1989, Phys. Rev. D, 39, 1013.
LoSecco, J. M., et al. 1989, Phys. Lett. A, 138, 5.
Loskutov, Yu. M. 1984a, Pis’ma Zh. Eksp. Teor. Fiz., 39, 438 (JETP
Lett., 39, 531).
Loskutov, Yu. M. 1984b, Dokl. Akad. Nauk. SSSR, 275, 1396 (Sov.
Phys. Dokl., 29, 322).
Lucio, J. L., Rosado, A., and Zepeda, A. 1985, Phys. Rev. D, 31, 1091.
Lyne, A. G., and Lorimer, D. R. 1994, Nature, 369, 127.
Lynn, B. W. 1981, Phys. Rev. D, 23, 2151.
Maalampi, J., and Peltoniemi, J. T. 1991, Phys. Lett. B, 269, 357.
Madsen, J., and Haensel, P. 1992, Strange Quark Matter in Physics and
Astrophysics, Nucl. Phys. B (Proc. Suppl.), 24B.
Maiani, L., Petronzio, R., and Zavattini, E. 1986, Phys. Lett. B, 175,
359.
Maki, Z., Nakagawa, M., and Sakata, S. 1962, Prog. Theor. Phys., 28,
870.
Malaney, R. A., Starkman, G. D., and Butler, M. N. 1994, Phys. Rev.
D, 49, 6232.
Malaney, R. A., Starkman, G. D., and Tremaine, S. 1995, Phys. Rev.
D, 51, 324.
Mann, A. K., and Zhang, W. 1990, Comm. Nucl. Part. Phys., 19, 295.
Manohar, A. 1987, Phys. Lett. B, 192, 217.
Marciano, W. J., and Sanda, A. I. 1977, Phys. Lett. B, 67, 303.
Marinelli, M., and Morpurgo, G. 1984, Phys. Lett. B, 137, 439.
Maruno, M., Takasugi, E., and Tanaka, M. 1991, Prog. Theor. Phys.,
86, 907.
Matese, J. J., and O’Connell, R. F. 1969, Phys. Rev., 180, 1289.
Matsuki, S., et al. 1995, in: Proc. XVth Moriond Workshop Dark Matter
in Cosmology, Clocks, and Tests of Fundamental Laws, Villars-sur-
Ollon, Switzerland, January 21–28, 1995.
Maurette, M. 1976, Ann. Rev. Nucl. Sci., 26, 319.
Maxwell, O., et al. 1977, Ap. J., 216, 77.
References 629
Mayle, R. W., and Wilson, J. R. 1993, Phys. Rep., 227, 97.

Mayle, R. W., Wilson, J. R., and Schramm, D. N. 1987, Ap. J., 318,
288.
Mayle, R., et al. 1988, Phys. Lett. B, 203, 188.
Mazurek, T. J. 1974, Nature, 252, 287.
Mazurek, T. J. 1975, Astrophys. Space Sci., 35, 117.
Mazurek, T. J. 1976, Ap. J., 207, L87.
Mazzitelli, I. 1989, Ap. J., 340, 249.
Mazzitelli, I., and D’Antona, F. 1986, Ap. J., 308, 706.
McCray, R. 1993, Ann. Rev. Astron. Astrophys., 31, 175.
McNaught, R. H. 1987, IAU Circular No. 4316.
Melott, A. L., McKay, D. W., and Ralston, J. P. 1988, Ap. J., 324, L43.
Melott, A. L., and Sciama, D. W. 1981, Phys. Rev. Lett., 46, 1369.
Melott, A. L., et al. 1994, Ap. J., 421, 16.
Mestel, L. 1952, Mon. Not. R. astr. Soc., 112, 583.
Mészáros, P. 1992, High-Energy Radiation from Magnetized Neutron
Stars (University of Chicago Press, Chicago).
Meyer, B. S. 1994, Ann. Rev. Astron. Astrophys., 32, 153.
Meyer, B. S. 1995, Ap. J., 449, L55.
Meyer, B. S., et al. 1992, Ap. J., 399, 656.
Midorikawa, S., Terazawa, H., and Akama, K. 1987, Mod. Phys. Lett.
A, 2, 561.
Migdal, A. B. 1956, Phys. Rev., 103, 1811.
Migdal, A. B., et al. 1990, Phys. Rep., 192, 179.
Mikaelian, K. O. 1978, Phys. Rev. D, 18, 3605.
Mikheyev, S. P., and Smirnov, A. Yu. 1985, Yad. Fiz., 42, 1441 (Sov.
J. Nucl. Phys., 42, 913).
Mikheyev, S. P., and Smirnov, A. Yu. 1986, Zh. Eksp. Teor. Fiz., 91, 7
(Sov. Phys. JETP, 64, 4).
Miller, D. S., Wilson, J. R., and Mayle, R. W. 1993, Ap. J., 415, 278.
Minakata, H., and Nunokawa, H. 1988, Phys. Rev. D, 38, 3605.
Minakata, H., and Nunokawa, H. 1995, Phys. Rev. D, 51, 6625.
Minakata, H., et al. 1987, Mod. Phys. Lett. A, 2, 827.
Moe, M. K. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 36.
Moffat, J. W. 1991, in: R. B. Mann and P. Wesson (eds.), Proc. Banff
Summer Institute on Gravitation, Banff, Alberta, 1990 (World Sci-
entific, Singapore).
Mohanty, S., and Nayak, S. N. 1993, Phys. Rev. Lett., 70, 4038; (E)
ibid., 71, 1117.
630 References
Mohanty, S., and Panda, P. K. 1994, Report hep-ph/9403205 (revised

version of 29 Oct. 1994).
Mohapatra, R. N., and Nussinov, S. 1989, Phys. Rev. D, 39, 1378.
Mohapatra, R. N., and Nussinov, S. 1992, Int. J. Mod. Phys. A, 7, 3817.
Mohapatra, R. N., Nussinov, S., and Zhang, X. 1994, Phys. Rev. D, 49,
3434.
Mohapatra, R. N., and Pal, P. 1991, Massive Neutrinos in Physics and
Astrophysics (World Scientific, Singapore).
Mohapatra, R. N., and Rothstein, I. Z. 1990, Phys. Lett. B, 593.
Morgan, J. 1981a, Phys. Lett. B, 102, 247.
Morgan, J. 1981b, Mon. Not. R. astr. Soc., 195, 173.
Mori, M., et al. 1992, Phys. Lett. B, 289, 463.
Morris, D. E. 1986, Phys. Rev. D, 34, 843.
Motobayashi, T., et al. 1994, Phys. Rev. Lett., 73, 2680.
Mourão, A. M., Bento, L., and Kerimov, B. K. 1990, Phys. Lett. B, 237,
469.
Mukhopadhyaya, B., and Gandhi, R. 1992, Phys. Rev. D, 3682.
Müller, J., et al. 1991, Ap. J., 382, L101.
Munakata, H., Kohyama, Y., and Itoh, N. 1985, Ap. J., 296, 197; (E)
1986, ibid., 304, 580.
Munakata, H., Kohyama, Y., and Itoh, N. 1987, Ap. J., 316, 708.
Murdin, P. 1990, End in Fire—The Supernova in the Large Magellanic
Cloud (Cambridge University Press, Cambridge).
Musolf, M. J., and Holstein, B. R. 1991, Phys. Rev. D, 43, 2956.
Muto, T., and Tatsumi, T. 1988, Prog. Theor. Phys., 80, 28.
Muto, T., Tatsumi, T., and Iwamoto, N. 1994, Phys. Rev. D, 50, 6089.
Myra, E. S., and Bludman, S. A. 1989, Ap. J., 340, 384.
Myra, E. S., and Burrows, A. 1990, Ap. J., 364, 222.
Nahmias, M. E. 1935, Proc. Camb. Phil. Soc., 31, 99.
Nakagawa, M., Kohyama, Y., and Itoh, N. 1987, Ap. J., 322, 291.
Nakagawa, M., et al. 1988, Ap. J., 326, 241.
Natale, A. A. 1991, Phys. Lett. B, 258, 227.
Natale, A. A., Pleitez, V., and Tacla, A. 1987, Phys. Rev. D, 36, 3278.
Nieves, J. F. 1982, Phys. Rev. D, 26, 3152.
Nieves, J. F. 1987, Report, Univ. Puerto Rico (unpublished).
Nieves, J. F., and Pal, P. B. 1989a, Phys. Rev. D, 39, 652; (E) ibid., 40,
2148.
Nieves, J. F., and Pal, P. B. 1989b, Phys. Rev. D, 40, 1350.
Nieves, J. F., and Pal, P. B. 1989c, Phys. Rev. D, 40, 1693.
References 631
Nieves, J. F., and Pal, P. B. 1994, Phys. Rev. D, 49, 1398.

Nieves, J. F., Pal, P., and Unger, D. G. 1983, Phys. Rev. D, 28, 908.
Nilles, H. P., and Raby, S. 1982, Nucl. Phys. B, 198, 102.
Nomoto, K., and Tsuruta, S. 1986, Ap. J., 305, L19.
Nomoto, K., and Tsuruta, S. 1987, Ap. J., 312, 711.
Nötzold, D. 1987, Phys. Lett. B, 196, 315.
Nötzold, D. 1988, Phys. Rev. D, 38, 1658.
Nötzold, D., and Raffelt, G. 1988, Nucl. Phys. B, 307, 924.
Nunokawa, H., and Minakata, H. 1993, Phys. Lett. B, 314, 371.
Nussinov, S., and Rephaeli, Y. 1987, Phys. Rev. D, 36, 2278.
Oakley, D. S., et al. 1994, Ap. J., 437, L63.
Oberauer, L. 1992 Nucl. Phys. B (Proc. Suppl.), 28A, 165.
Oberauer, L., von Feilitzsch, F., and Mössbauer, R. L. 1987, Phys. Lett.
B, 198, 113.
Oberauer, L., et al. 1993, Astropart. Phys., 1, 377.
O’Connell, R. F., and Matese, J. J. 1969a, Phys. Lett. A, 29, 533.
O’Connell, R. F., and Matese, J. J. 1969b, Nature, 222, 649.
Ögelman, H., and Finley, J. 1993, Ap. J., 413, L31.
Ögelman, H., Finley, J., and Zimmermann, H. 1993, Nature, 361, 136.
Okun, L. B. 1969, Yad. Fiz., 10, 358 (Sov. J. Nucl. Phys., 10, 206).
Olive, K. A., Schramm, D., and Steigman, G. 1981, Nucl. Phys. B, 180,
497.
Olive, K. A., et al. 1990, Phys. Lett. B, 236, 454.
Oraevskiı̆, V. N., and Semikoz, V. B. 1984, Zh. Eksp. Teor. Fiz., 86,
796 (Sov. Phys. JETP, 59, 465).
Oraevskiı̆, V. N., and Semikoz, V. B. 1985, Yad. Fiz., 42, 702 (Sov. J.
Nucl. Phys., 42, 446).
Oraevskiı̆, V. N., and Semikoz, V. B. 1987, Physica, 142A, 135.
Oraevskiı̆, V. N., and Semikoz, V. B. 1991, Phys. Lett. B, 263, 455.
Oraevskiı̆, V. N., and Semikoz, V. B., and Smorodinskiı̆, Ya. A. 1986,
Pis’ma Zh. Eksp. Teor. Fiz., 43, 549 (JETP Lett., 43, 709).
Otten, E. W. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 26.
Overduin, J. M., and Wesson, P. S. 1993, Ap. J., 414, 449.
Overduin, J. M., Wesson, P. S., and Bowyer, S. 1993, Ap. J., 404, 460.
Page, D., and Applegate, J. H. 1992, Ap. J., 394, L17.
Pakvasa, S., Simmons, W. A., and Weiler, T. J. 1989, Phys. Rev. D,
39, 1761.
Pal, P. B., and Pham, T. N. 1989, Phys. Rev. D, 40, 259.
Pal, P. B., and Wolfenstein, L. 1982, Phys. Rev. D, 25, 766.
632 References
Pantaleone, J. 1991, Phys. Lett. B, 268, 227.

Pantaleone, J. 1992a, Phys. Rev. D, 46, 510.
Pantaleone, J. 1992b, Phys. Lett. B, 287, 128.
Pantaleone, J. 1995, Phys. Lett. B, 342, 250.
Pantaleone, J., Halprin, A., and Leung, C. N. 1993, Phys. Rev. D, 47,
R4199.
Pantziris, A., and Kang, K. 1986, Phys. Rev. D, 33, 3509.
Papini, G., and Valluri, S. R. 1977, Phys. Rep., 33, 51.
Parfenov, K. V. 1989a, Yad. Fiz., 48, 1023 (Sov. J. Nucl. Phys., 48,
651).
Parfenov, K. V. 1989b, Yad. Fiz., 49, 1820 (Sov. J. Nucl. Phys., 49,
1127).
Parke, S. 1995, Phys. Rev. Lett., 74, 839.
Parker, P. D. 1966, Phys. Rev., 150, 851.
Parker, P. D. 1968, Ap. J., 153, L85.
Particle Data Group 1994, Phys. Rev. D, 50, 1173.
Paschos, E. A., and Zioutas, K. 1994, Phys. Lett. B, 323, 367.
Peccei, R. D. 1981, in: M. Konuma and T. Maskawa (eds.), Proc. Fourth
Kyoto Summer Institute on Grand Unified Theories and Related
Topics (World Science, Singapore).
Peccei, R. D. 1989, in: C. Jarlskog (ed.), CP Violation (World Scientific,
Singapore).
Peccei, R. D., and Quinn, H. R. 1977a, Phys. Rev. Lett., 38, 1440.
Peccei, R. D., and Quinn, H. R. 1977b, Phys. Rev. D, 16, 1791.
Peccei, R. D., Wu, T. T., and Yanagida, T. 1986, Phys. Lett. B, 172,
435.
Peebles, P. J. E. 1993, Principles of Physical Cosmology (Princeton
University Press, Princeton, N.J.).
Peierls, Sir R. 1976, Proc. R. Soc. London, A 347, 475.
Pérez, A., and Gandhi, R. 1990, Phys. Rev. D, 41, 2374.
Peterson, L. E., et al. 1966, J. Geophys. Res., 71, 5778.
Pethick, C. J., and Thorsson, V. 1994, Phys. Rev. Lett., 72, 1964.
Petschek, A. G. 1990, ed., Supernovae (Springer, New York).
Pinaev, V. S. 1963, Zh. Eksp. Teor. Fiz., 45, 548 (1964, Sov. Phys.
JETP, 18, 377).
Pines, D., Tamagaki, R., and Tsuruta, S. 1992, The Structure and Evo-
lution of Neutron Stars, Conf. Proc. (Addison-Wesley, Redwood
City).
Pines, D., and Nozières, P. 1966, The Theory of Quantum Liquids Vol. I:
Normal Fermi Liquids (Benjamin, New York).
Pisarski, R. D. 1989, Nucl. Phys. A, 498, 423.
References 633
Pochoda, P., and Schwarzschild, M. 1964, Ap. J., 139, 587.

Pogosyan, D., and Starobinsky, A. 1995, Ap. J., 447, 465.
Poincaré, H. 1892, Théorie Mathématique de la Lumière, Vol. 2 (Geor-
ges Carré, Paris).
Pontecorvo, B. 1948, Chalk River Laboratory Report PD-205 (quoted
after Davis, Mann, and Wolfenstein 1989).
Pontecorvo, B. 1957, Zh. Eksp. Teor. Fiz., 33, 549 (1958, Sov. Phys.
JETP, 6, 429).
JETP, 7, 172).
JETP, 9, 1148).
JETP, 26, 984).
Pontecorvo, B. 1983, Usp. Fiz. Nauk, 141, 675 (1983, Sov. Phys. Usp.,
26, 1087).
Prather, M. J. 1976, The Effect of a Brans-Dicke Cosmology upon Stel-
lar Evolution and the Evolution of Galaxies (Ph.D. thesis, Yale
University).
Preskill, J., Wise, M., and Wilczek, F. 1983, Phys. Lett. B, 120, 127.
Press, W. H., et al. 1986, Numerical Recipes—The Art of Scientific
Computing (Cambridge University Press, Cambridge).
Primakoff, H. 1951, Phys. Rev., 81, 899.
Proffitt, C. R. 1994, Ap. J., 425, 849.
Proffitt, C. R., and Michaud, G. 1991, Ap. J., 371, 584.
Pulido, J. 1992, Phys. Rep., 211, 167.
Pulido, J. 1993, Phys. Rev. D, 48, 1492.
Pulido, J. 1994, Phys. Lett. B, 323, 36.
Qian, Y.-Z., and Fuller, G. M. 1995, Phys. Rev. D, 51, 1479.
Qian, Y.-Z., et al. 1993, Phys. Rev. Lett., 71, 1965.
Raffelt, G. 1985, Phys. Rev. D, 31, 3002.
Raffelt, G. 1986a, Phys. Rev. D, 33, 897.
Raffelt, G. 1986b, Phys. Lett. B, 166, 402.
Raffelt, G. 1988b, Phys. Rev. D, 38, 3811.
Raffelt, G. 1990b, Ap. J., 365, 559.
Raffelt, G. 1990c, Mod. Phys. Lett. A, 5, 2581.
Raffelt, G. 1990d, Phys. Rep., 198, 1.
634 References
Raffelt, G., and Dearborn, D. 1987, Phys. Rev. D, 36, 2211.

Raffelt, G., and Dearborn, D. 1988, Phys. Rev. D, 37, 549.
Raffelt, G., Dearborn, D., and Silk, J. 1989, Ap. J., 336, 64.
Raffelt, G., and Seckel, D. 1988, Phys. Rev. Lett., 60, 1793.
Raffelt, G., and Seckel, D. 1991, Phys. Rev. Lett., 67, 2605.
Raffelt, G., and Seckel, D. 1995, Phys. Rev. D, 52, 1780.
Raffelt, G., and Sigl, G. 1993, Astropart. Phys., 1, 165.
Raffelt, G., Sigl, G., and Stodolsky, L. 1993, Phys. Rev. Lett., 70, 2363.
Raffelt, G., and Starkman, G. 1989, Phys. Rev. D, 40, 942.
Raffelt, G., and Stodolsky, L. 1982, Phys. Lett. B, 119, 323.
Raffelt, G., and Stodolsky, L. 1988, Phys. Rev. D, 37, 1237.
Raffelt, G., and Weiss, A. 1992, Astron. Astrophys., 264, 536.
Raffelt, G., and Weiss, A. 1995, Phys. Rev. D, 51, 1495.
Raghavan, R. S. 1991, in: K. K. Phua and Y. Yamaguchi (eds.), Proc.
25th Int. Conf. High-Energy Physics, 2–8 August 1990, Singapore
(South East Asia Theoret. Phys. Assoc. and Phys. Soc. of Japan).
Raghavan, R. S., He, X.-G., and Pakvasa, S. 1988, Phys. Rev. D, 38,
1317.
Raghavan, R. S., Pakvasa, S., and Brown, B. A. 1986, Phys. Rev. Lett.,
57, 1801.
Rajpoot, S. 1993, Mod. Phys. Lett. A, 8, 1179.
Raychaudhuri, P. 1971, Astrophys. Space Sci., 13, 231.
Raychaudhuri, P. 1986, Solar Phys., 106, 421.
Raychaudhuri, P. 1991, Mod. Phys. Lett. A, 6, 2003.
Reines, F., Gurr, H., and Sobel, H. 1976, Phys. Rev. Lett., 37, 315.
Reines, F., Sobel, H., and Gurr, H. 1974, Phys. Rev. Lett., 32, 180.
Renzini, A., and Fusi Pecci, F. 1988, Ann. Rev. Astron. Astrophys., 26,
199.
Renzini, A., et al. 1992, Ap. J., 400, 280.
Ressell, M. T. 1991, Phys. Rev. D, 44, 3001.
Ressell, M. T., and Turner, M. S. 1990, Comm. Astrophys., 14, 323.
Rich, J. 1993, Phys. Rev. D, 48, 4318.
Rich, J., Lloyd Owen, D., and Spiro, M. 1987, Phys. Rep., 151, 239.
Riisager, K., and Jensen, A. S. 1993, Phys. Lett. B, 301, 6.
Ritus, V. I. 1961, Zh. Eksp. Teor. Fiz., 41, 1285 (1962, Sov. Phys. JETP,
14, 915).
Rizzo, T. G. 1991, Phys. Rev. D, 44, 202.
Robertson, R. G. H., et al. 1991, Phys. Rev. Lett., 67, 957.
Roeder, R. C. 1967, Ap. J., 149, 131.
Roeder, R. C., and Demarque, P. R. 1966, Ap. J., 144, 1016.
Rosen, S. P. 1988, Phys. Rev. D, 37, 1682.
References 635
Ross, J. E., and Aller, L. H. 1976, Science, 191, 1223.

Rothman, T., and Matzner, R. 1982, Ap. J., 257, 450.
Rothstein, I. Z., Babu, K. S., and Seckel, D. 1993, Nucl. Phys. B, 403,
725.
Rudzsky, M. A. 1990, Astrophys. Space Sci., 165, 65.
Ruoso, G., et al. 1992, Z. Phys. C, 56, 505.
Ryan, J. J., Accetta, F., and Austin, R. H. 1985, Phys. Rev. D, 32, 802.
Sakuda, M. 1994, Phys. Rev. Lett., 72, 804.
Sakurai, J. J. 1967, Advanced Quantum Mechanics (Addison-Wesley,
Reading, Mass.).
Salati, P. 1994, Astropart. Phys., 2, 269.
Salpeter, E. E. 1960, Phys. Rev., 120, 1528.
Saltzberg, D. 1995, Phys. Lett. B, 355, 499.
Samuel, S. 1993, Phys. Rev. D, 48, 1462.
Sandage, A. 1986, Ann. Rev. Astron. Astrophys., 24, 421.
Sandage, A. 1990a, Ap. J., 350, 603.
Sandage, A. 1990b, Ap. J., 350, 631.
Sandage, A., and Cacciari, C. 1990, Ap. J., 350, 645.
Sarajedini, A., and Demarque, P. 1990, Ap. J., 365, 219.
Sato, K. 1975, Prog. Theor. Phys., 54, 1352.
Sato, K, and Sato, H. 1975, Prog. Theor. Phys., 54, 1564.
Sato, K., Shimizu, T., and Yamada, S. 1993, in: Y. Suzuki and K. Naka-
mura (eds.), Frontiers of Neutrino Astrophysics (Universal Acade-
my Press, Tokyo).
Sato, K., and Suzuki, H. 1987a, Phys. Rev. Lett., 58, 2722.
Sato, K., and Suzuki, H. 1987b, Phys. Lett. B, 196, 267.
Sawyer, R. F. 1988, Phys. Rev. Lett., 61, 2171.
Sawyer, R. F. 1989, Phys. Rev. C, 40, 865.
Sawyer, R. F. 1995, Phys. Rev. Lett., 75, 2260.
Schäfer, G., and Dehnen, H. 1983, Phys. Rev. D, 27, 2864.
Schechter, J., and Valle, J. W. F. 1981, Phys. Rev. D, 24, 1883.
Scherrer, R. J., and Spergel, D. N. 1993, Phys. Rev. D, 47, 4774.
Schinder, P. J., et al. 1987, Ap. J., 313, 531.
Schmidt, G. 1989, ed., The Use of Pulsating Stars in Fundamental Prob-
lems of Astronomy (Cambridge University Press, Cambridge).
Schneps, J. 1993, Neutrino 92, Nucl. Phys. B (Proc. Suppl.), 31, 307.
Schneps, J. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 220.
Schönfelder, V., Graml, F., and Penningfeld, F.-P. 1980, Ap. J., 240,
350.
Schramm, D. N. 1987, Comm. Nucl. Part. Phys., 17, 239.
Schramm, D. N., and Truran, J. W. 1990, Phys. Rep., 189, 89.
636 References
Schreckenbach, K., et al. 1985, Phys. Lett. B, 160, 325.

Schwarzschild, M. 1958, Structure and Evolution of the Stars (Princeton
University Press, Princeton, N.J.).
Sciama, D. W. 1990a, Phys. Rev. Lett., 65, 2839.
Sciama, D. W. 1990b, Ap. J., 364, 549.
Sciama, D. W. 1993a, Ap. J., 409, L25.
Sciama, D. W. 1993b, Modern Cosmology and the Dark Matter Problem
(Cambridge University Press, Cambridge).
Sciama, D. W. 1995, Ap. J., 448, 667.
Sciama, D. W., Persic, M., and Salucci, P. 1993, Publ. Astr. Soc. Pacific,
105, 102.
Seckel, D., Steigman, G., and Walker, T. 1991, Nucl. Phys. B, 366, 233.
Segretain, L., et al. 1994, Ap. J., 434, 641.
Sehgal, L. M., and Weber, A. 1992, Phys. Rev. D, 46, 2252.
Semertzidis, Y., et al. 1990, Phys. Rev. Lett., 64, 2988.
Semikoz, V. B. 1987a, Yad. Fiz., 46, 1592 (Sov. J. Nucl. Phys., 46, 946).
Semikoz, V. B. 1987b, Physica, 142A, 157.
Semikoz, V. B. 1992, Phys. Lett. B, 284, 337.
Semikoz, V. B., and Smorodinskiı̆, Ya. A. 1988, Pis’ma Zh. Eksp. Teor.
Fiz., 48, 361 (JETP Lett., 48, 399).
Semikoz, V. B., and Smorodinskiı̆, Ya. A. 1989, Zh. Eksp. Teor. Fiz.,
95, 35 (Sov. Phys. JETP, 68, 20).
Senatorov, A. V., and Voskresensky, D. N. 1987, Phys. Lett. B, 184,
119.
Shapiro, I. I. 1990, in: N. Ashby, D. F. Bartlett and W. Wyss (eds.),
General Relativity and Gravitation (Cambridge University Press,
Cambridge).
Shapiro, S. L., and Teukolsky, S. A. 1983, Black Holes, White Dwarfs,
and Neutron Stars (John Wiley, New York).
Shapiro, S. L., and Wasserman, I. 1981, Nature, 289, 657.
Shaviv, G., and Bahcall, J. N. 1969, Ap. J., 155, 135.
Shaviv, G., and Kovetz, A. 1976, Astron. Astrophys., 51, 383.
Shelton, I. 1987, IAU Circular No. 4316.
Shi, X., and Schramm, D. N. 1994, Particle World, 3, 109.
Shi, X., Schramm, D. N., and Dearborn, D. S. P. 1994, Phys. Rev. D,
50, 2414.
Shi, X., and Sigl, G. 1994, Phys. Lett. B, 323, 360.
Shi, X., et al. 1993, Comm. Nucl. Part. Phys., 21, 151.
Shifman, M. A., Vainshtein, A. I., and Zakharov, V. I. 1980, Nucl. Phys.
B, 166, 493.
Shklovskii, I. S. 1970, Astron. Zh., 46, 715.
References 637
Shlyakhter, A. I. 1976, Nature, 264, 340.

Shlyakhter, A. I. 1983, ATOMPKI Report A/1 (quoted after Scherrer
and Spergel 1993).
Shrock, R. E. 1981, Phys. Rev. D, 24, 1275.
Shrock, R. E. 1982, Nucl. Phys. B, 206, 359.
Shu, F. 1982, The Physical Universe—An Introduction to Astronomy
(University Science Books, Mill Valley, Calif.).
Sigl, G. 1995a, Phys. Rev. D, 51, 4035.
Sigl, G. 1995b, Fermilab-Pub-95/274-A, submitted to Phys. Rev. Lett.
Sigl, G., and Raffelt, G. 1993, Nucl. Phys. B, 406, 423.
Sigl, G., and Turner, M. S. 1995, Phys. Rev. D, 51, 1499.
Sikivie, P. 1983, Phys. Rev. Lett., 51, 1415.
Sikivie, P. 1984, Phys. Lett. B, 137, 353.
Sikivie, P. 1985, Phys. Rev. D, 32, 2988; (E) 1987, ibid., 36, 974.
Sikivie, P. 1987, in: E. Alvarez et al. (eds.), Cosmology and Particle
Physics (World Scientific, Singapore).
Sikivie, P. 1988, Phys. Rev. Lett., 61, 783.
Silin, V. P. 1960, Zh. Eksp. Teor. Fiz., 38, 1577 (Sov. Phys. JETP, 11,
1136).
Simpson, J. J. 1991, Phys. Lett. B, 269, 454.
Sitenko, A. G. 1967, Electromagnetic Fluctuations in Plasma (Aca-
demic Press, New York).
Skibo, J. G., Ramaty, R., and Leventhal, M. 1992, Ap. J., 397, 135.
Slad’, L. M. 1983, Pis’ma Zh. Eksp. Teor. Fiz., 37, 115 (JETP Lett.,
37, 143).
Slattery, W. L., Doolen, G. D., and DeWitt, H. E. 1980, Phys. Rev. A,
21, 2087.
Slattery, W. L., Doolen, G. D., and DeWitt, H. E. 1982, Phys. Rev. A,
26, 2255.
Smirnov, A. Yu. 1987, in: V. A. Kozyarivsky (ed.), Proc. Twentieth
International Cosmic Ray Conference (Nauka, Moscow).
Smirnov, A. Yu. 1991, Phys. Lett. B, 260, 161.
Smirnov, A. Yu., Spergel, D. N., and Bahcall, J. N. 1994, Phys. Rev.
D, 49, 1389.
Sneden, C., Pilachewski, C. A., and VandenBerg, D. A. 1986, Ap. J.,
311, 826.
Soares, J. M., and Wolfenstein, L. 1989, Phys. Rev. D, 40, 3666.
Sommerfeld, A. 1958, Optik (Akademische Verlagsgesellschaft, Leipzig).
Spergel, D. N., and Bahcall, J. N. 1988, Phys. Lett. B, 200, 366.
Spiro, M., and Vignaud, D. 1990, Phys. Lett. B, 242, 279.
Srednicki, M. 1985, Nucl. Phys. B, 260, 689.
638 References
Steigman, G., and Turner, M. S. 1985, Nucl. Phys. B, 253, 375.

Stewart, R. T., et al. 1993, Mon. Not. R. astr. Soc., 261, 593.
Stix, M. 1989, The Sun—An Introduction (Springer, Berlin).
Stodolsky, L. 1987, Phys. Rev. D, 36, 2273.
Stodolsky, L. 1988, Phys. Lett. B, 201, 353.
Stoeffl, W., and Decman, D. J. 1994, submitted to Phys. Rev. Lett.
Stothers, R. 1970, Phys. Rev. Lett., 24, 538.
Stothers, R. 1972, Ap. J., 175, 717.
Strobl, K., and Weiler, T. J. 1994, Phys. Rev. D, 50, 7690.
Stückelberg, E. C. G. 1941, Helv. Phys. Acta, 14, 51.
Subramanian, A. 1979, Current Science, 48, 705.
Sudbury Neutrino Observatory Collaboration 1987, Phys. Lett. B, 194,
321.
Sur, B., and Boyd, R. N. 1985, Phys. Rev. Lett., 54, 485.
Sutherland, P., et al. 1976, Phys. Rev. D, 13, 2700.
Suzuki, Y. 1993, in: Proc. Int. Symposium on Neutrino Astrophysics
(Takayama Kamioka, Japan, Oct. 19–22, 1992).
Suzuki, Y. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 54.
Sweigart, A. V. 1994, Ap. J., 426, 612.
Sweigart, A. V., Greggio, L., and Renzini, A. 1990, Ap. J., 364, 527.
Sweigart, A. V., and Gross, P. G. 1976, Ap. J. Suppl., 32, 367.
Sweigart, A. V., and Gross, P. G. 1978, Ap. J. Suppl., 36, 405.
Sweigart, A. V., Renzini, A., and Tornambè, A. 1987, Ap. J., 312, 762.
Takahara, M., and Sato, K. 1986, Phys. Lett. B, 174, 373.
Takahara, M., and Sato, K. 1987, Mod. Phys. Lett. A, 2, 293.
Takahashi, K., and Boyd, R. N. 1988, Ap. J., 327, 1009.
Takahashi, K., Witti, J., and Janka, H.-Th. 1994, Astron. Astrophys.,
286, 857.
Takasugi, E., and Tanaka, M. 1992, Prog. Theor. Phys., 87, 679.
Tammann, G. A., Löffler, W., and Schröder, A. 1994, Ap. J. Suppl., 92,
487.
Taylor, J. H., and Weisberg, J. M. 1989, Ap. J., 345, 434.
Teller, E. 1948, Phys. Rev., 73, 801.
Thompson Jr., R. J., et al. 1991, Ap. J., 366, L83.
’t Hooft, G. 1971, Phys. Lett. B, 37, 195.
’t Hooft, G. 1976a, Phys. Rev. Lett., 37, 8.
’t Hooft, G. 1976b, Phys. Rev. D, 14, 3432.
Thorsson, V. 1995, Report nucl-th-9502004.
Tinsley, B. M., and Gunn, J. E. 1976, Ap. J., 206, 525.
Totani, T., and Sato, K. 1995, Astropart. Phys., 3, 367.
Totsuka, Y. 1990, Report ICRR-Report-227-90-20, unpublished.
References 639
Totsuka, Y. 1993, Neutrino 92, Nucl. Phys. B (Proc. Suppl.), 31, 428.
Toussaint, D., and Wilczek, F. 1981, Nature, 289, 777.
Tsai, W., and Erber, T. 1975, Phys. Rev. D, 12, 1132.
Tsai, W., and Erber, T. 1976, Acta Phys. Austr., 45, 245.
Tsuruta, S. 1986, Comm. Astrophys., 11, 151.
Tsuruta, S. 1992, in: Pines, Tamagaki, and Tsuruta 1992.
Tsuruta, S., and Nomoto, K. 1987, in: A. Hewitt et al. (eds.), Obser-
vational Cosmology (IAU Sympsium No. 124).
Tsytovich, V. N. 1961, Zh. Eksp. Teor. Fiz., 40, 1775 (Sov. Phys. JETP,
13, 1249).
Tsytovich, V. N., et al. 1995, Phys. Lett. A, 205, 199.
Turck-Chièze, S., and Lopes, I. 1993, Ap. J., 408, 347.
Turck-Chièze, S., et al. 1993, Phys. Rep., 230, 57.
Turner, M. S. 1986, Phys. Rev. D, 33, 889.
Turner, M. S. 1987, Phys. Rev. Lett., 59, 2489.
Turner, M. S. 1988, Phys. Rev. Lett., 60, 1797.
Turner, M. S. 1992, Phys. Rev. D, 45, 1066.
Turner, M. S., Kang, H.-S., and Steigman, G. 1989, Phys. Rev. D, 40,
299.
Umeda, H., Nomoto, K., and Tsuruta, S. 1994, Ap. J., 431, 309.
Umeda, H., Tsuruta, S., and Nomoto, K. 1994, Ap. J., 433, 256.
Ushida, N., et al. 1986, Phys. Rev. Lett., 57, 2897.
Valle, J. W. F. 1987, Phys. Lett. B, 199, 432.
van Bibber, K., et al. 1987, Phys. Rev. Lett., 59, 759.
van Bibber, K., et al. 1989, Phys. Rev. D, 39, 2089
van Bibber, K., et al. 1992, Search for Pseudoscalar Cold Dark Matter,
Experimental Proposal, unpublished.
van Bibber, K., et al. 1994, Status of the Large-Scale Dark-Matter Axion
Search, Report UCRL-JC-118357 (Lawrence Livermore National
Laboratory).
van Buren, D., and Greenhouse, M. A. 1994, Ap. J., 431, 640.
VandenBerg, D. A., Bolte, M., and Stetson, P. B. 1990, Astron. J., 100,
445.
van den Bergh, S., and Tammann, G. A. 1991, Ann. Rev. Astron. As-
trophys., 29, 363.
van der Velde, J. C. 1989, Phys. Rev. D, 39, 1492.
van Horn, H. M. 1971, in: W. J. Luyten (ed.), White Dwarfs, IAU-
Symposium No. 42 (Reidel, Dordrecht).
Vassiliadis, G., et al. 1995, Proc. Int. Symp. Strangeness and Quark
Matter , Sept. 1–5, 1994, Crete, Greece (World Scientific, Singa-
pore).
640 References
Vaughn, F. J., et al. 1970, Phys. Rev. C, 2, 1657.

Vidyakin, G. S., et al. 1987, Zh. Eksp. Teor. Fiz., 93, 424 (Sov. Phys.
JETP, 66, 243).
Vidyakin, G. S., et al. 1990, Zh. Eksp. Teor. Fiz., 98, 764 (Sov. Phys.
JETP, 71, 424).
Vidyakin, G. S., et al. 1991, J. Moscow Phys. Soc., 1, 85.
Vidyakin, G. S., et al. 1992, Pis’ma Zh. Eksp. Teor. Fiz., 55, 212 (JETP
Lett., 55, 206).
Vila, S. C. 1976, Ap. J., 206, 213.
Vogel, P. 1984, Phys. Rev. D, 29, 1918.
Vogel, P. 1984, Phys. Rev. D, 30, 1505.
Vogel, P., and Engel, J. 1989, Phys. Rev. D, 39, 3378.
Voloshin, M. B. 1988, Phys. Lett. B, 209, 360.
Voloshin, M. B., and Vysotskiı̆, M. I. 1986, Yad. Fiz., 44, 845 (Sov. J.
Nucl. Phys., 44, 544).
Voloshin, M. B., Vysotskiı̆, M. I., and Okun, L. B. 1986a, Yad. Fiz.,
44, 677 (Sov. J. Nucl. Phys., 44, 440).
Voloshin, M. B., Vysotskiı̆, M. I., and Okun, L. B. 1986b, Zh. Eksp.
Teor. Fiz., 91, 754 (Sov. Phys. JETP, 64, 446); (E) 1987, ibid., 92,
368 (ibid., 65, 209).
von Feilitzsch, F., Hahn, A. A., and Schreckenbach, K. 1982, Phys. Lett.
B, 118, 162.
von Feilitzsch, F., and Oberauer, L. 1988, Phys. Lett. B, 200, 580.
Vorobyov, P. V., and Kolokolov, I. V. 1995, Report astro-ph/9501042.
WA66 Collaboration 1985, Phys. Lett. B, 160, 207.
Walker, A. R. 1992, Ap. J., 390, L81.
Walker, T. P., and Schramm, D. N. 1987, Phys. Lett. B, 195, 331.
Walker, T. P., et al. 1991, Ap. J., 376, 51.
Wang, J. 1992, Mod. Phys. Lett. A, 7, 1497.
Weidemann, V. 1990, Ann. Rev. Astron. Astrophys., 28, 103.
Weidemann, V., and Koester, D. 1984, Astron. Astrophys., 132, 195.
Weinberg, S. 1975, Phys. Rev. D, 11, 3583.
Weinberg, S. 1978, Phys. Rev. Lett., 40, 223.
Weinheimer, C., et al. 1993, Phys. Lett. B, 300, 210.
Weldon, H. A. 1982a, Phys. Rev. D, 26, 1394.
Weldon, H. A. 1982b, Phys. Rev. D, 26, 2789.
Weldon, H. A. 1989, Phys. Rev. D, 40, 2410.
Werntz, C. W. 1970, unpublished (quoted after Cisneros 1971).
Wheeler, J. C., Sneden, C., and Truran, J. W. 1989, Ann. Rev. Astron.
Astrophys., 252, 279.
White, M., Gelmini, G., and Silk, J. 1995, Phys. Rev. D, 51, 2669.
References 641
Wiezorek, C., et al. 1977, Z. Phys. A, 282, 121.

Wilczek, F. 1978, Phys. Rev. Lett., 40, 279.
Will, C. M. 1993, Theory and Experiment in Gravitational Physics—
Revised Edition (Cambridge University Press, Cambridge).
Wilson, J. R. 1983, in: J. Centrella, J. LeBlanc, and R. L. Bowers (eds.),
Numerical Astrophysics (Jones and Bartlett, Boston).
Wilson, J. R., and Mayle, R. W. 1988, Phys. Rep., 163, 63.
Winget, D. E., et al. 1987, Ap. J., 315, L77.
Winget, D. E., Hansen, C. J., and van Horn, H. M. 1983, Nature, 303,
781.
Winter, K. 1991, ed., Neutrino Physics (Cambridge University Press).
Winter, K. 1995, Neutrino 94, Nucl. Phys. B (Proc. Suppl.), 38, 211.
Wise, M. B., Georgi, H., and Glashow, S. L. 1981, Phys. Rev. Lett., 47,
402.
Witti, J., Janka, H.-Th., and Takahashi, K. 1994, Astron. Astrophys.,
286, 841.
Wolfenstein, L. 1978, Phys. Rev. D, 17, 2369.
Wolfenstein, L. 1986, Ann. Rev. Nucl. Part. Sci., 36, 137.
Wolfenstein, L. 1987, Phys. Lett. B, 194, 197.
Wood, M. A. 1992, Ap. J., 386, 539.
Woosley, S. E., and Hoffmann, R. D. 1992, Ap. J., 395, 202.
Woosley, S. E., and Weaver, T. A. 1986a, in: J. Audouze and N. Mathieu
(eds.), Nucleosynthesis and its Implications on Nuclear and Particle
Physics (Reidel, Dordrecht).
Woosley, S. E., and Weaver, T. A. 1986b, Ann. Rev. Astron. Astrophys.,
24, 205.
Woosley, S. E., et al. 1994, Ap. J., 433, 229.
Wuensch, W. U., et al. 1989, Phys. Rev. D, 40, 3153.
Xu, H. M., et al. 1994, Phys. Rev. Lett., 73, 2027.
Yanagida, T., and Yoshimura, M. 1988, Phys. Lett. B, 202, 301.
Yang, J., et al. 1979, Ap. J., 227, 697.
Yang, J., et al. 1984, Ap. J., 281, 493.
Yoshimura, M. 1988, Phys. Rev. D, 37, 2039.
Zacek, G., et al. 1986, Phys. Rev. D, 34, 2621.
Zaidi, M. H. 1965, Nuovo Cim., 40, 502.
Zatsepin, G. I. 1968, Pis’ma Zh. Eksp. Teor. Fiz., 8, 333 (JETP Lett.,
8, 205).
Zhang, W., et al. 1988, Phys. Rev. Lett., 61, 385.
Zhitnitskiı̆, A. P. 1980, Yad. Fiz., 31, 497 (Sov. J. Nucl. Phys., 31, 260).
Zisman, G. A. 1971, Uchenye Zapiski LGPI No. 386, pg. 80 (quoted
after Blinnikov et al. 1995).
Acronyms
The acronyms listed below do not include names of experiments such

as GALLEX or IMB, and also do not include names of laboratories
such as CERN or SLAC. These acronyms really play the role of proper
names; usually nobody quite remembers what they stand for.
AGB asymptotic giant branch

AU astronomical unit (distance to the Sun)
BBN big-bang nucleosynthesis
BC bolometric correction
BP Bahcall and Pinsonneault
BS blue straggler
CC charged current
CKM Cabbibo-Kobayashi-Maskawa
CL confidence level
CM center of mass
CMB cosmic microwave background
CMBR cosmic microwave background radiation
CNO carbon-nitrogen-oxygen
CO carbon-oxygen
D degenerate
DAV DA Variable
DFSZ Dine-Fishler-Srednicki-Zhitniskiı̆
EOS equation of state
FTD finite temperature and density
GIM Glashow-Iliopoulos-Maiani
GUT grand unified theory
HB horizontal branch
IAU International Astronomical Union
IMF initial mass function
KII Kamiokande II (detector)
pc parsec (3.08×1018 cm)
KSVZ Kim-Shifman-Vainshtein-Zakharov
642
Acronyms 643
L longitudinal
l.h. left-handed
l.h.s. left-hand side (of an equation)
ly light year
LMC Large Magellanic Cloud
LSP lightest supersymmetric particle
mfp mean free path
MS main sequence
MSW Mikheyev-Smirnov-Wolfenstein
NC neutral current
ND nondegenerate
T transverse
TL Turck-Chièze and Lopes
TO main-sequence turnoff
OPE one pion exchange
PQ Peccei-Quinn
QED quantum electrodynamics
QCD quantum chromodynamics
RG red giant
RGB red-giant branch
r.h. right-handed
r.h.s. right-hand side (of an equation)
SGB sub-giant branch
SN supernova
SNe supernovae
SNU solar neutrino unit
SNu supernova unit
SNBO Supernova Burst Observatory
UT universal time
VVO Voloshyn-Vysotskiı̆-Okun
WD white dwarf
Symbols
Symbols which have only a local meaning in, say, one paragraph are
not listed here. Four-momenta are usually denoted by uppercase italics
such as K, three-momenta are boldface lowercase letters such as k, the
modulus of a three-momentum is in lowercase italics such as k = |k|.
Lorentz indices are usually given in Greek letters (µ, ν, α, β, etc.),
three-indices and flavor indices in Latin letters (i, j, k, etc.).
Functions, Operators
Besides the usual functions and operators, the following convention may
be noteworthy.
∂x partial derivative ∂/∂x

ln natural logarithm
log logarithm base 10
Latin Symbols
a axion, axion field, annihilation operator, acceleration, radiation
constant (a = π 2 /15 in natural units)
ai ion-sphere radius
A electromagnetic vector potential, atomic mass number
B magnetic field, operator for “background” medium
Bω specific energy density of a radiation field
c speed of light (c = 1 in natural units)
cj speed of propagation of particle j
cp heat capacity at constant pressure
CV,A vector and axial-vector weak-coupling constants
Cj effective Peccei-Quinn charge of fermion j
d electric dipole moment
d, D distance
644
Symbols 645
dp differential in phase space, dp = d3 p/(2π)3

e electron, electric charge of electron, photon polarization four-
vector
ej charge of particle j
E electric field, energy of a particle
EF Fermi energy
f occupation numbers, dimensionless function, energy scale
fPQ Peccei-Quinn scale
fa axion decay constant
fπ pion decay constant (93 MeV)
fp occupation number of mode p
F electromagnetic field-strength tensor, number flux of particles,
dimensionless function of order unity
Fω specific energy flux in a radiation field
F fluence (time-integrated flux)
g dimensionless Yukawa coupling, gluon, graviton, number of spin
degrees of freedom
gaγ axion-photon coupling strength (units GeV−1 )
g10 gaγ /10−10 GeV−1
G color field-strength tensor, matrix of coupling constants, dimen-
sionless function
GN Newton’s constant (GN = m−2 Pl )
GF Fermi’s constant (GF = 1.166×10−5 GeV−2 )
H Hubble expansion parameter
H0 Hubble expansion parameter today (h × 100 km s−1 Mpc−1 )
H Hamiltonian density
k momentum of a particle, momentum transfer in a reaction
kB Boltzmann’s constant (kB = 1 in natural units)
kD Debye screening scale of the electrons
ki Debye screening scale of the ions
kS screening scale
kTF Thomas-Fermi screening scale
l momentum transfer in a reaction
& mean free path, flavor index
L luminosity of a star
L" solar luminosity (3.85×1033 erg)
L Lagrangian density
m mass
ma axion mass
me electron mass (0.511 MeV)
mπ pion mass (135 MeV)
646 Symbols
mN nucleon mass (938 MeV)

mu atomic mass unit (931 MeV)
mPl Planck mass (1.221×1019 GeV)
mW W -boson mass (80.2 GeV)
mZ Z-boson mass (91.2 GeV)
M absolute magnitude of a star, mass matrix of quarks or neutrinos
Mbol absolute bolometric magnitude of a star
M mass of a star, matrix element
M" solar mass (1.99×1033 g)
n particle density
nB baryon density
ne,p,n,ν electron, proton, neutron, neutrino density
nrefr index of refraction
N nucleon, number of degenerate Θ vacua in QCD
Nµν matrix element of neutrino current
p proton, momentum of a particle, pressure, dimensionless power
index
pF Fermi momentum
P four-momentum of a particle, stellar pulsation period
q momentum transfer, charge of a particle
Q energy loss or generation rate per unit volume
r radial coordinate
R radius of a star
R" solar radius (6.96×1010 cm)
s entropy density, dimensionless power index, square of CM en-
ergy, dimensionless spin-structure function
S static or dynamical structure function
t time
T temperature
T7 T /107 K
T8 T /108 K
T30 T /30 MeV
U total internal energy of a star, four-velocity
v velocity, vacuum expectation value of a Higgs-like field
V potential energy, volume
w up/strange quark mass ratio
W W -boson
WK,K ! Transition probability from K to K # .
x dimensionless energy (ω/T ), spatial coordinate, dimensionless
Fermi momentum (pF /m)
X mass fraction of hydrogen
Symbols 647
Xj Peccei-Quinn charge of a fermion j

Y mass fraction of helium
Yj number of fermions j per baryon
Ye mass fraction of helium in stellar envelope
YL lepton number per baryon
z up/down quark mass ratio, z-coordinate
Z mass fraction of “metals,” nuclear charge
Z, Z ◦ Z-boson
Greek Symbols
α fine-structure constant (e2 /4π = 1/137), asymmetry parameter
in angular distribution of decay photons
α# fine-structure constant for general bosons, α# = g 2 /4π with the
Yukawa coupling g
αa,χ fine-structure constant for axions, majorons
αs strong fine-structure constant
απ pionic fine-structure constant, απ = (f 2mN /mπ )2 /4π ≈ 15 with
f ≈ 1.0
β velocity of a particle, parameter of axion models, parameter in
nucleon-nucleon-axion bremsstrahlung rate
γ photon, Dirac matrix, adiabatic index, dimensionless fluctuation
rate γ = Γ/T
Γ rate for decay or fluctuations, plasma coupling parameter
Γσ spin-fluctuation rate in a nuclear medium
δ δ function, Kronecker δ, differential quantity
* energy loss or generation rate per unit mass, electric dipole mo-
ment, polarization vector
*ij neutrino transition electric moment
η degeneracy parameter, η = (µ − m)/T
θ mixing angle, scattering angle
Θ Θ parameter of QCD
ΘW weak mixing angle (sin2 ΘW = 0.2325)
κ opacity, dimensionless screening or momentum scale
κ∗ reduced opacity
λ wave length, mean free path
µ magnetic moment, chemical potential, muon, Lorentz index,
mean molecular weight
µ̂ nonrelativistic chemical potential, µ̂ = µ − m
µij neutrino transition magnetic moment
648 Symbols
µB Bohr magneton (e/2me )

µN nuclear magneton (e/2mp )
µe electron mean molecular weight (µe ≈ Ye−1 ), electron chemical
potential
ν neutrino, Lorentz index, dimensionless power index
ξ dimensionless correction factor, dimensionless axion-photon cou-
pling constant
π π = 3.1415 . . ., pion
ρ mass density, density matrix
ρ density matrix for antineutrinos
ρ0 nuclear density (3×1014 g cm−3 )
ρ14 ρ/1014 g cm−3
ρ15 ρ/1015 g cm−3
σ scattering cross section, Pauli matrix, spin operator
τ lifetime of a particle, duration of a stellar evolution phase, Pauli
matrix, optical depth
Φ Higgs field
Φ0 vacuum expectation value of Higgs field
χ majoron, majoron field
ψ fermion field
Ψ column vector of fermion fields (several flavors)
ω energy of a particle, energy transfer in reactions
ωP plasma frequency
ω0 zero-temperature plasma frequency
Ω cosmic density parameter, matrix of energies for mixed neutrinos
Subject Index
A direct search
decay photons of cosmic axions
α-Ori 189 491f
A665, A1413, A2218, A2256 (galaxy galactic (cavity search) 177f, 191f,
clusters) 490–92 544
adiabatic index 396 laboratory 182, 184f, 191f
adiabatic oscillations solar 100, 181, 191f, 462
→ neutrino oscillations SN 1987A 483
adiabatic temperature gradient 10–12 mass 527, 535
adiabaticity parameter 299 mixing with pion 527f, 534
models
AGB (asymptotic giant branch) 27, 35,
DFSZ 535–40
61
KSVZ 528–30, 542–44
aligned rotator 185
other 534–36
AMANDA xvi, 558
Nambu-Goldstone boson 190f, 528–30
anapole moment 268–70
quantum gravity 532f
Andromeda 446
solution to strong CP problem 527f
angular anomaly of SN 1987A neutrino
sphere 507
signal 422, 428–30
axio-recombination 100
antimatter supernova 497
arions 187–90
B
automatic symmetry 533
asymptotic giant branch 27, 35, 61 β decay 255
atmospheric neutrinos 290–3 ββ decay 257f, 560f
atomic mass unit 582, 591 Baade-Wesselink method 73
axio-electric effect 100 Baksan scintillator telescope (BST)
axion 414f, 417, 420
→ pseudoscalar boson band-structure effects 58, 106f
bounds 191f, 539–41 baryonic force 114
cavity experiment 177f, 191f, 544 beam-stop neutrinos 456f
cosmic density 491f, 528, 541–44 Betelgeuse 189
cosmic thermal production 541 big-bang nucleosynthesis: bounds
couplings majorons 259f, 561f
electron 536f millicharged particles 522f, 566f
fermions general 530–2 neutrino
gluon 159, 527, 530 Dirac dipole moment 277f
nucleon 536–38 mass 259f
photon 167f, 176, 191f, 536 time variation
decay 167f, 191f Fermi’s constant 546
decay constant 526f, 529, 535 Newton’s constant 547f
649
650 Subject Index
binary pulsar → PSR 1913+16 Cherenkov

birefringence effect 194, 216, 238f
neutrino flavor 195, 281 detectors → IMB, Kamiokande,
photon in media 180f, 194f, 208 Superkamiokande
vacuum 183–5, 187f, 190f, 554 chirality 160, 253
Bloch state 155 chiral symmetry 166, 529
blue loop 32, 39–41 chlorine detector (solar neutrinos) 342f,
bolometric correction 62 357–62, 371
bolometric magnitude: definition 28 CHORUS experiment 289f
Boltzmann’s constant 580 CKM matrix 262
Boltzmann collision equation (for mixed clump giants 76, 80
neutrinos) 323–31 CM energy 92
BOREXINO 343, 371, 393f CNO cycle 346f
bounce and shock 399, 401–5 COBE satellite 191
bound-free transition 100f coherent neutrino scattering 396, 446
branching ratio 449 collision equation for mixed neutrinos
Brans-Dicke theory 548 323–31
bremsstrahlung color-magnitude diagram
classical limit 143–47 evolutionary track 31
electronic 101-9 globular cluster M3 27
nucleonic globular-cluster observables 61
axial-vector vs. vector 127, 142 populations of stars 41
bubble phase 159 zero-age main sequence 26
meson condensate 155–57 Compton process 20f, 91–98, 108
neutrino pairs 127–30 conduction 11, 22, 78
pseudoscalars 119–26, 143–45, 148f convection
suppression at high density 148f description 12f
quark matter 157f helium dredge-up 66, 77
brown dwarfs 9, 24 mixing length 12, 16, 351, 549f
bubble phase 159 RGB bump 13, 61, 64f
bump on RGB 61, 64f superadiabatic 12
supernova 402–5, 444
C core
red giant → red giant core mass
Cabbibo mixing 261, 285 supernova → supernova core
Cabbibo-Kobayashi-Maskawa matrix correlations: Coulomb plasma 222–25
262 cosmic background fluxes
carbon-burning stars 2 γ-rays 487
cavity axion experiment 177f, 191f, 544 microwaves 191, 488, 546, 562
Cepheids 40f neutrinos
Chandrasekhar limit 9, 25, 37, 42, 396 core-collapse SNe 447, 487f
charged current hydrogen-burning stars 487
collision term (in neutrino primordial → neutrinos: cosmic
oscillations) 331 background
coupling constants 583 photons 489f
Hamiltonian 329, 583 cosmic strings 191, 542–44
reactions 1, 58f, 152–54 cosmic structure formation 258f
charge quantization 564–67 cosmion xvi
chemical potential 593–96, 600f Cotton-Mouton effect 180, 183
Subject Index 651
Coulomb gauge 204 dispersion

Coulomb propagator in plasma 221f, → neutrino dispersion, photon
226 dispersion
CP symmetry 241, 246, 262, 271, 453, general theory 193–202
524–26 double beta decay → ββ decay
Crab Nebula 38, 56f DUMAND xvi, 558
critical field strength 581 dynamical structure function →
crossing relation 323 structure function
current algebra 167, 535
E
D Einstein x-ray satellite 56–58, 189
electron charge 581
δ Cepheids 40f electro-Primakoff effect 172f
δ Scuti stars 41 energy conservation in stars 10f
damping → neutrino oscillations: energy-loss argument
damping by collisions analytic treatment 14–16
DA Variables 53 applied to
dark matter xv–xvi, xviii, 22, 24, 177f, HB stars 79–82, 98f, 107–9, 112,
258f, 528, 542–44, 558 176, 236
decoherence 311, 313f neutron stars 59f
degeneracy parameter 124, 595f, 603 red giants 83–87, 107–9, 236
delayed explosion mechanism 399, red supergiants 2
402–5, 426, 436f, 522 Sun 16f, 20f, 109f, 175f
Delbrück scattering 183 supernova core 501–23
deleptonization burst white dwarfs 47–52
→ prompt neutrino burst ZZ Ceti stars 52–54
deleptonization: supernova core 339, introduction 3–5, 21f
398, 401, 408 numerical treatment 16f, 47–52, 81,
density matrix (neutrino flavor) 83–87, 108f, 508–11, 513–16
collision equation 319, 323–31 energy-loss rate
definition 284, 315–16 bremsstrahlung
derivative coupling 119, 531f, 536, 563 electronic 102–5
nucleonic 121–6, 128–30, 147–49,
detailed balance 129, 137, 146
156f
DFSZ axions 535–40
Compton process 96–98
dielectric permittivity 173, 208f
plasma process 232–36
diffusion
Primakoff process 169, 174
elements → gravitational settling quark matter 157f
neutrinos → neutrino opacity URCA processes 153f
dipole moment energy-momentum conservation 121,
→ neutrino dipole moments 200
neutron electric 524–26 energy transfer
other particles 525 → convection, opacity
units 525, 581 equation 11–13
Dirac matrices 270 particle bounds
Dirac neutrinos → neutrino mass, general argument 4f, 17f
neutrinos: right-handed Sun 20f
direct explosion mechanism 399, 401f SN 1987A 503, 506–11
652 Subject Index
energy transfer (cont’d) GALLEX 343f, 357–59, 362–64, 371,

radiative 4, 17–19, 131 373
WIMPs 22 gallium detectors 343f, 357–59, 362–64,
equation of state 371
impact on SN neutrino signal 412f gauge
normal stars 6 bosons: new 114–16
nuclear 600 invariance 204–6
equivalence principle: tests 113, 498, General Relativity → gravitation
554f giant star
escape velocity → red giant
galaxy 26
definition 10
globular clusters 25
GIM suppression 266f
Euler-Heisenberg Lagrangian 183
globular cluster
EXOSAT 56f
explosion mechanism → red giants, HB stars, RR Lyrae
stars
→ supernova: type II
ages 62f, 66, 552–54
general properties 25f
F
M3 25, 27
Faraday effect 180 observables
Fermi interpretation 74–79
constant 545f, 583 in the color-magnitude diagram
energy 594 60–65
momentum 594 observational results 70–74
Fermi-Dirac distribution 593–96 theoretical relations 65–70
forward scattering 196–98 gluons 159, 525, 527, 530
fifth force 112–16, 500 g-mode → helioseismology
fine-structure constant gravitation
axion-nucleon 119 Brans-Dicke theory 548
electromagnetic 580 equivalence principle: tests 113, 498f,
new bosons 92 554f
pion-nucleon 119 graviton 116, 166, 191, 254
flavor-changing neutral current 263f, induced neutrino mixing 555
303, 326 nonmetric theory 498
flavor conversion
nonsymmetric extension 554
→ neutrino oscillations
quantum 532f
fluence: definition 463
Shapiro time delay 498, 554
foe 402
form factor → neutrino form factor time-varying GN 546–54
free-bound transition 100f gravitational settling
Fréjus experiment 291f description 13f
FTD field theory xviii–xix, 105, 174, globular-cluster ages 13, 66
241 red-giant envelope helium abundance
fundamental length scale 500 66, 74, 77
solar neutrino flux 14, 352, 354
G graviton 116, 166, 191, 254
group velocity 199
G117–B15A 53f GRS instrument 465f
Subject Index 653
H I, J
Hayashi line 32, 35, 40 IMB detector
HB (horizontal branch) atmospheric neutrinos 291f
location in color-magnitude diagram SN 1987A neutrinos 415–23
27, 41, 61 inflation 542
morphology 34 ions
HB stars contribution to screening 220f, 596,
→ RR Lyrae stars 599
energy-loss argument 79–82, 98f, correlations 106, 222–25
107–9, 112, 176, 236 ion-sphere radius 223f, 597
inner structure 29f, 80f, 592, 599 isochrone 61f
lifetime 70, 76, 79–82 Jupiter magnetic field 556
overview 34f
particle bounds: tabulation 82 K
helioscope 179, 181f
helioseismology xvi, 17, 113, 344f, κ mechanism 40
549–51 Kamiokande (Cherenkov detector)
helium abundance atmospheric neutrinos 291f
presolar 16f, 351, 549f dark-matter search xvi
primordial 24, 67 neutrino detection
red-giant envelope 66f, 74–78 efficiency 417
helium-burning stars → HB stars electron scattering 365–67, 416f
helium flash → helium ignition proton absorption 416f
helium ignition proton decay xvii
→ red giant core mass SN 1987A neutrinos 418–23
brightness 63f, 68–72 solar neutrinos 343f, 368–73, 383
description 33f Kelvin-Helmholtz cooling
delay by particle emission 63, 83–87, Sun 8
236 supernova core 397, 400, 407–13,
off-center 34, 78, 83f 501–23
Hertzsprung gap 32, 40 Kibble mechanism 542
Hertzsprung-Russell diagram kick velocity (neutron stars) 443–45
→ color-magnitude diagram Klein-Gordon equation for mixed
Higgs field 3, 165f, 194, 252f, 528, 534, neutrinos 282–84, 294
545, 559 Kramers-Kronig relation 198
Homestake (solar neutrino detector) KSVZ axions 528–30, 542–44
chlorine 342f, 357–62, 371, 373–76
iodine 343, 394 L
homologous models 14–16, 549, 552
horizontal branch → HB Lamb shift 565f
hot bubble 38, 397, 406 Landau damping 199, 214, 216f
hot plus cold dark matter 258 Landau-Pomeranchuk-Migdal effect
Hulse-Taylor binary pulsar 145, 148f
→ PSR 1913+16 Landau-Zener approximation 300
Hyades 48 Langmuir wave
hydrodynamic event 6 → plasmon: longitudinal
hydrogen-burning chains 341, 346f Large Magellanic Cloud 39, 74, 414, 446
hydrostatic equilibrium 5–7 large-numbers hypothesis 545, 549
654 Subject Index
left-right symmetric model 268 Majorana neutrinos → neutrino mass

leptonic force 114–16, 500 majorons
lepton → pseudoscalar bosons
fraction in nuclear matter 600, 602–5 ββ decay 560f
masses 252, 265 decay in media 195, 248–50, 259f
profile in SN core 398 interaction with supernova neutrinos
level crossing 295f, 298 495
linear-response theory 204–6, 208f motivation 558–62
long-range force → fifth force, neutrino decay: limits 562–64
scalar particles, photon nucleosynthesis constraints 259f
long-wavelength limit 121, 128f, 141–43
solar neutrino decay 389
Lorentz
supernova physics 562–64
addition of velocities 500f
mass fraction (of elements) 591f
gauge 204
mass loss 25, 34, 36
Lorentzian model for structure function
135, 144–47 mass-radius relationship 9, 42
LPM effect 145, 148f Maxwell’s equations 202–4
mean molecular weight 591
M meson condensate 59, 155–57, 243
metallicity
M1 transition 94, 102 α-enhanced 78f
M3 (globular cluster) 25, 27 globular cluster 26, 34, 60, 66f, 78f
magnetic dipole moment impact on opacity 12
→ neutrino dipole moments Sun 24
units 581 Mexican hat 529, 542
magnetic field Mikheyev-Smirnov-Wolfenstein effect
Earth 556 → neutrino oscillations: resonant
galactic 188–90, 484, 499 millicharged particles 227f, 232–36,
large-scale cosmic 191 564–67
Jupiter 556 misalignment mechanism 541f
neutron stars 185–88, 444f mixing angle in medium 294–7
primordial 278 mixing between particles
Small Magellanic Cloud 556 → neutrino mixing, neutrino
solar 186, 373–76, 387f, 554, 565 oscillations
supernova 522
axion-photon 179–81
twisting 308f
axion-pion 527f, 534
white dwarfs 185f
general theory 260–63
magnetic oscillations
graviton-photon 191
→ neutrino spin precession
magnetic permeability 208f gravitationally induced 555
magnetogram 375 mixing length 12, 16, 351, 549f
main sequence molecular chaos 319
helium 35, 592 monopoles xv
internal stellar conditions 592, 599 MSW
lifetime 27, 552 bathtub 301–3
location in color-magnitude diagram effect → neutrino oscillations
27, 41, 61 solution of solar neutrino problem
turnoff 27, 61–63 384–86
zero-age 26 triangle 301–3, 386, 434
Subject Index 655
multiple scattering neutrino dipole moments (cont’d)

Cherenkov detector 366 interaction structure 228, 265, 268–71
nuclear medium 135, 143–51, 511f medium induced 240
MUNU experiment 276 processes
muon decay 263 plasmon decay 229, 232–36, 274
scattering 272f
N summary 272–75
radiative decay
Nambu-Goldstone bosons 165f, 250,
→ neutrino radiative decay
529–31, 559f
summary of limits 275–79
NESTOR xvi, 558
theoretical prediction
neutral current
collision term in neutrino oscillations Dirac vs. Majorana 265, 271, 451
323f left-right symmetric model 268
coupling constants 584 standard-model 265f
flavor-changing 263f, 303, 326 structure of interaction 265, 268–71
Hamiltonian 95, 119, 136, 320, 531, neutrino dispersion
584 Cherenkov effect 238f
neutrality of matter 565 deflection 247f
neutrino absorption → neutrino opacity density-matrix treatment 321f
neutrino charge: electric 227f, 232–36, general theory 241–47
240, 499, 522f, 564–67 inhomogeneous medium 442f
neutrino cooling neutrino background 321f, 432f,
→ supernova core 440–42
red-giant core 83–85 nonisotropic medium 321, 432f,
white dwarfs 47–50 440–42
neutrino coupling constants: standard off-diagonal refractive index 322
583f relation for massive neutrinos in
neutrino cross section media 295f
electrons 365f, 416 spin-flip scattering 162
protons and nuclei 416 neutrino emission: standard
neutrino dark matter 258–60 → solar neutrino flux, supernova core
neutrino decay
astrophysical impact
→ neutrino radiative decay
red-giant core mass 83f
fast invisible 248–50, 389, 485, 559
red supergiant 2
standard-model 263–67
white-dwarf cooling 47–50
neutrino dipole moments
astrophysical impact and bounds historical perspective 1f
early universe 277f numerical rates 585–90
HB stars 82, 236 specific processes
neutron stars 59f bremsstrahlung (electronic) 105-7,
red-giant core mass 86f, 236 588
supernova 277, 477–79, 483, 521f bremsstrahlung (nucleonic) 127–30,
white dwarfs 48–50 157, 159
electric vs. magnetic 265f, 268–71, Compton (photoneutrino) 95f, 98,
273–75, 525 586f
experimental bounds free-bound transition 100f
→ neutrino radiative decay pair annihilation 99f, 586f
scattering experiments 267, 275f plasma 585f
656 Subject Index
neutrino form factor: electromagnetic neutrino oscillations (cont’d)

→ neutrino dipole moment experimental searches 289f
anapole moment 268–70 historical introduction and overview
charge 227f, 232–36, 240, 268, 564–67 280–82
charge radius 269f, 523 in media
general theory 267–72 adiabatic limit 297–99
in plasma 230f, 237–41 analytic results 299–300
neutrino lasing 389
homogeneous 296f
neutrino lightcurve → supernova core
mixing angle 294–7
neutrino mass
bounds neutrino background in supernovae
ββ decay 257f 432f, 440–42
cosmological 258–60 oscillation length 285f, 289, 294–97
experimental 252, 254–58 primordial 322
future supernovae 447f, 497 resonant
SN 1987A signal dispersion 426f analytic results 299f
SN 1987A signal duration 516–19 description 281, 298
Dirac 253f, 516–19 Landau-Zener approximation 300
effective → neutrino dispersion level crossing 295f, 298
inverted hierarchy 434 schematic model for the Sun 301–3
Majorana 253f, 257–59 solution of solar neutrino problem
neutrino mixing 384–86
→ neutrino oscillations supernova core 333f
bounds supernova mantle 432–42
atmospheric neutrinos 291
spin-precession picture 286f, 314f
decay experiments 456f, 461
supernova
oscillation experiments 289f
r-process nucleosynthesis 437–42 cooling phase 434–36
SN 1987A γ-rays 481 explosion mechanism 436f
SN 1987A prompt burst 432–34 overview 430f
SN 1987A signal duration 434–36 prompt burst 432–34, 496f, 498
supernova core: flavor conversion r-process nucleosynthesis 437–42
338 Sun
decay rates and dipole moments MSW solutions 384–86, 434f
264–67 schematic model of MSW effect
induced by 301–3
flavor-changing neutral currents vacuum solution 381–84
303 survival probability
gravity 555 general expression 284–86
mass matrix 262, 282
with energy distribution 288
neutrino opacity 132–35, 149–51, 329f,
with source distribution 287f
334f, 408, 514–16
neutrino oscillations temporal vs. spatial 283f
damping by collisions three-flavor 284f
kinetic equation 327–29, 332–34 two-flavor 284–88
simple picture 310–12 neutrino-photon scattering 245f
spin relaxation 313 neutrino radiative decay
time scale in SN core 335–38 → neutrino dipole moment, neutrino
equations of motion 282–84 two-photon coupling
Subject Index 657
neutrino radiative decay (cont’d) neutrinos: supernova → SN 1987A,

bounds from supernova core
beam stop 456f neutron stars
cosmic background neutrinos cooling 54–60
489–91 crust 58–60, 106f
diffuse flux from all stars 487f formation 38
reactors 453–55 kick velocities 443–45
solar positrons 457, 461 internal conditions 601
solar x-rays 459–61 magnetosphere 185–88
SN 1987A γ-rays 467f, 474f, 477–79 pseudoscalar field around 187
photon spectrum x-ray observations 56–58
stationary source 451–53 Newton’s constant
pulsed source 463–65, 469–73 value 6
standard-model predictions 264–67 time variation 546–54
summary of bounds 278f NOMAD experiment 289f
neutrino rocket engine 443–45 nonabelian Boltzmann collision
neutrino sphere 397, 407, 409 equation 323f
neutrino spin precession nuclear matter
supernova magnetic field 521f novel phases 155–59
theoretical description properties 151f, 600–5
equations of motion 304f nucleon mass: effective 151f, 602–5
in electric fields 305 nucleosynthesis → big-bang nucleosyn-
in medium 306 thesis, supernova (type II)
oscillation length 305 number fraction: definition 243
spin-flavor 306–9
time variation of solar neutrino flux O
374, 387f
neutrino trapping 396 off-diagonal refractive index 322
neutrino two-photon coupling 245, 265f, Oklo natural reactor 546
271f one-pion exchange potential 119, 122,
neutrinos: cosmic background 124f, 141
screening of leptonic force 114f, 500 opacity
mass density 258–60, 489–91 α-enhanced 78f
radiative decay 489–91 conductive 11, 22, 78
neutrinos: number of families 251, 260, definition 11
513 neutrino 132–35, 149–51, 329f, 334f,
neutrinos: right-handed 408, 514–16
component of Dirac 253f particle 11, 18f, 127–35
e+ e− decay limits 479, 482 radiative 18f
SN 1987A limit reduced 19
charge radius 523 Rosseland average 11, 18f
dipole moment 521f solar 21, 345, 353–55
Dirac mass 516–19 tables 12
mixing with sequential flavors OPAL 12
338–40 open clusters 60, 76
right-handed currents 519–20 OPE potential 119, 122, 124f, 141
spin-flip production 160–64 optical activity → birefringence
neutrinos: solar → solar neutrino flux optical theorem 198
neutrinos: strongly interacting 558f, 563 oscillation length 285f, 289, 294–97, 305
658 Subject Index
oscillations (particles) pion (cont’d)

→ neutrino oscillations decay 162, 167f, 256
axion-photon 179–82, 185f, 188–90 decay constant 166f, 527
graviton-photon 191 mass 119, 124f, 527
inhomogeneous media 181, 442f mixing with axion 527f, 534
pion-photon 183 Nambu-Goldstone boson 165f
oscillations (stars) → helioseismology, Planck mass: definition 6
stars: variable Planck’s constant 580
planetary nebulae 36, 43, 51
P plasma
conditions in stars 591–99
pair annihilation 99f coupling parameter Γ 224, 597–99
paraphoton 82, 92f, 99 crystallization 46, 598
Pauli matrices 286
fluctuations 172–74
Peccei-Quinn
frequency 198, 211f, 596, 599
mechanism 526–28, 532f
one-component 223–25
scale 166, 526f, 529, 535
strongly coupled 223f
peculiar velocity (neutron stars) 443–45
two-component 223
permittivity → dielectric permittivity
plasma process
phase velocity 199, 244
→ neutrino dipole moments
photon
astrophysical impact
baryonic 114
HB star lifetime 80–82
charge limits 555f
neutron-star crust 59f
decay → plasma process
leptonic 114–16 red-giant core mass 83–87
mass supernova core 523
→ plasma frequency general theory 227–36
effective 208 gluonic 159
vacuum 206, 555f plasmino 195
transverse 209, 214 plasmon
refraction → photon dispersion coalescence 170–72
splitting 184 e+ e− decay 195, 211
weakly interacting 18 longitudinal 170–72, 195, 202f, 213
photon dispersion νν decay → plasma process
approximate expressions 213f p-mode → helioseismology
external fields (vacuum birefringence) Poincaré sphere representation 286
183–5, 187f, 190f, 554 polarimetry of radio sources 190f
general theory 203–19 polarization tensor 205–12, 230f, 238
mixing with axions 180f polarization vector
transverse mass 209, 214 neutrino flavor
photoproduction definition 286f
→ Compton process degree of coherence 313
X-particles in HB stars 82, 429 individual modes 317
pinching of neutrino spectra 410f shrinking by collisions 314f
pion transverse 314, 325
→ one-pion exchange potential plasma excitations 206f
Compton scattering 157 positron flux: interplanetary 461, 487f
condensate 59, 155–57, 243 pp-chain 15, 341, 346f
coupling to nucleons 119, 151f, 531 pressure sources in stars 7–10
Subject Index 659
Primakoff process 116, 159, 168–75, PSR 0655+64 → 547

177, 357 PSR 1913+16 (Hulse-Taylor binary
prompt pulsar) xiv, 113, 116, 547
explosion mechanism 399, 401f PSR 1937+21 → 188, 190, 554
neutrino burst 397, 399f, 432–34, pulsar
496f, 498 → neutron star
proton pulse dispersion 554f
decay xvii pulsation
fraction in nuclear matter 601f constant 40
spin 537, 584 periods 40f
protoneutron star → supernova core period decrease 53f
pseudomomentum 200 PVLAS experiment 185, 192
pseudo Nambu-Goldstone bosons 166
pseudoscalar bosons
→ arions, axions, majorons Q
astrophysical impact and bounds
HB star lifetime 80–82, 98f, 176, QCD: CP problem 524–26
191f Q-nuclei (quarked nuclei) 557
neutron-star cooling 59 quark
red-giant core mass 85f, 107-9 free 556f
SN 1987A 501–4, 508–12 matter xv, 157–59, 557
Sun 16f, 20f, 176, 191f masses 252, 525
white-dwarf cooling 50–52 mass ratios 535
ZZ Ceti period decrease 54 mixing 261f
bounds on couplings to quantum gravity 532f
photons 17, 81f, 176, 191f
electrons 20f, 51f, 82, 85, 99f, R
107–11
nucleons 512
radiative energy transfer → energy
conversion to photons 179–82, 185f,
transfer, opacity
188–90
random media: particle oscillations 181,
derivative coupling 119, 531f, 563
442f
emission processes
rank-ordering statistics 374f
bremsstrahlung (electronic) 102–5,
reactor: nuclear
107f
natural 546
bremsstrahlung (nucleonic) 119–26,
143–45, 148f, 155f neutrino radiative decay limits 454f
bremsstrahlung (quarks) 157f neutrino spectrum 450, 453f, 455
Compton 20, 94, 98, 108f red-giant
energy spectrum 123, 175 branch → RGB
pair annihilation 99f core mass at helium ignition
Primakoff (gluonic) 159 increase by particle emission 83–87
Primakoff (photonic) 168–75, 177 observational 75–79
interaction Hamiltonian 119 theoretical 67f
long-range force 112 uncertainties 78f
opacity contribution 18f, 21, 131f, core rotation 78
506–8 properties and evolution 28–32, 592,
refractive index 250 599
solar direct search 100, 181, 191f reduced opacity 19, 131
660 Subject Index
refractive index scalar particles (cont’d)

→ photon dispersion, neutrino emission from binary pulsar 113
dispersion emission processes
definition 194 bremsstrahlung 103, 124f
forward scattering 196–98 Compton 93, 98
neutrino 242, 244 long-range force 112f, 500
relativistic limiting velocity 497f photon coupling 166
renormalization: wave function 200, Sciama’s neutrino 491f
217–19, 228, 233 screening effects
repulsion of energy levels 296 background neutrinos 114f
resonant oscillations Debye-Hückel scale 220f, 596f, 599
→ neutrino oscillations general theory 219–26
axion-photon 181, 186 in processes
RGB (red-giant branch) bremsstrahlung 102–6
brightest star 63f, 68–72 Primakoff 169
bump 61, 64f spin-flip scattering 277f
location in color-magnitude diagram ion contribution 220f, 596f, 599
27, 41 modification of Coulomb propagator
phase transition 37 221f, 226
right-handed Thomas-Fermi scale 221
currents 268, 519–20 secret neutrino interactions 558f, 563
neutrinos → neutrinos: right-handed see-saw mechanism 254
Ring Nebula in Lyra 36 self-energy 201, 209
R-method 65, 70, 72, 74–78 Shapiro time delay 498, 554
ROSAT 56–58 shining light through walls 182
Rosseland average 11, 18f shock wave 397, 399–405
r-process nucleosynthesis 405–6, 437–42 Small Magellanic Cloud 446, 556
RR Lyrae stars SMM satellite 465f
absolute brightness 62, 65, 69f, 73f SN → supernova
ages of globular clusters 62 SN 1987A
color 62 neutrino pulse
location in color-magnitude diagram analysis 423–26
40f anomalies 427–30
mass-to-light ratio 77 measurements 419–23
Rubakov-Callan effect xv optical lightcurve 414f
Rydberg atom 178 progenitor 39, 414
SN 1987A bounds
S antimatter supernova 497
arion-photon conversion 189
S17 factor 355–57, 377f, 386 axions
SAGE 343f, 357–59, 362f, 371 direct detection 495
Sanduleak −69 202 → 39, 414 neutrino signal duration 501–4,
scalar field 508–12
→ scalar particles, Higgs field fundamental length scale 500
around a neutron star 113 Lorentz addition of velocities 500f
scalar particles neutrino
bounds on coupling charge 499, 522f
electrons 82, 99, 107 decay 494, 496, 519
nucleons 82, 113 mass 426f, 499
Subject Index 661
SN 1987A bounds (cont’d) solar neutrino flux (cont’d)

neutrino (cont’d) modified by (cont’d)
neutrino-neutrino cross section 494f resonant oscillations 301–3, 384–86,
number of families 513 434f
secret interactions 494f temperature 354
neutrino oscillations WIMP energy transfer xvi
cooling phase 434–36 strange quark matter 557
prompt burst 432–34 Q-nuclear burning 557
pseudoscalar boson couplings 501–4, time-varying GN 549–51
508–12 vacuum oscillations 381–84
radiative particle decays 467f, 474f, radiative decay limits 458–62
477–79, 483 source reactions
relativistic limiting velocity 497f beryllium 343, 347–50, 355, 378f
right-handed currents 519–20 boron 343, 347–49, 355–57, 368–70,
right-handed neutrinos 377f, 383
charge radius 523 CNO 343, 347–49
dipole moment 521f electron capture vs. decay 350
Dirac mass 516–19 hep 347–49
mixing with sequential neutrinos pep 347, 350
338–40 pp 343, 347–49
secret neutrino interactions 563 time variation
supersymmetric particles 558 day-night 372, 385f, 390f
weak equivalence principle 498 semiannual 372f, 382f
SNBO 448 solar cycle 373–76, 387f
SNO 343, 371, 392f solar neutrino problem
SNu (supernova unit) 488 introduction and historical overview
SNU (solar neutrino unit) 358 341–45, 380f
solar axions 100, 181f, 191f flux deficits
solar maximum mission satellite 465f beryllium 378f
solar neutrino flux beryllium/boron branching ratio
antineutrino component 379f
from majoron decay 389 boron 377f
from spin-flavor oscillations 388 flux variation at Homestake 373–77
limits from Kamiokande 369 MSW solution 384–86
counting rate prediction for detection vacuum solution 381–84
Cherenkov 370 VVO solution (magnetic oscillations)
chlorine 359, 361 387f
gallium 359, 364 Sommerfeld parameter 355
future experiments 390–4 space-like excitations 194, 198f, 207,
measurements 216, 215f, 238f
Cherenkov 368–70 spectral density 173f
chlorine 360–62 speed of light 497f
gallium 362–64 spin-flavor oscillations → neutrino spin
modified by precession
electrically charged neutrinos 565 spin flip 160–64, 277f, 304–9, 317,
gravitational settling 14, 352, 354 516–23
opacities 353–55 spin-fluctuation rate 118, 121-23, 127,
neutrino decay 389 133, 144f
neutrino-neutrino scattering 495 spin relaxation 313
662 Subject Index
spin-spin interaction potential 141, 151 sum rules 139–41

starburst galaxies 446 Sun
stars → solar neutrino flux
ages 27, 35, 43, 62f, 552–54 activity cycle 373–76, 387f
formation 24–27 axion spectrum 175
initial mass function 24 bounds on
intermediate-mass 37
axion flux 100, 181, 191f
mass loss 25, 34, 36
time-varying GN 549–51
mass range 24
deflection of neutrinos 247f
massive 37–39
populations in color-magnitude energy-loss argument 16f, 20f, 109f,
diagram 41 175f
variable 12, 39–41, 52–54 global properties
statistical parallaxes 73 central temperature 7, 353
Stefan-Boltzmann law 409 convective layer 550
stellar collapse → supernova: type II distance 341, 372
stellar evolution Kelvin-Helmholtz time scale 8
bibliography 24 helium abundance 16f, 351, 549f
descriptive overview 23–41 luminosity 8, 341
evolutionary track 31 magnetic field 186, 373–76, 387f,
main phases 30 554, 565
stellar oscillations → helioseismology, mass 7
variable stars, ZZ Ceti stars plasma properties 599
stellar structure
radius 7, 351
convective 12f
opacity 21, 345, 353–55
equations 5–14
positron flux limits 457, 461
examples for models 29f
generic cases 7–10 spots 186, 373–76
homologous models 14–16, 549, 552 standard model 29, 351–53
long-range force: new 113 x- and γ-ray flux 458–62
sterile neutrinos superfluidity in neutron stars 58f
→ neutrinos: right-handed Superkamiokande (Cherenkov detector)
Stodolsky’s formula 313–15, 324f 343, 371, 390–2
Stokes parameters 286 supernova
strange quark matter xv, 157–59, 557 → SN 1987A
structure function burst observatory 448
Coulomb plasma 106, 222–25 core collapse → supernova: type II
dynamical energetics: particle bounds 483
classical 146f future 445–48
detailed balance 129, 137, 146
Kepler’s 39
formal definition 136–43, 323
galactic
long-wavelength limit 121, 128f,
positron flux 484f
141–43
Lorentzian model 135, 144–47 rate 39, 446, 487f
nuclear medium 128f, 135, 146–51, remnants 38f, 56–58
161–63 SN 1054 → 38
spin-density 138 Tycho’s 39
static 137, 222–25 type I → 36, 546
subgiant 27, 28 unit (SNu) 488
Subject Index 663
supernova core thermal broadening of beryllium line

binding energy 407 350f
characteristics of the medium 398, thermal equilibrium in stars 8
505f, 512, 602–5 Thomas-Fermi scale 221
matter accretion 397, 400, 425f, 503 transition moment → neutrino dipole
neutrino cooling moments
→ SN 1987A transverse current 204, 219f
analytic emission models transverse gauge 204
expected detector signal 418 transverse part of the flavor
schematic picture 397, 400, 407f polarization vector 314, 325
spectral characteristics 408–11 transverse photon mass 209, 214
time evolution 411–14
triangle condition 153f
neutrino flavor conversion 332–38
triangle loop 167f, 530, 536
particle cooling
triangle: MSW 301–3, 386, 434
axion emission 501–4, 508–12
general argument 501–8 triple-α reaction 33, 80
numerical studies 508–11, 513–16 tritium β decay 255
structure (numerical model) 505f, 512 twisting magnetic field 308f
supernova: type II two-photon coupling
→ supernova core neutrinos 245, 265f, 271f
description 37–39 various bosons 165–68
explosion mechanism 401–5, 436f, 522
neutrino oscillations U, V
cooling phase 434–36
explosion mechanism 436f
units
overview 430f
conversion factors 580–82
prompt burst 432–34
r-process nucleosynthesis 437–42 natural 6
nucleosynthesis 405–6, 437–42 rationalized 184, 580
stellar collapse 395–99 URCA process 1, 58f, 152–54
supersymmetric particles vacuum birefringence 183–5, 187f, 190f,
dark matter xv–xvi, 22, 558 554
emission from stars 81, 557f vector bosons
supersymmetry: flavor-changing neutral Compton process 91–93, 98
current 303 energy-loss bounds 82, 99, 110–12
SXT satellite 186 long-range force 112–16, 500
virial theorem
T introduction 6f
negative specific heat 7f
tau neutrino
VVO effect 387
charge 567
dipole moment 278, 455, 457, 476f
e+ e− decay: bounds W
galactic supernovae 484–86
reactors 455, 457 weak damping limit (of neutrino
SN 1987A γ-rays 480–82 oscillations) 326f, 330f
solar positrons 457, 461 weak decay spectrum 348
mass 256, 258–60, 485f weak mixing angle 583
theoretical decay rate 264 Weinberg angle 583
664 Subject Index
white dwarfs Whole-Earth Telescope 53

bounds on WIMP xv, 22
neutrino dipole moments 48–50
pseudoscalar bosons 50–52 X, Y, Z
time-varying GN 551
characteristics 9, 42–45, 599 x-rays
cooling theory 45–47 neutron stars 56–58
formation 35f particle bounds
inferred galactic age 43, 51 SN 1987A arion-photon conversion
location in color-magnitude diagram 189
40f SN 1987A axion decays 483
luminosity function 43–45 SN 1987A radiative neutrino
magnetic 185f decays 467f, 474f, 477–79
mass-radius relationship 9, 42 solar x-rays 458–62
neutrino cooling 47–50 satellites 56–58, 186, 465f
variable 41, 52–54 Yohkoh satellite 186
vs. red giant core 88 ZZ Ceti stars 41, 52–54

Raffelt - Stars As Laboratories For Fundamental Physics

Uploaded by

Copyright:

Available Formats

Raffelt - Stars As Laboratories For Fundamental Physics

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Raffelt - Stars As Laboratories For Fundamental Physics

Uploaded by

Copyright:

Available Formats

Georg G.

Final Manuscript, October 1995

Published by the University of Chicago Press, 1996

To be published by the University of Chicago Press, Chicago.

2. Anomalous Stellar Energy Losses Bounded

2.4 Globular-Cluster Stars (60)

3. Particles Interacting with Electrons

4. Processes in a Nuclear Medium

4.5 Neutrino Opacity (132)

5. Two-Photon Coupling of Low-Mass Bosons

6. Particle Dispersion and Decays in Media

6.3 Photon Dispersion (202)

9. Oscillations of Trapped Neutrinos

10. Solar Neutrinos

10.3 Observations (357)

11. Supernova Neutrinos

12. Radiative Particle Decays from Distant

13. What Have We Learned from SN 1987A?

13.7 Neutrino Opacity (514)

15. Miscellaneous Exotica

16. Neutrinos: The Bottom Line

16.3 New Interactions (574)

D. Characteristics of Stellar Plasmas (591)

All currently known phenomena of elementary particle physics are

boson of a new symmetry broken at some large energy scale. Such a

The complete or near masslessness of the particles studied here (neu-

cuts of higher-order graphs. These methods have been applied only

It has been challenging to hammer the multifarious and intertwined

quoted “astrophysical bounds on the axion mass” are really transformed

Parts of my presentation are devoted to theoretical and calcula-

In spite of this bittersweet insight, and in spite of some inevitable

Munich, May 1995.

The Energy-Loss Argument

Weakly interacting, low-mass particles such as neutrinos or axions con-

Still, neutrinos can be important in normal stars. This became

example is the axion (Chapter 14) which arises as the Nambu-Goldstone

For low-mass stars plenty of detailed observational data exist, allowing

trinos, to be harmless in stars. Of course, depending on whether the

1.2 Equations of Stellar Structure

impact on the overall picture of stellar structure and evolution, with

1.2.2 Generic Cases of Stellar Structure

forms us that a decrease of $Ekin +Egrav % causes the gravitational energy

The most salient feature of a normal stellar configuration is the in-

1.2.3 Energy Conservation

dLr /dr = 4πr2 (ρ, (1.3)

where Lr is the net flux of energy through a spherical shell of radius r

( = (nuc + (grav − (ν − (x , (1.4)

where (nuc is the rate by which nuclear energy is liberated, (ν is the

(grav = cp T (∇ad ṗ/p − Ṫ /T ) (1.5)

is the local energy gain when T and p change because of expansion or

against the force of gravity. In Eq. (1.5) cp is the heat capacity at

1.2.4 Energy Transfer

where aT 4 is the energy stored in the radiation field (a = π 2 /15 in

where κγ is the radiative opacity, κc the contribution from conduction

One of the main difficulties at calculating the opacity is that heavier

1.2.5 Gravitational Settling

age of the universe, this “cosmic age problem” is slightly alleviated by

1.3 Impact of Novel Particles