RADIO SCIENCE, Vol. 3 (New Series), No.

7, July 1968

D.iffusion Processes Influenced by Coniugate-Point Wave

Phenomena

John M. Cornwall
Aerospace Corporation, El Segundo, Calif. 90245, U.S.A.

and

Department of Physics, University of California at Los Angeles, Los Angeles, Calif. 90024, U.S.A.
(Received December 1, 1967)

Radial diffusion coefficients for low-energy protons in the magnetosphere are estimated for proc
esses involving Pc 1, 2, and 3 hydromagnetic waves. These diffusion coefficients are compared to
rough estimates of diffusion due to magnetic impulses and bays, to arrive at an overall diffusion rate.
An effective loss rate for ptich-angle diffusion (caused by Pc 1) is developed, and combined with radial
diffusion to yield a nonlinear differential equation. Some estimates of expected proton flux levels are
made on the basis of this equation.

1. Introduction We are interested in the effect of pitch-angle dif-

fusion on radial diffusion (see, e.g., Nakada and
In this paper, we discuss semiquantitatively the Mead (1965) for further references). Magnetic and
ways in which pitch-angle scattering caused by hydro- electric disturbances associated with the solar wind
magnetic waves can influence the inward radial dif- and other phenomena cause low-energy protons (and
fusion of low-energy protons. The problem of radial electrons) to diffuse radially inward from the magneto-
diffusion is an extremely complicated one, with pitch- sphere. These disturbances act on a time scale that
angle diffusion taken into account, and no definitive is long compared to the cyclotron and bounce periods
conclusions can be drawn without extensive computer of the particles, and therefore violate only the third
studies. The results of the present work are neces- adiabatic invariant. In addition, there are processes
sarily tentative, but we hope they are sufficiently that act on much shorter time scales to cause radial
reliable to use as guideposts. diffusion: (a) Pc l micropulsations can cause radial
A number of authors have discussed the generation diffusion in an axially asymmetric magnetosphere, as
of Pc 1 micropulsations (frequencies 0.2-5 Hz) (Corn- pointed out by Roederer (1967); (b) longer period
wall, 1965; Obayashi, 1965; Brice, 1964; Jacobs and micropulsations (Pc 2 or 3) can resonate with proton
Watanabe, 1964). When particles of the appropriate bounce time, violate the longitudinal invariant, and
energy have an anisotropic pitch-angle distribution, produce radial diffusion (Cornwall (1966) has evalu-
an instability results, with frequencies in the Pc 1 ated the same mechamism for electrons). It is impor-
band for the Davis-Williamson (1963) protons. (Elec- tant to evaluate the effect of these last two mechanisms
trons of ;::: 100 ke V produce a similar instability with quantitatively, since the diffusion coefficients proposed
whistlers.) Because hydromagnetic waves can bounce by Nakada and Mead, based on long-time scale dis-
between conjugate points of a magnetic field line, a turbances, is much too small to explain the observed
single wave packet can be amplified many times. The radial distribution of the Davis-Williamson protons.
effect is reminiscent of a Q-spoiled laser. The micro- Probably the most important effect of all is the loss
pulsations scatter the protons in pitch angle, which rate induced by pitch-angle diffusion, which competes
is best described as a diffusion process (Cornwall, strongly with radial diffusion. The loss rate is non-
1966; Kennel and Petschek, 1966). This pitch-angle linear when saturation effects set in (Cornwall, 1966;
diffusion is an effective loss mechanism for the Davis- Kennel and Petschek, 1966). When the proton flux
Williamson protons. tries to exceed a certain level, the loss rate becomes
740
 Conjugafe-Poinf Wave Phenomena 741

very large and limits the proton flux. When this loss steep radial dependence, as Fiilthammar has shown,
rate is combined with radial diffusion, a nonlinear comes from the very long decay time of the magnetic
differential equation results. impulses. This has the consequence that, at L = 6,
In this paper, we estimate the strength of various this type of diffusion is much too small to explain the
radial diffusion processes, and combine this informa- Davis-Williamson protons.
tion with nonlinear pitch-angle diffusion to estimate the A potentially much more important mechanism is
proton flux at various positions in the magnetosphere. related to magnetic bays, and their associated elec-
As a result, the proton flux level at saturation depends trostatic potential fields (Dungey, 1965; Fiilthammar,
only logarithmically on the radial diffusion constant, 1966a, b). Magnetic bays last from 10 to 30 min, and
which is fortunate since this diffusion coefficient is are caused by ionospheric currents that are driven
very poorly known. by electrostatic fields. Because magnetic field lines
are very nearly equipotential lines, these same elec-
trostatic fields persist to large radial distances, where
2. Radial Diffusion Processes they energize charged particles. The time scale
Table 1 depicts a quick survey of the results of this (10 to 30 min) is short compared to proton drift periods,
section, and is partly a review of other authors' work. hence (as Fiilthammar has shown) the radial diffusion
Certain diffusion processes are associated with electro- coefficient goes roughly like L 6 , at large (L?: 5) dis-
magnetic fields which are virtually pure magnetic, tances. For smaller distances, this kind of diffusion
and thus energy-conserving; table 1 carries a column is negligible because the electrostatic potential van-
for this information. ishes rapidly below subaurorallatitudes.
Fiilthammar's expression for the diffusion coef-
ficient is
2.1. Long-Time Scale Radial Diffusion Processes
These processes are reviewed by Dungey (1965), and
important work has been done by Falthammar (1965, (I)
1966a, b). We define long time as a time long com-
pared to the bounce period and cyclotron period of
the Davis-Williamson protons; thus, the primary where
effect of these processes is to violate the third adiabatic
invariant, while the first two are conserved. If these
processes were the only ones acting, then the energy E
Pn=4 L',. dr(En(L, t)En(L, t+r)) cos ruoT (2)
of an equatorial particle should be proportional to
the earth's field strength B - L- 3 by in variance of the is the power spectrum at the frequency TUDT of the
magnetic moment. azimuthal electric field E(L, t):
Many people have discussed diffusion by magnetic
impulses and sudden commencements, which have a E(L, t) = L En(L, t) cos (nwot+</>n). (3)
short rise time and long decay time (e.g., Fiilthammar, n=O
1965, 1966a, b; Nakada and Mead, 1965). The dif-
fusion coefficient as given by Nakada and Mead is In these expressions, B is the earth's magnetic field,
D- 3(L/l0) 10R~/day (Re is the earth's radius). The wo is the drift frequency, </>n a phase angle presumably

TABLE 1. A survey of processes liable to be effective in promoting radial diffusion

Type Time scale of fundamental Effectively Radial Effective diffusion

process energy-conserving? dependence coeff. (R:fday)

Magnetic impulses and sudden

commencements (betatron effect).
Fast rise; decay over several
hours to days.
No ....................... io......................... 3 (t
10

Magnetic bays (azimuthal

electric field).
1(}-30 min ............................ No ....................... .......................... (1-100) (fo)'
Electrostatic instabilities Very short to a few min ............ No ....................... ?........................... ?
(Bohm diffusion).

Pc 2-3 (violations of 2nd 5-45 s ................................. Yes ...................... Probably not as steep -0.5
invariant). as L.
Pc 1 (splitting of drift shells in 0.2-5 s ................................ Yes ...................... Ineffective below <1
axially asymmetric L=5.
magnetosphere).
.742 John M. Cornwall

randomly distributed, and the quantity in angular

brackets is the autocorrelation function of the Fourier D=4Xl()3 (
L )-112 (bB)2 R~lday (7)
10
components of the electric field. For fields generated
by magnetic bays, E points west on the nightside, where b is the hydromagnetic wave strength, and B is
east on the dayside, and so the term with n = 1 in the earth's field strength. For L;;::: 4, we might guess
eqs (1) through (3) is doubtless the most important. biB= I0- 2 (b = 5y at L=4), and thus
We shall in any event only keep the n= 1 term. For
the braekets in (2) we write L )-112
D = 0.4 ( R!lday. (8)
10
(4)
Of course, if biB changes with L very much, formula (7)
where T is a characteristic time for magnetic bays, instead of (8) should be used.
say 2 X 1()3 s. It is then easy to find D: Finally, Pc 1 hydromagnetic waves can cause radial
diffusion in an axially asymmetric magnetosphere
(Roederer, 1967). These waves, of frequency 0.2 to 5 Hz,
Bo=0.31 gauss. (5) are thought to be generated by the Davis-Williamson
protons, so in a sense the protons are causing their
radial diffusion. It is difficult to estimate the diffusion
Little is known about E(L), but extrapolation of iono- coefficient without extensive numerical analysis, but
spheric potentials shows that the associated potential we can argue as follows: A Davis-Williamson proton
over several earth radii might be 1 to 10 kV. We there- drifts around the earth on a time scale measured in
fore set E(L) = I0-8 -I0- 9 in CGS units, and get hours. During this time, strong pitch-angle diffusion
may change the pitch angle by a few tenths of a radian,
corresponding to a radial movement of a few tenths of
D= o-1oo) (foY R~lday. (6) an earth radius. We estimate D by

(This expression does not include the relative varia- D= l. (Llr)2 (9)
tion of E(L)2 with L, and eq (6) should not be used 2 TD
below L = 4 or 5.) This kind of diffusion can easily
dominate magnetic impulse diffusion. In going from where Llr is a fraction of an earth radius, and TD the
(5) to (6), we have set 1 + (wDTI2) 2 = l, since T is drift time (say 1 or 2 hr). An upper limit for D might be
considerably shorter than the drift period. In the op- lR; day- 1 This sort of diffusion cannot occur below
posite limit, TwD ~ 1, we find D- V 0 , since WD- L - 2 L = 5, where the magnetosphere begins to look axially
symmetric.
From the crudity of the estimates of this section,
2.2. Short-Time Scale Processes it is difficult to pin down a good value for the radial
We begin by discussing very briefly certain electro- diffusion coefficient. All things considered, it is
static instabilities that can promote the fast diffusion probably fair to take D at L = 6 of the order of 1 R~lday.
of plasma across a magnetic field. Such fast diffusion It would be not at all surprising to learn that other
we term generically Bohm diffusion. Most important workers have made different (and better) estimates of
are probably the universal instabilities (see, e.g., Krall the diffusion coefficient, but the author knows of no
and Rosenbluth, 1963). Density and temperature published values for processes other than magnetic
gradients in the magnetosphere, such as might be impulses.
associated with the ring current or the plasmapause,
or fluctuations in the particle injection source, are
responsible for these universal instabilities; the 3. Combined Diffusion and Nonlinear Loss Rate
sharper the gradient, the more severe the instability. Scattering of the Davis-Williamson protons by Pc 1
Unfortunately, at the present time there are not can be described as a diffusion process acting on the
quantitative estimates appreciable to the magneto- protons' pitch angles, which causes the protons to
sphere, but the diffusion coefficient can be quite large leak into the loss cone. Because the protons themselves
in principle. This kind of diffusion is essentially the produce Pc 1 waves, the pitch-angle diffusion process
only possible important mechanism below about L = 5. is nonlinear (Cornwall, 1966; Kennel and Petschek,
we then consider processes that violate the longi- 1966). For a sufficiently strong source of protons, sat-
tudinal invariant, i.e., processes with characteristic uration occurs and the proton flux is effectively
periods near the proton bounce period of about 30 s bounded. It would be virtually impossible to combine
(for a 75-ke V proton at L = 6). The Pc 2-3 (5 to 45 s) the nonlinear partial differential equations of pitch-
fits ideally. The formulas needed to calculate the dif- angle diffusion and wave growth with a radial diffusion
fusion coefficient have already been given (Cornwall, differential equation. There is, however, a simple
1966), in an application to electrons interacting with approximation that allows us to describe pitch-angle
Pc 1. The result is diffusion by an effective nonlinear loss term, whose
Conjugate-Point Wave Phenomena 743

magnitude has recently been calculated (Cocke and where J is the omnidirectional flux in cm-2 s-1. The
Cornwall, 1967). number of protons entering the loss cone/s is
Let l(r, v, t) be the Boltzmann distribution func-

I:t fodxx(l- x2)-1/21 I=x(l- x2)1/2De I

tion for the protons. The time rate of change of I by
pitch-angle diffusion can be written: iJI
iJx z=x 0

ijat a [x(l-x2)1/2De-
-=x-1(1-x2)1/2-
ax
ij]
ax (10)
(13)
where xo is the sine of the loss angle. Experimentally,
where x (not to be confused with the spatial coordinate) I- X 2 (roughly), so we carry out the differentiation in
is the sine of the pitch angle, D = 75 (L/4)-3 s- 1, and (13), multiply by a factor representing the fraction of
2 2 (b= Pc 1 magnetic field, B
E is essentially b /B
time micfopulsations are present (say, 0.3) and as-
=earth's field). (See Cocke and Cornwall, 1967, for sume it is sufficiently accurate to replace I by }. The
details.) We propose to replace (10) by a much simpler resulting equation of the type (II) is:
form:
aJ
-=-}A.(}) =-3XlQ2} (L)-3
-
al at 4
-=-A.(f)l. (II)
at
exp {
- 2 j lOS 4 day- 1.
X (L)-4}
(14)
To do so, we need to know some average behavior of
I and E. We can make an empirical fit to the calcula-
tions of Cocke and Cornwall; this fit is shown in figure Note the saturation effect: as j becomes large, the
l, where the data points are computer calculations, and loss rate grows exponentially.
Q is inversely proportional to the omnidirectional The next step is to add to (14) an effective source
flux j of protons. A good fit is: term from radial diffusion. On very general grounds
(Dungey, 1965; Falthammar, l%6a, b) the diffusion
4 term looks like
e(J) = 1.7 X 10- 4 exp {- 2 x
j lOS (L)- }
(12)
4
(15)

(again we replace the Boltzmann distribution function

I by }), where r is the radius variable: r= LRe. In
equilibrium,

aJ- = ra- (Dr-2

0= aJ) -JA.(J).
- (16)
at ar ar

This is a nonlinear equation in J, which a computer

could easily handle. The important nonlinear feature
means that, given the spatial dependence of}, the
magnitude of j can be determined at a particular
point. To illustrate, suppose j - rN, D- ,.S; (16)
becomes
N(N + 3)D}r-2=JA.(J). (17)

Take r=6Re, D= lR~ day-t, N --6 to find }(L=6) = 107

cm- 2 s- 1. This is quite comparable to observed fluxes.
Note that the value of j depends only logarithmically
on the details of the diffusion process, which is for-
tunate since these details are very poorly known.
Of course, there are other loss processes (e.g.,
o- charge exchange, Coulomb scattering) not considered
FicuRE ~- EJJ.ective loss rate (unnormalized) versus Q, Q is
here, but apparently in L = 4-6 pitch-angle diffusion
a vanable Inversely proportional to the omnidirectional is the main loss mechanism, on the basis of rough
./lw:; see text for details. estimates.
 744 John M. Cornwall

4. Conclusion 5. References
A wide variety of physical effects contribute to radial Brice, N. M. (1964), Fundamentals of very low frequency emission
diffusion, and these are difficult to estimate severally generation mechanisms, J. Geophys. Res. 69, No. 21,4515-4522.
Cocke, W. J., and J. M. Cornwall (1967), Theoretical simulation of
and collectively. The exponential dependence of loss micropulsations, J. Geophys. Res. 72, No. 11, 2843-2856.
rate on 1, however, minimizes the effect of this uncer- Cornwall, J. M. (1965), Cyclotron instabilities and electromagnetic
tainty.. We are left with the task of calculating accu- emission in the ULF and VLF frequency ranges, J. Geophys. Res.
rately the coefficient of }- 1 in the exponent in loss rate 70, No. 1, 61-69.
Cornwall, J. M. (1966), Micropulsations and the outer radiation zone,
(12). The work of Cocke and Cornwall (1967) may well J. Geophys. Res. 71, No.9, 2185-2199.
contain systematic errors of a factor of 2 or 3, which Davis, L. R., and J. M. Williamson (1963), Lowenergy trapped pro
might best be revealed by fitting the observed data to tons, Space Res. 3, 365-375.
the solution of (16) or a similar, more accurate equation. Dungey, J. W. (1965), Effects of electromagnetic perturbations on
particles trapped in the radiation belts, Space Science Rev. 4,
Little can be said about the radial dependence of 1 No. 1, 199-122.
until (16) is solved numerically. It is unlikely that J Fiilthammar, C-G. (1965), Effects of electric fields on trapped
can stray too far from a dependence like L - 4 , since radiation, J. Geophys. Res. 70, No. 11, 2503-2516.
J - L - 4 means that the exponential term in (14) Fiilthammar, C-G. (1966a), On the transport of trapped particles in
the outer magnetosphere, J. Geophys. Res. 71, No. 5, 1487-1491.
depends rather insensitively on L. This sort of radial Fiilthammar, C-G. (1966b), Coefficients of diffusion in the outer
dependence is reasonably consistent with experiment. radiation belt, Radiation Trapped in the Earth's Magnetic Field,
One major problem we have left untouched is that ed. B. M. McCormac (D. Reidel Co., Holland).
most of the processes listed in table 1 give quite small Jacobs, J. A., and T. Watanabe (1964), Micropulsation whistlers,
J. Atmosph. Terr. Phys. 26, No. 8, 825-829.
diffusion coefficients at L = 3-4, with at least the Kennel, C. F., and H. E. Petschek (1966), A limit on stably trapped
exception of electrostatic processes, about which we particle fluxes, J. Geophys. Res. 71, No. 1, 1.
know little. The Davis-Williamson protons are cer- Krall, N. A., and M. Rosenbluth (1963), Low-frequency stability of
tainly seen in this L range, so something must be nonuniform plasmas, Phys. Fluids 6, No. 2, 254.
Nakada, N. P., and G. D. Mead (1965), Diffusion of protons in the
diffusing them inward. We speculate that universal outer radiation belt, J. Geophys. Res. 70, No. 15, 3529-3532.
instabilities must be important near the density knee, Obayashi, T. (1965), Hydromagnetic whistlers, J. Geophys. Res. 70,
near L = 3-4, where plasma density gradients are No. 5, 1069-1078.
larger than elsewhere in the magnetosphere, and may Roederer, J. G. (1967), On the adiabatic motion of energetic particles
in a model magnetosphere, J. Geophys. Res. 72, No.3, 981-992.
well be important over a very wide range of L values.

I thank Professor C-G. Fiilthammar for valuable

discussion on the topics of this paper.
This work was completed at Aerospace Corporation.
Publication was supported by the U.S. Air Force under
contract AF 04(695}-1001. (Paper 3-7-421)