PHYSICAL REVIEW LETTERS 130, 071801 (2023)

Editors' Suggestion Featured in Physics

Measurement of the Electron Magnetic Moment

X. Fan ,1,2,* T. G. Myers,2 B. A. D. Sukra,2 and G. Gabrielse2,†
1
Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA
2
Center for Fundamental Physics, Department of Physics and Astronomy,
Northwestern University, Evanston, Illinois 60208, USA

(Received 27 September 2022; revised 7 December 2022; accepted 7 December 2022; published 13 February 2023)

The electron magnetic moment, −μ=μB ¼ g=2 ¼ 1.001 159 652 180 59 ð13Þ [0.13 ppt], is determined
2.2 times more accurately than the value that stood for fourteen years. The most precisely determined
property of an elementary particle tests the most precise prediction of the standard model (SM) to 1 part in
1012 . The test would improve an order of magnitude if the uncertainty from discrepant measurements of the
fine structure constant α is eliminated since the SM prediction is a function of α. The new measurement and
SM theory together predict α−1 ¼ 137.035 999 166 ð15Þ [0.11 ppb] with an uncertainty 10 times smaller
than the current disagreement between measured α values.

DOI: 10.1103/PhysRevLett.130.071801

The quest to find physics beyond the standard model of g S

μ ¼ − μB ; ð1Þ
particle physics (BSM) is well motivated because the SM 2 ℏ=2
is incomplete. No known CP violation mechanism [1] is
large enough to keep the matter and antimatter produced is proportional to its spin S, normalized to its spin
in the big bang [2] from annihilating as the universe eigenvalue ℏ=2. The energy levels are
cooled, dark matter [3,4] has not been identified,


and neither dark energy [5,6] nor inflation [7–10] has a 1
SM explanation. The most precise SM prediction is E ¼ hνs ms þ hνc n þ ; ð2Þ
2
the electron magnetic moment in Bohr magnetons,
−μ=μB ¼ g=2, with μB ¼ eℏ=ð2mÞ for electron charge
where h ¼ 2πℏ. The cyclotron frequency is νc ¼ eB=ð2πmÞ
−e and mass m, and the reduced Planck constant ℏ. It
and n ¼ 0; 1; …. The spin frequency is νs ¼ ðg=2Þνc and
affords great BSM sensitivity [11–19] in that BSM
ms ¼ 1=2. In terms of these frequencies, and the anomaly
particles and electron substructure could shift the mea-
frequency νa ≡ νs − νc,
sured value from what is now predicted (analogous to how
quark substructure shifts the proton moment). The SM μ g ν ν
sectors involved in the prediction include the Dirac − ¼ ¼ s ¼1þ a: ð3Þ
μB 2 νc νc
prediction [20], quantum electrodynamics (QED) [21–28]
with muon and tauon contributions [29], and also hadronic
An important feature of an electron measurement (not
[30–32] and weak interaction contributions [33–36]. The
available with muons [43], for example) is that its cyclotron
SM prediction is a function of the measured fine structure
frequency can be measured in situ. The electron thus serves
constant, α, displayed later in Eq. (7).
as its own magnetometer insofar as B cancels out of the
A new measurement, carried out blind of any measure-
frequency ratios. Choosing to measure νa =νc rather than
ment or prediction, determines μ=μB to 1.3 parts in 1013
νs =νc (called making a g − 2 measurement) significantly
(Fig. 1). Measured in a new apparatus in a lab at a different
university, the new value is 2.2 times more precise than, and
ppt
consistent with, the one that stood for fourteen years [37]. 1 0.5 0 0.5 1

In the most precise confrontation of theory and measure- g/2 2022

g/2 2008
ment, the SM prediction agrees to 1 part in 1012 . Our SM with D(Rb)
SM with D(Cs)
determination and the SM calculation are precise enough
179.5 180 180.5 181 181.5
for a test that is 10 times more precise, once the discrepancy (-P / P - 1.001 159 652 000) u1012
in measured α values [38,39] is resolved.
B

A one-electron quantum cyclotron is used. This is FIG. 1. This Northwestern determination (red) and our 2008
essentially a single electron suspended in a magnetic field Harvard determination (blue) [37]. SM predictions (solid and
B ¼ Bẑ and cooled to its lowest quantum states [42]. The open black points for slightly differing C10 [40,41]) are functions
magnetic moment operator for a spin-1=2 electron, of discrepant α measurements [38,39]. A ppt is 10−12 .

0031-9007=23=130(7)=071801(6) 071801-1 © 2023 American Physical Society

 PHYSICAL REVIEW LETTERS 130, 071801 (2023)

(a) (b) when B is chosen so ν̄c is far from resonance with cavity
radiation modes [52]. The cyclotron damping contributes
0.03 Hz to the cyclotron and anomaly linewidths (to
be discussed), a negligible 0.2 ppt and a very important
0.2 ppb, respectively. Blackbody photons that excite
n ¼ 0 to n ¼ 1 are eliminated by cooling the trap cavity
below 100 mK [42].
The Brown-Gabrielse invariance theorem [53],
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(c)
νc ¼ ν̄2c þ ν̄2z þ ν̄2m ð4Þ

provides the νc and νa ¼ νs − νc needed in Eq. (3) to

determine μ=μB . It is critical that Eq. (4) is invariant under
unavoidable misalignments of B and the axis of V, and
under elliptic distortions of V. The hierarchy ν̄c ≫ ν̄z ≫
FIG. 2. (a) Cryogenic system supports a 50 mK electron trap ν̄m ≫ δ allows an expansion of Eq. (4) that suffices for our
upon a 4.2 K solenoid to provide a very stable B. (b) Silver precision to be inserted in Eq. (3), so
electrodes of a cylindrical Penning trap. (c) Quantum spin and
cyclotron energy levels used for measurement. μ g ν̄a − ν̄2z =ð2f̄ c Þ Δg
− ¼ ≃1þ 2
þ cav ; ð5Þ
μB 2 f̄ c þ 3δ=2 þ ν̄z =ð2f̄ c Þ 2
reduces the effect of frequency measurement uncertainties.
However, it does not evade or reduce at all the largest g=2 with ν̄a and f̄ c [defined in Fig. 2(c)] to be deduced with ν̄z
measurement correction and its uncertainty, as we shall see. from measured line shapes. The added cavity-shift Δgcav =2
A stable magnetic field is still critical at our precision arises because the cyclotron frequency couples to radiation
because νa and νc are not measured simultaneously. Field modes of the trap cavity, shifting both ν̄c and ν̄a [54,55].
drift of 2 ppb=day [44] (4 times below that in [37]) makes This g=2 measurement correction and its uncertainty are
possible round-the-clock measurements, improved statis- not reduced or evaded by a g − 2 measurement. They must
tical precision, and better investigations of uncertainties. be determined and corrected at the full 10−13 precision
The apparatus in Fig. 2(a) achieves this by supporting a of μ=μB .
50 mK trap on a 4.2 K self-shielding solenoid [45] using a To measure the ν̄z needed in Eq. (5), the current induced
mixing chamber flexibly hanging from the rest of a dilution in the electrodes by the axial oscillation is sent through a
refrigerator [46]. resonant circuit that is the input of a cryogenic HEMT
An electron in the field Bẑ is trapped by adding an amplifier. The one-minute Fourier transform of the ampli-
electrostatic quadrupole potential V ∝ z2 − ρ2 =2, with ρ ¼ fier output in Fig. 3(c) illustrates the Johnson noise and
xx̂ þ yŷ [47]. Cylindrical Penning trap electrodes [48,49] electron signal canceling to make a dip that reveals ν̄z [56].
[Fig. 2(b)] with appropriately chosen relative dimensions Energy loss in the circuit damps the axial motion with a
and potentials produce such a potential for a centered time constant γ −1 z ¼ 32 ms. The amplifier heats the elec-
electron, which then oscillates nearly harmonically along ẑ tron axial motion to T z ¼ 0.5 K.
at the axial frequency ν̄z ≈ 114 MHz. For B ¼ 5.3 T, Small shifts in ν̄z provide quantum nondemolition
the trap-modified cyclotron and anomaly frequencies are detection (QND) of one-quantum spin and cyclotron
ν̄c ≈ 149 GHz and ν̄a ≈ 173 MHz [47]. A circular magnet- jumps, without the detection changing the cyclotron or
ron motion at ν̄m ¼ 43 kHz is cooled by axial sideband spin state. Saturated nickel rings [Fig. 2(b)] produce a
cooling [47,50] and its effect is negligible during the magnetic bottle gradient, ΔB ¼ B2 ½ðz2 − ρ2 =2Þẑ − zρρ̂
measurement. Figure 2(c) shows the lowest cyclotron with B2 ¼ 300 T=m2 . This couples spin and cyclotron
and spin energy levels and the frequency spacings, includ- energies to ν̄z , which then shifts by Δν̄z ≈ 1.3ðn þ ms Þ Hz.
ing a relativistic mass shift, δ, given by δ=νc ≡ (The B2 and Δν̄z are 5 and 3 times smaller than used
hνc =ðmc2 Þ ≈ 10−9 [47,51]. previously [37].) To rapidly detect jumps after the cyclotron
The lowest cyclotron states for each spin are effectively and anomaly drives are turned off, the amplified signal is
stable because the spin is so nearly uncoupled from its immediately fed back to the electron. This self-excited
environment [47]. Without a trap, the cyclotron state n ¼ 1 oscillator (SEO) [57] resonantly and rapidly drives itself to
has a lifetime γ −1
c ¼ 0.1 s. With a trap that is also a low-loss a large amplitude even if ν̄z shifts with amplitude, where-
microwave cavity, this rate for the spontaneous emission of upon the gain is adjusted to maintain the amplitude. A
synchrotron radiation is inhibited by a factor of 50 to 70. Fourier transform of the large signal reveals the small Δν̄z
Cyclotron excitations can then be detected before decay, that signals cyclotron and spin jumps.

071801-2
 PHYSICAL REVIEW LETTERS 130, 071801 (2023)

transition probability

transition probability
(a) (b) (c) scale on which the axial amplitude fluctuates, so the

noise power (a.u.)

0.4 0.4 0.8

0.3 0.3 0.6 predicted cyclotron line shape [dashed line in Fig. 3(a)]
0.2 0.2 0.4 approximates an exponential Boltzmann shape, centered
0.1 0.1 0.2 at frequency ν̄c ð1 þ ϵÞ. The anomaly transition time
0 4 2 0 2 4 0 4 2 0 2 4 0 40 20 0 20 40 ðϵν̄a Þ−1 ¼ 1.1 s is much slower than the axial amplitude
drive detuning (ppb) drive detuning (ppb) frequency - Q z (Hz) fluctuations, whereupon the predicted thermal anomaly line
is essentially symmetric about ν̄a ð1 þ ϵÞ and is negligibly
FIG. 3. Quantum jump cyclotron (a) and anomaly (b) line narrow. The observed anomaly linewidth of 0.06 Hz
shapes that are measured (points), predicted (dashed line) and fit
(0.35 ppb) in Fig. 3(b) is from other sources. Half is from
(solid line) vs fractional drive detunings from f̄c ð1 þ ϵÞ and
ν̄a ð1 þ ϵÞ (defined later in the text). (c) A dip in Johnson noise the cyclotron decay lifetime and half is from applying the
reveals ν̄z . anomaly drive for only 30 s.
The anomaly line shape is consistent with what is
predicted but the cyclotron line shape is not. Presumably
Quantum jump spectroscopy produces anomaly and this is due to unwanted magnetic field fluctuations that are
cyclotron resonances [Figs. 3(a) and 3(b)] from which to averaged differently in the anomaly and cyclotron line
extract ν̄a and f̄ c to use in Eq. (5). Cyclotron and anomaly shapes. Such fluctuations, with a 200 Hz bandwidth, were
quantum jump trials are alternated. The magnetic field drift observed with a superconducting solenoid being jostled by
of 0.2 ppb=hr in the new apparatus is slow enough to its environment [60]. The anomaly line shape would
correct using a quadratic fit to the lowest cyclotron drive average away such fluctuations to yield the narrow line
frequencies that produce excitations. Each cyclotron and observed [e.g., Fig. 3(b)]; the cyclotron line shape would
anomaly quantum jump trial starts with resonant anomaly not, giving a possible explanation for the observed
and cyclotron drives that prepare the electron in the spin-up 0.5–0.8 ppb broadening [e.g., Fig. 3(a)].
ground state, jn ¼ 0; ms ¼ 1=2i, followed by 1 s of axial Both ν̄a and f̄ c are extracted from such line shapes.
magnetron sideband cooling [47,50]. Cyclotron line shapes are fit to the predicted line shape
Cyclotron jumps to n ¼ 1 are driven by a 5 s microwave [dashed line in Fig. 3(a)], convoluted with a Gaussian
drive injected between trap electrodes [Fig. 2(b)], with an function to accommodate the broadening. Such a fit,
off-resonance anomaly drive also applied. Jumps occur in illustrated by the solid curve in Fig. 3(a), typically gives
less than 20% of the trials to avoid saturation effects. a 2 ppb cyclotron linewidth, a Gaussian broadening width
Cavity-inhibited spontaneous emission [52] makes the
of about 0.5 ppb, T z ¼ 0.55  0.11 K, and f̄ c with an
excitation persist long enough so that self-excitation feed-
uncertainty of about 0.08 ppb. For anomaly line shapes
back [57] can be turned on in the next 1 s to detect the
[e.g., Fig. 3(b)], nearly symmetric and fractionally narrower
1.3 Hz shift that signals a cyclotron quantum jump.
by about a factor of 4, the uncertainty in ν̄a is thus not very
Anomaly quantum jumps are driven by an oscillatory
significant for the final uncertainty. Fitting with or without
potential applied to trap electrodes for 30 s to drive an off-
Gaussian broadening makes little difference [e.g., solid
resonance axial oscillation of the electron through the radial
curve in Fig. 3(b)].
magnetic gradient B2 zρ. A cyclotron drive remains applied
The cavity-shift Δgcav =2 in Eq. (5), the only correction to
but is off resonance. The electron sees the oscillating
what is directly measured, arises because the cyclotron
magnetic field perpendicular to ẑ as needed to flip its spin,
oscillator couples to radiation modes of the trap cavity and
with a radial gradient that allows a simultaneous cyclotron
transition [47]. A spontaneous decay to the spin-down shifts ν̄c [54,55]. It is the downside of the cavity-inhibited
ground state, jn ¼ 0; ms ¼ −1=2i, would be detected spontaneous emission that desirably narrows resonance
during the 60 s (more than 10 cyclotron decay times) after lines, and makes it possible to observe a cyclotron
the drives are turned off. A maximum jump rate of 40% excitation before it decays. The cylindrical trap was
suggests a slight power broadening, but ν̄a is still deter- invented [48] to allow cavity modes and shifts to be
understood and calculated. Nonetheless, the mode frequen-
mined far more precisely than f̄ c .
cies and Q values must still be measured because of energy
Well-understood, asymmetric cyclotron and symmetric
losses in induced surface currents, imperfect cavity machin-
anomaly line shapes are predicted [58] for thermal
axial motion at temperature T z within a magnetic gradient ing, slits that make cavity sections into separately biased
trap electrodes, and dimension changes as the cavity cools
B2 z2 . To this, the effect of cyclotron decay has been
below 100 mK from 300 K. Three consistent methods are
added [59]. The average oscillation amplitude squared is
used: (1) parametrically pumped electrons [59,61,62],
z2 ¼ kB T z =ð4π 2 ν̄2z mÞ, where kB is the Boltzmann constant. (2) measuring how long one electron stays in its first
The average field for the electron is shifted by ϵB ¼ B2 z2 excited cyclotron state [37,59], and (3) a new method of
and broadened by the same amount. The cyclotron band- observing the decay time of an electron exited to nc ≈ 10.
width ϵν̄c corresponds to a time ðϵν̄c Þ−1 ¼ 1.3 ms needed A renormalized calculation [54,55] with added cyclotron
to establish ν̄c . This is much faster than the γ −1
z ¼ 32 ms damping [37,59] avoids the infinite cavity shifts that result

071801-3
 PHYSICAL REVIEW LETTERS 130, 071801 (2023)
magnetic field (T)
3 3.5 4 4.5 5 5.5
TABLE I. Largest uncertainties for g=2.
(a) gg
3
' = 1.001 159 652 180 59
2 2
Source Uncertainty × 1013
(10 )

2
-12

1 Statistical 0.29
0 Cyclotron broadening 0.94
g
2
'

1 Cavity correction 0.90

2 Nuclear paramagnetism 0.12
3 Anomaly power shift 0.10
Magnetic field drift 0.09
(b)
J /(2 S) (Hz)

10
1 Total 1.3

2
c

10
80 90 100 110 120 130 140 150
Several SM sectors together predict
cyclotron frequency (GHz)

2
3
4
g α α α α
¼ 1 þ C2 þ C4 þ C6 þ C8
FIG. 4. (a) Measured g=2 − 1.001 159 652 180 59 before 2 π π π π
(white) and after (red) cavity-shift correction. (b) Measurements
5
α
take place in valleys of the cyclotron damping rate where þ C10 þ    þ aμτ þ ahadronic þ aweak : ð7Þ
spontaneous emission is inhibited. π

The Dirac prediction [20] is first on the right. QED provides

from summing all mode contributions. This calculation the asymptotic series in powers of α, along with the muon
assumes the mode frequencies of a perfect cylinder, one Q and tauon contributions aμτ [40]. The constants C2 [21], C4
for TE modes, and another for TM modes. We calculate [22,23], C6 [24,25], and C8 [26] are calculated exactly, but
with dimensions chosen to best match observed frequencies require measured lepton mass ratios as input [29]. The
and a single Q value for all modes. After shifts from the 72 measurements are so precise that a numerically calculated
observed modes using the ideal frequencies and the one Q tenth order C10 [27,28] is required and tested. A second
value are subtracted out, contributions for these modes evaluation of C10 [41] differs slightly for reasons not yet
using measured frequencies and Q values are added back understood and the open points in Figs. 1 and 5 use this
in. The leading contribution to cavity shift uncertainties alternative. Hadronic and weak interaction contributions
comes from modifications of the field that an electron sees are ahadronic [30–32] and aweak [33–36]. The exact C8 and
from imperfections and misalignments of the trap cavity. the numerical C10 are remarkable advances that reduce the
Figure 4(a) shows the consistency of μ=μB determinations calculation uncertainty well below the uncertainties
at 11 different magnetic fields, after each receives a reported for the measured μ=μB and α.
different cavity shift. The most precise α measurements [38,39], needed for the
A weighted average of the 11 determinations gives SM prediction of g=2 in Eq. (7), disagree by 5.5σ, about
μ g 10 times our measurement uncertainty (Fig. 1). Until
− ¼ ¼ 1.001 159 652 180 59ð13Þ ½0.13 ppt; ð6Þ the discrepancy is resolved, the best that can be said is
μB 2 that the predicted and measured μ=μB agree to about
δðg=2Þ ¼ 0.7 × 10−12 , half of the α discrepancy. A generic
with 1σ uncertainty in the last two digits in parentheses.
chiral symmetry model [63] thenpsuggests ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi that the
Figure 1 shows the good agreement of this 2022 determi-
nation at Northwestern with our 2008 determination at electron radius is less than Re ¼ jδðg=2Þjℏ=ðmcÞ ¼
Harvard [37] and an uncertainty that is improved by a factor 3.2 × 10−19 m, and that the mass of possible elec-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

of 2.2. Because uncertainty correlations from similar tron constituents must exceed m ¼ m= jδðg=2Þj ¼
measurement methods are difficult to determine, we do 620 GeV=c2 . If δðg=2Þ would equal our μ=μB deter-
not recommend averaging our two determinations. Table I mination uncertainty, then Re ¼ 1.4 × 10−19 m and
lists uncertainty contributions to the final result. The m ¼ 1.4 TeV=c2 .
statistical uncertainty is from the fits that extract f̄ c and A 2.2 times reduced δðg=2Þ would bring us to the level of
ν̄a . The two dominant uncertainties have been discussed— the intriguing 4.2 standard deviation discrepancy between
cyclotron broadening and cavity shifts (treated as correlated the measured and predicted muon magnetic moment
for nearby fields). The nuclear paramagnetism uncertainty [43,64]. The muon’s BSM sensitivity, expected to be 40
is based upon the measured temperature fluctuations of the 000 times higher (the ratio of muon and electron masses), is
silver trap electrodes. The anomaly power shift uncertainty largely offset by our 3150 times smaller uncertainty.
comes from the measured frequency dependence on drive The fine structure constant α is the fundamental measure
strength. of the strength of the electromagnetic interaction in the low

071801-4
 PHYSICAL REVIEW LETTERS 130, 071801 (2023)
ppb
1 0.5 0 0.5 development is supported by the Templeton Foundation, and
g/2(2022) with SM low-loss trap cavity development is supported by the DOE
g/2(2008) with SM
Rb SQMS Center.
Cs
0 50 100 150 200
9
(D -1 - 137. 035 999 000) u10
*
[email protected]
†
FIG. 5. SM prediction of α using μ=μB from this Northwestern [email protected]
measurement (red), and from our 2008 Harvard measurement [1] A. D. Sakharov, Sov. J. Exp. Theor. Phys. Lett. 5, 24 (1967).
(blue), with solid and open points for slightly differing C10 [2] G. Gamow, Phys. Rev. 70, 572 (1946).
[40,41]. The α measurements (black) were made with Cs at [3] F. Zwicky, Helv. Phys. Acta 6, 110 (1933).
Berkeley [38] and Rb in Paris [39]. A ppb is 10−9 . [4] V. C. Rubin, J. Ford, W. K., and N. Thonnard, Astrophys. J.
238, 471 (1980).
[5] S. Perlmutter et al., Astrophys. J. 517, 565 (1999).
[6] A. G. Riess et al., Astron. J. 116, 1009 (1998).
energy limit. For the SI system of units, α ¼ e2 =ð4πϵ0 ℏcÞ [7] A. Linde, Phys. Lett. 108B, 389 (1982).
is a measure of the vacuum permittivity ϵ0, given that and e, [8] A. Starobinsky, Phys. Lett. 91B, 99 (1980).
ℏ, and the speed of light c are now defined [65]. Our μ=μB [9] A. Starobinsky, Phys. Lett. 117B, 175 (1982).
and the SM give [10] A. H. Guth, Phys. Rev. D 23, 347 (1981).
[11] P. Brax, A.-C. Davis, B. Elder, and L. K. Wong, Phys. Rev.
α−1 ¼ 137.035 999 166 ð02Þ ð15Þ ½0.014 ppb½0.11 ppb; D 97, 084050 (2018).
[12] E. J. Chun, J. Kim, and T. Mondal, J. High Energy Phys. 12
¼ 137.035 999 166 ð15Þ ½0.11 ppb; ð8Þ (2019) 068.
[13] G. F. Giudice, P. Paradisi, and M. Passera, J. High Energy
with theoretical and experimental uncertainties in the Phys. 11 (2012) 113.
first and second brackets. Figure 5 compares to the α [14] A. Aboubrahim, T. Ibrahim, and P. Nath, Phys. Rev. D 89,
measurements (black) that disagree with each other by 093016 (2014).
5.5σ. Our value differs by 2.1 standard deviations [15] M. Endo, K. Hamaguchi, and G. Mishima, Phys. Rev. D 86,
from the Paris Rb determination of α [39] and by 3.9 095029 (2012).
[16] H. Davoudiasl and W. J. Marciano, Phys. Rev. D 98, 075011
standard deviations from the Berkeley Cs determination
(2018).
[38]. The C10 in [41] would change only “66” to “59” [17] M. Bauer, M. Neubert, S. Renner, M. Schnubel, and A.
in Eq. (8). Thamm, Phys. Rev. Lett. 124, 211803 (2020).
For the future, a measurement is underway to realize the [18] J. Liu, C. E. M. Wagner, and X.-P. Wang, J. High Energy
new precision with a positron, to improve the test of the Phys. 03 (2019) 008.
fundamental CPT symmetry invariance of the SM by a [19] M. Endo and W. Yin, J. High Energy Phys. 08 (2019) 122.
factor of 40 [66]. Much larger improvements in the [20] P. A. M. Dirac, Proc. R. Soc. A 118, 351 (1928).
precision of μ=μB now seem feasible given the demon- [21] J. Schwinger, Phys. Rev. 73, 416 (1948).
stration of more stable apparatus, improved statistics, and [22] A. Petermann, Helv. Phys. Acta 30, 407 (1957).
better understood uncertainties. Detectors being tested, [23] C. M. Sommerfield, Ann. Phys. (N.Y.) 5, 26 (1958).
[24] T. Kinoshita, Phys. Rev. Lett. 75, 4728 (1995).
more harmonic and lower loss trap cavities, and detector
[25] S. Laporta and E. Remiddi, Phys. Lett. B 379, 283 (1996).
backaction circumvention methods [67,68] should enable [26] S. Laporta, Phys. Lett. B 772, 232 (2017).
much more precise measurements to come. [27] T. Aoyama, T. Kinoshita, and M. Nio, Phys. Rev. D 97,
In conclusion, an electron magnetic moment measure- 036001 (2018).
ment is carried out blind to previous measurements and [28] T. Aoyama, M. Hayakawa, T. Kinoshita, and M. Nio,
predictions. A PhD thesis [69] and a longer publication in Phys. Rev. Lett. 109, 111807 (2012).
preparation give fuller accounts. In new apparatus at a [29] A. Kurz, T. Liu, P. Marquard, and M. Steinhauser,
different university, the measured μ=μB is consistent with Nucl. Phys. B879, 1 (2014).
our 2008 measurement, with a factor of 2.2 improved [30] D. Nomura and T. Teubner, Nucl. Phys. B867, 236 (2013).
precision. The most precise prediction of the SM agrees [31] A. Kurz, T. Liu, P. Marquard, and M. Steinhauser,
Phys. Lett. B 734, 144 (2014).
with the most precise determination of a property of an
[32] Jegerlehner Fred, EPJ Web Conf. 218, 01003 (2019).
elementary particle to about 1 part in 1012 . When discrepant [33] K. Fujikawa, B. W. Lee, and A. I. Sanda, Phys. Rev. D 6,
α measurements are resolved, the new measurement 2923 (1972).
uncertainty of 1.3 parts in 1013 is available for a more [34] A. Czarnecki, B. Krause, and W. J. Marciano, Phys. Rev.
precise test for BSM physics. Lett. 76, 3267 (1996).
[35] M. Knecht, M. Perrottet, E. de Rafael, and S. Peris, J. High
Early contributions were made by S. E. Fayer. NSF Energy Phys. 11 (2002) 003.
1903756 and 2110565 provided the support, with X. Fan [36] A. Czarnecki, W. J. Marciano, and A. Vainshtein, Phys. Rev.
supported by the Masason Foundation. Detector D 67, 073006 (2003).

071801-5
 PHYSICAL REVIEW LETTERS 130, 071801 (2023)

[37] D. Hanneke, S. Fogwell, and G. Gabrielse, Phys. Rev. Lett. [53] L. S. Brown and G. Gabrielse, Phys. Rev. A 25, 2423
100, 120801 (2008). (1982).
[38] R. H. Parker, C. Yu, W. Zhong, B. Estey, and H. Müller, [54] L. S. Brown, G. Gabrielse, K. Helmerson, and J. Tan, Phys.
Science 360, 191 (2018). Rev. Lett. 55, 44 (1985).
[39] L. Morel, Z. Yao, P. Cladé, and S. Guellati-Khélifa, Nature [55] L. S. Brown, G. Gabrielse, K. Helmerson, and J. Tan, Phys.
(London) 588, 61 (2020). Rev. A 32, 3204 (1985).
[40] T. Aoyama, T. Kinoshita, and M. Nio, Atoms 7, 28 (2019). [56] D. J. Wineland and H. G. Dehmelt, J. Appl. Phys. 46, 919
[41] S. Volkov, Phys. Rev. D 100, 096004 (2019). (1975).
[42] S. Peil and G. Gabrielse, Phys. Rev. Lett. 83, 1287 (1999). [57] B. D’Urso, R. Van Handel, B. Odom, D. Hanneke, and G.
[43] B. Abi et al., Phys. Rev. Lett. 126, 141801 (2021). Gabrielse, Phys. Rev. Lett. 94, 113002 (2005).
[44] X. Fan, S. E. Fayer, and G. Gabrielse, Rev. Sci. Instrum. 90, [58] L. S. Brown, Ann. Phys. (N.Y.) 159, 62 (1985).
083107 (2019). [59] D. Hanneke, S. Fogwell Hoogerheide, and G. Gabrielse,
[45] G. Gabrielse and J. Tan, J. Appl. Phys. 63, 5143 (1988). Phys. Rev. A 83, 052122 (2011).
[46] G. Gabrielse, S. Fayer, T. Myers, and X. Fan, Atoms 7, 45 [60] J. W. Britton, J. G. Bohnet, B. C. Sawyer, H. Uys, M. J.
(2019). Biercuk, and J. J. Bollinger, Phys. Rev. A 93, 062511
[47] L. S. Brown and G. Gabrielse, Rev. Mod. Phys. 58, 233 (2016).
(1986). [61] J. Tan and G. Gabrielse, Phys. Rev. A 48, 3105 (1993).
[48] G. Gabrielse and F. C. MacKintosh, Int. J. Mass Spectrom. [62] J. Tan and G. Gabrielse, Phys. Rev. Lett. 67, 3090 (1991).
Ion Process. 57, 1 (1984). [63] S. J. Brodsky and S. D. Drell, Phys. Rev. D 22, 2236 (1980).
[49] J. Tan and G. Gabrielse, Appl. Phys. Lett. 55, 2144 (1989). [64] T. Aoyama et al., Phys. Rep. 887, 1 (2020).
[50] R. S. Van Dyck, Jr., P. B. Schwinberg, and H. G. Dehmelt, in [65] M. Stock, R. Davis, E. de Mirandés, and M. J. T. Milton,
New Frontiers in High Energy Physics, edited by B. Metrologia 56, 022001 (2019).
Kursunoglu, A. Perlmutter, and L. Scott (Plenum, New [66] S. Fogwell Hoogerheide, J. C. Dorr, E. Novitski, and G.
York, 1978), p. 159. Gabrielse, Rev. Sci. Instrum. 86, 053301 (2015).
[51] G. Gabrielse, H. Dehmelt, and W. Kells, Phys. Rev. Lett. 54, [67] X. Fan and G. Gabrielse, Phys. Rev. Lett. 126, 070402 (2021).
537 (1985). [68] X. Fan and G. Gabrielse, Phys. Rev. A 103, 022824 (2021).
[52] G. Gabrielse and H. Dehmelt, Phys. Rev. Lett. 55, 67 [69] X. Fan, Ph.D. thesis, Harvard University, 2022, thesis
(1985). advisor: G. Gabrielse.

071801-6