PHYSICAL REVIEW D, VOLUME 70, 083501

Dark-matter electric and magnetic dipole moments

Kris Sigurdson,1,* Michael Doran,2 Andriy Kurylov,1 Robert R. Caldwell,2 and Marc Kamionkowski1
1
California Institute of Technology, Mail Code 130-33, Pasadena, California 91125, USA
2
Department of Physics and Astronomy, Dartmouth College, 6127 Wilder Laboratory, Hanover, New Hampshire 03755, USA
(Received 17 June 2004; published 6 October 2004)
We consider the consequences of a neutral dark-matter particle with a nonzero electric and/or
magnetic dipole moment. Theoretical constraints, as well as constraints from direct searches, precision
tests of the standard-model, the cosmic microwave background and matter power spectra, and cosmic
gamma rays, are included. We find that a relatively light particle with mass between an MeV and a few
GeVand an electric or magnetic dipole as large as
3  1016 e cm (roughly 1:6  105 B ) satisfies all
experimental and observational bounds. Some of the remaining parameter space may be probed with
forthcoming more sensitive direct searches and with the Gamma-Ray Large Area Space Telescope.

DOI: 10.1103/PhysRevD.70.083501 PACS numbers: 95.35.+d, 13.40.Em, 14.80.–j, 95.30.Cq

I. INTRODUCTION some have studied whether dark-matter might be charged

[9] or have a millicharge [10,11].
A wealth of observational evidence indicates the exis-
Our investigation follows in spirit the latter possibility.
tence of considerably more mass in galaxies and clusters
In particular, dark-matter is so called because the cou-
of galaxies than we see in stars and gas. The source of the
pling to photons is assumed to be nonexistent or very
missing mass has been a problem since Zwicky’s 1933
measurement of the masses of extragalactic systems [1]. weak, or else we would have presumably seen such par-
Given the evidence from galaxy clusters, galaxy dynam- ticles either through absorption or emission of radiation or
ics and structure formation, big-bang nucleosynthesis, in laboratory experiments. In this paper, we ask the
and the cosmic microwave background that baryons can question, ‘How dark is ‘dark’?’’ In other words, how
only account for
1=6 of this matter, most of it must be weak must the coupling of the dark-matter particle to
nonbaryonic. Although neutrinos provide the cosmologi- the photon be in order to be consistent with laboratory
cal density of dark-matter if their masses sum to
12 eV, and astrophysical constraints? In the work on milli-
such particles cannot (essentially from the Pauli princi- charged particles, a dark-matter coupling to photons
ple) have a sufficiently high phase-space density to ac- was assumed to arise from a tiny charge.
count for galactic dark-matter halos [2]; moreover, such In this paper we consider the possibility that the dark-
masses are now inconsistent with neutrino-mass mea- matter possesses an electric or magnetic dipole moment.
surements [3]. Theorists have thus taken to considering The punch line, illustrated in Fig. 1, is that a Dirac
for dark-matter candidates new physics beyond the particle with an electric or magnetic dipole moment of
standard-model. To date, the most promising candi- order
1017 e cm with a mass between an MeVand a few
dates —those that appear in fairly minimal extensions GeV can provide the dark-matter while satisfying all
of the standard-model and which coincidentally have a experimental and observational constraints.1
cosmological density near the critical density— are a In the following Section, we introduce the effective
weakly-interacting massive particle (WIMP), such as Lagrangian for the dipolar-dark-matter (DDM) interac-
the neutralino, the supersymmetric partner of the photon, tion with photons. We discuss the relic abundance in
Z0 boson, and/or Higgs boson [4], or the axion [5]. A Section III. Section IV presents constraints on dark-
considerable theoretical literature on the properties and matter dipole moments and masses that arise from direct
phenomenology of these particles has arisen, and there searches at low-background detectors as well as con-
are considerable ongoing experimental efforts to detect straints from high-altitude experiments. Section V dis-
these particles. cusses constraints due to precision tests of the standard-
In the absence of discovery of such particles, it may be model, while Section VI discusses constraints due to the
well worth exploring other possibilities. Thus, an alter- cosmic microwave background and the growth of large-
native line of investigation takes a more model- scale structure. We provide some concluding remarks in
independent approach and seeks to explore phenomeno-
logically the possible properties of a dark-matter particle. 1
Throughout, we will quote numbers for both the electric and
Along these lines, for example, constraints to strongly- magnetic dipole moments in units of e cm, where e is the
interacting dark-matter were considered in Ref. [6]; self- electron charge. For reference, the Bohr magneton B 
interacting dark-matter has been considered [7,8], and eh=2me  1:93  1011 e cm in these units. Also note that we
work in rationalized Gaussian units so that the fine-structure
constant  e2 =4 hc  1=137, and in particle-physics units
*Electronic address: ksigurds@tapir.caltech.edu (h  c  1) e2  4 =137 and e  0:303.

1550-7998= 2004=70(8)=083501(15)$22.50 70 083501-1  2004 The American Physical Society

KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501

FIG. 1 (color online). The constraints on the dipolar-dark-matter parameter space m ; D;M that come from present day
searches and experiments. Viable candidates must lie in the shaded region, below the solid lines and outside the long-dashed lines.
The short-dashed ‘‘relic abundance’’ curve shows where the dark-matter would have a cosmological density  h2  0:135,
assuming standard freeze-out, no particle-antiparticle asymmetry, and no interactions with standard-model particles apart from
the dipole coupling to photons. Note that the EGRET and GLAST curves constrain the combination D4  M4 1=4 , the
perturbative and unitarity curves apply to the stronger of D;M , while all other curves restrict D2  M2 1=2 .

Section VII. The Appendix provides details of the calcu- On dimensional grounds, we expect the electric dipole
lation of the drag-force between the baryon and DDM moment to satisfy the limit D & em1 ’ 2 
fluids used in Section VI. 1014 mp =m e cm, where mp is the proton mass.
Similar arguments also apply to the magnetic dipole
II. THEORY OF DIPOLE MOMENTS moment.2 This limit is shown as the perturbative bound
in Fig. 1, as violation of this bound would signal some
A particle with a permanent electric and/or magnetic nontrivial or nonperturbative field configuration. As we
dipole moment must have a nonzero spin; we thus con- will see below, more rigorous but slightly weaker upper
sider spin-1/2 particles. Moreover, Majorana particles limits can be set with unitarity arguments.
cannot have permanent dipole moments, so we consider These upper limits already simplify our analysis. The
Dirac fermions. Since the spin and the magnetic dipole phenomenology of charged dark-matter particles is de-
are both axial vectors, a magnetic dipole moment can termined largely by the ability of these particles to form
arise without violating any discrete symmetries. atom-like bound states with electrons, protons, or each
However, the electric dipole moment is a vector and other. However, dipolar-dark-matter cannot form such
thus requires time-reversal and parity violation. bound states. A neutral particle with a magnetic-moment
The effective Lagrangian for coupling of a Dirac fer- will not form bound states with charged particles.
mion with a magnetic dipole moment M and an Curiously enough, a neutral particle with an electric
electric dipole moment D to the electromagnetic field dipole moment (EDM) can indeed form a bound state
F is with an electron, as first noted by no less than Fermi and
i Teller [12], but only if the dipole moment is greater than
L    M  5 D F : (1) 0:639ea0  3:4  109 e cm (assuming m me ) where
2
a0 is the Bohr radius. For smaller values of the dipole, the
At energies low compared to the dark-matter mass, the
photon is blind to distinctions between M and D (unless 2
The limit is satisfied if m is the lightest scale relevant for
time-reversal-violating observables are considered). the DDM sector (see Section V for discussion). Note, however,
Hence, we can discuss limits to D which equally apply that the actual magnitude of the dipole moments in a particular
to M, except where noted. theory can be significantly below this limit.

083501-2
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
electron ‘‘sees’’ both poles of the dipole and finds no
stable orbit. This critical electric dipole moment scales
inversely with the dipole-electron reduced mass, so a
bound state with a proton can occur if the dipole mass
is mp and D * 1:8  1012 e cm. As we will see
below, such values for the EDM cannot occur for a point-
like DDM. Likewise, the weakness of the dipole-dipole
interaction prevents the formation of any stable dark-
matter atoms.
The first cosmological constraint is that from big-bang
nucleosynthesis (BBN). BBN requires that the effective
number of relativistic degrees of freedom at T
MeV FIG. 2. Feynman diagrams for annihilation of a DDM–anti-
DDM pair to two photons.
does not exceed the equivalent of roughly 0.2 neutrino
species [13]. Since the particles we are considering are
Dirac particles, they act like two effective neutrino spe-   D4  M4 m2 =2
!2 v
cies and thus cannot be relativistic and in equilibrium at
 1:0  1033 m2GeV D417  M417 cm3 sec1 ;
BBN. Generally, such considerations rule out m & MeV,
and so we restrict our attention in Fig. 1 to masses above  !ff v  Neff D2  M2
an MeV. Strictly speaking, if the dipole moment is  2  1027 Neff D217  M217 cm3 sec1 ;
D; M & 1022 e cm, and if the particle has no other (4)
interactions with standard-model particles, then a parti-
cle of mass & MeV can decouple at a temperature * where D17 ; M17  D;M =1017 e cm , and mGeV 
10 GeV, and if so, it will evade the BBN limit. P
m =GeV. Here, Neff  f Q2f Ncf is the effective number
of fermion-antifermion pairs with mass mf < m , Qf is
III. DARK-MATTER ANNIHILATION AND
RELIC ABUNDANCE the charge of fermion f, and Ncf is the number of color
degrees of freedom for fermion f. (Ncf  1 for elec-
DDM particles can exist in thermal equilibrium in the trons.) In the standard-model, annihilation can also occur
early Universe when the temperature T m , and their to W  W  pairs above threshold. For D17 ; M17 
interactions will freeze out when T drops below m m =mp & 5000, fermions are the dominant final-state.
resulting in some cosmological relic abundance. The The present day mass density of DDM particles thus
mass density of relic DDM particles is fixed by the depends primarily on the dipole moment. If such particles
cross-section ann for annihilation to all lighter particles are to account for the dark-matter, then  h2  0:135
times the relative velocity v through (see, e.g., Eq. (5.47)
[15], and D2  M2 1=2 ’ 1:0  1017 e cm for m

in Ref. [14]),

 1 GeV. The full mass dependence of this result is shown
m p
 h2 ’ 3:8  107 lnA= lnA =A in Fig. 1.
mp The cross sections in Eq. (4) are s-wave cross sections.

1 According to partial-wave unitarity, the total s-wave
1=2 ann v
 0:135g =10 annihilation cross-section must be  & 4 =m2 [16], so
5:3  1026 cm3 sec1
p that D;M m & 3, fixed by the cross-section for anni-
ln A= lnA hilation to two photons. This limit is shown in Fig. 1, as is
 ; (2)
21 the more stringent, but less rigorous, limit D;M &
where e=m .
p Of course, the present day mass density of DDM par-
A  0:038 g mpl m ann v

 ticles could differ from the estimates obtained above. To
6:3  109 g =10 1=2 ann v
 (3)
m =GeV 5:3  1026 cm3 sec1
assuming that annihilation takes place (as it does in our
case) through the s wave. Here, g is the effective number
of relativistic degrees of freedom at the temperature Tf

m =A of freeze-out. For the interaction of Eq. (1),
DDM–anti-DDM pairs can annihilate to either two pho-
tons or to charged particle-antiparticle pairs through the
diagrams shown in Figs. 2 and 3. The cross sections for FIG. 3. Feynman diagram for DDM–anti-DDM annihilation
these two processes (to lowest-order in v) are to fermion-antifermion pairs.

083501-3
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
obtain these results, we assumed (i) that the dipole inter-
action with photons is the only interaction of these par-
ticles; and (ii) that there is no particle-antiparticle
asymmetry. It is reasonable to assume that in any realistic
model, a dark-matter dipole interaction will arise from
loop diagrams involving other standard-model and new
FIG. 4. Feynman diagram for elastic-scattering of an elec-
particles. If so, then there may be other contributions to
tron from a DDM-particle.
the annihilation cross sections. In this case, the relic
abundance will be smaller than we have estimated above.
We thus conclude that if there is no particle-antiparticle the DDM might produce a signal in a direct-detection
asymmetry, our estimates should be treated as an upper experiment. Although the total cross-section is formally
bound to the relic abundance, and the  h2 curve in infinite, the divergence comes from the small-momen-
Fig. 1 should thus be considered an upper limit to the tum-transfer scatterings that will be screened by atomic
desired values of D and M. On the other hand, the relic electrons. Roughly speaking, then, the DDM-nucleus
abundance could also be increased if exotic processes elastic-scattering cross-section will be 
Ze 2 D2 
increase the expansion rate during freeze-out [17]. M2 =2 v2 ’ 6:4  1032 Z2 D217  M217 cm2 , using
If there is an asymmetry between and , then the v
103 c. Current null searches in germanium detectors
relic abundance is increased relative to our estimate. In [A; Z  76; 32 ] correspond for masses m
10 GeV
this case, however, the present day Universe should con- to a rough upper limit to the cross-section
tain predominantly either particles or antiparticles.
1042 cm2 [19], thus ruling out any dipole moment
Although there is no a priori reason to expect there to D217  M217 1=2 * 107 . This is shown in Fig. 1 as the
be a particle-antiparticle asymmetry, the observed horizontal dashed line at D  1024 e cm. Note that the
baryon-antibaryon asymmetry might lead us to expect cross-section limit depends (and increases) with mass at
an analogous dark-matter asymmetry, should the dark- higher masses; the curve appears as a horizontal line
matter be composed of Dirac particles. It is possible such simply because of the break in scale on the y axis.
asymmetries have a common origin. This seems like a very stringent limit, especially con-
Finally, we have assumed above that the particles sidering the value, D17
1, favored for the correct cos-
freeze out when they are nonrelativistic. However, as the mological density. However, if the dipole moment
dipole moment is lowered for a given mass, freeze-out becomes large enough, the particles will be slowed in
occurs earlier. If the dipole moment is reduced beyond a the rock above the detector and thus evade detection in
certain value, and if there are no other couplings to these underground experiments. In order to determine the
standard-model particles, then freeze-out will be relativ- magnitude of the dipole moment for this to occur, we
istic. These particles will then be roughly as abundant as need to calculate the stopping power dE=dx for the
photons, and they will overclose the Universe by huge particle as it passes through the atmosphere and then
margins unless their masses are & few eV; even in this the rock. Since elastic-scattering takes place through
case they will violate constraints to hot dark-matter from exchange of a photon, it leads to a long-range interaction
the CMB and large-scale structure, and they will also be and, as we have seen above, a formally divergent elastic-
unable, from the Tremaine-Gunn argument, to make up scattering cross-section. The calculation of the stopping
the dark-matter in Galactic halos. The transition from power thus parallels that for ionization loss due to
nonrelativistic to relativistic freeze-out occurs (again, Coulomb collisions, with two important differences.
assuming no nondipole interactions with standard-model First of all, since the long-range force is / r3 , as op-
particles) for m D217 & 1010 GeV for m * MeV, and posed to r2 for Rutherford scattering, stopping occurs
for m & MeV, at m D4=3 17 & 200 MeV.
via scattering from nuclei, rather than electrons. Second,
since this interaction falls more rapidly with radius than
the Coulomb interaction, the stopping power is due pri-
IV. DIRECT-DETECTION marily to hard scatters at small impact parameter, rather
The diagram for scattering of a DDM-particle with a than soft scatters at a wide range of impact parameters.
particle of charge Ze occurs through the exchange of a Our result for the stopping power due to scattering
photon, as shown in Fig. 4 (not unlike the electron- from nuclei of charge Z is
neutron interaction [18]). In the nonrelativistic limit, the Z
differential cross-section for this process is given by, dE Ze 2 D2  M2 2 c2
 nN Td  nN ; (6)
dx 2 mN
d Z2 e2 D2  M2
 2 2 ; (5) where the kinetic-energy transfer in a single collision is
d 8 v 1  cos!
T  p2 1  cos! =mN , x is the depth, and  
where v is the relative velocity. Through this interaction  m ; mN  m mN m  mN 1 is the reduced mass.

083501-4
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
For very weak WIMP interactions with nuclei, the most situated at a depth of 16 mwe. With a detector energy
restrictive limits on the WIMP-proton cross-section (the threshold of Eth  5 keV, it is sensitive to DDM masses
smallest upper bounds) are obtained from null searches down to m
10 GeV. Near this threshold we find that
from experiments that are deepest underground (so as to DDM particles are stopped by the shielding for D17 *
reduce the background). However, the most restrictive 20. This bound grows more prohibitive with increasing
constraints on the cross-section at the upper end of the mass, as illustrated in Fig. 1. The Cryogenic Rare Event
excluded range of cross sections will come from the Search with Superconducting Thermometers (CRESST)
shallowest underground experiment with a null result. [21], though at a depth of 3800 mwe, extends to slightly
From Eq. (6), we find that the dark-matter particles lower masses, having a detector threshold energy Eth 
will only penetrate to a depth x  Ei  Ef =jdE=dxj 0:6 keV. Near m
1 GeV the minimum dipole strength
where Ei  12 m v2 is the initial dark-matter kinetic- is D17 * 2. However, there are no limits from under-
energy and Ef is the final energy. For a stopped particle, ground experiments for DDM masses below 1 GeV.
Ef  0. However, the particle only needs to lose enough Two airborne experiments—unobscured by the atmo-
energy for Ef to drop below the detection threshold for a sphere or rock—which have closed the windows on some
particular experiment. Equating the maximum kinetic- forms of strongly-interacting dark-matter [6,22], also
energy transferable in a collision (!  ) to the threshold restrict dark-matter dipole moments. To determine the
detectable nuclear-recoil energy (Eth ), we find the veloc- predicted signal at a detector, we recast Eq. (5) as the
ity must be slowed to v2f  md Eth =2 m ; md 2 , where cross-section per energy transfer, whereby d=dT 
md is the mass of the nuclei in the detector, and  m ; md Z2 e2 D2  M2 =4 v2 T. The event rate (per time, en-
ergy, and unit mass of detector) is
is the DDM-nucleus reduced mass. Hence, the final dark-
matter energy must be Ef  m md Eth =4 m ; md 2 . v d
So far we have assumed that the particle loses energy R  NN 0:3 GeV cm2
m dT
but is not significantly deflected in each scatter; this will

mp keV
be a good approximation if m mN . However, when  2:3D217  M217 sec1 keV1 g1 ; (8)
m & mSi ’ 26 GeV, the dark-matter particle may be m T
backscattered upon encountering a terrestrial nucleus,
rather than simply being slowed without deflection. In where NN is the number of nuclei per gram of material.
this case, the particle will diffuse, undergoing N
The silicon semiconductor detector flown on a balloon
mN =2m scatters before coming to rest. If so, the pene- in the upper atmosphere by Rich et al. [23] observed an
event rate of
0:5 counts sec1 keV1 g1 nuclear re-
tration depth will be reduced by an additional factor of
coils in the lowest energy bin at 2 keV. For dark-matter

N 1=2 . We thus replace dE=dx ! dE=dx 1 
masses above the threshold
7 GeV, we thus require
mN =2m 1=2 in our expression for the penetration depth. D217  M217 mp =m < 0:2.
Then, for a given shielding thickness L, in meters The X-ray Quantum Calorimeter (XQC) detector
water equivalent (mwe), we invert the expression for the flown on a rocket by McCammon et al. [24], was designed
stopping distance to obtain the following bound on the to probe the soft x-ray background. However, it serendip-
dipole strength: itously provides a tight constraint on dipolar-dark-matter.
1 m md
v2  14  m To predict the expected number of events, we start by
2m 2 Eth
;md
D 2  M2 > ; (7) computing the number of DDM particles that could im-
P 2
2  m ;mi 1  m =2m
e2
2 L f i Zi i
1=2 pinge on the detector: N  n vAt  3  107 mp =m ,
m2 i
i where n is the galactic number density of dark-matter
where the index i sums over the composition of the particles, v is their velocity, the cross-sectional area of
shielding material, and fi is the fractional composition the XQC detector is A  0:33 cm2 , and the rocket flight
by weight. We use a realistic model of the composition of was about t
120 seconds. The chief property of the
the Earth (chemical composition by weight: O [46:6%], Si XQC detector is the 14 m thick Si substrate above the
[27:7%], Al [8:1%], Fe [5%], Ca [3:6%], Na [2:8%], K thermistor, providing a target of NN
6:5 
[2:6%], Mg [2:1%]) and atmosphere (10 mwe consisting 1019 nuclei =cm2 . Thus, the event/energy count
of a 4:1 ratio of nitrogen to oxygen), although the result- NN N d=dT integrated over the 25–100 eV energy bin
ing bounds do not change substantially if we ignore the gives a predicted
0:38D217  M217 mp =m events
atmosphere and approximate the Earth’s crust as entirely compared to the
10 observed events. Since the detector
composed of Si. We take the initial DDM velocity to be has a 25 eV threshold, energy transfer by dark-matter
300 km sec1 . particles as light as
1 GeV can be detected.
The shallowest underground experiment with a strong Altogether, the balloon and rocket experiments exclude
null result is the Stanford Underground Facility run of the a wide range of dipole strengths and masses, as illustrated
Cryogenic Dark-Matter Search (CDMS) [20], which was in Fig. 1.

083501-5
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
V. CONSTRAINTS FROM PRECISION
MEASUREMENTS
We now consider the limits placed on DDM due to
precision tests of the standard-model. Our use of pertur-
bation theory is valid provided the energy scale of the
interaction E satisfies D;M E  1. In addition, we re-
quire that the DDM mass satisfies D;M m & 1,
equivalent to the unitary bound [16], which ensures the
self-consistency of the local operator in Eq. (1). Indeed, if
( is the energy scale at which a dipole is generated then
one generally expects D;M (
1. In L we assume FIG. 6. Lowest-order correction to the muon anomalous
that all interacting particles with masses greater than ( magnetic-moment induced by dipole moments of the dark-
have been integrated out. Consequently, one must have at matter particle.
least m < ( for the dark-matter to be dynamical, which
also yields D;M m & 1. 6  1015 e cm. The order of magnitude of this result
can be obtained on dimensional grounds, if we consider
A. Muon Anomalous Magnetic-Moment that the DDM dipole moment contributes to a via at least
a two-loop graph (see Fig. 6), with two electromagnetic
The interaction in Eq. (1) contributes to the photon
couplings and two dipole couplings. Including a factor
propagator via the diagram shown in Fig. 5. The photon-
1=16 2 per loop, one obtains the estimate,
DDM interaction vertices are either both electric or mag-
netic dipolar; the mixed diagram where one vertex is e2
magnetic and the other is electric is proportional to 0a
D2  M2 E 2 ; (11)
16 2 2
+,- F F,-  0 for photons with equal momenta. The
sum of the diagrams produces the following contribution where E is the characteristic energy scale for the process.
to the photon vacuum-polarization tensor: In the case of the muon, E
m , which reproduces the
rigorous result to within an order of magnitude.
) q2  )q2 q2 g  q q
 /q2 q2 g  q q ; B. Electric Dipole Moments

 Furry’s theorem tells us that in evaluating radiative
D2  M2 1 m2 corrections to a process one should only keep the dia-
/ 1  ln ; (9)
8 2 3 2 grams with an even number of photons attached to a
where the photon momentum is taken to be small, q2  closed loop.3 Contributions with an odd number of pho-
tons attached sum to zero. On the other hand, one must
m2 (resulting in /q2  1), and  is the renormalization
have an odd number of time-reversal-odd (T-odd) EDM
scale, which should be smaller than (. We take  & vertices in the DDM loop to generate a T-odd operator.
1 TeV for our estimates. With this self-energy correction, These considerations show that the lowest-order nonvan-
the photon propagator for /q2  1 can be written as ishing diagram must have four photons attached to the


1 1 DDM loop; diagrams with two photons attached, similar
ig 2  2 : (10) to the one in Fig. 6, vanish (see above). Out of the four
q q  1=/
photons attached to the DDM loop, either one or three can
The second term above generates a correction to the muon have EDM coupling to DDM. Note that in this scenario
gyromagnetic ratio 0a   m2 /=3 . Interestingly, both electric and magnetic DDM moments are necessary
this contribution is not explicitly suppressed by the to generate a dipole moment for a SM fermion. With these
DDM mass. In view of recent measurements [25] and considerations in mind, the lowest-order three-loop dia-
comparison with the SM predictions, we require that gram that induces an EDM for a charged fermion is
0a does not exceed 109 , whereby D2  M2 1=2 < shown in Fig. 7. One obtains the following estimate for
the induced electric dipole moment:
e3 m3f m
D f
DMD2  M2 2 3
ln2 ; (12)
16 mf

3
Since the theorem is valid for interactions that preserve
FIG. 5. One-loop correction to the photon self-energy in- charge conjugation invariance we can apply it to electric and
duced by dipole moments M;D of the dark-matter particle. magnetic dipole moment interactions.

083501-6
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
1
m2W  p ; (14)
2GF 1  mW =m2Z 1  ,r
2

where ,r is a correction calculable in a given theory. The

interaction in Eq. (1) modifies the standard expression for
,r, whereby ,rNew  )m2Z  )0 . In the standard-
model ,rSM  0:0355  0:0019  0:0002. On the other
FIG. 7. Three-loop contributions to the EDM of a charged hand, one can use experimentally measured values for ,
fermion f. Either one or three of the DDM-photon interaction mW;Z , and GF in Eq. (14) to infer ,rexp  0:0326 
vertices must be of EDM type. The dots indicate all other 0:0023, which gives ,rNew < 0:003 at the 95% confidence
diagrams which can be obtained from the one shown by level. Therefore, we obtain the limit D2  M2 1=2 &
permutation of the interaction vertices. 3  1016 e cm. A full calculation of the vacuum-
polarization yields the constraint shown in Fig. 1. This
where a possible double-logarithmic enhancement is in- turns out to be the strongest constraint due to precision
cluded. For the electron, the present limit is De < 4  tests of the standard-model.
1027 e cm, which implies D;M < 3  1013 e cm for
m
100 GeV and D
M. For smaller m the limit D. Z-Pole Observables
becomes weaker. The DDM will contribute to various Z0 -pole observ-
There are constraints on the EDMs of other systems, ables through two-loop diagrams similar to the one
such as the neutron and the mercury atom. It is nontrivial shown in Fig. 8, at the level D2  M2 m2Z =64 3 .
to translate such constraints into limits on the underlying Requiring that these contributions do not exceed the
interaction. In the case of the neutron, one may attempt to
0:1% precision to which Z0 -pole observables are typi-
treat the neutron as a point particle for virtual-photon cally known [26] results in the constraint D2 
energies below 1 GeV. For higher loop momenta, photons
M2 1=2 < 1014 e cm. Note that in order for perturbation
begin seeing the quarks and the contribution to the EDM
theory to apply for energies E
mZ , one must have
becomes suppressed by the quark masses. In this case one
D;M mZ < 1, which means D;M & 7 
may use the above equation D
M with mf ! mn and no
1016 e cm. Interestingly, consistency with a perturbative
logarithmic enhancement, in order to estimate the neu- treatment at the Z0 pole imposes much stronger con-
tron EDM: straints on the DDM than the Z0 -pole observables
themselves.
e3 j3n j3 m3n
D n  DMD2  M2 < 6  1026 e cm; E. Direct Production
16 2 3
(13) If kinematically allowed, DDM can be directly pro-
duced in various scattering and decay experiments. In this
case one may use the ‘‘missing-energy’’ signature to
which results in the limit D;M & 4  1015 e cm (as- constrain the DDM couplings. Here, we consider
suming D
M. In the above equation, 3n  1:91 is missing-energy constraints from both low-energy (B
the magnetic-moment of the neutron. It appears because and K  meson decays) as well as collider (LEP, CDF)
the neutron is neutral, and couples to the photon in Fig. 7 experiments.
via a magnetic-moment interaction. The limit for the
EDM of the mercury atom is much stronger than the
neutron, DHg & 1028 e cm. Unfortunately, the mercury
atom is a complicated system for which the EDM is
influenced by many sources. Therefore, we leave the Hg
limit for future study.

C.W Boson Mass

The DDM can contribute to the running of the fine-
structure constant for momenta ranging up to the Z0
mass. Such running will affect the relationship between
the Fermi constant GF , the mass of the W  boson, and the FIG. 8. Lowest-order correction to Z0 -pole observables in-
fine-structure constant at zero momentum: duced by dipole moments of the dark-matter particle.

083501-7
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
 
1. B and K Decays 2. Collider Experiments
Searching for light (m & 1 GeV) dark-matter using A typical example of a process where DDM can be
missing-energy signatures in rare B meson decays was directly produced in a collider experiment is shown in
originally suggested in Ref. [27]. There, data from Fig. 10. Here, two fermions f scatter to produce a final-
BABAR [28] and CLEO [29] were used to set a limit state containing some set of visible particles X (photon,
BrB ! K   invisible & 104 [derived from multiple jets, etc.) along with particles that are not de-
BrB ! K   ]. This limit can be used to constrain tected. In the SM, the latter are neutrinos. Limits on the
the dipole moments of dark-matter. The diagram for rate for such processes have been set by, e.g., the L3 and
B ! K  decay is shown in Fig. 9(a). The rate for CDF collaborations [32]. At LEP, X consisted of a single
this decay can be related to the photon-exchange contri- photon whereas at CDF it consisted of one or more had-
bution to B ! K  l l j shown in Fig. 9(b). Since the ronic jets.
graphs have identical topologies, the difference in rates In order to translate constraints from collider experi-
will come from the difference in effective couplings and ments into limits on DDM couplings one needs an ana-
the final-state phase-space integrals. One can estimate, lytical expression for the rate for ff ! X . Naive
application of the effective Lagrangian in Eq. (1) would
BrB ! K  D2  M2 m2B result in upper limits from these missing-energy searches

of roughly 1017 e cm. However, this constraint does not
BrB ! K  l l j e2
actually exclude larger values of the dipole moments.
PSK  Indeed, as discussed above, perturbation theory will
 ; (15)
PSK  l l break down when the energy scale for the process E
satisfies D;M E * 1. This means that missing-energy
where PS   stand for the corresponding final-state searches from L3 (E  200 GeV) and CDF (E 
phase-space integrals, and mB  5:279 GeV is the B 1:8 TeV) cannot be used to probe effective dipole mo-
mass. Belle [30] and BABAR [31] find BrB ! ments D;M > 1016 e cm and D;M > 1017 e cm;
K  l l & 106 . Since the ratio of the phase-space in- respectively, unless the underlying physics that gives
tegrals is of order unity, and since in the absence of rise to the dipole moment is specified.
accidental cancellations BrB ! K  l l j &
   
BrB ! K l l , one obtains the constraint F. Other Laboratory Constraints

2 2 1=2 m   Important constraints can be obtained for millicharged
6 D  M B
2  10 & 104 ; (16) particles from the Lamb shift [10,33] and from a targeted
e
experiment at SLAC [34]. We have checked, however, that
which leads to D2  M2 1=2 & 3:8  1014 e cm. This due to the different energy dependence of the photon-
constraint is relevant for m < mB  mK =2  dipole vertex, as opposed to the photon-millicharge ver-
tex, the DDM-induced correction to the Lamb shift is
2:38 GeV.
small for dipole moments not eliminated by other preci-
Rare K  decays can be treated in a similar manner.
sion measurements, such as the running of the fine-
The relevant branching ratios are BrK  !  e e 
structure constant. Likewise, although the SLAC experi-
2:880:13
0:13  10
7
and BrK  !    0:1570:175 0:082  ment is in principle sensitive to neutral particles with a
109 [26]. The resulting constraint on the dipole moment
dipole, the production and energy deposition of dipole
is D2  M2 1=2 & 1:5  1015 e cm. This constraint ap- particles is sufficiently small, for dipole moments consis-
plies for m < mK  m  =2  0:18 GeV. We see that
constraints from B and K  decays are not competitive
with other constraints shown in Fig. 1.

FIG. 10. A typical process that would produce a missing-

FIG. 9. Photon-exchange contributions to (a) BrB ! energy signature in a collider experiment. Here, X stands for
K and (b) BrB ! K  l l . The blobs collectively the visible portion of the final-state. Neutrinos or DDM may
represent quark flavor-changing interactions. carry a large fraction of the energy but are not detected.

083501-8
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
tent with accelerator experiments, to evade detection in We next estimate ab , the drag-force per unit mass due
the SLAC experiment. to DDM-proton scattering. We first note that the peculiar
velocity of the baryon-photon fluid (obtained from the
VI. CONSTRAINTS FROM LARGE-SCALE continuity equation) in the early Universe will be V

STRUCTURE AND THE CMB H=k c0, where H is the Hubble parameter, k is the
physical wavenumber of the mode in question, and 0

We now consider the effects of the interaction L on 105 is the amplitude of the fractional density perturba-
the evolution of cosmological perturbations and their tion. Since H=k & 1 for modes inside the horizon, we
resulting imprints on the matter power spectrum and must have V & 105 c. On the other hand, the proton
the CMB. A dipole moment can induce a coupling of thermal velocity dispersion is vp
T=mp 1=2 c *
the dark-matter to the baryon-photon fluid by scattering 104:5 c before recombination. Thus, for the early times
from photons through the diagrams shown in Fig. 11, or of interest to us here, the relative velocity between the
by scattering from protons, helium nuclei, and/or elec- DDM and the baryon-photon fluid is small compared with
trons through the diagram shown in Fig. 4. What we will the thermal proton velocities. Thus, the appropriate rela-
show below is that the dark-matter is coupled to the tive velocity to use in Eq. (5) in estimating the proton-
baryon-photon fluid at early times, and decouples at later DDM cross-section is vp , resulting in a DDM-proton
times. When the dark-matter is coupled to the photon-
cross-section  b
e2 D2  M2 =v2p . The momentum
baryon fluid, the pressure of the plasma resists the growth
transfer per scatter is
vp , where  is the proton-DDM
of gravitational potential wells. Thus, the short-
wavelength modes of the density field that enter the reduced mass, and the difference of the fluxes of protons
horizon at early times will have their growth suppressed moving with as opposed to against the DDM fluid is np V,
relative to the standard calculation resulting in a suppres- where np is the proton density. The drag-force per unit
sion of small-scale power. The evolution of the longer- mass on the DDM fluid due to scattering with protons is
wavelength modes that enter the horizon after the dark- thus ab
e2 D2  M2 =m V=vp np . We also con-
matter has decoupled remain unaffected. Before present- clude from the appearance of  in this result that drag due
ing the results of our detailed analysis, we begin with to scattering from electrons is negligible compared with
some simple estimates. baryon drag.
We first show that DDM-photon scattering is negligible Since np / T 3 and vp / T 1=2 , we find ab / T 2:5 as
compared with DDM-baryon scattering in providing the opposed to a / T 6 . Thus, at early times, photon drag
drag-force between the DDM fluid and the baryon-photon dominates while baryon drag dominates at later times.
fluid. To do so, we first estimate the drag-force per unit The transition occurs at a temperature T
GeV for
mass (i.e., the deceleration) on a DDM-particle that values of m and D;M of interest to us, and such
moves with a velocity V with respect to the rest frame high temperatures correspond to (comoving) horizon
of a blackbody at temperature T. The diagrams in Fig. 11 scales considerably smaller than the distance scales ( *
will lead to a photon-DDM scattering cross-section Mpc) probed by large-scale structure. We can thus ne-
 
D4  M4 E2 . Considering that the momentum glect photon drag. From ab / T 2:5 we infer a deceleration
transfer to the DDM-particle in each scatter is
E and time for the DDM fluid tdec  V=ab / T 2:5 . Since this
that the difference of the fluxes of photons moving in the decreases more rapidly than the Hubble time tU
mPl T 2
same versus opposite direction to the DDM-particle is (where mPl
1019 GeV is the Planck mass), we conclude

T 3 V=c , we conclude that the deceleration due to pho- that DDM particles are tightly coupled to the plasma at
ton scattering is a
D4  M4 T 6 V=c =m . early times and then are decoupled at later times. With
these rough estimates, the transition temperature is T

10 keVD215  M215 2 1  m =mp 2 suggesting that
power on scales smaller than -
102 D215  M215 2 
1  m =mp 2 Mpc will be suppressed. The T 0:5 depen-
dence of the ratio of the deceleration and expansion times
suggests furthermore that the small-scale suppression
will change gradually, rather than exponentially, with
wavenumber k. Knowing that the linear-theory power
spectrum is measured and roughly consistent with scale
invariance down to distances
Mpc leads us to conclude
that dipole moments D2  M2 1=2 * 5  1015 1 
FIG. 11. The photon-DDM scattering diagram, the analogue m =mp 1=2 e cm will be ruled out. Strictly speaking,
of the Compton-scattering diagrams for electric or magnetic when m < mp , the detailed calculation must take into
moments. account the velocity dispersion of the DDM particles; our

083501-9
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
detailed calculation below includes these effects. As seen baryons by DDM through the interaction L influences
below, the detailed analysis leads to a slightly stronger the growth of cosmological perturbations by introducing
constraint. additional collision terms to the Boltzmann equations,
which ultimately result in a drag-force between the DDM
and the colliding species in the equations describing the
A. Exact Equations cosmological fluid (see, e.g., Refs. [37,38], which consider
The standard calculation of perturbations in an ex- similar effects). Below we present the exact perturbation
panding universe requires the solution of the combined equations including the effects of dark-matter with elec-
Einstein and Boltzmann equations for the distribution tric or magnetic dipole moments. Since solutions to these
functions of the dark-matter, baryons, photons, and neu- equations are numerically intensive when photons and
trinos including all relevant standard-model interactions baryons are tightly coupled through Compton-scattering,
(see, e.g., Refs. [35,36] and references therein). Since the we also discuss the equations appropriate for solving for
perturbations are initially very small, linear perturbation the DDM, photon, and baryon perturbations during the
theory is an excellent approximation; this allows us to epoch of tight coupling.
solve the perturbation equations in Fourier space at each In the synchronous gauge the equations describing the
wavenumber k independently of all other wavenumbers evolution of baryons, photons, and dark-matter with an
(modes are uncoupled). The scattering of photons and electric or magnetic dipole moment are

4 2 1 1
0_    !  h;_ 0_ b  !b  h;_ 0_  !  h;_
3 3 2 2


1
!_   k2 0  4  ane T !b  !  an hi  !  ! ;
4
a_ 4,
!_ b   !b  c2sb k2 0b  an  !  !b  an hvi b !  !b ;
a 3,b e T 
a_ , 4,
!_   !  c2s k2 0  b an hvi b !b  !  an hi  !  ! : (17)
a , 3,

While the evolution equations for the density contrast photon-DDM drag term is small, and we consider it no
0j  0,j =,j for each species j 2 f; b; g are as in the further in Eq. (17).
standard case [35], as discussed above, the evolution The quantity
equations for the fluid-velocity perturbations have addi-
tional drag-force terms due to the photon-DDM interac- 41  AY m
hvi b  q e2 D2  M2
tion. Note that in these equations and what follows the 3 2 2
hvp i  hv i 2 m  mp
variable !j  ikVj is the divergence of the fluid-velocity
in Fourier space, 4j is the shear, csj is the intrinsic sound (20)
speed, and nj and ,j are the background number and is the appropriate thermally-averaged cross-section times
energy densities of a particular species j, respectively. relative velocity for the baryon-DDM coupling, and
The variable h is the trace of the scalar metric perturba- v
u
tion in Fourier space (not to be confused with the Hubble m  mp u 2
t hvp i  hv i  1
2
parameter), a is the cosmological scale factor, and an A8 (21)
m  4mp hvp i2  4hv i2
overdot represents a derivative with respect to the con-
formal time @. Furthermore, T is the Thomson cross- is the relative efficiency for coupling to helium nuclei
section, while compared to protons. The Appendix provides a derivation
80 of this collision coefficient. In these expressions, Y 
hi   D4  M4 T2 ; (18) ,He =,b ’ 0:24 is the cosmological helium mass fraction
21 q
is the appropriately thermally-averaged DDM-photon (approximating mHe ’ 4mp ), hvp i  8Tb = mp is the
cross-section, which can be obtained from the differential average thermal speed of the protons, hv i 
cross-section [39,40], q
8T = m is the average thermal speed of the DDM,
d  D4  M4 E2
 3  cos2 ! ; (19) and T , Tb , and T are the photon, baryon, and DDM
d 8 2 temperatures, respectively. The dark-matter temperature
for photon-DDM scattering. As argued above, the evolves according to

083501-10
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501

a_ 2a,b hviT b where

T_  2 T  Tb  T
a m  mp
1 @c
8a, hi / (25)
 1  Rb @
 T  T ; (22)
3m
is the parameter that controls how strongly the new
where hviT b is the same as the expression given in interaction affects the evolution of the slip. In terms of
Eq. (20) with the replacement of A by AT which is given these definitions the baryon-velocity evolution equation
by the expression in Eq. (21) with the factor m  is
mp =m  4mp replaced by m  mp =m 
4mp 2 . The final term, describing the dark-matter heat- 1
!_ b 
ing by photons, is important at very early times. For the 1  Rb  /Rb
dipole strength and mass range considered, the influence


a_ 1
is manifest only on very small length scales, below the   !b  k2 c2sb 0b  Rb 0  4
range of interest. a 4

At early times, the DDM temperature T ’ Tb , but at _ _ 1 a_
Rb Sb  / !  !  !b : (26)
later times, when the DDM decouples, T drops relative to 2a
Tb . The DDM-proton cross-section is / v2 , which leads
to hvi b / hvp i1 . As a result, we cannot directly apply The photon-evolution equation is then given by the exact
the results of Ref. [37], wherein a velocity-independent expression
dark-matter-baryon interaction was assumed. However,
we have verified that we recover their results if we take a 1 a_ 1
!_    !_ b  !b  c2sb k2 0b  !  !b
velocity-independent cross-section as the source of dark- Rb a @
matter–baryon drag.

1
k2 0  4 : (27)
4
B. Tightly Coupled Equations
At early times when @1 c  ane T _
a=a the rapid We use these equations to follow the initial evolution of
scattering of baryons and photons forces these species the baryon and photon fluid variables and switch to the
to have nearly equal fluid velocities, and consequently the exact equations of Eq. (17) at later times. For the evolu-
solution of the equations shown in Eq. (17) is numerically tion of the DDM fluid variables we always use the exact
intensive. Following standard procedures [35,41] we de- form of Eq. (17).
rive a set of equations to leading order in the (conformal)
Compton-scattering time @c that are appropriate for
evolving the fluid variables through this epoch of tight
coupling. We first write down an equation for the time
derivative of !b  ! which is usually termed the baryon-
photon ‘‘slip’’ to leading order in @c ,
2Rb a_ @c
!_ b  !_   !b  ! 
1  Rb a 1  Rb


a5 1 1 a_
  !b  k2 c2sb 0_ b  0_   0
a 4 2a
1 _ 1 _
a 1
 !  !_ b  !  !b ; (23)
@ 2a@

where we have introduced the (conformal) DDM-baryon

scattering time @1  an hvi b , and Rb  4, =3,b .
It is useful to separate this equation as a sum of the terms
FIG. 10 (color online). Matter power spectra including
not containing @ (this is just the time derivative of the baryon-DDM drag. The solid curve is for is for D2 
standard slip, which we denote Sb ), and the new terms M2 1=2  1:4  1015 e cm The short-dash is for D2 
introduced by the DDM coupling, M2 1=2  1:0  1016 e cm The long-dash curve is for D2 
M2 1=2  5  1015 e cm. These are all for a mass of 1 GeV.
1 a_ The curves are all for the standard concordance cosmological
!_ b  !_   S_ b  / !_  !_ b  !  !b ; (24)
2a parameters, and the data points are from SDSS [43].

083501-11
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
As the effect of DDM on the CMB may be partially
degenerate with other cosmological parameters, we have
explored a parameter space that allows us to constrain m
and D; M after marginalizing over other cosmological
parameters. We consider flat (CDM models and our
chosen parameter space is the dark-matter density
m h2 , the baryon density b h2 , the Hubble parameter
h in units of 100 km sec1 Mpc1 , the optical depth @CMB
to the last-scattering surface, and the primordial spectral
index n. We have employed the Markov chain Monte
Carlo technique (see, e.g., Ref. [42]) to efficiently explore
this parameter space, taking the most recent results from
SDSS [43], WMAP [44], CBI [45], VSA [46], and SNe Ia
[47] as our data. Note that although DDM has no effect on
1500 observations of Type Ia supernovae, we include these data
because the other parameters we allow to vary are con-
strained by these observations. We conclude using a
FIG. 11 (color online). CMB power spectra including DDM- relative-likelihood test that cosmological measurements
baryon drag. The labeling of the curves is the same as in lead to the bound shown in Fig. 1. The numerical calcu-
Fig. 10, and the data points are those from WMAP [44]. lations confirm the qualitative behavior discussed above.
Dipole moments as large as D2  M2 1=2
1017 e cm,
C. Effects on the Matter and CMB Power Spectra near the upper end of our allowed parameter space, are
In Fig. 10 we show the linear matter power spectrum thus cosmologically viable.
and in Fig. 11 we show the angular power spectrum of the
CMB for several values of the dipole moment and for VII. GAMMA RAYS
DDM mass m  1 GeV. Physically, the effects of DDM
on the matter power spectrum and CMB can be simply DDM particles in the Galactic halo can annihilate to
understood. Prior to matter-radiation equality the pho- two photons through the diagrams shown in Fig. 3. Since
tons have a much larger density than the baryons or the halo particles move with velocities v ’ 300 km sec1 
DDM and so to a first approximation completely drive the c, the photons produced will be very nearly monoener-
behavior of the baryon perturbations through Compton- getic with energies equal to the DDM-particle rest mass.
scattering. In turn, the baryon perturbations drive the The intensity at Earth of such gamma rays is obtained by
behavior of the DDM perturbations, very efficiently be- integrating the emissivity, n2 h !2 vi, where n is the
fore DDM decoupling so that the DDM density contrast DDM number density, along the given line of sight. The
0 on scales that enter the horizon during this epoch track intensity is largest toward the Galactic center, where the
the oscillations of the baryon-photon fluid before grow- dark-matter density is largest. In this direction, the
ing, and less efficiently after DDM decoupling so that the gamma-ray intensity is then [4],
baryons simply cause a drag on the growth of 0 . In either dF  !2 v
case the matter power spectrum is suppressed relative to  1:0  1010 30 3 1 m2
GeV Icm
2
sec1 sr1 ;
d 10 cm sec
the standard case. The behavior of the CMB angular (28)
power spectrum can be similarly understood. Roughly
speaking, the coupling of the DDM and baryons increases where I is a scaled integral of n2 along a line of sight
the effective baryon loading of the plasma at early times toward the Galactic center. The numerical coefficient is
so that the CMB power spectra look similar to those from one-half that from Ref. [4] since we have here particle-
high-baryon models. This is of course an imperfect cor- antiparticle annihilation rather than Majorana annihila-
respondence as modes of larger wavelength enter the tion. Roughly speaking, I ’ 3  30 for cored-isothermal-
horizon when the coupling is weaker, and so at later and sphere models of the Galactic halo, while I can extend up
later times the evolution of the photon perturbations to
300 for Navarro-Frenk-White profiles [48]; i.e., un-
becomes more and more like the standard-CDM case. certainty in the dark-matter distribution in the inner
But due to geometrical projection effects modes of wave- Galaxy leads to an uncertainty of 2 orders of magnitude
number k contribute to all l & kdA where dA is the in the predicted flux. We thus expect
angular-diameter distance to the last-scattering surface,
dF
and so the effects of DDM on small length scales can be  3  1013  3  1011 D417
noticed even on relatively large angular scales in the d
CMB.  M417 cm2 sec1 sr1 : (29)

083501-12
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
To constrain dipole moments from nonobservation of a We have restricted our attention to particles with
gamma-ray line, we choose to use the most conservative masses m * MeV, with the notion that lower-mass
estimate, I ’ 3 for the dimensionless line integral. particles will violate BBN limit, as discussed toward
Moreover, we are not aware of any EGRET analysis that the end of Section 2. We also consider masses m *
places limits, in particular, to a line-flux. We thus obtain MeV, as particles of lower-mass will almost certainly
very conservative limits by using the binned continuum undergo relativistic freeze-out and thus lead to unaccept-
fluxes for the total gamma-ray flux listed in Table 1 of able dark-matter candidates. However, as also noted
Ref. [48] and noting that a line-flux in that bin cannot above that if an m & MeV particle has a dipole moment
exceed the measured continuum flux. The EGRET limits D;M & 1022 e cm and no other interactions with or-
apply for masses 0:1 & m =GeV & 10, and range from dinary matter, then it might still be consistent with BBN.
D17 & 180 for m & GeV to D17 & 100 for m ’ Of course, such a particle will, assuming standard freeze-
10 GeV, as shown in Fig. 1. out, have a mass density many orders of magnitude larger
Again, a few caveats are in order. First of all, our limit than the dark-matter density. But suppose we were to
is quantitatively conservative —we chose the halo model surmise that the dark-matter density was fixed by some
that produces the lowest flux, and a detailed EGRET other mechanism; e.g., suppose the dipole was sufficiently
analysis would probably yield a line-flux limit lower weak that it never came into equilibrium. In this case, an
than what we have assumed. On the other hand, the strong additional constraint to the dark-matter dipoles can be
dependence / D4 of the predicted flux on the dipole obtained from energy-loss arguments applied to stars in
moment guarantees that the upper limit to the acceptable globular clusters. Such arguments eliminate dipole mo-
dipole moment will not depend quite so strongly on these ments D;M & 6  1023 e cm for masses m & 5 keV
details. Second, if D17 * 5 in the mass range 100 MeV to [51]. We have also considered constraints from astrophys-
1 GeV, then the correct cosmological abundance most ical phenomena such as the stability of the Galactic disk,
likely requires a particle-antiparticle asymmetry. If so, lifetime of compact objects, and annihilations in the solar
then the annihilation rate in the halo could be reduced far neighborhood [6], and find that these constraints on the
below the values we have obtained above. We conclude by mass and interaction strengths are not competitive with
noting that with the increased sensitivity of the Gamma- those presented here.
Ray Large Area Space Telescope (GLAST), a detailed It would be of interest to attempt to embed this sce-
search for a line-flux, and the possibility that the actual nario in a consistent particle-physics model. We might
halo model provides a more generous annihilation rate, find links between baryonic and nonbaryonic matter
an observable GLAST signature may exist for masses abundances, the dark-matter electric dipole moment and
0:1  1 GeV and dipole moments as low as D17
10. the CP violation needed for baryogenesis, and the mag-
netic moments of dark-matter and baryons. (E.g.,
VIII. DISCUSSION Ref. [52] considered a fermionic technibaryon with elec-
tromagnetic dipole interactions as a dark-matter candi-
In this paper we have considered the cosmology and date.) However, such model building is beyond the scope
phenomenology of dark-matter particles with a nonzero of the present study. Our approach throughout has been
magnetic or electric dipole moment. We have found that entirely phenomenological, as we have been motivated by
information from precision tests of the standard-model, the desire to answer the question,‘‘How dark is dark?’’
direct dark-matter searches, gamma-ray experiments,
and the CMB and large-scale structure restrict the dipole
moment to be D;M & 3  1016 e cm for masses m & ACKNOWLEDGMENTS
few GeV and D;M & 1024 e cm for larger masses. We thank S. Golwala, J. Albert, M. Zaldarriaga, and R.
(This improves on an earlier limit on WIMP electromag- Shrock for useful discussions. K. S. acknowledges the
netic dipole moments based on direct-detection [49].) support of the Canadian NSERC. This work was sup-
Some of the allowed regions of parameter space may ported in part by NASA NAG5-9821 and DoE DE-FG03-
soon be probed with GLAST and with future more sensi- 92-ER40701 (at Caltech) and NSF PHY-0099543 (at
tive direct-detection experiments. The electromagnetic Dartmouth). M. D. and R. C. thank Caltech for hospitality
interactions of these particles with nuclei are coherent. during the course of this investigation.
Moreover, these particles cannot annihilate directly to
neutrinos. Therefore, searches for energetic neutrinos
APPENDIX: DERIVATION OF THE BARYON-
from decays of the products of annihilation in the
DARK-MATTER COLLISION TERM
Sun or Earth are thus likely to provide less sensitive
probes than direct searches [50]. Moreover, if there is a To determine how the cosmological perturbation equa-
particle-antiparticle asymmetry, then the energetic- tions for baryons and dark-matter are altered when we
neutrino flux could be reduced without altering the bestow the dark-matter with a magnetic or electric dipole
direct-detection rate. moment, we must formally evaluate the collision operator

083501-13
KRIS SIGURDSON et al. PHYSICAL REVIEW D 70 083501
of the general-relativistic Boltzmann equation in a given 1 Z 3 Z
gauge [14,35,36] for the dipole interaction of Eq. (1). We ax  d v f v~ d3 vp fp v~ p jv~ p  v~ j
n
have completed this calculation in detail, and find that the Z d
dipole interaction produces a drag-force proportional to  d v xf  v xi : (A3)
d
the relative velocity V  v  vb of the dark-matter and
baryon fluids. As the relative velocity is gauge invariant Here   !; E is the scattering angle in the center-of-
in linear perturbation theory and all scatterings are local mass frame, and v xf  v xi is the difference between the
processes, we may thus take a simpler, more physically final and initial x component of the dark-matter–particle
transparent approach and just evaluate this drag-force velocity; the difference is the same in the center-of-mass
using nonrelativistic statistical mechanics. It is this ap- and laboratory frames. The differential cross-section
proach we now present. d=d is that given in Eq. (5).
We wish to calculate the drag-force per unit mass, or Consider an individual scattering event. Let be the
deceleration, due to collisions with protons to the dark- angle that v~ p  v~ makes with the x^ direction; this is
matter fluid as it passes through the baryon-photon fluid. then the angle that v~ makes with the x^ axis in the center-
Comoving scales - * Mpc enter the horizon when the of-mass frame, and the magnitude of the initial and final
cosmological temperature is T & 10 eV, when the DDM dark-matter velocities in the center-of-mass frame is
particles (which are restricted to m * MeV) are non- vcm  mp v=mp  m , where v  jv~ p  v~ j is the rela-
relativistic. We may thus consider thermal velocity dis- tive velocity. The initial x^ component of the dark-matter
tributions for nonrelativistic baryons and dark-matter. velocity in the center-of-mass frame is then v xi 
Since the drag-force can only depend on the dark-matter– vcm cos . The scattering angles ! and E are then the
baryon relative velocity, we take the baryon fluid to be at polar and azimuthal angles that the scattered dark-matter
rest and the dark-matter fluid to have a velocity of mag- velocity makes with the initial velocity in the center-of-
nitude V in the x^ direction. Then, the proton phase-space mass frame. By rotating this coordinate system by an
distribution is angle about the z^ axis to align it with the laboratory
np 2 2
x^ axis, we find v xf  vcm cos cos!  sin sin! sinE .
fp v~ p  evp =2vp ; (A1)
2 3=2 v3p Thus,
where vp  kTp =mp 1=2 is the proton velocity dispersion Z d mp Z2 e2 D2  M2 cos
d v xf v xi  :
and np the proton number density, and d 2 mp  m v
n v~  V x^ 2
(A4)
f v~  exp  ; (A2)
2 3=2 v3 2v2 Completing the integral in Eq. (A3) in the limit V 
vp ; v , we find
is the dark-matter phase-space distribution, with v 
kT=m 1=2 . Recall also that we expect V  vp , as dis- 2Ze 2 D2  M2 mp np V
cussed above. ax  1=2 q : (A5)
3 2 mp  m v2p  v2
The drag-force per unit mass is obtained by integrating
the momentum transfer per collision over all collisions q
between protons and dark-matter particles. From the Taking into account the definition hvp i  8Tp = mp ,
symmetry of the problem, the deceleration of the dark- this drag-force leads to the drag-force term in Eqs. (17),
matter fluid will be in the x^ direction, and it will have a (20), and (21) when including the simple corrections for a
magnitude, mass fraction Y of helium.

[1] F. Zwicky, Helv. Phys. Acta 6, 110 (1933). [6] G. D. Starkman, A. Gould, R. Esmailzadeh,
[2] S. Tremaine and J. E. Gunn, Phys. Rev. Lett. 42, 407 and S. Dimopoulos, Phys. Rev. D 41, 3594
(1979). (1990).
[3] See, e.g., http://pdg.lbl.gov. [7] E. D. Carlson, M. E. Machacek, and L. J. Hall,
[4] G. Jungman, M. Kamionkowski, and K. Griest, Phys. Astrophys. J. 398, 43 (1992).
Rep. 267, 195 (1996). [8] D. N. Spergel and P. J. Steinhardt, Phys. Rev. Lett. 84,
[5] M. S. Turner, Phys. Rep. 197, 67 (1990); G. Raffelt, Phys. 3760 (2000).
Rep. 198, 1 (1990); K. van Bibber and L. Rosenberg, [9] A. Gould, B. T. Draine, R.W. Romani, and S. Nussinov,
Phys. Rep. 325, 1 (2000). Phys. Lett. B 238, 337 (1990).

083501-14
DARK-MATTER ELECTRIC AND MAGNETIC DIPOLE MOMENTS PHYSICAL REVIEW D 70 083501
[10] See, e.g., S. Davidson, S. Hannestad, and G. Raffelt, J. [31] BABAR Collaboration, B. Aubert et al., hep-ex/0107026.
High Energy Phys. 05, (2000) 003. [32] L3 Collaboration, P. Achard et al., Phys. Lett. B 587, 16
[11] S. L. Dubovsky, D. S. Gorbunov, and G. I. Rubtsov, JETP (2004); CDF Collaboration, D. Acosta, Phys. Rev. Lett.
Lett. 79, 1 (2004); S. L. Dubovsky, D. S. Gorbunov, and 92, 121802 (2004); CDF Collaboration, T. Affolder et al.,
G. I. Rubtsov, Pis’ma Zh. Eksp. Teor. Fiz. 79, 3 (2004). Phys. Rev. Lett. 88, 041801 (2002).
[12] E. Fermi and E. Teller, Phys. Rev. 72, 399 (1947). [33] S. Davidson, B. Campbell, and D. Bailey, Phys. Rev. D
[13] S. Burles et al., Phys. Rev. Lett. 82, 4176 (1999). 43, 2314 (1991).
[14] E.W. Kolb and M. S. Turner, The Early Universe [34] A. A. Prinz et al., Phys. Rev. Lett. 81, 1175 (1998).
(Addison-Wesley, Reading, MA, 1990). [35] C. P. Ma and E. Bertschinger, Astrophys. J. 455, 7
[15] P. de Bernardis. et al., Nature (London) 404, 955 (2000); (1995).
S. Hanany et al., Astrophys. J. Lett. 545, 5 (2000); N.W. [36] Scott Dodelson, Modern Cosmology (Academic, New
Halverson et al., Astrophys. J. 568, 38 (2002); B. S. York, 2003).
Mason et al., Astrophys. J. 591, 540 (2003); A. Benoit [37] X. L. Chen, S. Hannestad, and R. J. Scherrer, Phys. Rev. D
et al., Astron. Astrophys. L 399, 25 (2003); D. N. Spergel 65, 123515 (2002).
et al., Astrophys. J. Suppl. Ser. 148, 175 (2003). [38] K. Sigurdson and M. Kamionkowski, Phys. Rev. Lett. 92,
[16] K. Griest and M. Kamionkowski, Phys. Rev. Lett. 64, 615 171302 (2004).
(1990). [39] F. E. Low, Phys. Rev. 96, 1428 (1954).
[17] M. Kamionkowski and M. S. Turner, Phys. Rev. D 42, [40] M. Gell-Mann and M. L. Goldberger, Phys. Rev. 96, 1433
3310 (1990). (1954).
[18] L. L. Foldy, Phys. Rev. 83, 688 (1951). [41] P. J. E. Peebles and J. T. Yu, Astrophys. J. 162, 815 (1970).
[19] CDMS Collaboration, D. S. Akerib et al., astro-ph/ [42] M. Doran and C. M. Mueller, astro-ph/0311311.
0405033. [43] SDSS Collaboration, M. Tegmark et al., Astrophys. J.
[20] CDMS Collaboration, D. S. Akerib et al., Phys. Rev. D 606, 702 (2004).
68, 082002 (2003). [44] G. Hinshaw et al., Astrophys. J. Suppl. Ser. 148, 135
[21] G. Angloher et al., Astropart. Phys. 18, 43 (2002). (2003).
[22] P. C. McGuire and P. J. Steinhardt, astro-ph/0105567. [45] A. C. S. Readhead et al., Astrophys. J. 609, 498 (2004).
[23] J. Rich, R. Rocchia, and M. Spiro, Phys. Lett. B 194, 173 [46] C. Dickinson et al., astro-ph/0402498.
(1987). [47] A. G. Riess et al., Astrophys. J. 607, 665 (2004); J. L.
[24] D. McCammon et al., Astrophys. J. 576, 188 (2002). Tonry et al., Astrophys. J. 594, 1 (2003); R. A. Knop
[25] Muon g-2 Collaboration, G.W. Bennett et al., Phys. Rev. et al., Astrophys. J. 598, 102 (2003); B. J. Barris et al.,
Lett. 89, 101804 (2002); Phys. Rev. Lett. 89, 129903(E) Astrophys. J. 602, 571 (2004).
(2002). [48] L. Bergstrom, P. Ullio, and J. Buckley, Astropart. Phys. 9,
[26] Particle Data Group, K. Hagiwara et al., Phys. Rev. D 66, 137 (1998).
010001 (2002). [49] M. Pospelov and T. ter Veldhuis, Phys. Lett. B 480, 181
[27] C. Bird, P. Jackson, R. Kowalewski, and M. Pospelov, (2000).
hep-ph/0401195. [50] M. Kamionkowski et al., Phys. Rev. Lett. 74, 5174
[28] BABAR Collaboration,B. Aubert et al., hep-ex/0304020. (1995).
[29] CLEO Collaboration, T. E. Browder et al., Phys. Rev. [51] G. Raffelt, Stars as Laboratories for Fundamental
Lett. 86, 2950 (2001). Physics (University of Chicago, Chicago, 1996).
[30] BELLE CollaborationK. Abe et al., Phys. Rev. Lett. 88, [52] J. Bagnasco, M. Dine, and S. Thomas, Phys. Lett. B 320,
021801 (2002). 99 (1994).

083501-15