arXiv:0903.3116v4 [astro-ph.HE] 13 Jul 2009
Phenomenology of Dark Matter annihilation into a long-lived
Ira Z. Rothstein,1, ∗Thomas Schwetz,2, †and Jure Zupan3, 4, ‡
1Carnegie Mellon University, Dept. of Physics, Pittsburgh PA 15213, USA
2Max-Planck-Institute for Nuclear Physics,
PO Box 103980, 69029 Heidelberg, Germany
3Theory Division, Physics Department,
CERN, CH-1211 Geneva 23, Switzerland
4Faculty of mathematics and physics, University of Ljubljana,
Jadranska 19, 1000 Ljubljana, Slovenia
We propose a scenario where Dark Matter (DM) annihilates into an intermediate state which
travels a distance λ ≡ v/Γ on the order of galactic scales and then decays to Standard Model
(SM) particles. The long lifetime disperses the production zone of the SM particles away from
the galactic center and hence, relaxes constraints from gamma ray observations on canonical an-
nihilation scenarios. We utilize this set up to explain the electron and positron excesses observed
recently by PAMELA, ATIC and FERMI. While an explanation in terms of usual DM annihilations
seems to conflict with gamma ray observations, we show that within the proposed scenario, the
PAMELA/ATIC/FERMI results are consistent with the gamma ray data. The distinction from
decay scenarios is discsussed and we comment on the prospects for DM production at LHC. The
typical decay length λ>
∼10 kpc of the intermediate state can have its origin from a dimension
six operator suppressed by a scale Λ ∼ 1013GeV, which is roughly the seesaw scale for neutrino
∗Electronic address: izr˙AT˙andrew.cmu.edu
†Electronic address: schwetz˙AT˙mpi-hd.mpg.de
‡Electronic address: jure.zupan˙AT˙cern.ch
An exciting and very plausible possibility is that dark matter (DM) interacts non-
gravitationally. A very active ongoing program exists to search for such interactions both
directly and indirectly. Indirect evidence could arise from cosmic ray signals originating from
areas of high dark matter density — the galactic halos and cores. Decaying or annihilating
DM can act as an additional source of cosmic ray fluxes on Earth, and could be seen as an
excess above the expected flux. The channel in which the excess could show up depends on
how DM couples to standard model (SM) particles.
DM decays and annihilations into charged particles inevitably lead to enhanced gamma
ray fluxes due to internal and final state bremsstrahlung. In both cases the gamma ray
spectrum will be correlated with the energy spectrum of charged particle, and will be the
same (up to overall energy scale) for the same final states. However, the angular distribution
of gamma rays in the sky will differ between the two cases. Since the annihilation rate scales
with ρ2(ρ being the DM density), the gamma ray signal will be more highly peaked toward
the galactic center due to the rise in the density profile for small r. The decay rate, on the
other hand, scales linearly with ρ and therefore the gamma ray signal is less peaked.
In this paper we consider an alternative scenario, where the DM annihilates into long
lived particles (LLP). This setup will lead to a gamma ray signature which can interpolate
between the signatures of DM decays and annihilation. The prolonged decay lifetime of LLP
effectively smears out the distribution of gamma rays. To be concrete we will consider this
scenario in the context of DM explanation of the positron excess found by PAMELA ,
ATIC  and FERMI . The PAMELA experiment which sees a rise in the positron-to-
electron ratio in cosmic rays for energies between 10 and 100 GeV corroborates and extends
the range of earlier data , while ATIC sees a rise and then a striking fall near 800 GeV in
the power-law for the combined electron and positron flux. The rise at lower energies, below
the bump, corroborates previous data  while the bump at 800 GeV had never been seen
before. Recent data from the FERMI satellite do not confirm this bump, while an excess
of the electron and positron flux over the expectations from common cosmic ray models is
still there. FERMI data indicate a smooth spectrum falling approximately as E−3till about
800 GeV. Above this energy data from the HESS telescope [6, 7] indicate a much steeper
slope of the electron/positron flux. Although it seems quite reasonable to believe that this
data has an explanation in terms of nearby young pulsars [8, 9, 10, 11, 12, 13, 14, 15] (or
some of the data could be spurious), it will serve as a useful tool in highlighting the features
of the LLP scenario. In particular, the data exemplify a case where there is a need for an
increased annihilation cross section ?σv? ∼ 10−22cm3/s (relative to the typical cross section
?σv? ∼ 3 × 10−26cm3/s needed for ΩDM∼ 0.2), while there is no evidence of excess in the
gamma ray flux in the direction of the galactic center. Standard annihilation scenarios with
such large cross sections seem to conflict [16, 17, 18, 19] with the results from the HESS
telescope [20, 21] for commonly assumed DM profiles. We will show that, by allowing the
annihilation products to be long lived, the constraints from gamma rays can be avoided.
To be specific, let us consider the situation where a DM particle χ annihilates into an
intermediate state φ which subsequently decays into standard model particles
χχ → φφ → 2SM2SM. (1)
As an example we will assume in the following that φ decays into muons,
φ → µ+µ−, (2)
though our arguments do not rely on the specific decay mode of φ. We concentrate on this
mode since it is the relevant one for the PAMELA/ATIC data (while not changing our main
conclusions other annihilation modes may be preferred when including new FERMI data
). In contrast to [23, 24, 25, 26, 27, 28, 29] we now assume that φ is a long-lived particle
(LLP), such that it will propagate over galactic distances.1For the moment we will not
concern ourselves with whether or not the DM particle χ is a thermal relic. We will return
to this issue at the end of the paper in section IV. In Sec. II we present a general discussion
of the LLP scenario and discuss the effects on the production of Standard Model particles
from DM annihilations focusing on gamma rays. In Sec. III we apply this idea to the recent
electron and positron flux data from PAMELA, ATIC, FERMI and HESS and show that
they can be made consistent with gamma ray observations from HESS even for rather cuspy
DM profiles. In Sec. V we summarize our results, discuss possible distinguishing features of
the scenario, mention prospects to see DM at LHC, and speculate on a possible connection
between the high scale physics responsible for the decay of φ and the suppression of neutrino
masses via the seesaw mechanism.
II. GAMMA RAYS FROM DM ANNIHILATIONS VIA A LONG-LIVED STATE
A. General discussion
Let a φ be produced at ? r = 0. Then the probability that it decays in a volume element
d3r around ? r is
where Γ is the decay rate of φ, λ the corresponding decay length in the laboratory frame,
while the velocity of φ in the laboratory frame, β = v/c, and the relativistic gamma factor
where mχ,φare the masses of χ and φ. In the last equalities above we used the fact that χ
Let us now calculate the photon flux resulting from final state radiation of the process
shown in Eqs. 1 and 2. The differential photon flux Φγin a solid angle ∆Ω is given by
|? r −?r′|2
1Phenomenologically, annihilations into rapidly decaying particles are challenging to distinguish from pure
annihilation [19, 30] and thus we will not consider them distinct in this paper.
Here ρ(r) is the energy density of DM particles χ, ?σv? is the averaged annihilation cross
section of χ, and we have assumed that χ is a Majorana particle, a choice we shall adopt
throughout this work. The factor of 1/2 accounts for the fact that the incoming particles
are identical, as it has not been included in the definition of ?σv?. d2Nγ/(dEγdcosθ) is
the double differential gamma spectrum per φ decay. It can be obtained by boosting the
isotropic spectrum in the φ rest frame (RF), where
The factor 2 in Eq. 5 accounts for the fact that we will have two φ decays per annihilation.
In Eq. 5 we use coordinates with the origin at the galactic center, and χχ annihilate at
the point?r′and φ decays at ? r. Then θ is the angle between the line of sight (l.o.s.) and the
direction of motion of φ:
cosθ = −ˆ s · (? r −?r′)
|? r −?r′|
where ˆ s is the unit vector along the line of sight. In the following we are going to compare
the predicted photon flux to data from the HESS telescope, using observations of the galactic
center (GC)  within a cone of solid angle 10−5. In this case we have
∆ΩGC= 2π(1 − cos∆ψ) = 10−5,
⊙+ s2− 2r⊙scosψ ,ˆ s = (cosψ,cosϕsinψ,sinϕsinψ), (9)
where r⊙≃ 8.5 kpc is the distance of the solar system from the galactic center. Furthermore,
we use the HESS observation of the galactic ridge (GR) . This is a rectangular region of
the sky with galactic latitude and longitude of |l| ≤ ∆l = 0.8◦, |b| ≤ ∆b = 0.3◦, where the
inner region corresponding to the GC defined above has been subtracted. For the GR we
∆ΩGR≈ 4∆l∆b − ∆ΩGC≈ 2.8 × 10−4,
dΩ ≈ 4
1 −l2+ b2
⊙+ s2− 2r⊙scoslcosb,ˆ s ≈
To compare with HESS observations , one additionally needs to subtract from the photon
flux observed in the GR region |l| ≤ 0.8◦, |b| ≤ 0.3◦the average photon flux in the region
0.8◦< |b| < 1.5◦, a procedure that was used for the background subtraction in .
Before we specialize to the case of non-relativistic φ and proceed with our discussion we
mention briefly some limiting cases of Eq. 5. First, we consider the limit Γ → ∞ (or λ → 0).
In this case the integrand in Eq. 5 will be non-zero only for ? r =?r′. Hence we can replace
ρ(r′) → ρ(r), which can be pulled out of the d3r′integral. Using for the remaining d3r′
As expected we recover the standard expression for the bremsstrahlung photon flux from
annihilation χχ → ℓ+ℓ−+ nγ, apart from a factor 2, since in our case there are twice as
many photons because of two φ decays per χ annihilation (see Eq. 2).
Second, we consider the case of highly relativistic φ. Then one has γ ≫ 1 and β ≈ 1,
while the photons are collimated with φ within an opening angle of size 1/γ. This implies
that there is only a contribution to the integral when (? r −?r′) is aligned with the line of
sight, and the d3r′integration reduces to another integration along the line of sight:
1 − e−s′/λ?
As a reference, we also give the expression for the photon flux from decaying dark matter
(see, e.g. )
where Γ is the DM decay width. The important difference between decaying and annihilating
DM is that the DM density enters linearly in the first case, and quadratically in the second,
which has important observational implications discussed below.
B.The effective DM profile for non-relativistic intermediate state φ
In order to simplify the calculations we specialize from now on to the case of non-
relativistic φ, as the generic effects will not change for the case of a relativistic φ. The
effect of an intermediate LLP with a suitable decay length λ will always be to flatten out
the distribution of SM particle production with respect to the DM distribution, indepen-
dently of its γ. This follows from the structure of Eq. 14 or the general expression Eq. 5.
The general case would deserve a dedicated quantitative study which is beyond the scope of
the present work.
For non-relativistic φ one has γ ≈ 1, β ≪ 1, and the decay of φ is isotropic. In particular,
the photon spectrum is given by Eq. 6. The spectrum can be pulled out of the integration
and the effect is a smearing of the density distribution on the scale λ. One may define an
effective density distribution:
|? r −?r′|2
−r + r′
−|r − r′|
with the exponential integral
Ei(x) ≡ −PV
where PV denotes principal value. Using the effective DM profile in Eq. 16 we can write the
photon flux from a given solid angle ∆Ω as
where the dimensionless J-factor is defined by
For the numerical calculations in this work we use ρ⊙= 0.3GeVcm−3and r⊙= 8.5 kpc,
and for the DM density ρ(r) we always assume a NFW profile 
ρ(r) = ρ⊙r⊙
?1 + r⊙/rs
1 + r/rs
with rs= 20 kpc. Eq. 18 has the same form as the photon flux from standard annihilations
χχ → µ+µ−+ nγ, apart from a factor 2 from the decay of two φ’s, as mentioned above.
In the non-relativistic case the two parameters mφand Γ appear only in the particular
combination λ = γβ/Γ, see Eqs. 3 and 4. Therefore, apart from the two parameters ?σv? and
mχof the standard annihilation scenario, we have now effectively one additional parameter
corresponding to the decay length of φ in the rest frame of the galaxy. Numerically one has
with δ ≡ (mχ− mφ)/mχ≪ 1 and τ = 1/Γ.
In the LLP scenario the source of SM particles from DM annihilations is proportional to
via the propagation of the intermediate state φ over galactic distances we decouple to some
extent the production of SM particles and associated gammas from the DM distribution.
Fig. 1 shows the effective DM profile for various values of λ. From the figure we find that
∼λ we suppress ρeff(r) with respect to ρ(r) since the φ had no time to decay yet, while
∼λ we obtain a ρeff slightly larger than the original DM profile. The relative size of
this over-production increases with λ. Finally, at large distances r ≫ λ all φ’s have decayed
and we recover the NFW profile.
One can understand this behavior qualitatively from the definition of ρeffin Eq. 16. From
the first line of this equation it follows that for r ≫ λ the exponential is non-zero only for
? r ≈?r′and one obtains ρeff(r) → ρ(r). On the other hand, for r ≪ λ and for profiles
ρ(r) ∝ r−γ(γ>
where the last relation follows just from dimensional analysis. Hence, the slope of the photon
source term at r ≪ λ is reduced by one power with respect to ρ2(r). Since for a NFW profile
γ = 1, we find for r ≪ λ that ρ2
eff(r), while in the case of standard annihilations it is proportional to ρ2(r). This means that
∼1) one has roughly
r′2γ|? r −?r′|2=
r′2γ−2|? r −?r′|2∝
eff(r) ∝ ρNFW(r) ∝ 1/r. Note that for DM decay the source of
eff [cm-3 GeV]
0.010.11 10 100
ρeff / ρNFW
values for λ in kpc
FIG. 1: Upper panel: effective DM density profiles as defined in Eq. 16 for various values of λ. Lower
panel, effective DM profiles relative to the NFW profile (corresponding to λ = 0).
0.11 10 100
λ = 10 kpc
FIG. 2: Interpolation between the SM particle source terms for DM annihilation and decay. The dashed
curves show ρ2
⊙for annihilation and ρNFW(r)/ρ⊙for decay, whereas the solid curve corresponds
⊙for DM annihilation into a long lived intermediate state with a decay length of λ = 10 kpc.
gamma rays is proportional to ρ(r), in contrast to the ρ2(r) for annihilations. Therefore, for a
NFW profile our scenario exactly interpolates between DM decay at r ≪ λ and annihilation
at r ≫ λ. This behavior is shown in Fig. 2. Note that the parameter λ controls the absolute
value of ρ2
the particular (somewhat arbitrary) normalization used in Fig. 2. However, for such large λ
effat small r. For λ ≃ 40 kpc one would match exactly the profile for DM decay in
J - Factor
FIG. 3: J-factors for GC and GR (background subtracted as in ) as a function of λ. Dashed lines
correspond to the J-factors for λ = 0.
TABLE I: J-factors for GC and GR (background subtracted as in ) for various values of λ.
the asymptotic ρ2(r) behavior at large r is reached only at distances much larger than the
size of our galaxy. We emphasize that the slope at small r is independent of λ, and hence
it is a generic prediction of our scenario that for profiles with γ ≃ 1 at r<
ray signal looks like DM decay from the inner region of the galaxy (cf. Eq. 15), whereas it
mimics DM annihilation when looking away from the galactic center.
Let us now discuss the dependence of the J-factor, Eq. 19, on λ. In contrast to the usual
annihilation case, J encodes not only the astrophysical DM profile, but it depends now also
on the particle physics parameter λ via ρeff. This dependence is shown in Fig. 3 for the GC
and GR regions observed by HESS, see Eqs. 8 and 10. The figure shows that J gets reduced
as soon as λ becomes larger than the size x of the observed region, since then most of the
φ decay outside of the observed region. For the GC we have x ≃ r⊙
whereas for the GR we have x ≃ r⊙
where the curves in Fig. 3 start to deviate from the J-factor for λ = 0. In Tab. I we give
J for the GC and GR for some values of λ. In the case of GR the J factors also take into
account that HESS subtracted the background by comparing to a region outside the center,
as mentioned after Eq. 11. The quoted J factors are then a difference of the average fluxes
from the two regions. For nonzero λ this can lead to a further suppression of a factor of few.
?∆ΩGC/π ≈ 15 pc,
√∆l∆b ≈ 73 pc, in agreement with the values of λ
III. APPLICATION TO RECENT COSMIC RAY DATA
In this section we apply our LLP scenario to the recent data from cosmic ray experiments
and discuss various other constraints. We describe our fit to the electron and positron data
from PAMELA, ATIC, FERMI and HESS in Sec. IIIA, and show that such a fit can be
made consistent with the HESS GC and GR observations in Sec. IIIB. The bounds from
neutrinos are discussed in Sec. IIIC.
A. The electron-positron signals
The electron flux from DM annihilations (which is equal to the positron flux) at r⊙is
with b(Ee) = E2
the injection spectrum for electrons (equal to the one for positrons) per φ decay, which we
calculate by assuming that φ decays into muons. Then dNe/dE is calculated from the decay
of the muons by using pythia-6.4.19  taking into account final state radiation. We
provide analytic parameterizations of the injection spectra in Appendix A. The diffusion
length λD is given by λ2
throughout this work we assume the so-called MED propagation model from , where
K0 = 0.0112kpc2/Myr and δ = 0.70. The halo function I(λD) is obtained as a series of
Bessel- and Fourier transforms of ρ2
partial differential equation for I(λD) and solve for it numerically, which speeds
up the computation greatly. Details are relegated to appendix B. Note that the
flux in Eq. 23 is a factor 2 larger than in the case of χ annihilations directly into muons,
since we obtain 2 φ’s for each χχ annihilation, each of them giving a µ+µ−pair.
We consider the measurement of the positron fraction Φe+/(Φe++Φe−) from PAMELA ,
where we use only the 9 data points above 6.8 GeV where the effect of solar modulation is
expected to be small. Then we use data on the sum of electrons and positrons (Φe+ + Φe−)
from ATIC , FERMI , and HESS [6, 7] (the effect of older data with much larger
errors, e.g. PPB-BETS , is expected to be small, therefore they are not included in the
fits). Since ATIC and FERMI data are not consistent with each other we do not combine
them but present results using only either of the two. For ATIC we use the combined data
from ATIC 1, 2, and 4, see last reference in . The two HESS measurements  and 
overlap in the intermediate energy range. Therefore we use only the first 4 data points from
, while in the overlap region we take only the measurements of , which have smaller
errors. The overall energy scale of FERMI and the two HESS data sets is varied within the
quoted uncertainties. When we fit “PAMELA only” also the lowest three data points from
ATIC are included in order to constrain the normalization of the background fluxes. For the
astrophysical electron and positron background fluxes we use the parameterization from 
for the fluxes from Galprop . Following , these fluxes are multiplied by Ce±Eαe±,
where in the fit we allow free normalization constants Ce± and assume αe± = 0±0.05 (1σ),
independently for e−and e+.
e/(GeVτE), τE = 1016s, and the electron velocity ve ≈ c. dNe/dE is
D(E,E′) = 4K0τE(Eδ−1− E′δ−1)/(1 − δ) with E,E′in GeV and
eff(r⊙) in .We instead write down a
<σv> [cm3 s-1]
DM mass mχ [GeV]
<σv> [cm3 s-1]
DM mass mχ [GeV]
λ = 0
λ = 1 kpc
λ = 10 kpc
λ = 100 kpc
Allowed regions at 3σ for PAMELA (gray), PAMELA+HESS+FERMI (red),and
PAMELA+HESS+FERMI (dark red) and the constraints from HESS photon observations of the galac-
tic center (GC) and galactic ridge (GR) for λ = 0,1,10,100 kpc. The solid (dashed) curves show HESS
photon constraints at 90% CL with (at 3σ without) including a power law background in the fit, see text
for details. The regions above the curves are excluded.
λ [kpc] mχ[TeV] ?σv?[cm3s−1] χ2
minmχ[TeV] ?σv?[cm3s−1] χ2
minmχ[TeV] ?σv?[cm3s−1] χ2
1.2 × 10−22
1.1 × 10−22
7.9 × 10−23
6.3 × 10−22
7.9 × 10−23
7.9 × 10−23
6.9 × 10−23
5.0 × 10−22
3.2 × 10−24
2.5 × 10−24
2.5 × 10−24
10 × 10−24
TABLE II: Best fit values of mχand ?σv? and the corresponding χ2
for some representative values of the φ decay length λ. The number of degrees of freedom are 38, 43, 10 for
minvalues for electron-positron data
PAMELA+HESS+ATIC, PAMELA+HESS+FERMI, PAMELA-only, respectively.
λ = 10 kpc
λ = 0
E3 dN / dE [GeV2m-2s-1sr-1]
χ2 = 83 / 43
χ2 = 77 / 43
λ = 10 kpc
λ = 0
E3 dN / dE [GeV2m-2s-1sr-1]
χ2 = 112 / 38
χ2 = 102 / 38
FIG. 5: PAMELA data on the positron fraction (left) and the electron-positron data (right) compared to
the predicted spectra for λ = 0 (blue curves) and λ = 10 kpc (black curves) at the best fit values given in
Tab. II. Solid curves correspond to signal + background, whereas with the dashed curves we show background
and signal (right panels only) components separately. Upper panels are for PAMELA+FERMI+HESS, lower
panels for PAMELA+ATIC+HESS. The green curves in the left plots show the spectrum at the best fit to
only PAMELA data for λ = 10 kpc.
The results of our fit to electron data are shown as the shaded regions in Fig. 4 in the
plane of the DM mass mχand the χ annihilation cross section ?σv? for four choices of the φ
decay length λ. These regions are defined by contours of ∆χ2= 11.8 with respect to the χ2
minimum (3σ for 2 dof). Since FERMI and ATIC data are inconsistent at about 3 σ level,
we perform separate fits where one of the two data sets is excluded. The corresponding best
fit values are given in Tab. II, and the fit to the data is shown in Fig. 5.
Irrespective of whether FERMI or ATIC data is used, the steep slope of the high-energy
HESS data provides a strong constraint on the DM mass. This then leads to only a small
change for the best fit values of mχand ?σv? in the two cases. If we fit only PAMELA data,
however, there is a degeneracy between mχand ?σv?, and as long as the peak in the electron
spectrum is above the last PAMELA data point at 83 GeV a good fit is obtained. As shown
in Fig. 4 these results are basically independent of the φ decay length λ, as long as this is
not much larger than the distance from us to the center of the galaxy. The reason is that
electrons and positrons are trapped in the turbulent galactic magnetic field and the observed
signal is dominated by sources “near by”, within several kpc. Therefore, they are not very
sensitive to the change of the source distribution from ρ2(r) to ρ2
important close to the galactic center. Only if λ becomes much larger than r⊙≈ 8.5 kpc
the total electron production close to us will be suppressed, which would require an increase
in the annihilation cross section to maintain the signal. This effect is visible in Fig. 4 and
Tab. II from the results for λ = 100 kpc.
eff(r), which is mostly
B. Gamma ray constraints from HESS GC and GR observations
The photon fluxes predicted for given particle physics parameters mχ,?σv?,λ by Eq. 18
can be compared with observations. We use the gamma ray data from HESS observations
of the GC  and the GR . These data range from about 200 GeV to 20 TeV and are
consistent with a power law spectrum ∝ E−α
parameters we adopt two different strategies. Most conservative bounds can be obtained by
requiring that the signal predicted by DM must not exceed any data point of the observed
flux. We obtain these bounds (denoted by “without background”) by excluding points in
the parameter space where the prediction exceeds x+3σ for any data point, where x is the
observed flux and σ its error bar. This leads to very conservative bounds, since it requires
that in the signal region there is no astrophysical background, whereas in order to account
for the observed flux at energies where DM does not contribute some astrophysical source
has to be assumed. Therefore, we show also bounds by using a second method, called “with
background”. Here we fit the data with a power law background (with free normalization
and power) + the signal from DM. For given λ exclusion limits in the plane of mχand ?σv?
at 90% CL are obtained by the contours with ∆χ2= 4.6 with respect to the χ2minimum.
The photon spectrum dNγ/dEγper φ decay used in Eq. 18 is calculated with pythia-6.4.19
, assuming that φ decays into µ+µ−+ nγ, see Appendix A.
The bounds from the HESS gamma ray observations are shown in Fig. 4 together with
the regions favored by electron-positron data. Clearly, for λ = 0 the electron data are
inconsistent with the gamma ray constraints. The photon flux from the galactic center
gets reduced for finite λ, see Fig. 3, and the bounds shift to larger values of ?σv? as λ is
increased. Fig. 4 shows that for λ ≃ 10 kpc the regions favored by electron-positron data
are consistent with the HESS gamma ray constraints, even for the 90% CL bounds with
background. As one might expect the bounds are also consistent with purely decaying dark
matter [31, 39, 40] (and even in the case where decaying dark matter is also allowed to
Apart from these high energy gamma ray observations additional information can be
obtained from less energetic photons. At energies around 10 GeV the photon flux comes
mainly from inverse Compton scattering on CMB photons, star-light and dust-light, both
for annihilating [42, 43, 44, 45] and decaying DM [46, 47, 48]. This leads to a diffuse gamma
with α ≈ 2.3. In order to obtain bounds on DM
ray signal, potentially observable by FERMI in the near future. The recent diffuse gamma
flux measurement by FERMI in the 10◦− 20◦band above the galactic plane leads to a
bound an order of magnitude weaker than needed to probe the mχ− ?σv? region preferred
by electron/positron data even for the case of λ = 0 . Since the effective density is
smaller in the inner galactic region for λ > 0, cf. Fig. 1, this bound is expected to become
even weaker for the LLP scenario. Looking away from the inner galactic region the signal
can, however, be stronger than for the standard annihilating DM. To asses the impact of
FERMI a more detailed analysis is called for.
In addition to constraints from gamma rays, there are also strong constraints due to
synchrotron radiation of radio waves. These bounds were extensively studied in [16, 18,
49, 50] where it was found that annihilation explanations of the positrons seem to be ruled
out, unless the DM profiles are made effectively less steep as in decays or as in LLP. Thus
it seems very plausible that for sufficiently large λ LLP can satisfy the bounds, though
calculations along the lines of [16, 18] are in order. Constraints using potential gamma ray
signals from dwarf galaxies, on the other hand, are less powerful [51, 52]. Constraints from
galaxy clusters have been discussed in .
C. Neutrinos from the galactic center
Super-Kamiokande (SK) provides an upper limit on the upward going muon flux from
various extra-terrestrial sources , see  for similar recent results from Ice Cube. Since
the LLP scenario predicts 8 neutrinos for each DM annihilation (from the decay of the four
muons from χχ → φφ → 2µ+2µ−) the bound on upward going muons from the galactic
center is potentially relevant. The neutrino induced muon flux can be calculated as, see
is the range of a muon with energy Eµuntil its energy drops below Ethr, for which we take the
SK analysis threshold of 1.6 GeV, with αµ= 2×10−3GeVcm2g−1, βµ= 4.2×10−6cm2g−1,
and ρ is the density of the material in gcm−3. Further, na ≈ raρ/mp are the number
densities of neutrons and protons with rp ≈ 5/9, rn ≈ 4/9, and for the detection cross
section we use
= 0.25,0.15, Bn,p
= 0.06,0.04, and An,p
dΦν/dEνis the flux of muon neutrinos arriving at the earth within a solid angle ∆Ω. In
our case the flux is equal for neutrinos and anti-neutrinos, and it is given by
cone half-opening angle [degree]
muon flux [10-15 cm-2s-1]
λ = 0
mχ = 3.2 TeV, <σv> = 10-22 cm3s-1
SK bound at 90% CL
FIG. 6: Muon flux predicted for mχ= 3.2 TeV, ?σv? = 10−22cm3/s and various values of λ, compared to
the 90% CL upper limit from Super-Kamiokande .
The oscillation probabilities in terms of the lepton mixing matrix elements Uαiare Pνe→νµ=
given in Appendix A. The J-factor in Eq. 27 is defined in the same way as for photons, see
Eq. 19, and therefore the neutrino flux will be reduced with increasing the φ decay length λ
similar to the case of photons.
SK provides an upper bound on the muon flux from a cone around the galactic center
as a function of the cone half opening angle up to 30◦. In Fig. 6 we compare the SK upper
bound to the predicted muon fluxes for fixed mχand ?σv? as a function of λ. We conclude
that λ = 0 is close to the present bound, while for λ>
relaxed. For SK neutrino bounds for decaying DM see [58, 59, 60].
Let us also mention that in the LLP scenario we do not expect any observable neutrino flux
from DM annihilations in the sun (or in the earth), since φ is to a very good approximation
stable at the scale of the solar system. These neutrino fluxes will be exponentially suppressed
by the ratio of the sun–earth distance (or the earth radius) to λ.
i=1|Uei|2|Uµi|2≈ 0.21, Pνµ→νµ=?3
i=1|Uµi|4≈ 0.395. dNνe(νµ)/dEν is the spectrum of
electron (muon) neutrinos per φ decay (and integrates to 1). A parameterization for it is
∼5 kpc the SK bound is significantly
IV.CAN χ BE A THERMAL RELIC?
Let us now address the issue of whether or not χ can be a thermal relic. The LLP
scenario, where χ annihilates first to 2 φ’s and these then after some time decay to SM
particles, faces similar challenges as direct χ’s annihilation into SM particles.
First of all, for χ to be a thermal relic, the required χ annihilation cross section at the time
of freeze-out in the early universe is several orders of magnitudes too small to explain the
PAMELA/ATIC/FERMI cosmic ray anomaly. For χ with TeV mass, the cross section at the
freeze-out should be about ?σv?FO∼ 3×10−26cm3s−1, while as we have seen in section IIIA,
the annihilation cross section that explains the cosmic ray anomaly, is much larger, ?σv?PA∼
10−22cm3s−1. One intriguing possibility which can explain the mismatch is that there exists
an attractive long range force between the dark matter particles 2. This leads to an
enhancement of the cross section at small velocities by a factor of 1/v. With a typical
velocity of χ’s in the galactic halo v/c ∼ 10−3this so called “Sommerfeld enhancement” is
roughly of the right size. In the LLP scenario for the case of a non-relativistic φ, there is
a final phase space suppression of the cross section for χ’s annihilating in the galactic halo
compared to the early universe. However, larger Sommerfeld enhancements are attainable,
if there exists a bound state very close to threshold [62, 63].
The Sommerfeld enhancement solution has several potential phenomenological problems.
For instance, if the Sommerfeld enhancement worked to arbitrarily small velocities, then
annihilation in proto-halos could lead to a too large contribution to the diffuse gamma ray
background . For massive enough attractive force carriers this bound does not apply
(such as GeV mass force carriers of ) as the enhancement saturates. More importantly,
the highly energetic leptons and photons originating from χ annihilations at T<
the early universe could lead to photo-dissociation of light elements destroying the successful
predictions of standard big bang nucleosynthesis (BBN) . The resulting bound on the
annihilation cross section for χχ → µ+µ−is ?σv?FO< 2.0 × 10−23cm3s−1× (mχ/1TeV)−1.
This bound is already somewhat smaller than needed for the explanation of the cosmic ray
anomaly, see Fig 4. The bound itself was obtained assuming time independent cross section,
so a more detailed study in the framework of models giving Sommerfeld enhancement (and
in the LLP scenario) may be warranted. However, since in our case φ is very long lived
compared to BBN time scales, one may expect that within the LLP scenario there is no
threat for BBN from rapid χ annihilations, while there are constraints from the late decays
of φ’s, see below.
There are other ways to “boost” the annihilation cross section in the galactic halo with-
out running into these phenomenological problems. For instance the annihilation could go
through a resonance with mass of order 2mχ[66, 67]. This could arise naturally in models of
extra dimensions with linearly spaced KK modes. Another possibility would be a kination
model [68, 69], where the expansion rate at the time of decoupling is increased due to a
rolling scalar field, leading to a reduced relic density. Yet another possibility is that χ is
a product of a decay of a different meta-stable thermal relic . Clearly, no boost factor
is needed for decaying DM, but the large decay time leads to interesting model-building
implications [59, 71, 72, 73].
In the LLP scenario, where χ is a thermal relic and φ a meta-stable thermal relic, there
are other constraints coming from the late decay of the φ particle. Phenomenologically, the
LLP scenario is interesting if φ travels a distance λ>
impact on the expected photon flux from the galactic center as discussed in section IIIB.
This means that the φ life time is
∼10 keV in
∼10 kpc, because then it has a significant
Such late decaying relics can lead to modifications of the light element abundances [74, 75].
2Long range means here that the mass of the force carrier is smaller than Mv.
For TeV masses and lifetimes in this range, the φ particle must have a relic density which
is three orders of magnitude smaller than the one for χ, see figures 5 and 6 of . For
∼1013s there is an even stronger bound on the φ abundance due to constraints on
the diffuse gamma ray background [74, 76] (see figure 7 of , where the bounds were
derived for the case of radiative decays, while the bounds for leptonic decays are expected
to be slightly weakened). The ratio σann
Consequently, if the couplings to quarks are not excessively suppressed, φ pair production
should be observable at LHC. We leave it as an open exercise in model building to embed
the above hierarchy of cross sections in a concrete model.
must therefore be of order 10−3or smaller.
V.DISCUSSION AND CONCLUSIONS
In this paper we introduced a scenario for cosmic ray production due to dark matter
(DM) which is an alternative to annihilations or decays3. The scenario softens the angular
distribution of the gamma ray spectrum by having the DM annihilate into long lived particles
(LLP), that subsequently decay to SM particles. We have demonstrated within the context of
the PAMELA/ATIC/FERMI positron and electron cosmic ray anomaly, that while standard
DM annihilation scenarios seem to be in conflict with the gamma ray data due to large
required cross sections (?σv? ∼ 10−23cm3s−1), the prolongation of the final state lifetimes in
the LLP scenario can accommodate the data. It should be emphasized that the gamma ray
bounds are sensitive to the dark matter profiles. For less cuspy profiles the bounds would
be weakened. Here we have only considered the NFW profile which grows as 1/r for small
r. Recent simulations [77, 78] that include baryons seem to indicate an even faster rise for
small r which would only strengthen the gamma ray bounds.
An attractive feature of DM annihilations is the direct relation between the annihilation
cross section for indirect DM detection and the production cross section at collider exper-
iments. In the LLP scenario this connection is lost to some extent, since the annihilation
cross section of χ is not directly related to the production cross section of φ. However,
assuming that χ and φ are thermal relics, bounds from light element abundances and diffuse
gamma rays on decaying relics imply that the annihilation cross section of φ has to be some
orders of magnitude larger than the one of χ in order to suppress its relic abundance, see
discussion in section IV. Under the assumption of a non-negligible coupling to hadrons this
implies good prospects for φ production at LHC (if kinematically accessible).
It is interesting to ask how one would differentiate the LLP scenario from DM decay. The
FERMI satellite should be able to accurately measure the angular distribution of the gamma
ray flux. This distribution can be used to differentiate between decays and the LLP scenario
introduced in this paper. Figure 7 shows the gamma ray flux at fixed galactic latitude b = 15
degrees as a function of the galactic longitude for the decay and the annihilation profiles
together with the one for the LLP scenario with λ = 10 kpc. We see that while the long
lifetime reduces the slope compared to annihilation it still cuts off faster than the decay
profile. An analysis of the angular distribution for decaying dark matter was performed in
 and it would be interesting to extend this analysis to our scenario.
3A scenario where the dark matter both anhinilates and decays was propsosed in .
?500 50 100150
FIG. 7: Angular distribution of the gamma ray flux for the LLP scenario with λ = 10 kpc in comparison
to the ones for DM decay and annihilation. We show the gamma ray flux for fixed galactic latitude b = 15◦
as a function of the longitude l, normalized to its value at 30◦.
Another distinguishing feature of LLP will be the smearing out of clumps. If the only
boost factor were due to clumps then this would be a striking signal for dark matter. Typi-
cally the over-densities lie towards the edges of galaxies where they are not tidally disrupted.
The long lived particle scenario would smooth out this striking signal.
Here we have explored only the case of non-relativistic intermediate states, which implies
a near degeneracy between the DM and the LLP. However, the softening of the angular dis-
tribution demonstrated here, which allows one to avoid constraints from gamma ray bounds,
should be active in the case of relativistic annihilation products as well. An investigation of
this case is warranted.
Finally, let us speculate on the possible origin of the energy scale corresponding to the
typical life times of the LLP in our scenario, which is set by our distance to the galactic
center of order 10 kpc, see Eq. 28. If we assume that the decay of the LLP φ is governed
by a dimension six operator suppressed by a high scale Λ one has roughly τ ≃ 16πΛ4/m5
Then we find from Eq. 28 for the scale Λ
Λ ∼2 × 1012GeV
We observe that the scale 1013GeV corresponds roughly to the seesaw scale for neutrino
masses, since v2/Λ ∼ 1 eV for v ∼ 100 GeV and Λ ∼ 1013GeV.4This is particularly
intriguing in light of the leptonic nature of the currently observed cosmic ray features and
might indicate some common high scale physics responsible for light neutrino masses and
the LLP decay in our scenario.
4We mention that a similar argument for a dimension five operator would point to a scale Λ above the
We thank Christopher van Eldik for useful discussion on the HESS GR analysis. T.S.
acknowledges support from the Transregio Sonderforschungsbereich TR27 “Neutrinos and
Beyond” der Deutschen Forschungsgemeinschaft.
APPENDIX A: INJECTION SPECTRA
Here we provide injection spectra for photons, electrons, and neutrinos resulting from the
decay φ → µ+µ−in φ’s rest frame.5Since we work in the limit where φ’s are non-relativistic,
the provided injection spectra can be directly used in the numerical analysis in section III.
The spectra are obtained by fitting analytic functions to a Monte Carlo sample generated
with pythia-6.4.19 . In the following we define x = 2E/mφand 0 ≤ x ≤ 1.
Photon injection spectra. The photon spectrum for the decay (or similarly annihi-
lation) of a heavy particle with mass mφ into charged particles is well described by the
Weizs¨ acker-Williams form (see e.g. [17, 18])
Eγπ[(1 − x)2+ 1]ln
φ(1 − x)
To be specific we have assumed here decay into a lepton pair l. Though in our work we
consider only the muon final state we list here for completeness also the photon spectrum
for electron and tau final states. The results from Pythia for muon and electron final states
are very close to Eq. A1. In the range for Eγfrom 10 GeV to 500 GeV for 1 TeV decaying
DM the Pythia spectrum is 0.8 to 0.6 of the Weizs¨ acker-Williams form. We parameterize
the spectrum by
φ(1 − x)
[(1 − x)2+ C]. (A2)
The coefficients in the above formula depend on the value of mφ. We parameterize this
K(mφ) = K0
where K = A,B,C. Fitting Eq. A2 to the spectrum generated with Pythia we find for
5The spectra apply also for the case of standard annihilations χχ → µ+µ−by replacing mφ→ 2mχ.
The photon spectrum is then described to better than 13% (18% for µ+µ−) everywhere in
the fitted range (mostly to better than 5%). This was obtained from 6 values of mφ, for
each of the values with 107simulated decay events.
For the decay into τ+τ−we have produced 107events in Pythia. Since the taus are heavy
and can decay, producing hadrons and leptons, and these further decay or bremsstrahl
photons, the Weizs¨ acker-Williams form Eq. A1 is no longer a good description of the final
photon spectrum. A form that gives a good description is
Eγπexp[−(A1x + A2x2)] +
with Ai,Bidepending as in Eq. A3 on mφwith
The error on the predicted spectra is below 15% (in the first bin it is 30%).
Electron injection spectra. Consider a two-body decay φ → µ+µ−and neglect the
photon emission. Then the electron spectrum can be obtained by boosting the isotropic
spectrum from the muon rest frame into the rest frame of φ. In our work we do include final
state radiation and therefore, we prefer to use the electron spectrum generated with Pythia.
We use the function
exp[−(A1x + A2x2)] + B0+ B1x, (A7)
with Ai,Bidepending as in Eq. A3 on mφ. Fitting this form to the generated spectrum we
The error on the predicted spectra is below 7%.
Neutrino injection spectra. In each decay φ → µ+µ−we have two neutrinos—νµand
νe—and two antineutrinos with the same spectra. It turns out to be convenient to fit for
the full neutrinos spectrum (νe+νµ) and for the electron neutrino spectrum separately. The
muon neutrino spectrum is then the difference between the total neutrino spectrum and the
electron neutrino spectrum. For the total νe+ νµneutrino spectrum (normalized to 2) we
use the parameterization
exp?−(A1x + A2x2)?+ B , (A9)
with Ai,B depending as in Eq. A3 on mφand
with the error on the predicted spectra below 4% (but mostly below 2%). And for the
electron neutrino spectrum we use
exp?−(A1x + A2x2+ A1/2
√1 − x)?+ B,(A11)
with again Ai,B depending as in Eq. A3 on mφ, and
The νespectrum is described to better than 5%.
APPENDIX B: DIFFERENTIAL EQUATION FOR I(λD)
The propagation of positrons in the galaxy is described by a transport equation (see, e.g.
∂tΨ − ∇?K(x,E)∇Ψ) − ∂E
with Ψ(? x,E) the positron number density per energy interval and q(? x,E) the positron source
term. The loss coefficient due to Compton scattering is b(E) = E0ǫ2/τE where ǫ = E/E0
with E0 = 1 GeV and τE = 1016s, while the transport through the turbulent galactic
magnetic field is approximated by a spatially constant diffusion coefficient K(? x,E) = K0ǫδ.
For annihilating dark matter the positron injection source term is
?b(E)Ψ?= q(x,E), (B1)
q(? x,E) = κ
while for LLP scenario ρeff instead of ρ should be used (for decaying dark matter, on the
other hand, the dependence on ρ is linear). The positrons flux at radial distance from
galactic center r and distance z from the galactic plane is then given by
Ψ(r,z,ǫ) = κτE
where the diffusion length is λD=√4K0˜ τ, with ˜ τ =˜t −˜tSand˜t = τE(ǫδ−1/(1 − δ)). The
function˜I(λD,r,z) depends only on astrophysics, i.e. only on the DM halo profile. The
function I(λD) in Eq. 23 is then equal to˜I(λD,r,z) at the position of the solar system
In  the function˜I(λD,r,z) is given in terms of an infinite sum over Bessel and cosine
functions, however this expansion converges very slowly.
Alternatively, using the ansatz Eq. B3 in Eq. B1, one can write down a partial differential
λD∂λD˜I(λD,r,z) = 0. (B5)
The boundary conditions
˜I(λD,Rgal,z) = 0,
˜I(λD,r,±L) = 0,
follow from a requirement that Ψ(r,z,ǫ) vanishes on the boundaries of the galactic disk with
radius Rgaland thickness 2L (the first boundary condition in Eq. B6 has to be multiplied by
a smooth step function near the galactic disk boundaries to make it consistent with the next
two boundary conditions). The above partial differential equation can be solved numerically
easily using standard methods, for instance the method of lines. The solution is obtained on
a personal computer in less then a minute for the whole galaxy, compared to several hours
for each point, if the expansion in Bessel and cosine functions is used. With appropriately
modified boundary conditions the above partial differential equation can be used for LLP
scenario as well as for decaying dark matter.
 O. Adriani et al., arXiv:0810.4995 [astro-ph].
 J. Chang et al., Nature 456 (2008) 362; preliminary results from the ATIC-4 flight were shown
by J. Isbert at the workshop ”Tango in Paris”, May 4–6, 2009, Institut d’Astrophysique de
 A. A. Abdo et al. [The Fermi LAT Collaboration], arXiv:0905.0025 [astro-ph.HE].
 A. Buffington, C. D. Orth and G. F. Smoot, Astrophys. J. 199, 669 (1975); D Mueller
and K.K Tang, Astrophys. J. 312, 183 (1987); R. L. Golden et al., Astrophys. J. 436,
769 (1994); S. W. Barwick et al. [HEAT Collaboration], Astrophys. J. 482, L191 (1997)
 S. Torrii et al., Astrophys. J. 559, 973 (2001); J. Nishimura et al., Advances in Space Research
19, 767 (1997); J. Nishimura et al. Astrophys. J. 238, 394 (1980).
 F. Aharonian et al. [H.E.S.S. Collaboration], Phys. Rev. Lett. 101, 261104 (2008)
 H. E. S. Aharonian, arXiv:0905.0105 [astro-ph.HE].
 J. H. Taylor and D. R. Stinebring, ARA&A 24, 285 (1986).
 D. C. Backer, Pulsars (Galactic and Extragalactic Ra- dio Astronomy, 1988), pp. 480521.
 D. Hooper, P. Blasi and P. D. Serpico, JCAP 0901, 025 (2009) [arXiv:0810.1527 [astro-ph]].
 H. Yuksel, M. D. Kistler and T. Stanev, arXiv:0810.2784 [astro-ph].
 S. Profumo, arXiv:0812.4457 [astro-ph].
 N. J. Shaviv, E. Nakar and T. Piran, arXiv:0902.0376 [astro-ph.HE].
 D. Malyshev, I. Cholis and J. Gelfand, arXiv:0903.1310 [astro-ph.HE].
 J. Hall and D. Hooper, arXiv:0811.3362 [astro-ph].
 G. Bertone, M. Cirelli, A. Strumia and M. Taoso, arXiv:0811.3744 [astro-ph].
 N. F. Bell and T. D. Jacques, arXiv:0811.0821 [astro-ph].
 L. Bergstrom, G. Bertone, T. Bringmann, J. Edsjo and M. Taoso, arXiv:0812.3895 [astro-ph].
 P. Meade, M. Papucci and T. Volansky, arXiv:0901.2925 [hep-ph].
 F. Aharonian et al. [H.E.S.S. Collaboration], Phys. Rev. Lett. 97, 221102 (2006) [Erratum-
ibid. 97, 249901 (2006)] [arXiv:astro-ph/0610509].
 F. Aharonian et al. [H.E.S.S. Collaboration], Nature 439, 695 (2006) [arXiv:astro-ph/0603021].
 P. Meade, M. Papucci, A. Strumia and T. Volansky, arXiv:0905.0480 [hep-ph].
 M. Pospelov and A. Ritz, Phys. Lett. B 671, 391 (2009) [arXiv:0810.1502 [hep-ph]].
 N. Arkani-Hamed, D. P. Finkbeiner, T. Slatyer and N. Weiner, Phys. Rev. D 79, 015014
(2009) [arXiv:0810.0713 [hep-ph]].
 I. Cholis, D. P. Finkbeiner, L. Goodenough and N. Weiner, arXiv:0810.5344 [astro-ph].
 Y. Nomura and J. Thaler, arXiv:0810.5397 [hep-ph].
 I. Cholis, G. Dobler, D. P. Finkbeiner, L. Goodenough and N. Weiner, arXiv:0811.3641 [astro-
 P. J. Fox and E. Poppitz, arXiv:0811.0399 [hep-ph].
 F. Chen, J. M. Cline and A. R. Frey, arXiv:0901.4327 [hep-ph].
 J. Mardon, Y. Nomura, D. Stolarski and J. Thaler, arXiv:0901.2926 [hep-ph].
 E. Nardi, F. Sannino and A. Strumia, JCAP 0901 (2009) 043 [arXiv:0811.4153 [hep-ph]].
 J. F. Navarro, C. S. Frenk and S. D. M. White, Astrophys. J. 462, 563 (1996)
 T. Sjostrand, S. Mrenna and P. Skands, JHEP 0605, 026 (2006) [arXiv:hep-ph/0603175].
 T. Delahaye, R. Lineros, F. Donato, N. Fornengo and P. Salati, Phys. Rev. D 77 (2008) 063527
 S. Torii et al., arXiv:0809.0760 [astro-ph].
 E. A. Baltz and J. Edsjo, Phys. Rev. D 59, 023511 (1999) [arXiv:astro-ph/9808243].
 I. V. Moskalenko and A. W. Strong, Astrophys. J. 493 (1998) 694 [arXiv:astro-ph/9710124].
 M. Cirelli, R. Franceschini and A. Strumia, Nucl. Phys. B 800, 204 (2008) [arXiv:0802.3378
 G. Bertone, W. Buchmuller, L. Covi and A. Ibarra, JCAP 0711, 003 (2007) [arXiv:0709.2299
 A. Ibarra and D. Tran, arXiv:0811.1555 [hep-ph].
 K. Cheung, P. Y. Tseng and T. C. Yuan, arXiv:0902.4035 [hep-ph].
 C. R. Chen, M. M. Nojiri, S. C. Park, J. Shu and M. Takeuchi, arXiv:0903.1971 [hep-ph].
 E. Borriello, A. Cuoco and G. Miele, arXiv:0903.1852 [astro-ph.GA].
 M. Lattanzi and J. I. Silk, arXiv:0812.0360 [astro-ph].
 R. Harnik and G. D. Kribs, arXiv:0810.5557 [hep-ph].
 K. Ishiwata, S. Matsumoto and T. Moroi, arXiv:0903.0242 [hep-ph].
 S. Shirai, F. Takahashi and T. T. Yanagida, arXiv:0902.4770 [hep-ph].
 P. Grajek, G. Kane, D. Phalen, A. Pierce and S. Watson, arXiv:0812.4555 [hep-ph].
 J. Zhang, X. J. Bi, J. Liu, S. M. Liu, P. f. Yin, Q. Yuan and S. H. Zhu, arXiv:0812.0522
 K. Ishiwata, S. Matsumoto and T. Moroi, Phys. Rev. D 79, 043527 (2009) [arXiv:0811.4492
 R. Essig, N. Sehgal and L. E. Strigari, arXiv:0902.4750 [hep-ph].
 L. Pieri, M. Lattanzi and J. Silk, arXiv:0902.4330 [astro-ph.HE].
 Q. Yuan, X. Bi, F. Huang and X. Chen, arXiv:0902.4294 [astro-ph.CO].
 S. Desai et al. [Super-Kamiokande Collaboration], Phys. Rev. D 70, 083523 (2004) [Erratum- Download full-text
ibid. D 70, 109901 (2004)] [arXiv:hep-ex/0404025].
 R. Abbasi, arXiv:0902.2460 [astro-ph.CO].
 C. Delaunay, P. J. Fox and G. Perez, arXiv:0812.3331 [hep-ph].
 V. Barger, W. Y. Keung, G. Shaughnessy and A. Tregre, Phys. Rev. D 76, 095008 (2007)
 J. Hisano, M. Kawasaki, K. Kohri and K. Nakayama, arXiv:0812.0219 [hep-ph].
 A. Arvanitaki, S. Dimopoulos, S. Dubovsky, P. W. Graham, R. Harnik and S. Rajendran,
 J. Liu, P. f. Yin and S. h. Zhu, arXiv:0812.0964 [astro-ph].
 J. Hisano, S. Matsumoto, M. Nagai, O. Saito and M. Senami, Phys. Lett. B 646, 34 (2007)
 J. D. March-Russell and S. M. West, arXiv:0812.0559 [astro-ph].
 M. Pospelov, A. Ritz and M. B. Voloshin, Phys. Lett. B 662, 53 (2008) [arXiv:0711.4866
 M. Kamionkowski and S. Profumo, Phys. Rev. Lett. 101, 261301 (2008) [arXiv:0810.3233
 J. Hisano, M. Kawasaki, K. Kohri, T. Moroi and K. Nakayama, arXiv:0901.3582 [hep-ph].
 M. Ibe, H. Murayama and T. T. Yanagida, arXiv:0812.0072 [hep-ph].
 W. L. Guo and Y. L. Wu, arXiv:0901.1450 [hep-ph].
 P. Salati, Phys. Lett. B 571, 121 (2003) [arXiv:astro-ph/0207396].
 F. Rosati, Phys. Lett. B 570, 5 (2003) [arXiv:hep-ph/0302159].
 M. Fairbairn and J. Zupan, arXiv:0810.4147 [hep-ph].
 M. Pospelov and M. Trott, arXiv:0812.0432 [hep-ph].
 P. f. Yin, Q. Yuan, J. Liu, J. Zhang, X. j. Bi and S. h. Zhu, Phys. Rev. D 79, 023512 (2009)
 C. R. Chen, M. M. Nojiri, F. Takahashi and T. T. Yanagida, arXiv:0811.3357 [astro-ph].
 J. R. Ellis, G. B. Gelmini, J. L. Lopez, D. V. Nanopoulos and S. Sarkar, Nucl. Phys. B 373,
 M. Kawasaki and T. Moroi, Astrophys. J. 452, 506 (1995) [arXiv:astro-ph/9412055].
 G. D. Kribs and I. Z. Rothstein, Phys. Rev. D 55, 4435 (1997) [Erratum-ibid. D 56, 1822
 O. Y. Gnedin, A. V. Kravtsov, A. A. Klypin and D. Nagai, Astrophys. J. 616, 16 (2004)
 S. Kazantzidis et al., Astrophys. J. 623, L67 (2005) [arXiv:astro-ph/0407407].