On The Gamma-Ray Emission From Reticulum II and Other Dwarf Galaxies
The recent discovery of ten new dwarf galaxy candidates by the Dark Energy Survey (DES) and the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS) could increase the Fermi Gamma-Ray Space Telescope’s sensitivity to annihilating dark matter particles, potentially enabling a definitive test of the dark matter interpretation of the long-standing Galactic Center gamma-ray excess. In this paper, we compare the previous analyses of Fermi data from the directions of the new dwarf candidates (including the relatively nearby Reticulum II) and perform our own analysis, with the goal of establishing the statistical significance of any gamma-ray signal from these sources. We confirm the presence of an excess from Reticulum II, with a spectral shape that is compatible with the Galactic Center signal. The significance of this emission is greater than that observed from 99.84% of randomly chosen high-latitude blank-sky locations, corresponding to a local detection significance of 3.2. We improve upon the standard blank-sky calibration approach through the use of multi-wavelength catalogs, which allow us to avoid regions that are likely to contain unresolved gamma-ray sources.
Over the past several years, a bright and statistically significant excess of gamma-rays has been reported from the region surrounding the Galactic Center Goodenough and Hooper (2009); Hooper and Goodenough (2011); Hooper and Linden (2011); Abazajian and Kaplinghat (2012); Gordon and Macias (2013); Hooper and Slatyer (2013); Abazajian et al. (2014); Daylan et al. (2014); Calore et al. (2014). The spectral and morphological characteristics of this signal are each in good agreement with that predicted from annihilating dark matter particles with a mass of 35-60 GeV and a cross section of cms (for the representative case of annihilations to ). And although the proposed astrophysical explanations for this excess have been shown to face considerable challenges, it is not currently possible to entirely rule out the possibility that these photons originate from a large population of unresolved point sources Cholis et al. (2014); Hooper et al. (2013); Petrovic et al. (2015) or from a series of cosmic ray outbursts Petrovic et al. (2014); Carlson and Profumo (2014); Calore et al. (2015). In light of this situation, gamma-ray observations of the Milky Way’s dwarf spheroidal galaxies play a critical role, being potentially able to provide a confirmation or refutation of the dark matter interpretation of the Galactic Center excess.
Searches for gamma-rays from known dwarf galaxies (Ackermann et al., 2015; Geringer-Sameth et al., 2014; Ackermann et al., 2014) have yielded stringent constraints on the dark matter parameter space. They have not yet, however, been sufficiently sensitive to cover the full range of cross sections favored to explain the Galactic Center excess. It has been anticipated that ongoing and planned optical surveys will discover a significant number of presently unknown Milky Way dwarf spheroidal galaxies He et al. (2013); Tollerud et al. (2008); Hargis et al. (2014); Rossetto et al. (2011). If one of more of these objects happens to be nearby and/or contain a high density of dark matter, it could constitute an important target for the Fermi Gamma-Ray Space Telescope, significantly strengthening their sensitivity to annihilating dark matter.
Very recently, optical imaging data from the Dark Energy Survey (DES) was used to discover nine new dwarf galaxy candidates Bechtol et al. (2015); Koposov et al. (2015). Shortly thereafter, yet another dwarf candidate (Triangulum II) was discovered from within the data from the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS) Laevens et al. (2015). Of particular interest is the object Reticulum II (also known as DES J0335.6-5403) whose proximity (30-32 kpc) and spatial extent (half-light radius of pc) make it likely to be a dwarf galaxy (rather than a globular cluster) and a very promising target for gamma-ray searches for annihilating dark matter. Although spectroscopic follow-up will be required to measure the dark matter distributions of these systems, it is plausible that Recticulum II (or perhaps Triangulum II) could provide a gamma-ray signal from annihilating dark matter that is brighter than that from any other known dwarf galaxy.
Two independent groups have already reported the results of their analyses of Fermi Gamma-Ray Space Telescope data from the directions of Reticulium II and DES’s other new dwarf galaxy candidates.
The first of these analyses, presented jointly by the Fermi and DES collaborations, identified a modest gamma-ray excess from the direction of Reticulium II, with a test statistic (TS) of 6.7 Drlica-Wagner et al. (2015).
In this article, we present our own analysis of the (Pass 7) Fermi data from the directions of Reticulum II and other Milky Way dwarf galaxies with the goal of assessing the characteristics and statistical significance of any excess that might exist. We confirm the existence of a gamma-ray signal from Reticulum II, and assess the (local) significance of this excess to be 3.2. And although spectroscopic follow-up of Reticulum II will be required before this observation can be used to constrain or infer the value of the dark matter annihilation cross section, for a plausible range of dark matter profiles, this result appears to be consistent with dark matter interpretations of the Galactic Center signal and the null results from other dwarf galaxies.
Ii Fermi Data Analysis
In order to calculate the significance of any gamma-ray emission observed from a given dwarf galaxy (or dwarf galaxy candidate), we examine approximately 6.5 years of Fermi-LAT data,
In our analysis, we follow a prescription as similar as possible to that employed by the Fermi-LAT collaboration (Ackermann et al., 2014, 2015; Drlica-Wagner et al., 2015). Specifically, we first set the global normalization of background sources and the dwarf spheroidal over the entire 500 MeV – 500 GeV energy range, utilizing the Fermi-LAT gtlike code and the MINUIT algorithm. In this phase, we seed the dwarf spheroidal spectrum as a simple power-law with an index of -2.0. We then fix the normalization of all background components, and employ the pyLikelihood package to scan the flux of the dwarf spheroidal in each energy bin, calculating the delta-log-likelihood () as a function of the source flux. In order to calculate the TS for a given dark matter model, we minimize the total log-likelihood summed over all energy bins after constraining the photon flux by the spectral shape of the dark matter model.
To validate the results of this method, we perform two tests. First, we randomly select 100 Fermi-LAT 3FGL point sources Acero et al. (2015) with 30, a “curve significance” smaller than 2 (indicating consistency with a power-law spectrum) and a TS smaller than 100 in the energy range of 300 MeV to 100 GeV. In order to compare our results to those given in the 3FGL catalog, we employ the above technique with the following modifications. We restrict our analysis to four years of Fermi-LAT data,
Secondly, we apply the “blank-sky” null-test employed in previous dwarf spheroidal studies. Specifically, we select 1905 sky locations with 30, which are 1 removed from any 3FGL source and 5 removed from any extended 3FGL source. In this case, we employ the full 6.5 years of data, adopt the default energy range, and test the comparison to a 49 GeV dark matter model annihilating to (corresponding to the best-fit value of the mass for the spectrum of the Galactic Center excess Calore et al. (2014)). In Fig. 2 we show the resulting distribution of our blank-sky test locations. While the existence of systematic errors in the modeling of the gamma-ray background drives this distribution far from that expected from Poisson variations, the result is in good agreement with all previous studies. In this figure, we show results corresponding to the case in which no additional requirements are placed on the blank sky locations (blue), and to when the blank sky locations used are further required to lie no closer than from any source listed in the BZCAT, CRATES, CGRaBS, or ATNF catalogs (red). This will be discussed in more detail in Sec. IV.
|TS (Point-Like)||TS (NFW-Like)|
|Dwarf Name||Distance (kpc)||Latitude ()||Ref. Martinez (2013)||Ref. Geringer-Sameth et al. (2015b)||GeV||any||GeV||any|
|Reticulum II||32 (30)||-49.7||–||–||17.4||18.1||–||–|
|Tucana II||58 (69)||-52.4||–||–||1.44||1.82||–||–|
|Indus I||69 (100)||-42.1||–||–||0.0||0.0||–||–|
|Horologium I||87 (79)||-54.7||–||–||0.09||0.17||–||–|
|Phoenix II||95 (83)||-59.7||–||–||0.0||0.55||–||–|
|Eridanus III||95 (87)||-59.6||–||–||0.0||0.53||–||–|
|Pictoris I||126 (114)||-40.6||–||–||0.0||0.0||–||–|
|Eridanus II||330 (380)||-51.6||–||–||0.0||0.61||–||–|
|Segue 1||23||50.4||19.5 0.29||1.07||1.18||1.10||1.72|
|Ursa Major II||32||37.4||19.3 0.28||0.0||0.32||0.02||0.88|
|Willman 1||38||56.8||19.1 0.31||–||3.94||4.47||5.70||5.89|
|Coma Berenices||44||83.6||19.0 0.25||0.0||0.0||0.0||0.04|
|Bootes I||66||69.6||18.8 0.22||0.0||0.75||0.09||0.50|
|Ursa Minor||76||44.8||18.8 0.19||0.0||1.36||0.0||0.99|
|Ursa Major I||97||54.4||18.3 0.24||0.0||0.07||0.0||0.24|
|Leo IV||154||56.5||17.9 0.28||0.0||0.0||0.0||0.0|
|Canes Venatici II||160||82.7||17.9 0.25||0.27||1.56||0.50||1.62|
|Canes Venatici I||218||79.8||17.7 0.26||0.39||0.47||0.35||0.42|
|Leo II||233||67.2||17.6 0.18||0.0||0.0||0.0||0.0|
|Leo I||254||49.1||17.7 0.18||0.0||1.67||0.0||1.91|
In Fig. 3, we show the delta-log-likelihood () distribution for our analysis of Fermi data from the direction of Reticulum II. As in both Ref. Drlica-Wagner et al. (2015) and Ref. Geringer-Sameth et al. (2015a), we find an excess of events in the bins covering approximately 2-10 GeV. For a spectral shape corresponding to a 49 GeV dark matter particle annihilating to (the best-fit mass for the Galactic Center excess Calore et al. (2014)), we find a value of TS=17.4 from Reticulum II, corresponding to a significance of 3.2 (see Fig. 2). If we do not impose this choice of the dark matter mass, but rather allow the mass to float as a free parameter, the value of the TS increases only slightly (to 18.1), illustrating the compatibility between this signal and that observed from the Galactic Center.
In Table 1, we list the TS values found in our analysis for each of the previously known Milky Way dwarf spheroidal galaxies, and for the ten newly discovered dwarf galaxy candidates. Values are given assuming either a spectrum corresponding to the best-fit mass for the Galactic Center excess, or for any dark matter mass. For a Galactic Center-like spectrum, Reticulum II yields the highest significance (TS=17.4), followed by Willman 1 (3.94), Hercules (3.09), Sagittarius (2.13), Tucana II (1.44), and Segue 1 (1.07). Other than that from Reticulum II, no statistically significant excesses are observed.
Also given in Table 1 are the values of the -factors for each dwarf galaxy with sufficient kinematic information (from spectroscopic data) to obtain a determination. This quantity is defined as follows:
where is taken to be a circle of radius around the given dwarf, describes the dark matter density profile of the dwarf, and the second integral is performed over the observed line-of-sight (los). We provide the -factor values as reported by two groups: Martinez et al. Martinez (2013) and Geringer-Sameth et al. Geringer-Sameth et al. (2015b). In general, the dark matter profiles of the classical dwarfs are well constrained by stellar kinematics, resulting in relatively well determined -factors. In contrast, the ultra-faint dwarfs (Segue 1, Ursa Major II, Willman 1) contain far fewer stars, and exhibit much larger -factor error bars. Deeper measurements, capable of detecting more numerous faint stars, will ultimately improve this situation. Although no spectroscopic information exists for any of the ten new dwarf galaxy candidates, we expect such follow-up measurements to occur in the near future.
For those dwarfs with profiles constrained by stellar kinematics, we also list in Table 1 the values of the TS found when the source is treated as a spatially extended object, rather than as a point-like source. In particular, we adopt an NFW-profile for these systems, with a scale radius equal to the central value reported in Refs. Ackermann et al. (2014); Martinez (2013). In Fig. 4, we plot the TS as a function of the halo’s scale radius (in degrees, ) for five of the dwarf galaxies under consideration. No strong evidence for (or against) spatial extension is observed. The significances of Willman 1 and Ursa Major II marginally increase if an extended halo is assumed, while the significances of Reticulum II and Sagittarius marginally decrease.
Iv Controlling Backgrounds With Multi-Wavelength Source Catalogs
In Ref. Carlson et al. (2015), it was pointed out that Fermi’s sensitivity to dark matter annihilation in dwarf spheroidal galaxies could be increased by taking into account information available in multi-wavelength source catalogs. In particular, a significant fraction of the highest TS points in the “blank sky” correspond to the locations of unresolved blazars, radio galaxies, and starforming galaxies. By making use of only regions of the “blank sky” which are not near sources listed in multi-wavelength catalogs, it is possible to reduce the contamination from such sources.
In Fig. 2, the TS distribution of the high-latitude blank-sky is shown without utilizing multi-wavelength information (blue), and after avoiding all locations located within of any source listed in the Roma-BZCAT Multi-Frequency Catalog of Blazars (BZCAT) (Massaro et al., 2008), the Combined Radio All-Sky Targeted Eight-GHz Survey (CRATES) catalog Healey et al. (2007), the Candidate Gamma-Ray Blazar Survey (CGRaBS) catalog Healey et al. (2008), or the Australia Telescope National Facility (ATNF) pulsar catalog Hobbs et al. (2004) (red). The application of this cut significantly reduces the fraction of the sky with large TS values.
To take this multi-wavelength information into account, we apply the following procedure in our analysis. For a given dwarf galaxy (or dwarf galaxy candidate), we check the catalogs described in the previous paragraph for any sources located within . If any are found, we re-run our analysis, including in the background model a source at that location. We then take the new TS of the dwarf, and see how many locations on the blank sky yield a higher TS, when a background source is included at the location of the nearby catalog source. We then use the fraction of high-TS blank-sky locations to calculate the -value and significance of any dwarf galaxy excess.
This procedure is most important in the case of Reticulum II, which is located 0.44 from the source CRATES J033553-543025.
As we were finalizing this paper, it was pointed out to us that the faint radio source PMN J0335-5406 in the Parkes-MIT-NRAO catalog is located from the location of Reticulum II.
V A Self-Consistent Interpretation
In this section, we consider the excess observed from Reticulum II, along with the lack of significant detections from other Milky Way dwarf galaxies, and ask whether these results are mutually consistent. Focusing on the case in which annihilating dark matter is responsible for the Galactic Center excess ( GeV, for the case of annihilatiions to ), the lack of significant excess emission from the known dwarf galaxies constrains cm/s Ackermann et al. (2015). And while this constraint is compatible with dark matter interpretations of the Galactic Center excess, the normalization of the Galactic Center signal implies that the cross section is unlikely to be smaller than this value by more than a factor of a few. More specifically, if we allow the overall normalization and the scale radius of the Milky Way’s dark matter halo profile to vary within the range allowed by dynamical constraints ( GeV/cm, kpc Iocco et al. (2011); Catena and Ullio (2010)), we find consistency with an annihilation cross section as small as cm/s. From this perspective, the prospects for the future detection of a gamma-ray signal from one or more dwarf galaxies appears encouraging.
For an annihilation cross section at the upper limit of Ref. Ackermann et al. (2015) ( cm/s), the normalization of the signal from Reticulium II requires approximately –20.1. Noting the empirical (and approximate) relationship between the distances and -factors of ultra-faint dwarfs, Ref. Drlica-Wagner et al. (2015) points out that Reticulium II might be expected to have a somewhat smaller value, , although even a value as high as of 20.1 would not be a particuarly significant outlier. The necessity of a large -factor for Reticulum II (if its gamma-ray excess is from annihilating dark matter) can also be seen from the results of our analysis, as shown in Table 1. Roughly speaking, the predicted value for the TS of a given dwarf is proportional to its gamma-ray flux, and thus to its -factor. The modest TS values observed from Segue 1 and Ursa Major II suggest significantly lower -factors for these systems than for Reticulum II. Given this situation, we eagerly await the spectroscopic follow-up of Reticulium II. If the gamma-ray excess from this source in fact originates from annihilating dark matter, we should anticipate a large value for its -factor, likely in excess of GeV cm.
Vi Are Statistical Fluctuations Sufficient To Explain The Differences Between The Results Found Using Pass 7 and Pass 8 Data?
At face value, our determination of TS=18.1 from the direction of Reticulum II appears to be in conflict with the more modest value of 6.7 quoted by the Fermi Collaboration Drlica-Wagner et al. (2015). The most significant difference between these two analyses is in the data sets that are being considered: our analysis makes use of the publicly available Pass 7 data, whereas the Fermi Collaboration paper utilizes the more recent Pass 8 data set. Intriguingly, a similar discrepancy can be seen in a comparison of the Fermi’s Collaboration’s Pass 7 and Pass 8 studies of known dwarf galaxies. In particular, whereas the Fermi Collaboration’s Pass 7 analysis revealed an excess from three ultra-faint dwarfs (Segue 1, Ursa Major II, and Willman 1), at a level of TS (Ackermann et al., 2014), no excess was observed according to their more recent Pass 8 paper Ackermann et al. (2015). This would-be conflict between the Fermi Pass 7 and Pass 8 dwarf papers is somewhat surprising in light of the fact that these two analyses make use of data taken over significantly overlapping time periods (over the first four years of Fermi’s mission; the more recent analysis adds two more years of data to this set). The Fermi Collaboration points out, however, that after taking into account the new event selection associated with the transition from Pass 7 to Pass 8, the overlap between these two data sets is not particularly large; only 30% of the 1-10 GeV photons used in the most recent analysis were also employed in their earlier study (put another way, approximately 65% of those events in Fermi’s Pass 7 dwarf analysis were also included in the recent Pass 8 study) Ackermann et al. (2015). Assuming that the previous TS 10 excess was the result of a statistical fluctuation, we estimate that there was an approximately 10% chance that the more recent data set would yield TS (after taking into account differences in effective area and exposure time). So in this respect, we concur with the conclusion of the Fermi Collaboration.
In order to compare our results directly to those of Ref. (Ackermann et al., 2014), we reanalyzed the three ultra-faint dwarf spheroidal galaxies (Segue 1, Willman 1, and Ursa Major II) using the P7V6 Reprocessed data, employing in this test only four years
Vii Summary and Conclusions
In this article, we have revisited the gamma-ray emission from known Milky Way dwarf galaxies, and from the dwarf galaxy candidates recently discovered in the data from DES and Pan-STARRS. Of particular interest are the new dwarf candidates Reticulum II and Triangulum II, which are each located at a distance of only 30 kpc from Earth, making them promising targets for dark matter searches. Our analysis of Fermi data from the direction of Reticulum II identifies an excess of gamma-rays with a local statistical significance of 3.2. This is slightly higher than, but not dissimilar to, that reported in the previous studies of other groups Drlica-Wagner et al. (2015); Geringer-Sameth et al. (2015a). We also confirm that Reticulum II’s gamma-ray excess is most prominent at energies between 2-10 GeV, in good agreement with the spectral shape of the excess previously reported from the region surrounding the Galactic Center. We do not observe any significant -ray emission from the direction of Triangulum II.
Looking forward, spectroscopic follow-up of Reticulum II and the other new dwarf galaxy candidates will be important for interpreting this data. In order for this excess to be compatible with the lack of significant gamma-ray detections from other dwarf galaxies (most importantly, Segue 1 and Ursa Major II), Reticulum II must contain a high density of dark matter, corresponding to GeVcm. A measurement of Reticulum II’s -factor that is much smaller than this value would place serious doubt as to any dark matter interpretation of its excess. Additional data from Fermi will also have much to bear on this question. With 50% more data, such as could be acquired over the next few years, we estimate that the detection of Reticulum II’s gamma-ray emission could exceed TS=25, corresponding to approximately 4 significance, and comparable to the threshold for membership in Fermi’s point source catalogs.
Acknowledgements. We would like to thank Andrea Albert and Alex Drlica-Wagner for helpful comments and discussions, as well as Eric Carlson for pointing out the proximity of the nearby CRATES source. DH is supported by the US Department of Energy under contract DE-FG02-13ER41958. Fermilab is operated by Fermi Research Alliance, LLC, under Contract No. DE- AC02-07CH11359 with the US Department of Energy. TL is supported by the National Aeronautics and Space Administration through Einstein Postdoctoral Fellowship Award No. PF3-140110.
- TS is defined as twice the difference in the global log-likelihood between the null and alternative hypotheses.
- MET range: 239557417 - 447078115
- MET range: 239557417 - 365467563
- The presence of this source was pointed out to us by Eric Carlson.
- We note that the spectral shape absorbed by the CRATES source is very hard, and quite unlike that of the gamma-ray emission from typical radio sources. We consider it unlikely that this source contributes significantly to the gamma-ray flux observed from the direction of Reticulum II.
- We thank Alex Drlica-Wagner for bringing this source to our attention.
- MET range: 239557417 – 365817602
- L. Goodenough and D. Hooper (2009), eprint 0910.2998.
- D. Hooper and L. Goodenough, Phys.Lett. B697, 412 (2011), eprint 1010.2752.
- D. Hooper and T. Linden, Phys.Rev. D84, 123005 (2011), eprint 1110.0006.
- K. N. Abazajian and M. Kaplinghat, Phys.Rev. D86, 083511 (2012), eprint 1207.6047.
- C. Gordon and O. Macias, Phys.Rev. D88, 083521 (2013), eprint 1306.5725.
- D. Hooper and T. R. Slatyer, Phys.Dark Univ. 2, 118 (2013), eprint 1302.6589.
- K. N. Abazajian, N. Canac, S. Horiuchi, and M. Kaplinghat, Phys.Rev. D90, 023526 (2014), eprint 1402.4090.
- T. Daylan, D. P. Finkbeiner, D. Hooper, T. Linden, S. K. N. Portillo, et al. (2014), eprint 1402.6703.
- F. Calore, I. Cholis, and C. Weniger (2014), eprint 1409.0042.
- I. Cholis, D. Hooper, and T. Linden (2014), eprint 1407.5625.
- D. Hooper, I. Cholis, T. Linden, J. Siegal-Gaskins, and T. Slatyer, Phys.Rev. D88, 083009 (2013), eprint 1305.0830.
- J. Petrovic, P. D. Serpico, and G. Zaharijas, JCAP 1502, 023 (2015), eprint 1411.2980.
- J. Petrovic, P. D. Serpico, and G. Zaharijas, JCAP 1410, 052 (2014), eprint 1405.7928.
- E. Carlson and S. Profumo, Phys.Rev. D90, 023015 (2014), eprint 1405.7685.
- F. Calore, I. Cholis, C. Evoli, D. Hooper, T. Linden, and C. Weniger, in preparation (2015).
- M. Ackermann et al. (Fermi-LAT) (2015), eprint 1503.02641.
- A. Geringer-Sameth, S. M. Koushiappas, and M. G. Walker (2014), eprint 1410.2242.
- M. Ackermann et al. (Fermi-LAT Collaboration), Phys.Rev. D89, 042001 (2014), eprint 1310.0828.
- C. He, K. Bechtol, A. P. Hearin, and D. Hooper (2013), eprint 1309.4780.
- E. J. Tollerud, J. S. Bullock, L. E. Strigari, and B. Willman, Astrophys.J. 688, 277 (2008), eprint 0806.4381.
- J. R. Hargis, B. Willman, and A. H. G. Peter, Astrophys.J. 795, L13 (2014), eprint 1407.4470.
- B. M. Rossetto et al. (Dark Energy Survey Collaboration), Astron.J. 141, 185 (2011), eprint 1104.4718.
- K. Bechtol et al. (The DES Collaboration) (2015), eprint 1503.02584.
- S. E. Koposov, V. Belokurov, G. Torrealba, and N. W. Evans (2015), eprint 1503.02079.
- B. P. Laevens et al. (2015), eprint 1503.05554.
- A. Drlica-Wagner et al. (Fermi-LAT, DES) (2015), eprint 1503.02632.
- A. Geringer-Sameth, M. G. Walker, S. M. Koushiappas, S. E. Koposov, V. Belokurov, et al. (2015a), eprint 1503.02320.
- F. Acero et al. (Fermi-LAT Collaboration) (2015), eprint 1501.02003.
- G. D. Martinez (2013), eprint 1309.2641.
- A. Geringer-Sameth, S. M. Koushiappas, and M. Walker, Astrophys.J. 801, 74 (2015b), eprint 1408.0002.
- E. Carlson, D. Hooper, and T. Linden, Phys.Rev. D91, 061302 (2015), eprint 1409.1572.
- E. Massaro, P. Giommi, C. Leto, P. Marchegiani, A. Maselli, et al. (2008), eprint 0810.2206.
- S. E. Healey, R. W. Romani, G. B. Taylor, E. M. Sadler, R. Ricci, et al., Astrophys.J.Suppl. 171, 61 (2007), eprint astro-ph/0702346.
- S. E. Healey, R. W. Romani, G. Cotter, P. F. Michelson, E. F. Schlafly, et al., Astrophys.J.Suppl. 175, 97 (2008), eprint 0709.1735.
- G. Hobbs, R. Manchester, A. Teoh, and M. Hobbs, IAU Symp. 218, 139 (2004), eprint astro-ph/0309219.
- F. Iocco, M. Pato, G. Bertone, and P. Jetzer, JCAP 11, 029 (2011), eprint 1107.5810.
- R. Catena and P. Ullio, JCAP 8, 004 (2010), eprint 0907.0018.