The Likely Cause of the EGRET GeV Anomaly and its Implications
aa r X i v : . [ a s t r o - ph ] N ov The Likely Cause of the EGRET GeVAnomaly and its Implications
F.W. Stecker
NASA Goddard Space Flight Center, Greenbelt, MD 20771
S. D. Hunter
NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
D. A. Kniffen
NASA Goddard Space Flight Center, Greenbelt, MD 20771, USAUniversity Space Research Association, Columbia, MD 21044, USA
Abstract
Analysis of data from the EGRET γ -ray detector on the Compton Gamma RayObservatory indicated an anomaly in the form of an excess diffuse galactic fluxat GeV energies over that which was theoretically predicted. Various explanationsfor this anomaly have been put forth, including the invocation of supersymmetricdark matter annihilation. We reexamine these explanations here, including a newdiscussion of the possible systematic errors in the sensitivity determination of theEGRET detector. We conclude that the most likely explanation of the EGRET“GeV anomaly” was an error in the estimation of the EGRET sensitivity at energiesabove ∼ γ -raybackground spectrum. There are important implications of our analysis for theupcoming Gamma Ray Large Area Telescope (GLAST) mission. Key words: γ -rays , background radiation The EGRET γ -ray detector, a spark chamber telescope flown aboard theCompton Gamma Ray Observatory satellite, provided detailed measurements Preprint submitted to Elsevier Preprint 27 October 2018 f astrophysical γ -ray sources and galactic and extragalactic diffuse γ -rayfluxes. The reported galactic diffuse flux was measured and mapped over thewhole sky.Theoretical studies of the physics and astrophysics of galactic γ -ray productionprovided predictions of the expected fluxes and energy spectra [1]. In theenergy range above ∼ ∼
60% higher than the theoretical predictions [2]. This apparent discrepancybetween the reported fluxes and the theoretical calculations will be referredto here as the “GeV anomaly”.Unresolved galactic point sources can be ruled out as an explanation for theGeV anomaly for two reasons: (1) they would be strongly concentrated in thegalactic plane and the GE anomaly, as we will show, is seen isotropically overthe whole sky, and (2) the largest class of galactic point sources are pulsarsand such sources are concentrated in the inner galaxy and they make up lessthan 15% of the total galactic flux [3].There have been three approaches for accounting for the GeV anomaly as adiffuse phenomenon, viz. either (1) invoking a cosmic-ray electron source spec-trum proportional to E − with a resulting significant Compton γ -ray com-ponent above 1 GeV increasing the flux [4], (2) making modifications to theassumptions regarding both the primary cosmic-ray nucleon and electron spec-tra in numerical models in order to push up the total theoretical γ -ray fluxes[5], or (3) postulating new physics, namely supersymmetric dark matter an-nihilation, to account for the anomaly [6].The first approach postulates that the cosmic-ray electron spectrum observedat Earth is much steeper than the average spectrum in the Galaxy, but thatthis ad hoc situation can result as an effect of the distribution of the supernovaremnants (SNR) which produce the electrons and electron propagation effects.A prediction of this model is a center-anticenter asymmetry in the anomalyowing to the strong galactocentric radial distribution of SNR. A reduced bumpat high galactic latitudes would also be expected, owing to the steeper localcosmic-ray electron spectrum.Approaches (2) and (3) have significant implications. Increasing the galactic γ -ray production rate at GeV energies in the model of Ref. [5] produced a re-duction in the implied extragalactic diffuse flux in the followup calculation [7],with a marked dip at energies near ∼ γ -ray component fromsupersymmetric dark matter annihilation has, of course, much greater implica-tions. It is therefore important to reexamine the issue of the GeV anomaly bypaying careful attention to the collateral implications of deviations from the2anonical predictions [1] and to examine the more prosaic possibility that theassumed high-energy sensitivity of the EGRET detector may not have beencorrect. In this paper, we will make the case that a problem in the analysisof the EGRET sensitivity calibration is the most likely explanation for the“GeV anomaly” and that the dark matter hypothesis can be ruled out by anexamination of the GeV spectrum over the whole sky. γ -rays Galactic γ -rays are produced by interactions of relativistic electrons and pro-tons with interstellar gas and photons. The physical processes involved areelectron bremsstrahlung, neutral pion production, and Compton interactionsof cosmic ray electrons [9],[10],[1]. Of these processes, the decay of neutralpions produced by cosmic rays interacting with interstellar gas is expected todominate at energies above 0.1 GeV [1],[2]. The vast majority of these γ -raysare produced by cosmic rays with energies below ∼
20 GeV [11]. In this energyrange the cosmic ray spectrum, particularly in the local galactic neighborhoodthat accounts for the γ -ray production at high galactic latitudes, is well mea-sured [12]. The pion production cross section at these energies is also very wellknown [11],[13]. At γ -ray energies up to ∼ ∼ γ -rayspectrum will have the same spectral index as the primary proton spectrum[15],[16]. More recent calculations [18] are in good agreement with the resultspresented in Refs. [15] and [16], confirming that the physics of pion productionat GeV energies is well determined. A modification of this physics involvingscaling violation has been suggested [17], however the resulting effect is toosmall to explain the GeV anomaly by scaling violation alone.It has been suggested that the GeV anomaly can be explained by postulatingthat the average primary proton spectrum in the galaxy is harder than thatobserved locally. With scaling assumed, the required proton spectral indexis Γ = 2 .
45 [18],[19]. Even with some scaling violation [17], one requires aspectral index Γ ≃ .
5. The problem with this assumption is that the localproton spectrum has a significantly steeper measured index Γ = 2 . ± . ∼ o from the plane, the diffuse γ -raysare produced by these locally measured protons (and a smaller number ofcosmic-ray α particles which have a similar spectral index [12]). Even shouldthe proton spectrum in the inner galaxy be harder than the local spectrumat high galactic latitudes, it is the steeper local spectrum which produces thehigh-latitude pion-decay γ -rays. Therefore, the predicted spectrum should bethe canonical one. In other words, there should be no GeV anomaly at highgalactic latitudes. This conclusion would also result if the anomaly is strictly3roduced by a harder cosmic-ray electron spectrum averaged over the wholegalactic disk [4].On the contrary, it is generally accepted that the GeV anomaly is uniformover the entire sky , a result which was obtained from the EGRET data andwhich was implicit in Figs. 3 and 5 of Ref. [8] by the EGRET group. We willshow here, in a very quantitative way, that the anomaly is seen at all galacticlatitudes independent of the line-of-sight density of galactic hydrogen gas. Wewill further argue that such complete isotropy is what would be expected ifthis anomaly is most likely traceable to the detector itself. Our detailed analysis of the galactic γ -ray flux has now been expanded tocover the entire sky [22]. This was accomplished by extending the latituderange of the galactic plane analysis to ± o , chosen to match the latituderange which includes all significant emission by molecular hydrogen clouds [21],and further extending the model to the galactic poles [22] by using the all-sky Leiden-Dwingaloo [23] and Instituto Argentino de Radioastronoma (IAR)Southern Hemisphere HI survey [24]. The three dimensional cosmic-ray den-sity for the latitude range | b | < o was derived from the Galactic plane matterdistribution on the assumption of dynamic balance [25]. The γ -ray productionfunction per H atom in the 22 o < | b | < o latitude range was extended to thepoles on the assumption that γ -ray production in this range is dominated byinteractions with cosmic rays having the locally measured spectrum. The anal-ysis approach used in Ref. [8] to determine the extragalactic diffuse emissionwas then repeated using the new all-sky model for each of the ten standardenergy ranges as well as the four broad energy ranges discussed in Ref. [2].This analysis, using the EGRET phase 1 and phase 2 source-subtracted data,yields an extragalactic γ -ray spectrum which is consistent with the one derivedin Ref. [8]. This consistency is expected because the extragalactic flux is theresidual observed flux as the galactic model flux is extrapolated to zero. Theslope of the extrapolation does not affect the intercept, as shown in Figure 1.Furthermore, this all-sky analysis confirms that the GeV anomaly is uniformover the entire sky , a result which was implicit in Figs. 3 and 5 Ref. [8], butwhich is now more quantitatively shown in Fig. 1. The lack of any structure inthe anomaly related to the Galactic plane, galactic center, anti-center or halo,strongly indicates that the GeV anomaly is due to a systematic error in theEGRET calibration (see below) rather than being a real astronomical effect.The excellent 1:1 correlation of the EGRET data with the expanded all-skymodel over the entire sky and the entire range of emission, after multiplica-tion by a single renormalization factor, further corroborates this conclusionas also shown in Fig. 1. Table 1 and Fig. 2 breaks down this effect as a func-4ion of energy which indicates a miscalibration which gets worse at the higherenergies.The overall effect of applying this renormalization correction for point sourcesis slight over a large energy range on a log-log plot and does not conflict withany theoretical models within the large observational error bars of the dataabove 1 GeV. For γ -ray blazars observed by EGRET, the fit to power-lawspectra would result in slightly steeper spectra, with the error bars on specificpoints generally large in the energy range above 1 GeV.To specifically illustrate the effect of our renormalization factors on the spectraof point sources, we have applied our renormalization to the EGRET data onboth the pulsed and unpulsed nebula spectrum of the Crab Nebula as shownin Figure 3 based on the compilation in Ref. [26]. The renormalized spectrumis consistent with previous theoretical models of γ -ray emission from the CrabNebula [26],[27]. It has been suggested that the GeV anomaly is the result of a new componentto the diffuse γ -ray spectrum, one that is in addition to the pion-decay compo-nent, namely a component from supersymmetric dark matter annihilation [6].However, there are serious problems with this hypothesis: (1) The same processof dark matter annihilation would produce a flux of cosmic-ray antiprotonswhich is incompatible with the measured value [28]. (2) The celestial γ -raydistribution from dark matter annihilation would be both highly asymmetricand clumped [29]. This is again in direct contradiction with isotropy of theanomaly. It was pointed out in the EGRET calibration paper [30] that at high energiesthe EGRET sensitivity is poorly known, partly due to the uncertain effectof self-vetoing caused by charged-particle backsplash impinging onto the an-ticoincidence scintillator following shower generation in the detector [8]. Thiseffect eliminates good γ -ray events. Any uncertainty in the self-veto correctionwould alter the sensitivity of the detector at energies above ∼ γ -ray beam ex-posure. The beam was scanned across the face of the tracking detector at aseries of off-axis angles from 0 o to 40 o with three different roll-axis angles and10 different energies. An EGS (Electron Gamma Shower) Monte Carlo (MC)code was used to study the effect of the backsplash under differing conditions,but the fidelity of the MC model was low. It did indicate that the backsplashwas strongly dependent on the charged-particle threshold, which was known tobe strongly variable across the anti-coincidence detector. The photomultipliergains were reduced after the SLAC calibration to mitigate the higher thananticipated backsplash effect. The effect of lowering the gains is substantial,and is not reflected in the sensitivity calibration results as given in Ref. [30].The all-sky pervasiveness of the apparent excess flux above 1 GeV energy(see above) strongly indicates that this anomaly is intrinsic to the detector. Asystematic error in the understanding of the detector response would be onelikely source. In an effort to investigate this possibility, we have reexaminedthe EGRET archival data carefully, selecting only the highest quality EGRETdata. The selection was based on: (1) choosing γ -ray events with the most reli-able energy determination, (2) selecting γ -ray events within 30 o of the detectoraxis for which the instrument response is most reliably determined, and (3) ex-cluding γ -ray events within a 4 σ angular uncertainty from the Earth’s horizon.We have examined data on the pulsed and unpulsed emission from the Crab,Vela, and Geminga and the diffuse flux from the Lockman Hole for phases1 – 4 of the EGRET data for energies > GeV , We find that for differentobservations of these sources, there were variations in their measured fluxesin excess of 40%. Since the flux from these sources is expected to be constant,this indicates systematic errors caused apparent time variations over differentobserving periods. This is not surprising, since the effect of degraded perfor-mance with the aging of the spark chamber gas was calibrated in flight byfixing the >
100 MeV flux and not taking account of the energy dependenceof this effect [31], thus producing a false variability at GeV energies.While we find anomalous behavior pointing to an imprecise knowledge of thedetector response, we are unable to explain the GeV anomaly as caused by anysingle systematic instrument effect that we have studied. Other contributorsbesides an imprecise knowledge of the detector response that could explainan apparent trans-GeV anomaly include (1) an error in the analysis whichmight cause non- γ -ray events to be identified as good high-energy events, (2)an incorrect assignment of energies and (3) an induced detector backgroundin the harsh environment of a low-Earth orbit. It is difficult to eliminate anyof these possibilities, but no specific evidence has been found that they exist.Given the systematic energy dependent uncertainties in the sensitivity cal-ibration for the EGRET instrument discussed in this paper, the publishedEGRET results above 1 GeV cannot be reliably depended upon. We suggest6 able 1INVERSE NORMALIZATION FACTOR VS.ENERGYEnergy Bin Inverse Normalization Factor30-50 MeV 0.815 ± ± ± ± ± ± ± ± ± ± that one should incorporate the energy dependent correction factor shown inFigure 2. This suggestion applies to possible use by the GLAST LAT (LargeArea Telescope) collaboration for comparison as a sensitivity calibration checkon their extensive Monte Carlo simulations and beam tests. We must awaitresults from GLAST in order to pursue this question further. Our new all-sky analysis of the EGRET data confirms a systematic anomalyin the form of an apparent excess galactic emission at energies above 1 GeVover what is expected from galactic cosmic ray interactions with gas nucleiin atomic and molecular clouds and interstellar photons, as previously re-ported [2]. We find further quantitative support that this anomaly is constantover the whole sky, not being correlated with any astronomical or galacticfeatures. It is therefore most likely caused by a systematic error in the cali-bration of the effective detector sensitivity. Although a detailed reanalysis ofthe calibration data is impossible at this time, we have shown that plausiblesystematic uncertainties in the calibration of the EGRET sensitivity for γ -rayenergies above 1 GeV can readily account for the universal anomalous excessflux. We conclude that neither making modifications to the observed primarycosmic-ray nucleon and electron spectra in order to push up the predictedtheoretical γ -ray flux [5] nor postulating a component from supersymmetricdark matter annihilation [6] are required to account for the “GeV anomaly”.7ur new analysis is consistent with the extragalactic flux derived in Ref. [8].This reevaluation of the EGRET sensitivity also has implications for calibra-tion checks of the GLAST γ -ray detector to be launched in the near future. Acknowledgments
We thank Michael Salamon and Dave Thompson for their comments and sug-gestions regarding this work and Zev Gurman for his analysis of EGRETarchival data.
References [1] F. W. Stecker, Astrophys. J. , 60 (1977).[2] S. D. Hunter et al. , Astrophys. J. , 205 (1997).[3] A. K. Harding and F. W. Stecker, Nature , 316 (1981).[4] M. Pohl and J. A. Esposito, Astrophys. J. , 327 (1998).[5] A. W. Strong, I. V. Moskalenko, and O. Reimer, Astrophys. J. , 962 (2004).[6] W. De Boer, Astron. and Astrophys. , 51 (2005).[7] A. W. Strong, I. V. Moskalenko, and O. Reimer, Astrophys. J. , 956 (2004).[8] P. Sreekumar et al. , Astrophys. J. , 523 (1998).[9] F. W. Stecker, Astrophys. and Space Sci. , 377 (1970).[10] F. W. Stecker, Cosmic Gamma Rays (Baltimore: Mono Pub.) (1971).[11] F. W. Stecker, Astrophys. J. , 499 (1973).[12] Y. Shikaze et al. , Astropart. Phys. , 154 (2007).[13] V. de Ata´ıde Malafaia Lopos dos Santos, e-print astro-ph/nucl-th/0602026. [14] C. D. Dermer, Astron. and Astrophys. 157 (1986) 223.[15] F. W. Stecker, in Cosmic Gamma Rays, Neutrinos, and Related Astrophysics ed. M. M. Shapiro and J. P. Wefel (Dordrecht: Kluwer), 85 (1989).[16] F. W. Stecker, Astrophys. J. , 919 (1979).[17] T. Kamae et al. , Astrophys. J. , 692 (2006).[18] M. Mori, Astrophys. J. , 225 (1997).
19] P. Gralewicz, J. Wdowczyk, A. W. Wolfendale and L. Zhang, Astron. andAstrophys. , 925 (1997).[20] In Ref.[5] the approach is taken of arbitrarily raising the cosmic-ray protonand electron spectra by different amounts in order to optimize the fit to thereported EGRET GeV anomaly. We see no basis for these ad hoc assumptions,particularly as the proton flux is well measured [12].[21] T. M. Dame et al. , Astrophys. J. , 706 (1987).[22] S. D. Hunter, D. A. Kniffen and F. W. Stecker, in preparation.[23] D. Hartmann and B. W. Burton 1997, Atlas of Galactic Neutral Hydrogen,Cambridge, UK: Cambridge U. Press[24] R. Morras, E. Bajaja, E. M. Arnal, and W. G. L. P¨oppel, Astron. andAstrophys. Sup., , 25 (2000).[25] D. L. Bertsch et al. , Astrophys. J. , 587, (1993).[26] L. Kuiper et al. , Astron. and Astrophys. , 918 (2001).[27] O. C. de Jager et al. , Astrophys. J. , 253 (1996).[28] L. Bergstr¨om, J. Edsj¨o, M. Gustafsson and P. Salati, Jour. Cosm. and Astropart.Phys. 05(2006)006.[29] J. Diemand, M. Kuhlen, and P. Madau, Astrophys. J. , 262 (2007).[30] D. J. Thompson et al.
Astrophys. J. Suppl. , 629 (1993).[31] J. A. Esposito et al. Astrophys. J. Suppl. , 203 (1999).[32] Y. C. Lin, EGRET internal document, unpublished (1992). .0 0.2 0.4 0.6 0.8 1.0 1.2Model Flux [10 -4 ph cm -2 s -1 sr -1 ]0.00.20.40.60.81.01.2 O b s e r v e d F l ux [ - ph c m - s - s r - ] (b) O b s e r v e d F l ux [ - ph c m - s - s r - ] (a) Fig. 1. (a) Plot of integral (
E > o < l < o , | b | < o . (b) A similar plot with a renormal-ization factor of (1 . − applied to the observed flux. In both plots, the integralextragalactic diffuse flux of 1.5 × − cm − s − sr − has been added to the diffusemodel. The dotted line indicates the expected 1:1 relationship between the modeland observed fluxes. Contours show the number of 0.5 o × o pixels containingthe flux indicated within a bin of width 1 . × − cm − s − sr − . The contour val-ues are 10 , , , , ... These plots clearly demonstrate that the GeV anomalyexists uniformly over the whole sky and extends from high to low intensity galacticflux emission. -1 Energy [GeV]0.00.51.01.52.0 R e no r m a li za ti on F ac t o rI nv e r s e Fig. 2. Required inverse renormalization factor for different energy bins, given asthe ratio of observed-to-predicted flux vs. energy. -4 -2 Energy [MeV]10 -6 -5 -4 -3 -2 E x F l ux [ M e V c m - s - ] Nebula
Egret10 -4 -2 Energy [MeV]10 -6 -5 -4 -3 -2 E x F l ux [ M e V c m - s - ] Total Pulse
Egret