Deep XMM Observations of Draco rule out at the 99% Confidence Level a Dark Matter Decay Origin for the 3.5 keV Line
aa r X i v : . [ a s t r o - ph . H E ] M a r MNRAS , 000–000 (0000) Preprint 16 August 2018 Compiled using MNRAS L A TEX style file v3.0
Deep XMM Observations of Draco rule out at the 99% ConfidenceLevel a Dark Matter Decay Origin for the 3.5 keV Line
Tesla Jeltema ⋆ and Stefano Profumo † Department of Physics and Santa Cruz Institute for Particle Physics University of California, Santa Cruz, CA 95064, USA
16 August 2018
ABSTRACT
We searched for an X-ray line at energies around 3.5 keV in deep, ∼ . Msec XMM-Newtonobservations of the dwarf spheroidal galaxy Draco. No line was found in either the MOSor the PN detectors. The data in this energy range are completely consistent with a single,unfolded power law modeling the particle background, which dominates at these energies,plus instrumental lines; the addition of a ∼ . keV line feature gives no improvement to thefit. The corresponding upper limit on the line flux rules out a dark matter decay origin for the3.5 keV line found in observations of clusters of galaxies and in the Galactic Center at greaterthan 99% C.L.. Key words:
X-rays: galaxies; X-rays: galaxies: clusters; X-rays: ISM; line: identification;(cosmology:) dark matter
The detection of a line with an energy between 3.50 – 3.57keV (hereafter indicated as “the 3.5 keV line” for brevity)in the X-ray data from individual and stacked observationsof clusters of galaxies (Bulbul et al. 2014), from the Galac-tic center (Jeltema & Profumo 2015) and, tentatively, from M31(Boyarsky et al. 2014) (see however Jeltema & Profumo 2015;Jeltema & Profumo 2014) has triggered widespread interest: theline might be associated with a two-body radiative decay includ-ing one photon of a dark matter particle with a mass of around 7keV and a lifetime of about − × sec. Such a particle has anatural theoretical counterpart in sterile neutrino models, a class ofdark matter candidates whose motivation goes beyond that of ex-plaining the missing non-baryonic matter in the universe (see e.g.Boyarsky et al. 2009, for a review).Jeltema & Profumo (2015) pointed out early on that atomicde-excitation lines from He-like Potassium ions (K XVIII) are aplausible counterpart to the 3.5 keV line both in clusters of galax-ies and in the Milky Way. This possibility was initially discardedby Bulbul et al. (2014) based on estimates of the required K abun-dance that relied on photospheric K solar abundances, and on multi-temperature models biased towards high temperatures. The latter,as demonstrated in Jeltema & Profumo (2014), artificially suppressthe brightness of the K XVIII de-excitation lines by up to morethan one order of magnitude. Additionally, coronal K abundancesare larger by about one order of magnitude than photospheric Ksolar abundances, as recently pointed out by Phillips et al. (2015). ⋆ [email protected] † [email protected] As a result, the case for K XVIII as the culprit for the 3.5 keV lineappears at present quite plausible.Additional circumstantial evidence against a dark matter de-cay origin for the 3.5 keV line has also emerged. Malyshev et al.(2014) searched for the line in stacked, archival XMM observa-tions of dwarf spheroidal galaxies, reporting a null result that highlyconstrained a dark matter decay origin for the line. Anderson et al.(2015) analyzed stacked observations of galaxies and galaxygroups, systems where the thermal emission would be too faint toproduce a detectable line from e.g. K XVIII, and also failed to findany evidence for a 3.5 keV line. Urban et al. (2015) studied Suzakudata from X-ray-bright clusters, confirming that the 3.5 keV signalcould naturally be ascribed to K, and questioning the compatibil-ity of the line morphology with the dark matter decay hypothesis.Finally, Carlson et al. (2015) studied in detail the morphology ofthe 3.5 keV emission from the Perseus cluster of galaxies and fromthe Galactic center, finding a notable correlation with the morphol-ogy of bright elemental emission lines, and excluding a dark matterdecay origin even for cored Galactic dark matter density profiles.A recent study of charge exchange processes indicates that an ad-ditional possibility is that the 3.5 keV line originates from a setof high- n S XVI transitions (populated by charge transfer betweenbare sulfur ions and neutral hydrogen) to the ground state (Gu et al.2015).It is important to acknowledge that null results obtained sofar are still compatible with a non-standard origin for the 3.5keV line. Notably, axion-like particle conversion in magnetic fields(Cicoli et al. 2014; Alvarez et al. 2015) could reproduce the mor-phology of the 3.5 keV line in Perseus reported in Carlson et al.(2015); other possibilities include, for example, inelastic exciteddark matter (Finkbeiner & Weiner 2014). In all such instances, thesignal strength scales non-trivially with the integrated dark matter c (cid:13) T. Jeltema and S. Profumo no r m a li z ed c oun t s s − k e V − −3 −3 no r m a li z ed c oun t s s − k e V − Energy (keV) 00.10.20.3 no r m a li z ed c oun t s s − k e V − −3 −3 no r m a li z ed c oun t s s − k e V − Energy (keV)
Figure 1.
Left : Combined MOS spectrum and residuals in the 2.5-5.0 keV energy range fit to an unfolded, single power law. For visual effect here the spectrumhas been binned by a factor of five.
Right:
Combined PN spectrum and residuals (with a similar factor of five binning) in the 2.5-5.0 keV energy range fit to anunfolded single power law plus an instrumental line due to Ti K α emission at 4.51 keV. A weaker Ti K β line can be seen at 4.93 keV but has no effect on the3.5 keV line constraints. mass along the line of sight, or it depends sensitively on astrophys-ical conditions such as the magnetic field strength.Lovell et al. (2015) used N-body simulations from the Aquar-ius project (Springel et al. 2008) to estimate the flux ratio for astandard dark matter decay process across different targets, includ-ing for the Draco dwarf spheroidal galaxy (dSph) and the Galacticcenter (GC). This ratio has a certain statistical distribution, whichdepends on the choice of the placement of the observer. The centralfinding of that study is that a 1.3 Ms long XMM-Newton observa-tion of the Draco dSph would enable the discovery or exclusion atthe 3 σ level of a dark matter decay interpretation of the 3.5 keVsignal.Here, we utilize recent, deep archival XMM-Newton observa-tions of the Draco dSph to test a dark matter decay origin for the3.5 keV line. We find no evidence of a line in either the MOS orPN data, and we are able to rule out a dark matter decay origin atgreater than the 99% confidence level.The remainder of this manuscript has the following structure:we describe the XMM observations and data reduction in the fol-lowing section 2; we then describe our flux calculation and comparewith the flux limits from the XMM MOS and PN data in section 3,and we present our conclusions in the final section 4. Draco was observed by XMM-Newton in 31 separate observations,5 in 2009 (PI Dhuga) and 26 in 2015 (PI Boyarsky), with individ-ual exposure times ranging from 17 to 87 ksec and a total time inall observations of 1.66 Msec. We reprocessed all 31 observations using standard procedures and utilizing the XMM SAS and ESAS(Snowden et al. 2008; Kuntz & Snowden 2008) software packages.Starting from the Observation Data Files, the raw EPIC data waspipeline-processed with the emchain and epchain tasks. Flarefiltering was carried out with the ESAS tasks mos-filter and pn-filter ; these time periods of increased particle backgrounddue to soft protons can lead to background levels elevated by twoorders of magnitude and are thus removed from the data. Unfortu-nately, in the case of Draco particle background flaring was signif-icant in many of the observations. For the two MOS detectors, twoobservations (ObsID 0603190401 and 0770190601) were almostentirely contaminated by flaring, and we removed these from ourfinal data set; the other observations had reduced usable exposuretimes. The net exposure time after filtering was a little over oneMsec for each MOS detector with a total time for both detectorsof 2.1 Msec. The PN detector is typically more effected by particleflaring than the MOS detectors, and we found that only 20 of 31observations had flares satisfactorily removed by pn-filter ; forthese observations the net usable exposure time for PN was 0.58Msec.Point sources were detected and removed separately from eachobservation using the ESAS task cheese ; point source detectionwas run on broad-band images (0.4-7.2 keV) with a flux limit of − erg cm − s − and a minimum separation of 10 arcsec. Lowexposure regions are likewise masked by cheese . Spectra wereextracted from the full field-of-view from each detector in eachflare-filtered observation; however, for the MOS1 detector CCDs3 and 6 were excluded due to micrometeroid damage. Spectra and http://xmm.esac.esa.int/sas/ MNRAS , 000–000 (0000) eep XMM Observations of Draco rule out a dark matter decay origin for the 3.5 keV line Line Energy [keV] × -7 × -6 × -6 × -6 × -6 L i n e F l ux [ c t s c m - s - ] Lovell+Wolf+Geringer-Smith+ Bulbul+ PN Bulbul+ MOSBoyarsky+ M31Boyarsky+ Perseus99% C.L.90% C.L.68% C.L. 3.3 3.4 3.5 3.6 3.7
Line Energy [keV] × -7 × -6 × -6 × -6 × -6 × -6 × -6 L i n e F l ux [ c t s c m - s - ] Lovell+Wolf+Geringer-Smith+ Bulbul+, PN Bulbul+, MOSBoyarsky+, M31Boyarsky+, Perseus
Figure 2.
Left : Limits on the flux of a line in the energy range between 3.3 and 3.7 keV from MOS observations of the Draco dSph, at the 68%, 90% and99% C.L. (green, red and black lines, respectively) and predictions for the flux of a 3.5 keV line assuming a dark matter decay origin for the line detected atthat energy from stacked clusters of galaxies and from the Milky Way center (see text for details). The horizontal black lines indicate the 1 σ energy rangefor the line position as inferred by Boyarsky et al. 2014 for Perseus ( . ± . keV) and for M31 ( . ± . keV) and by Bulbul et al. 2014 fromcluster observations ( . ± . and . ± . keV for their “full sample” MOS and PN results, resepctively); Right : same, for PN observations (note thedifference in vertical scale). corresponding redistribution matrix files (RMF) and ancillary re-sponse files (ARF) for the 0.4-7.2 keV range were created using mos-spectra and pn-spectra in the ESAS package. The in-dividual spectra and response files were co-added using the rou-tines mathpha , addrmf , and addarf in the FTOOLS package(Blackburn 1995). Combined RMF and ARF files were weighed bythe relative contribution of each observation to the total exposuretime. The spectra and responses for the MOS1 and MOS2 cam-eras were combined in to a single summed MOS spectrum, whilethe spectra and responses for the PN detector were combined sepa-rately.Spectral modeling employed the energy range between 2.5keV and 5 keV. This energy range was chosen to exclude stronginstrumental emission lines while being much, much broader thanthe energy resolution of the detectors ( ∼ eV). At these en-ergies, the X-ray background is dominated by the quiescent parti-cle background (Kuntz & Snowden 2008) which we model with anunfolded, power law (no vignetting) in XSPEC (version 12.8.1p,Arnaud 1996). As shown in Fig. 1, the combined MOS spectrumin the 2.5-5 keV range is well fit by an unfolded, single powerlaw alone with reduced χ = 0 . ( χ = 475 / degrees offreedom). Adding a Gaussian line between 3.4 and 3.6 keV givesno improvement to the fit, and a line at these energies with a fluxgreater than ∼ − photons cm − s − is excluded, as shown inFig. 2. The confidence contours in Fig. 2 are determined based onthe change in the fit statistic when stepping over the relevant pa-rameters using the steppar command in XSPEC . The combinedPN spectrum is well fit by an unfolded power law plus an instru-mental line due to Ti K α emission (4.51 keV), which we model asa narrow Gaussian (Fig. 1, right). The reduced χ for this fit is 0.99( χ = 490 / degrees of freedom). Again, adding a Gaussianline between 3.4 and 3.6 keV gives no significant improvement tothe fit. The fit is somewhat improved by adding a second instru-mental line, Ti K β , at 4.93 keV, but this feature has no effect on the3.5 keV line constraints. As can be seen from Fig. 2, right the upperlimit on the flux of a line near 3.5 keV from the PN data is weakerthan from the MOS data given the shorter usable exposure time butdoes serve as additional confirmation of the lack of a 3.5 keV linefrom Draco. We utilize three distinct predictions for the 3.5 keV line flux thatshould have been observed with the XMM observations describedabove for a dark matter decay origin . The first one makes use of theresults of Lovell et al. (2015), which calculated the flux expectedfrom a 14 arcmin angular region around Draco given the brightnessof the 3.5 keV line observed from the Galactic Center and the ratioof the flux from Draco-like halos and from the Galactic center asextrapolated from the Aquarius simulation. The resulting distribu-tion in predictions is bracketed by the range F = (1 . − . × − cts cm − s − , where the lower and upper values bracket 95% of the predic-tions, and with the most-probable value being F = 2 . × − cts cm − s − (see especially their Appendix C3 for addi-tional details on assumptions and method). We calculated that thepoint source masking we adopt and the non-uniform coverage (e.g.from the lost MOS CCDs and chip gaps) described in the previoussection suppress the predicted flux to 77% of its un-masked value(we neglect the additional signal from the annulus between 14 and15 arcmin). We calculated the fraction of masked signal employ-ing the dark matter density profile and distance to Draco quotedin Abdo et al. (2010). We verified that the impact on the maskingfraction of varying the parameters in the dark matter density profileand the distance within their 2- σ range is in all cases smaller than0.1%. We show the resulting range of expected fluxes in Fig. 2 withthe vertical blue bar, and indicate the central value with a full bluesquare.We additionally considered two alternate predictions for theflux expected in the data we analyze for the decaying dark mat-ter scenario. We followed the procedure outlined in Malyshev et al.(2014). There, two alternate determinations of the mass within half-light radius were adopted to estimate the integrated line of sightdark matter column density for Draco, based on the results of theanalyses of Wolf et al. (2010) and Geringer-Sameth et al. (2015).Malyshev et al. (2014) then proceeded to add, for the direction ofDraco, the flux from dark matter within the Milky Way, and addedto it the prediction for the component from the Draco dSph itself. MNRAS000
Left : Limits on the flux of a line in the energy range between 3.3 and 3.7 keV from MOS observations of the Draco dSph, at the 68%, 90% and99% C.L. (green, red and black lines, respectively) and predictions for the flux of a 3.5 keV line assuming a dark matter decay origin for the line detected atthat energy from stacked clusters of galaxies and from the Milky Way center (see text for details). The horizontal black lines indicate the 1 σ energy rangefor the line position as inferred by Boyarsky et al. 2014 for Perseus ( . ± . keV) and for M31 ( . ± . keV) and by Bulbul et al. 2014 fromcluster observations ( . ± . and . ± . keV for their “full sample” MOS and PN results, resepctively); Right : same, for PN observations (note thedifference in vertical scale). corresponding redistribution matrix files (RMF) and ancillary re-sponse files (ARF) for the 0.4-7.2 keV range were created using mos-spectra and pn-spectra in the ESAS package. The in-dividual spectra and response files were co-added using the rou-tines mathpha , addrmf , and addarf in the FTOOLS package(Blackburn 1995). Combined RMF and ARF files were weighed bythe relative contribution of each observation to the total exposuretime. The spectra and responses for the MOS1 and MOS2 cam-eras were combined in to a single summed MOS spectrum, whilethe spectra and responses for the PN detector were combined sepa-rately.Spectral modeling employed the energy range between 2.5keV and 5 keV. This energy range was chosen to exclude stronginstrumental emission lines while being much, much broader thanthe energy resolution of the detectors ( ∼ eV). At these en-ergies, the X-ray background is dominated by the quiescent parti-cle background (Kuntz & Snowden 2008) which we model with anunfolded, power law (no vignetting) in XSPEC (version 12.8.1p,Arnaud 1996). As shown in Fig. 1, the combined MOS spectrumin the 2.5-5 keV range is well fit by an unfolded, single powerlaw alone with reduced χ = 0 . ( χ = 475 / degrees offreedom). Adding a Gaussian line between 3.4 and 3.6 keV givesno improvement to the fit, and a line at these energies with a fluxgreater than ∼ − photons cm − s − is excluded, as shown inFig. 2. The confidence contours in Fig. 2 are determined based onthe change in the fit statistic when stepping over the relevant pa-rameters using the steppar command in XSPEC . The combinedPN spectrum is well fit by an unfolded power law plus an instru-mental line due to Ti K α emission (4.51 keV), which we model asa narrow Gaussian (Fig. 1, right). The reduced χ for this fit is 0.99( χ = 490 / degrees of freedom). Again, adding a Gaussianline between 3.4 and 3.6 keV gives no significant improvement tothe fit. The fit is somewhat improved by adding a second instru-mental line, Ti K β , at 4.93 keV, but this feature has no effect on the3.5 keV line constraints. As can be seen from Fig. 2, right the upperlimit on the flux of a line near 3.5 keV from the PN data is weakerthan from the MOS data given the shorter usable exposure time butdoes serve as additional confirmation of the lack of a 3.5 keV linefrom Draco. We utilize three distinct predictions for the 3.5 keV line flux thatshould have been observed with the XMM observations describedabove for a dark matter decay origin . The first one makes use of theresults of Lovell et al. (2015), which calculated the flux expectedfrom a 14 arcmin angular region around Draco given the brightnessof the 3.5 keV line observed from the Galactic Center and the ratioof the flux from Draco-like halos and from the Galactic center asextrapolated from the Aquarius simulation. The resulting distribu-tion in predictions is bracketed by the range F = (1 . − . × − cts cm − s − , where the lower and upper values bracket 95% of the predic-tions, and with the most-probable value being F = 2 . × − cts cm − s − (see especially their Appendix C3 for addi-tional details on assumptions and method). We calculated that thepoint source masking we adopt and the non-uniform coverage (e.g.from the lost MOS CCDs and chip gaps) described in the previoussection suppress the predicted flux to 77% of its un-masked value(we neglect the additional signal from the annulus between 14 and15 arcmin). We calculated the fraction of masked signal employ-ing the dark matter density profile and distance to Draco quotedin Abdo et al. (2010). We verified that the impact on the maskingfraction of varying the parameters in the dark matter density profileand the distance within their 2- σ range is in all cases smaller than0.1%. We show the resulting range of expected fluxes in Fig. 2 withthe vertical blue bar, and indicate the central value with a full bluesquare.We additionally considered two alternate predictions for theflux expected in the data we analyze for the decaying dark mat-ter scenario. We followed the procedure outlined in Malyshev et al.(2014). There, two alternate determinations of the mass within half-light radius were adopted to estimate the integrated line of sightdark matter column density for Draco, based on the results of theanalyses of Wolf et al. (2010) and Geringer-Sameth et al. (2015).Malyshev et al. (2014) then proceeded to add, for the direction ofDraco, the flux from dark matter within the Milky Way, and addedto it the prediction for the component from the Draco dSph itself. MNRAS000 , 000–000 (0000)
T. Jeltema and S. Profumo
For the former, Malyshev et al. (2014) presents a “mean” flux basedon the “favored NFW” Milky Way dark matter halo of Klypin et al.(2002), as well as a very conservative “minimal” model based ona “maximal disk” halo structure. We choose to show our predic-tions for the flux in Fig. 2 for the mean flux from the Milky Way,but it is a straightforward exercise (and one that does not impactour results or conclusions qualitatively) to rescale them for theminimal Milky Way model. The predicted flux was normalized,as in Malyshev et al. (2014) and in Lovell et al. (2015), to the pa-rameters corresponding to the best-fit point of the cluster observa-tions of Bulbul et al. (2014). We show the predictions for the Dracohalo determination in Wolf et al. (2010) with brown diamonds andGeringer-Sameth et al. (2015) with purple triangles. The uncertain-ties we show in the figure reflect the uncertainties from the determi-nation of the halo parameters as in Malyshev et al. (2014). As wedid for the predictions from Lovell et al. (2015), we corrected thepredicted flux quoted in Malyshev et al. (2014) for masking in ourobservations, and for the larger, in this case, angular region we uti-lize compared to the flux predictions in Malyshev et al. (2014). Weillustrate our results in Fig. 2. We indicate with green, red and blacklines the 68%, 90% and 99% Confidence Level (C.L.) limits on themaximal allowed flux associated with a line at the energy indicatedby the x-axis. We also show the predictions for the line flux de-scribed above as well as the range of energies for the line reportedfrom cluster observations as described in Bulbul et al. (2014) andthe range obtained by Boyarsky et al. (2014) from observations ofthe Perseus cluster and of M31.As can be seen in Fig. 2, the lack of a detected line in theMOS data rule out at higher than 99% confidence level a linewith even the most conservative predicted fluxes based on a con-servative range of possible density profiles for the Draco dwarf.Specifically, the lower limit on the predicted Draco line flux ofLovell et al. (2015) based on the brightness of the line observedin the Galactic Center is excluded at 99.1%; the lower limit on thepredicted flux given the observed stacked cluster line flux are ex-cluded at better than 99.999% for either the Wolf et al. (2010) or theGeringer-Sameth et al. (2015) profiles. Therefore, a generic darkmatter decay origin of the 3.5 keV line feature is highly unlikely.In Fig. 3, we show constraints on the sterile neutrino param-eter space in terms of the particle’s mass m s and mixing anglewith active neutrinos θ given the line flux limits from the MOSDraco observations, in the relevant mass range for a dark mat-ter decay interpretation of the 3.5 keV line. The cyan shaded re-gion is excluded at the 2 σ level, and assumes the central valuefor the Geringer-Sameth et al. (2015) dark matter halo parame-ters for Draco, and the most conservative Milky Way halo con-sidered in Malyshev et al. (2014) (corresponding to the “maximaldisk model” of Klypin et al. (2002)). The blue shaded region, in-stead, adopts the default “favored NFW” Milky Way dark matterhalo density profile (Klypin et al. 2002).Taking the most conservative possible assumptions both forthe flux from Draco and from the Milky Way Galactic halo (corre-sponding to the predictions of Wolf et al. (2010) for the flux fromDraco and the most conservative Milky Way halo of Klypin et al.(2002)), we are able to set a lower limit on the lifetime of a 7 keVsterile neutrino decaying into a 3.5 keV line of τ > . × s (95% C.L.), corresponding to a mixing angle sin (2 θ ) < . × − (95% C.L.). Our most conservative limits are thus more thana factor 4 below the favored mixing angle predicated by a darkmatter decay interpretation of the 3.5 keV line signal ( sin (2 θ ) ≈ × − , Bulbul et al. 2014).After our paper appeared, Ruchayskiy et al. (2015) analyzed essentially the same data set as this work. They come to a similarconclusion that no 3.5 keV line is detected in Draco, though theirlimit on the flux of the line is less stringent than ours. The maindifference appears to stem from the fact that they find a mild ex-cess in the PN data which we do not see. Here we comment brieflyon some of the differences in the two analyses. Ruchayskiy et al.(2015) make three comments on our analysis. 1) We do not includean extragalactic background power law in addition to the particlebackground; 2) the ∼ σ excesses in the MOS spectrum near 3.35keV and 3.7 keV might be due to un-modeled weak instrumentallines (from K K α and Ca K α at 3.31 and 3.69 keV); 3) we do notjointly fit the MOS and PN data. We have now tried all three ofthese variations and find that they all have a negligible effect onour results. Adding additional lines or background components asin points 1) and 2) is not warranted by the data nor does it signifi-cantly improve our fits when done. In fact, we find that when we doadd these components our flux limit on a line near 3.5 keV in theMOS data is actually strengthened , albeit slightly, lowering the fluxlimit by 10%. We note that when adding an additional power lawfor the extragalactic background in addition to the unfolded powerlaw for the particle background, we expanded the energy range to2.5-7 keV and added lines for the strong Cr, Mn, and Fe instrumen-tal lines in this range, but again we find no excess near 3.5 keV.Here the power law components are free to vary; the best fit pho-ton indices are 1.5 for the extragalactic power law and 0.33 for theunfolded power law with a reduced χ of 0.99 ( χ = 876 / degrees of freedom).Results from combined fits to the MOS and PN data mustbe interpreted with care as there are known offsets in the relativecalibration of the two camera types. In particular, fits to a stackof bright sources show an offset of 5-8% between MOS and PN(Read et al. 2014) above 3 keV, which is similar to or larger thanthe ratio of the predicted 3.5 keV line flux to the neighboring con-tinuum. In performing a joint fit to the MOS and PN spectra, weseparately fit the continuum models for the two instruments, butjointly fit the 3.5 keV line energy and flux. We again find no excessnear 3.5 keV. Our limit on the flux of a line near 3.5 keV is weak-ened only slightly (by 25-30%) from the MOS-only limit shown inthe left panel of Fig. 2, and we still exclude the most conservativepredictions from Lovell et al. (2015) at the 99% confidence level.The primary reason that Ruchayskiy et al. (2015) quote aweaker limit on the flux of a line from Draco appears to be dueto the fact that they find a ∼ σ excess in the PN which is notpresent in our analysis. It is unclear why they find an excess wherewe do not, but we note that in their spectral fits they include dataup to 10 keV where the effective area is both rapidly dropping anda factor of ∼ lower than at 3 keV for PN, and more than an orderof magnitude lower for MOS. In addition, their background model,given the large energy range, includes 13 line components and twopower laws compared to our single power law. Using ∼ . Msec observations of the Draco dSph with XMM-Newton we were able to obtain one of the most stringent constraintson a dark matter decay origin for the 3.5 keV line observed fromclusters of galaxies and from the Milky Way center. Our results ruleout a dark matter decay interpretation with greater than 99% C.L.,and, under very conservative assumptions on the relevant dark mat-ter density profiles, imply a lower limit on the dark matter lifetime
MNRAS , 000–000 (0000) eep XMM Observations of Draco rule out a dark matter decay origin for the 3.5 keV line m s [keV] -11 -10 s i n ( θ ) MOS, full sample (Bulbul+)PN, full sample (Bulbul+)
Figure 3.
Constraints on the parameter space of sterile neutrinos, defined bythe particle’s mass m s and mixing angle with active neutrinos, θ . The cyan-shaded region is excluded, at the 2 σ level ( ∼
95% C.L.), by Draco MOSobservations, using the most conservative Milky Way dark matter densityprofile considered in Malyshev et al. (2014), while the blue-shaded regionemploys the nominal “favored NFW” profile, which we also use for Fig. 2. of τ > . × s at 95% C.L. for a dark matter mass of 7 keVradiatively decaying to a two-body final state with one photon.In view of the results presented here, and in view of the recentre-assessment of the potassium abundance (Phillips et al. 2015),we conclude that the most probable counterpart to the 3.5 keVline observed towards the Milky Way center and from individualand stacked observations of clusters of galaxies are atomic de-excitation lines of the K XVIII ion. Charge-exchange processesmight also provide an alternate astrophysical explanation (Gu et al.2015). Scenarios advocating new physics where a 3.5 keV signal issuppressed in dwarf galaxies, such as an axion-like particle conver-sion to 3.5 keV photons in the presence of a magnetic field, are notruled out unless Occam’s razor is advocated.Future observations of clusters and of the Galactic center withAstro-H remain a priority to pinpoint the physical origin and thenature of the 3.5 keV line, while, in view of our results, additionaldeep observations of local dwarf galaxies with current or futuretelescopes are unlikely to advance our understanding of this partic-ular feature. ACKNOWLEDGMENTS
We would like to thank the referee for their valuable commentson our paper. TJ is partly supported by NSF AST 1517545. SP ispartly supported by the US Department of Energy, Contract DE-SC0010107-001.
REFERENCES
Abdo A. A., et al., 2010, Astrophys. J., 712, 147Alvarez P. D., Conlon J. P., Day F. V., Marsh M. C. D., Rummel M., 2015,JCAP, 1504, 013Anderson M. E., Churazov E., Bregman J. N., 2015, MNRAS, 452, 3905Arnaud K. A., 1996, in Jacoby G. H., Barnes J., eds, Astronomical Society of the Pacific Conference Series Vol. 101, Astronomical Data AnalysisSoftware and Systems V. p. 17Blackburn J. K., 1995, in Shaw R. A., Payne H. E., Hayes J. J. E., eds,Astronomical Society of the Pacific Conference Series Vol. 77, Astro-nomical Data Analysis Software and Systems IV. p. 367Boyarsky A., Ruchayskiy O., Shaposhnikov M., 2009,Ann.Rev.Nucl.Part.Sci., 59, 191Boyarsky A., Ruchayskiy O., Iakubovskyi D., Franse J., 2014,Physical Review Letters, 113, 251301Bulbul E., Markevitch M., Foster A., Smith R. K., Loewenstein M., RandallS. W., 2014, ApJ, 789, 13Carlson E., Jeltema T., Profumo S., 2015, JCAP, 1502, 009Cicoli M., Conlon J. P., Marsh M. C. D., Rummel M., 2014, Phys. Rev. D,90, 023540Finkbeiner D. P., Weiner N., 2014, preprint, ( arXiv:1402.6671 )Geringer-Sameth A., Koushiappas S. M., Walker M., 2015, ApJ, 801, 74Gu L., Kaastra J., Raassen A. J. J., Mullen P. D., Cumbee R. S., Lyons D.,Stancil P. C., 2015, preprint, ( arXiv:1511.06557 )Jeltema T., Profumo S., 2014, preprint, ( arXiv:1411.1759 )Jeltema T. E., Profumo S., 2015, Mon. Not. Roy. Astron. Soc., 450, 2143Klypin A., Zhao H., Somerville R. S., 2002, Astrophys. J., 573, 597Kuntz K. D., Snowden S. L., 2008, Astronomy & Astrophysics, 478, 575Lovell M. R., Bertone G., Boyarsky A., Jenkins A., Ruchayskiy O., 2015,MNRAS, 451, 1573Malyshev D., Neronov A., Eckert D., 2014, Phys. Rev., D90, 103506Phillips K. J. H., Sylwester B., Sylwester J., 2015, Astrophys. J., 809, 50Read A. M., Guainazzi M., Sembay S., 2014, A&A, 564, A75Ruchayskiy O., et al., 2015, preprint, ( arXiv:1512.07217 )Snowden S. L., Mushotzky R. F., Kuntz K. D., Davis D. S., 2008,Astronomy & Astrophysics, 478, 615Springel V., et al., 2008, Mon. Not. Roy. Astron. Soc., 391, 1685Urban O., Werner N., Allen S. W., Simionescu A., Kaastra J. S., StrigariL. E., 2015, MNRAS, 451, 2447Wolf J., Martinez G. D., Bullock J. S., Kaplinghat M., Geha M., Mu˜nozR. R., Simon J. D., Avedo F. F., 2010, Mon. Not. Roy. Astron. Soc.,406, 1220MNRAS000