The Fermi GeV excess and beyond: New techniques for indirect DM searches Christoph Weniger GRAPPA @ UvA (University of Amsterdam) 13th Rencontres du Vietnam Exploring the Dark Universe 27th July 2017
Overview GeV excess evidence GeV excess interpretations Millisecond pulsars Searches in gamma-ray data Searches in radio SkyFACT: Template regression & image reconstruction 2
The Fermi Galactic center/bulge GeV excess Five years of Fermi LAT data > 1 GeV? Initial claims by Goodenough & Hooper (2009) (see also Vitale & Morselli (2009)) Approved by Fermi LAT collaboration in 2015 Controversial discussion about origin continuous Is this a dark matter annihilation signal? 3
Literature overview Papers that looked at data Goodenough & Hooper, arxiv:0910.2998 Vitale & Morselli, 2009 Hooper & Goodenough, Phys. Lett. B697 (2011) 412 Hooper & Linden, Phys. Rev. D84 (2011) 123005 Boyarsky, Malyshev & Ruchayskiy, Phys. Lett. B705 (2011) 165 Abazajian & Kaplinghat, PRD 86 (2012) 083511 Hooper & Slatyer, Phys. Dark Univ. 2 (2013) 118 Gordon & Macias, Phys. ReV. D88 (2013) 083521 Macias & Gordon, PRD 89 (2014) 063515 Abazajian, Canac, Horiuchi, Kaplinghat, Phys. Rev. D90 (2014) 023526 Cholis, Evoli, Calore, Linden, Weniger, Hooper, JCAP 1512 (2015) 12 Calore, Cholis & Weniger, JCAP 1503 (2015) 038 Zhou, Liang, Huang, Li, Fan, Chang, Phys. Rev. D91 (2015) 123010 Gaggero, Taoso, Urbano, Valli & Ullio, JCAP 1512 (2015) 056 Daylan, Finkbeiner, Hooper, Linden, Portillo et al., Physics of Dark Universe 12 (2016) 1 De Boer, Gebauer, Neumann, Biermann, arxiv:1610.08926 (ICRC 2016 proceedings) Huang, Ensslin & Selig, JCAP 1604 (2016) 030 Carlson, Linden, Profumo, Phys. Rev. D94 (2016) 063504 Bartels, Krishnamurthy, Weniger, Phys. Rev. Lett. 116 (2016) 5 Macis, Gordon, Crocker, Coleman, Paterson, arxiv:1611.06644 Lee, Lisanti, Safdi, Slatyer, Xue, Phys. Rev. Lett. 116 (2016) 5 Ajello et al. 2016, Astrophys. J. 819, 44 Ackermann et al., 2017, Astrophys. J. 840, 43 Ajello et al., 2017, arxiv:1705.00009 (+ a few that I must have missed) Excess is likely DM Excess is there Excess is likely not DM Excess is not there + hundreds of DM theory papers 4
Most analyses are based on template regression + Neutral pion + Bremsstrahlung Inverse Compton + + + + = Point sources DM signal Fermi bubbles, isotropic background, Loop I, Earth limb, Sun,... Data Free parameters: 5
How to get the templates 1) Inject primary CR at sources Carlson+ 2015 2) Propagate them with the code of your choice DRAGON See Carmelo Evoli s talk 3) Interaction with gas & ISRF Strong+ 2000; Porter & Strong 2005; Moskalenko+ 2006; Porter+ 2008 Cold neutral medium Traced by 21 cm line Molecular clouds Traced by CO line 6
Studying background systematics List of caveats: We looked at 60 different Galactic diffusion models Variations of diffusion properties, magnetic fields, ISRF, source distributions, reacceleration, diffusion zone height We throw away spectrum information + CR sources at GC 7
Residuals Calore+ 2015 40 deg x 40 deg Left: Point source mask clearly visible Middle: Residuals at the level of <20% are observed Right: Re-adding the DM template clearly shows an extended excess around the GC 8
Spectra from template fits Calore+ 2015 Syste m plane atics fro m contr ol reg Galactic ions 9
Morphology from spectral fits Huang+ 2015 (using D3PO) Cloud-like component Bubble-like component DM-like component Pixel-by-pixel spectral decomposition: But: other spectra lead to different results De Boer, Gebauer, et al. 2016 10
KISS (Keep it Simple and Stupid) version of the GeV excess Idea: At each latitude, use rescaled 300 MeV emission (- point sources) as bkg template: Bartels & CW Residuals depend only on longitudinal changes in the spectrum Blue: stacked MSP spectrum 11
Results from different groups Ackermann+ 2017 Different groups, different ROIs Daylan+ 14 (GC analysis) Ajello+15 Calore+14 Summary as of 2015 Calore+ 2015 Macias+ 2016, see also Ackermann+17, Casandjian+ 15 12
DM annihilation Calore+ 2014 13
DM fits are self-consistent in different sky regions Calore+ 2014 14
Other contributions to the bulge emission Star formation in central molecular zone Fermi GeV excess: 3x1037 erg/s (e.g. Calore+ 2015) ~5% of star formation in CMZ (e.g. Kruijssen+ 2014) O(1000) SN per Myr 1051 erg/sn & 10-3 lepton efficiency (e.g. Lemoine- Carlson+ '15 Goumard 2012) 3x1037 erg/s injected by SN in the CMZ! Significant challenges: spectrum and morphology Potentially, past activity of the central SMBH Cooling compresses electron spectrum electrons peak at ~30 GeV ICS peaks around 1-3 GeV But: Cooling timescale < Diffusion timescale Cholis+ '15 Diffusion kernel is Gaussian, not inv. Power-law Significant challenges: spectrum and morphology Petrovic+14; Cholis+15 15
Millisecond pulsars MSPs proposed as explanation for GeV excess by Abazajian 11 (anticipated by Wang 05) Contra: (Millisecond) pulsars cannot account for the Hooper+ 2013 Inner Galaxy's GeV Excess See also: Cholis+14; Linden 15; Petrovic+14; Abazajian+14 They are not abundant enough in the Galactic bulge Only MSPs at O(1 kpc) distances can be observed easily. Dynamical models actually suggest that MSPs are distributed similar to what the GeV excess suggests (Brandt & Kocsis '15) Their progenitor systems (LMXB) are not abundantly observed in the bulge Life-cycle of LMXBs are far from understood. Their observed gamma-ray spectrum is not compatible with the GeV excess We demonstrated (Calore+13) that systematics are too large for this statement. Bulge MSPs should have been seen as individual sources, but they haven't We showed (Bartels+15) that gamma-ray observations are not only compatible with the MSP hypothesis, but that they prefer it with high statistical significance. The wavelet fluctuation signal is just gas Masking problematic sky regions with strong gas contributions does not alter results significantly The brightest bulge MSPs should have been seen in radio We showed (Calore+15) that they are instead just around the corner. 16
An observational challenge A signal composed of point sources would appear more speckled than a purely diffuse signal (e.g. DM annihilation) See also Clark+ 16 Credit: Lee+ 2014 Idea: Search for wavelet fluctuations (Bartels+15) Data convolution x Kernel = PSF Wavelet transform See also Lee+15 17
Wavelet transform of inner Galaxy data MSP model used in Monte Carlo Free parameters Total number of sources N Cutoff luminosity L max General observation: enhancement of >3 sigma fluctuations suppression of <2 sigma fluctuations Bartels+ 15 18
Strong support for MSP hypothesis More bulge MSPs 10 sigma detection! 1-4 GeV Expected for bulge MSPs Bartels+ 15 Maximum MSP luminosity [erg/s] Results Can explain ~100% of the GeV excess with MSP population with reasonable cutoff luminosity Resolved component of modeled emission accounts for ~10% of the GeV excess, 90% are extrapolated based on reasonable luminosity function Bartels+ 15 19
Similar results by other groups Ajello+ 17 Lee+ 2015 Standard PSC search for pulsar candidates in Fermi data suggests ~2.7 times as many bulge pulsars as disk pulsars However, ~5x larger cutoff luminosity Non-Poissonian NFW template preferred over Poissonian one suggests that excess emission is due to unresolved sources However, ~5x lower cutoff luminosity 20
Previous searches & current situation Radio searches: Observations since 1980s (mostly Parkes, Arecibo), since 2002 GBT Today*: ~370 MSPs (~240 field, ~130 in globular clusters) [e.g., Stovall+13] From surveys (e.g. Parkes HTRU) From deep observations of globular clusters From radio follow-ups of Fermi LAT sources (~70 MSPs) [Ray+12] MPS searches at the Galactic center are very hard [Marcquart & Kanekar 15] *As of Jan 2016 Gamma-ray searches: Discovery of numerous gamma-ray MSPs came as surprise, but now well established (Abdo+10) MSPs usually appear as unassociatd sources in Fermi LAT data (spectral curvature, non-variable) Follow-up searches required to (1) discover associated radio pulsation and (2) fold ephemerides back into gamma rays At least one MSP found by blind search for gamma-ray pulsation alone For a review see Grenier & Harding 15 [Abdo+ 2013, 2nd Fermi Pulsar catalog] 21
Modeling the radio properties of bulge MSPs Modeled pulsars in x-y plane Predict enhancement of MSP density by several orders of magnitude in the Galactic bulge w.r.t disk Surface density of radio-bright bulge MSPs Varies from ~100 deg-2 to ~1 deg-2, depending on the distance from the GC. Earth Bulge Comparison with globular clusters suggests Calore+ 15 22
Radio searches for bulge MSPs Radio detection prospects (Calore+ '15) (Bulge population is just below sensitivity of Parkes HTRU mid-lat survey) GBT targeted searches ~100h: ~3 bulge MSPs MeerKAT mid-lat survey ~300h: ~30 bulge MSPs Plans for the near future We teamed up with MeerKAT TRAPUM plans for dedicated survey in the next 1-2 years GBT searches targeting Fermi unassociated sources or VLA steep-spectrum radio souces Bhakta+ 17 Detections Calore+ '15 Bulge MeerKAT Thick disk Distance 23
Yes, Mod but... NONE of the diffuse emission models gives an acceptable fit to the data 1. Even the best models are excluded by many hundred sigmas Goodness-of-fit tests typically return p-value < 10-300 2. Many excess along the Galactic disk Some of the excesses have same size as Galactic center excess (Calore+15) 3. Bracketing uncertainties by looking at many wrong models does not give the right answer e.g. Ajello+15 Model parameters But everybody is doing it. Set of tested models Real model? Model parameters We need better models and/or massively enlarge the parameter space. 24
Accounting for systematics with SkyFACT SkyFACT (Sky Factorization with Adaptive Constrained Templates) Based on penalized likelihood estimation Storm, CW, Calore, 2017 Hybrid between template fitting & image reconstruction Spatial template Spectral template Nuisance parameters Poisson likelihood Penalization terms Notes Typically >105 parameters Problem typically convex only one minimum We adopt a maximum-entropy prior 25
Data and templates Data Gas ring I Gas ring II Gas ring III Inverse Compton 26
Residuals ~2 GeV Regular template fit Templates with 10%-30% uncertainty + GeV excess 27
Residuals ~ 6 GeV Regular template fit Templates with 10%-30% uncertainty + GeV excess 28
Selected results Dark gas reconstructed Extended sources Fermi bubbles 29
Results for the GeV excess PRELIMINARY Bartels+, in prep. 40 x 40 deg2 Best fit to data obtained with MSP spectrum * (boxy bulge + NB) This suggests that the GeV excess simply traces stellar mass in the bulge... 30
Conclusions Mostly agreement about the existence of an excess of GeV photons in the inner Galaxy Excess is extremely suggestive of a dark matter annihilation signal However, numerous astrophysical processes can produce an excess with the observed luminosity, although only a few reproduce the observed spatial & morphological characteristics Best candidates are arguably millisecond pulsars in the Galactic bulge Supported by photon clustering in inner Galaxy, which matches the expectations Morphology of the excess compatible with distribution of stellar mass in the bulge (boxy bulge + nuclear star cluster) At least two formation scenarios exist that could probably account for the suggested density & distribution of MSPs in the bulge Future progress: radio searches Pulsation searches with GBT (pointed) or MeerKAT (survey) Imaging searches with VLA (archived and new) 31