UNASSOCIATED GAMMA-RAY SOURCES AS TARGETS FOR INDIRECT DM DETECTION - - PowerPoint PPT Presentation

1

• J. Coronado-Blázquez
• M. Sánchez-Conde, A. Domínguez, M. di Mauro, E. Charles, N. Mirabal

for the Fermi-LAT Collaboration

Halo Substructure & Dark Matter Searches Madrid - June 2018

UNASSOCIATED GAMMA-RAY SOURCES AS TARGETS FOR INDIRECT DM DETECTION WITH FERMI-LAT

DARK MATTER (DM) SUBHALOS AS TARGETS

❑ΛCDM cosmological model predicts lots of substructure → DM subhalos ❑Subhalo with masses below ~107𝑁⨀ do not retain gas (baryons) → no emission ❑BUT, if they annihilate (WIMP model) → DM-induced gamma-ray emission ❑Fermi-LAT (2008-) → We have gamma-ray source catalogs ❑Lots of unidentified sources (unIDs) in catalogs → Some of them may be subhalos ❑N-body cosmological simulations → What do we expect? ❑We do not have an unequivocal signal of DM annihilation → constraints on 𝝉𝒘 , 𝒏𝝍

2

DM ANNIHILATION IN THE WIMP MODEL

𝜓𝜓 → ൞ 𝜐+𝜐− 𝑐ത 𝑐 𝑋+𝑋− ?1 ?2 → ⋯ → 𝛿𝛿 𝐺 𝐹 > 𝐹𝑢ℎ = 𝐾𝑔𝑏𝑑𝑢𝑝𝑠 ∗ 𝑔

𝑞𝑞(𝐹 > 𝐹𝑢ℎ)

3

Astrophysics (Density profile, distance…) Particle Physics (channel, annihilation spectra…)

𝐾𝑔𝑏𝑑𝑢𝑝𝑠 = න

𝛦𝛻

𝑒𝛻 න

𝑚.𝑝.𝑡

𝜍𝐸𝑁

2

𝑠 𝜇 𝑒𝜇 𝑔

𝑞𝑞 = ෍ 𝑔

𝐶𝑔 1 4𝜌 𝑒𝑂

𝑔

𝑒𝐹

𝑔

𝜏𝑤 2𝑛𝜓

2 Branching ratio taken as 1

DM ANNIHILATION IN THE WIMP MODEL

4

𝜏𝑤 ∝ 𝑛𝜓

2 · 𝐺 𝑛𝑗𝑜

𝐾𝑔𝑏𝑑𝑢𝑝𝑠 · ׬

𝐹𝑢ℎ 𝐹

𝑒𝑂 𝑒𝐹 𝑒𝐹 = 𝑛𝜓

2 · 𝐺 𝑛𝑗𝑜

𝐾𝑔𝑏𝑑𝑢𝑝𝑠 · 𝑂𝛿 We want to probe the lowest possible 𝝉𝒘 values to rule out WIMP models

Instrument Theory Simulations

𝜓𝜓 → ൞ 𝜐+𝜐− 𝑐ത 𝑐 𝑋+𝑋− ?1 ?2 → ⋯ → 𝛿𝛿 𝐺 𝐹 > 𝐹𝑢ℎ = 𝐾𝑔𝑏𝑑𝑢𝑝𝑠 ∗ 𝑔

𝑞𝑞(𝐹 > 𝐹𝑢ℎ) Astrophysics (Density profile, distance…) Particle Physics (channel, annihilation spectra…)

FIRST INGREDIENT: DM INTEGRATED SPECTRA

5

• From Cirelli PPPC4 (PYTHIA8), including

electroweak corrections

• ‘Usual’ annihilation channels

𝑐ത 𝑐, 𝜐+𝜐−, 𝑋+𝑋−, 𝑓𝑢𝑑.

• Wimp masses from 5 GeV up to 100 TeV
• Parametric fit to Power Law with

SuperExponential Cutoff: 𝑒𝑂 𝑒𝐹 = 𝐿 · 𝐹 𝐹0

−Γ

𝑓

− 𝐹 𝐹𝑑𝑣𝑢

𝛾

(Later used for the minimum flux)

FIRST INGREDIENT: DM INTEGRATED SPECTRA

6

• We want the integrated spectra,

𝑂𝛿 = න

𝐹𝑢ℎ 𝐹

𝑒𝑂 𝑒𝐹 𝑒𝐹

• Dependance on the catalog

energy threshold

SECOND INGREDIENT: J-FACTOR

• We search for DM subhalos in Fermi-LAT unassociated (unIDs) sources
• N-body cosmological simulations predict a number of subhalos
• We could have the brightest of them already hidden among the LAT unIDs

𝜏𝑤 ∝ 𝐺𝑛𝑗𝑜 𝐾𝑔𝑏𝑑𝑢𝑝𝑠 · 𝑂𝛿 𝐾𝑔𝑏𝑑𝑢𝑝𝑠 = න

𝛦𝛻

𝑒𝛻 න

𝑚.𝑝.𝑡

𝜍𝐸𝑁

2

𝑠 𝜇 𝑒𝜇

7

SECOND INGREDIENT: J-FACTOR

• We use Via Lactea II (VL-II) simulation (0805.1244), DM only,

Milky Way size, resolving subhalo masses down to ~105𝑁⊙

• Radial distribution well described by antibiased NFW or Einasto
• Subhalo mass function a power law with index close to -2
• But subhalos below the resolution limit can be very important

Diemand+08 (0805.1244)

SECOND INGREDIENT: J-FACTOR

A low mass subhalo close enough to the Earth can have a bigger J-factor than a further, massive subhalo

SECOND INGREDIENT: J-FACTOR

• The less massive the subhalo, the nearest

must be to have a relevant flux

• Also, 𝐾 ∝ 𝑑3 ∝ 𝑁−3 (𝑑 ≡concentration,

bigger for lower masses)

SECOND INGREDIENT: J-FACTOR

• We repopulate the original VL-II simulation in

a realistic yet computationally feasible way below its mass resolution limit.

• Currently a work in progress with A.

Aguirre-Santaella

FERMI-LAT UNASSOCIATED SOURCES (UNIDs)

12

• Obs. Time (yr)

Energy Range Total UnIDs 2FHL (1508.04449) 6.7 50 -2000 GeV 360 48 3FHL (1702.00664) 7 10 - 2000 GeV 1556 177 3FGL (1501.02003) 4 0.1 – 300 GeV 3033 1010

UNIDs “FILTERING”

13

• 𝜏𝑤 ∝ 𝐾−1 → less DM subhalo

candidates among unIDs means better constraints

• Exponential rise in our constraining

power below ~20% of sources in every catalog

• 20% = 202 sources in 3FGL, 10 in

2FHL and 35 in 3FHL

• From these numbers down, every

source we remove has a big impact

DM SUBHALO FILTERS

• 1. Source associations
• 2. Flux variability
• 3. Latitude
• 4. Machine learning identification
• 5. Multiwavelength emission
• 6. Complex regions

14

We adopt a conservative approach

SOURCE ASSOCIATIONS: OBSERVATIONAL CAMPAIGNS

• Observational campaigns allowed us to remove 66 AGNs

and 18 pulsars

• We remove 15 additional sources found to be Millisecond

Pulsars (MSPs)

15

SOURCE FLUX VARIABILITY

• Apart from the official catalogs, we use FAVA (Fermi All-Sky Variability Analysis)
• Does not relay on any diffuse model – just relative deviations
• Some sources behave as steady on average for long periods, but can present a short

flare → FAVA weekly binning important

16

…but seen in flare Null variability…

3FHL J0500.6+1903

SOURCE FLUX VARIABILITY

• Note: in some cases, the flare occurs after the time period used for the

catalog (e.g. 2FHL – 320 weeks, while a given source may flare in week 400)

• We discard 2 2FHL, 12 3FHL and 13 3FGL sources
• We also identify source “duplicities”: two sources that actually are

the same – 2 cases confirmed

17

LATITUDE

There are subhalos at low latitudes, but 1) the diffuse emission makes them very difficult to detect and 2) most of these unIDs are expected to be pulsars, which can fake DM

We cut the Galactic plane ±10°, removing 14 sources in 2FHL, 70 in 3FHL, 429 in 3FGL

18

Much less diffuse emission – Life gets easier

MACHINE LEARNING IDENTIFICATION

• Salvetti+17 (AKA 3FGLzoo) – “4FGLzoo” on the way
• Lefaucheur+Pita17

Assign probabilities to sources based on ML algorithms to derive their physical nature

Applied only to the 3FGL (due to higher available statistics) – 186 rejections 2FHL & 3FHL rejections only if also present in the 3FGL (29 in 3FHL)

19

MACHINE LEARNING IDENTIFICATION

• Low false positive rate: <4% (3FGLzoo), 2-4% Lefaucheur+Pita
• Most of the classified sources are common to both papers, but

some are not

• 162 unIDs classified as AGN in the 3FGLzoo + 24 extra in

Lefaucheur+Pita

• Numbers refer already to the pool of unIDs that remained after

the previous cuts

20

MACHINE LEARNING IDENTIFICATION

• Mirabal+16 (1605.00711) applies machine learning for pulsar searches – 34 high-

latitude pulsar candidates (only 3 remaining in our sample)

• However, pulsars can mimic a DM signal

Ackermann+12

• Therefore, this work is specially useful for us, not for rejecting unIDs but to point out VIP

candidates

21

MULTIWAVELENGTH EMISSION

22

• DM subhalos expected to

bright only in gammas

• Should they exhibit other

wavelength emission (without any other kind of source within 5 - 10 arcmin, depending on the catalog) we eliminate them

tools.asdc.asi.it

IR+Optical (WISE, 2MASS, USNO) Gamma (2FHL) Search in 5 arcmin

MULTIWAVELENGTH EMISSION

23

ASDC Stroh+13 www.swift.psu.edu /unassociated/ SWIFT (HEASARC) Total 2FHL 4 2 6 3FHL 10 2 5 17 3FGL 7 13 207 227 IR-Optic X-Ray

COMPLEX REGIONS

• 3FGL sources flagged with “c” (e.g. 3FGL J0342.3+3148c)
• See Sec. 3.8 of 1501.02003 (3FGL paper) – Considered potential artifacts
• 78, most on the Galactic plane (i.e. already excluded from our analysis before); we

discard 5 high-latitude sources (only in 3FGL)

24

FAMOUS (EX-)CANDIDATES

25

▪ 3FGL J2212.5+0703 (Bertoni+16) – actually 2 sources ▪ 3FGL J1924.8-1034 (Xia+17) – classified as AGN by machine learning ▪ 3FGL J1119.9-2204 (Hooper+17) – seen with SWIFT ▪ 3FGL J0318.1+0252 (Hooper+17) – seen with SWIFT ▪ 3FGL J2212.5+0703 (Hooper+17) – FAVA correlation, seen with SWIFT

All 3FGL (low energy) sources

UNIDS FILTERING RESULTS

26

Original Result 2FHL 48 4 3FHL 177 24 3FGL 1010 16

THIRD INGREDIENT: LAT SENSITIVITY TO DM SUBHALOS

• Minimum flux to have a 5-sigma detection over background
• Normally taken as the threshold flux of the catalog
• BUT, important dependance on annihilation channel,

source sky position and catalog setup

27

THIRD INGREDIENT: LAT SENSITIVITY TO DM SUBHALOS

• We use the fermipy analysis software to simulate sources mimicking

the catalog setup (observation time, energy range, diffuse+isotropic templates…)

• A putative dark matter source is simulated for each position,

catalog setup, annihilation channel and DM mass

• All-sky maps with this information

28

COMPARISON BETWEEN DIFFERENT DM MASSES

30

𝑛𝐸𝑁 = 10 𝐻𝑓𝑊 𝑛𝐸𝑁 = 1 𝑈𝑓𝑊 3FGL setup, 𝜐+𝜐− channel

31

𝐺

𝑛𝑗𝑜 vs. WIMP

mass, all latitudes 𝐺

𝑛𝑗𝑜 vs. Gal. latitude

𝐺

𝑛𝑗𝑜 vs. WIMP mass

32

+ + =

DM annihilation spectra J-factor Minimum flux

DARK MATTER CONSTRAINTS

33

PRELIMINARY PRELIMINARY

SENSITIVITY REACH OF THE METHOD

34

PRELIMINARY PRELIMINARY

CONCLUSIONS & FUTURE

• The method proves to be complementary and competitive to other

indirect searches

• Conservative yet realistic constraints
• The constraints can be improved via new associations, potentially ruling
• ut thermal WIMPs up to ~400 GeV (𝑐ത

𝑐) and ~250 GeV (𝜐+𝜐−)

35

CONCLUSIONS & FUTURE

36

• The future 4FGL will be the deepest and most complete gamma-ray

catalog, providing new analysis targets

• This analysis was blind to spectral information and a dedicated spectral

analysis is ongoing already to improve the limits further

Thank you very much

BACKUP SLIDES

37

SIGNIFICANCE = FLUX?

38

LATITUDE

• UnIDs strongly clustered on the Galactic plane
• We cut out the region 𝑐 < 10° from our analysis

39

VARIABILITY

• We require at least a flare at 5𝜏, and not spatially or temporally coincident

with a known flare (to avoid PSF "spill over“)

40

3FGL J2043.8-4801 3FGL J2056.2-4714

VARIABILITY

• We require at least a flare at 5𝜏, and not spatially or temporally coincident

with a known flare (to avoid PSF "spill over“)

41

TS = 17 Flares correlation→caused by the other source

VARIABILITY – EXAMPLE OF SOURCE “DUPLICITY”

42

(Almost) same position and spectra, but not associated with each other

3FHL 3FGL

VARIABILITY – EXAMPLE OF SOURCE “DUPLICITY”

43

WHY NO OTHER SOURCE IN 5/10 ARCMIN?

44

• Data on the left refer to 2 sources at 2.3

and 2.6 arcmin

• Due to Fermi PSF, it is uncertain whether

any of these sources correspond to the gamma source or to another thing

• To be sure, we require 5 or 10 arcmin

(depending on the source positional uncertainty) to be completely empty of

• ther sources
• Should we have any multiwavelength

emission in there, we reject the unID from

• ur “clean” list
• We discard 4 2FHL, 10 3FHL and 7

3FGL sources

3FGL J0953.7-1510@5 arcmin

MULTIWAVELENGTH EMISSION

• We also use the SWIFT observational

follow-up of Fermi unIDs (Stroh+13)

• We search for X-ray excesses (at 3𝜏)
• Updated in (almost) real time in

www.swift.psu.edu/unassociated/

• Removes 13 3FGL sources, 2 of them also

in 3FHL catalog

• Now exploring the possibility to request

time for observing VIP candidates

45

Blue – error ellipse; cyan – detected excess

MULTIWAVELENGTH EMISSION

• Finally, we search in the HEASARC archive

for SWIFT observations of Fermi unIDs

• Only discarded if they are reported as

associated with 3FGL or 2/3FHL sources in the SWIFT Master Catalog

46

We also search in XMM-Newton, ROSAT, Suzaku, Integral, NuStar, Chandra, Hitomi, NICER, RXTE… but they do not firmly associate any source. We discard 207 3FGL, 2 2FHL and 5 3FHL sources, 4 of them Mirabal+16 pulsar candidates

47

48

49

CONSTRAINTS DEPENDING ON THE NUMBER OF UNIDs

More improvement removing the last 60 sources than the first 950!

WHY NO OTHER SOURCE IN 5/10 ARCMIN?

51

PSF “spill over”

52

This kind of “striped” emission (2MASS, SDSS & USNO) is systematically repeated – large source density in

• ptical/near IR

We require other discrete points (VizieR, WISE, NVSS, FIRST…) and other WL to conservatively discard them

53