Recent evidence for gamma-ray line emission from Fermi-LAT data: dark matter or artifact? Meng Su Pappalardo/Einstein Fellow MIT/CfA Collaborators: Douglas Finkbeiner, Christoph Weniger Fermilab Astrophysics Seminar, March 4 th 2013

Fritz Zwicky's pioneering work in 1933 Vera Rubin and her colleagues: galaxies rotation curve of nearly galaxies (1970s)

Standard Dark Matter Scenario: Thermal WIMP (Weakly Interacting Massive Particle) Dark matter particles are produced and annihilating in thermal equilibrium after big bang (according to a given particle physics model). Universe is expanding and cooling. Universe becomes too cool to continue producing DM particles. Remaining DM annihilates to standard model particles. Universe’s rate of expansion exceeds rate of particle annihilations, making annihilations rare. “Dark Matter Freezes Out”

Indirect Detection: gamma-ray

Techniques for the Indirect Detection of Dark Matter Morphological Differentiation Galactic Center Search Milky Way Halo Dwarf Spheroidal Galaxies Anisotropy Power Spectrum Gamma-Rays from Galaxy Clusters Cosmic Ray Electrons/Gammas from the Sun Spectral Differentiation Gamma-Ray Line Search The two-body annihilation of DM into photons produces monochromatic gamma rays e+e- ratio and electron + positron Spectrum (Recall previous talks by Alexander Belikov and Lawrence Rudnick)

Gamma-Ray Line Search GOAL: Search for a spectral signal from an annihilation (or decay) of dark matter directly into gamma rays. Targets: Any annihilating or decaying dark matter particle which produces photons directly, or possibly produces a very hard spectrum of single photons through internal bremsstrahlung Benefits: No astrophysical process is expect to produce a gamma-ray line. Resulting constraints can fall far below thermal cross-section Difficulties: Direct production of gamma-gamma is not a generic prediction of dark matter annihilation or decay.

From Tomi Ylinen’s thesis

Atwood et al. 2009

Atwood et al. 2012

The Fermi-LAT three year maps Su & Finkbeiner (2012)

Galactic Diffuse Gamma-ray Emission (from Tsunefumi Mizuno)

Fermi Bubble from three year maps

Bubbles as foreground for DM search

Fermi-LAT Collaboration, arXiv: v1

Best limits so far: Fermi-LAT Collaboration, arXiv: v1

How to define a search region? Estimate the signal from DM, e.g. Einasto profile squared, projected along line of sight. Estimate background from lower energy (1-20 GeV) photons. Estimate S/N as DM signal/background. Make a cut on estimated S/N.

There have been hints of something at ~ 130 GeV!

Weniger (2012)

Profumo and Linden (2012)

Why the reluctance to call a sharp spectral feature a line? Perhaps because 2 WIMPs -> 2 gammas is usually loop suppressed (in, e.g., the MSSM), so you expect a continuum ,000x brighter. This was not observed, making a line impossible. But there are other theories where a line is OK. Let’s ignore theoretical prejudice and simply ask if there is a line in the Fermi LAT data.

Timeline of 130 GeV line: 12 April - Weniger (looks like a line at 130 GeV) 26 April - Profumo & Linden (is it the Fermi bubbles?) 10 May - Tempel et al., (No, it’s not a bubble, could be DM) 21 May - Boyarsky (lots of blobs, probably not DM) 25 May - Acharya, Kane... (It’s a Wino) 29 May - Bergstrom (reviews claims as part of larger review) 30 May - Jim Cline (two lines) 30 May - Buckley & Hooper (theoretical models) 5 June - Geringer-Sameth & Koushiappas (Line search in dwarfs) 7 June - Su & Finkbeiner (Off center 1.5 deg, Einasto, 6.5sigma, use high energy-resolution events) ( As of March 2013, Weniger paper has 120+ citations )

Fermi-LAT maps at 100–180 GeV Su & Finkbeiner (arXiv: )

Smoothed maps

A cusp structure!

Null tests Subtraction of other energy maps Cosmic ray contamination?

Spectrum: Make maps in each of 16 energy bins, assume that emission in each bin is a linear combination of template maps, and plot the template coefficients. Coefficients are determined by maximizing the Poisson likelihood of observing the observed counts given the model. Templates choice corresponds to hypothesis to be tested.

Templates for spectrum fitting

Energy spectrum of the cusp

Not from large scale Fermi bubbles 1) χ+χ  γ+γ 2):χ+χ  Z 0 +γ E γ=m x -M z 2 /4m x A pair of lines at 110.8±4.4 GeV and 128.8±2.7 GeV Consistent with single line at ± 2.7 GeV

Galactic longitude and latitude profile Even though the high- incidence-angle photons ( θ > 40 ◦ ; right) panels have half the exposure (9.7% vs. 19% for the left panels), they have more than half of the photons, and nearly the same TS due to lower off-line background leaking in. Offset from the GC?

High incidence angle events with better energy resolution Yvonne Edmonds PhD thesis

Galactic longitude and latitude profile Even though the high- incidence-angle photons ( θ > 40 ◦ ; right) panels have half the exposure (9.7% vs. 19% for the left panels), they have more than half of the photons, and nearly the same TS due to lower off-line background leaking in. Offset from the GC?

Tests: We do not see the signal elsewhere in the Galactic plane:

Offset dark matter profile fits better

The detection significance of the gamma- ray cusp for various models We do the fit in many ways. Off-center Einasto is the best.

Two lines model

Assessment of line profile The 129 GeV feature shape is strikingly similar to that expected for a line. The 111 GeV feature is unconvincing, but is also compatible with a line. In some cases, fluctuations appear, but are not present in both low and high incidence spectra. This test did not have to succeed. The fact that the high- incidence photon sample has sharper spectral features is important.

A MODIFIED SURVEY STRATEGY FOR FERMI The scan strategy of Fermi -LAT could be altered for 1 year to confirm the 130 GeV line! The scan strategy of Fermi -LAT could be altered for 1 year to confirm the 130 GeV line! This current strategy is excellent for uniformity of full-sky coverage, but is far from optimal for collecting high- incidence-angle photons from the GC. This current strategy is excellent for uniformity of full-sky coverage, but is far from optimal for collecting high- incidence-angle photons from the GC. The exposure time of our (40 ◦ < θ < 60 ◦ ) sample exceeds the current strategy (observed 9.7% of the time) by more than a factor of 4. Require GC have an incidence angle of 45 ◦ < θ < 55 ◦. After 1 year of altered observing, we would have a sample of high incidence photons equal to the current sample, and could evaluate their significance directly, in the absence of any trials factor! After 1 year of altered observing, we would have a sample of high incidence photons equal to the current sample, and could evaluate their significance directly, in the absence of any trials factor!

What can go wrong?

Peculiarities of the Galactic center observation?

Atwood et al Vela pulsar The Crab

Incidence angle distribution

Hypothesis 1: The Galactic center is bright, so instrumental artifacts are more significant there

Credit: HESS Collaboration; NASA/UMass/D.Wang et al.

Hypothesis 1: The Galactic center is bright, so instrumental artifacts are more significant there At E > 100 GeV, the Galactic center is only modestly brighter than the surrounding regions, away from the GC or from low/high energy (bright X-ray source?) At E > 100 GeV, the Galactic center is only modestly brighter than the surrounding regions, away from the GC or from low/high energy (bright X-ray source?) It is difficult to see how this could happen. It is difficult to see how this could happen. Regions with a high gamma ray-to-CR ratio, which are used for calibration purposes: enhance the impact of an instrumental effect, like e.g. energy reconstruction or acceptance anomalies that Regions with a high gamma ray-to-CR ratio, which are used for calibration purposes: enhance the impact of an instrumental effect, like e.g. energy reconstruction or acceptance anomalies that

No lines in other samples

Hypothesis 2: The Galactic center has a hard spectrum, making energy mapping errors more significant. The GC spectrum is not much harder than the rest of the Inner Galactic plane, but the latter shows no sign of a feature at 130 GeV. The GC spectrum is not much harder than the rest of the Inner Galactic plane, but the latter shows no sign of a feature at 130 GeV. Not enough photon for >300 GeV to mimic the line!

Hypothesis 3: the GC observations have a restricted range of incidence angles on the instrument. Instrumental problems could be projected onto the Galactic center simply for geometric reasons Instrumental problems could be projected onto the Galactic center simply for geometric reasons In Dec (Jun) the Sun passes near the Galactic center (anticenter), solar panels orientation determines the GC direction is close to the Sun!

How reconstruction works?

Parameterize the shower profile Energy leakage requires extensive modeling of showers for energy reconstruction From Tomi Ylinen’s thsis

Modeling of the longitudinal and transverse profiles of electromagnetic showers and on the modeling of the development of the showers through the LAT calorimeter. The shower maximum is not well contained is ∼ 25% for photons at 100 GeV Philippe Bruel et al. (2012)

Shower longitudinal (transverse) profile parameterization Philippe Bruel et al. (2012) Simulation of CsI calorimeter with GEANT4

Performance of the shower profile fit: bias (left), resolution (center), shower containment Philippe Bruel et al. (2012)

How to test?

“Earth limb” photons The Fermi-LAT collaboration (2009)

Subsample from high incident angle events What’s that?

So it looks like there a 3.3 sigma (4.7 sigma?) detection of an artifact in the Fermi data, but only for incidence angles degrees Dip in energy response?

The Earth limb line and correlations with the GC signal The majority of high-incidence limb events appear near the orbital pole

Timing correlation

But we see no way this can be the explanation. Even if we discard all events with 30 < theta < 45, we get 4 sigma:

How about the offset from the Galactic center?

Kuhlen et al., arXiv: “Eris” simulation, run with Gasoline n-body code Cosmological zoom-in simulation of a MW-like galaxy. The region of highest DM density (X) is away from the dynamical center of the simulation, by 2.5 smoothing Lengths. This offset is typical over billions of years.

The region of highest DM density (X) is away from the dynamical center of the simulation, by 2.5 smoothing lengths. This offset is typical over billions of years.

But: a central core?

Eris conclusions: The Kuhlen et al. paper suggests that the distribution of DM in the center of a barred spiral is nontrivial, and the point of maximum density may not (always) be at the center. However, it is not clear at all that this solves our problem with the 130 GeV feature.

Lines at the same energy from other regions of the sky might help…

DOUBLE GAMMA-RAY LINES FROM UNASSOCIATED FERMI-LAT SOURCES DOUBLE GAMMA-RAY LINES FROM UNASSOCIATED FERMI-LAT SOURCES Su & Finkbeiner (arXiv: )

Background estimation Su & Finkbeiner (arXiv: )

No such feature from faint AGNs

The line ratio is consistent

What’s the future?

HESS-II expected performance in the tens of GeV Image Credit: H.E.S.S. Collaboration, Frikkie van Greunen 600 m^2 mirror area and a very high resolution camera. lower the energy threshold from 100 GeV to about 30 GeV and enhance the HESS sensitivity. HESS collaboration, ICRC2009

Bergstrom et al. ( arXiv: ) Line detection sensitivity

Pass 7 reprocessed data LAT team have reprocessed the data with updated calorimeter response calibration – Updated light yield calibration in each crystal -> affects energy reconstruction – Light asymmetry calibration -> affects position resolution, thus improves PSF at high energies Impacts and changes – PSF at high energies is improved – Up to 5% shift in the energy scale (time and energy dependent) – Spectrum of background contamination has changed

Galper et al. (2012) Gamma-400 (planned launch )

DAMPE (2015) The detector will be composed of a telescope (red layers in left figure) and an EM calorimeter. The EM calorimeter will be composed of 576 BGO crystal bars with dimensions of 2.5 cm×2.5 cm×30cm. The BGO crystals form 12 layers with an area of about 60cm×60 cm each. The r.l is ~27. There will be also Silicon detector on the top for charge and position measurement Total thickness is 34.5 r.l. G.F. 0.5m2.sr Total weight 1.5 tons J. Chang (DSU 2011)

CALET (2014)

Conclusions: The line signal is not a discovery yet. - need more data (trials factors!) - can change survey strategy to get it fast Want to know: - Is the cusp really off center? - are there two lines (or more)? Doubling the data will address these questions...