Galaxy-cluster-stacked Fermi-LAT II: extended central hadronic signal

TL;DR

This work analyzes faint gamma-ray signatures in galaxy clusters by stacking Fermi-LAT data scaled to $R_{500}$ for 75 MCXC clusters, updating prior results with the 4FGL-DR4 catalog. A robust, spatially extended central excess emerges in the $1-100$ GeV band, favoring hadronic emission from a flat CRI distribution with $p\approx2$ and a nearly homogeneous spatial index $\sigma\approx0.1$, in agreement with the CRI population needed to power diffuse intracluster radio emission. The central signal is inconsistent with unresolved central point sources and yields a quantified CRI energy density $A\approx10^{-13.6}$ erg cm$^{-3}$ per $d\ln E$ at 100 GeV, while constraining residual signals from exotic processes such as dark-matter annihilation to $dF/d\ln\epsilon<10^{-14}$ erg s$^{-1}$ cm$^{-2}$ sr$^{-1}$. Across model variants, the results remain robust, reinforcing the hadronic scenario for both the central gamma-ray excess and the diffuse radio halos/relics in clusters. The findings also align with gamma-ray measurements in Coma, supporting a unified hadronic framework for CRI in the intracluster medium.

Abstract

Faint $γ$-ray signatures emerge in Fermi-LAT data stacked scaled to the characteristic $θ_{500}$ angles of MCXC galaxy clusters. After Paper I of this series thus discovered virial shocks, later supported in other bands, this second paper focuses on cluster cores. Stacking $1$-$100$ GeV source-masked data around clusters shows a significant ($4.7σ$ for 75 clusters) and extended central excess, inconsistent with central point sources. The resolved signal is best fit ($3.7σ$ TS-test) as hadronic emission from a cosmic-ray ion (CRI) distribution that is flat both spectrally ($p\equiv1-d\ln u/d\ln E=2.0\pm0.3$) and spatially (CRI-to-gas index $σ=0.1\pm0.4$), carrying an energy density $du(0.1θ_{500})/d\ln E=10^{-13.6\pm0.5}$ erg cm$^{-3}$ at $E=100$ GeV energy; insufficient resolution would raise $p$ and $σ$. Such CRI match the long-predicted distribution needed to power diffuse intracluster radio emission in its various forms (mini-halos, giant halos, standard relics, their transitional forms, and mega-halos), disfavoring models that invoke electron (re)acceleration in weak shocks or turbulence. Stringent upper limits on residual $γ$-ray emission, e.g. from dark-matter annihilation, are imposed.

Figures (10)

• Figure 1: The Fermi-LAT signal of ReissKeshet18, stacked over 112 clusters (dot-dashed blue curve), is updated (nominal analysis: magenta diamonds), showing equally significant central and virial-shock signals with the same data using only 75 clusters, thanks to the improved 4FGL-DR4 catalog. Each signal is presented as the significance $S$ (symbols with lines to guide the eye, in standard-error units) of the excess above the background $S_0$ (labelled horizontal lines, shifted vertically for visibility) as a function of the normalized radius $\tau$. Follow-up MCXC stacking analyses (bottom to top labels specify sample sizes) have shown additional signals well-localized in the same region ($2.2<\tau<2.5$; vertical yellow band), in diffuse emission (empty symbols), discrete sources (filled), or without separating the two (intermittent empty and filled symbols): in LWA HouEtAl23 and polarized GMIMS Keshet24GMIMS, and in NVSS (four-pointed stars) and 2XRS (five-stars) source catalogs IlaniEtAl24a; the co-added MCXC excess (black squares) is consistent with a stacked cylindrical shock model Keshet24GMIMS. Also shown (triangles) are virial shock signals in individual clusters (bottom to top) Coma, in LAT keshet2018evidence and VERITAS KeshetEtAl17 data as a function of $\tau_b$, and A2319 (right triangles) and A2142 (left triangles) in LAT data keshet20coincident, all coincident with a drop in Sunyaev-Zel'dovich (SZ) signal.
• Figure 2: Mass and redshift in our nominal sample of 75 high-latitude clusters (disks, with median values given by green dashed lines, colored light cyan for $\theta_{500}<0{\overset{^\circ}{.}}25$ and dark purple for $\theta_{500}>0{\overset{^\circ}{.}}3$) and of the ReissKeshet18 sample of 112 clusters (red circles). High-redshift clusters are absent due to the $\theta_{500}>0{\overset{^\circ}{.}}2$ cut. The present sample is smaller than in ReissKeshet18 due to the updated point-source catalog (4FGL DR4 instead of 3FGL), the more stringent latitude cut ($|b|>30^{\circ}$), and avoiding the two cluster-overlap regions pointed out in ReissKeshet18.
• Figure 3: Radially-binned brightness stacked for nominal parameters and sample: 75 high-latitude, $|b|>30^{\circ}$ clusters, using a second-order fitted foreground based on $\tau<15$ data. Top panel: separate energy channels 1 (red down-triangles), 2 (green left-triangles, with illustrative solid line to guide the eye), 3 (blue right-triangles), and 4 (black up-triangles) shown at low ($\Delta\tau=1/2$) resolution (top; error bars slightly shifted horizontally for visibility). Bottom panel: averages over the four channels shown at both low (red circles with error bars and line to guide the eye) and high ($\Delta\tau=1/8$; blue error bars) resolutions. Low resolution results also shown (greed squares with error bars) for epochs I+II. The virial-shock region is highlighted as in Fig. \ref{['fig:VSummary']}. The central ($\tau<0.5$) emission is detected at a nominal $4.6\sigma$ confidence level.
• Figure 4: Same as Fig. \ref{['fig:Flux']} (top panel), but zooming into the central $\tau<1$ region, for the nominal sample ($\theta_{500}>0{\overset{^\circ}{.}}2$; left panel), the $\theta_{500}>0{\overset{^\circ}{.}}25$ sub-sample (31 non-compact clusters; middle), and the $\theta_{500}>0{\overset{^\circ}{.}}3$ sub-sample (18 extended clusters; right), along with best-fit point-source (dotted curves) and diffuse (dashed curves) nominal models (channels distinguished by colors and filled triangles in central bin). The high-energy channel is shown even where its photon statistics are insufficient for stacking, presenting as a fixed negative brightness (outside $\tau>0.5$ for the nominal sample, and for additional bins for the smaller, more extended samples).
• Figure 5: Orthogonal projections (see labels) of best-fit (blue disk with $1\sigma$, $2\sigma$, and $3\sigma$ multi-variate confidence-level contours) $\mathsf{n}=3$ CRI parameters in the nominal analysis (without channel $j=4$) of the non-compact ($\theta_{500}>0{\overset{^\circ}{.}}25$ cluster) sub-sample. Predictions based on a joint compilation of radio MHs, GHs, and relics Keshet10 are also shown (parallelogram or error bars for $p=2.0, 2.2$ as down, up triangles with $\sigma=0$ slightly offset vertically for visibility).