Analysis and implications of the spatio-spectral morphology of the Fermi Bubbles

TL;DR

This work conducts a pixel-by-pixel spectral analysis of the Fermi Bubbles using ten years of Fermi/LAT data and a template-free reconstruction to distinguish hadronic from leptonic gamma-ray emission. Forward-modeling with Naima and spatially varying ISRF enables fits with three spectral forms (PL, EPL, BPL) across ~37,000 pixels, revealing that EPL and BPL forms provide substantially better fits than simple PL for both emission channels and that spectral hardening occurs toward the southern tip. In the leptonic case, the electron energy density increases toward the bubble edges, while Klein-Nishina effects suppress starlight contributions, making dust-dominated ISRF the key driver; cooling times are extremely short (~1 Myr at the caps), challenging a central, advection-based origin and favoring in-situ peripheral acceleration or a hadronic-dominated scenario. The results place strong constraints on the FBs’ energy budget and origin, suggesting distributed acceleration or sustained hadronic CRs as viable explanations and linking the gamma-ray observations to multi-wavelength ISRF properties and CR transport/acceleration physics.

Abstract

The Fermi Bubbles are gamma-ray structures extending from the center of the Milky Way to +/-50 degree Galactic latitude that were discovered in data obtained by the Fermi/LAT instrument. Their origin and power source remain uncertain. To help address this uncertainty, here we use a template-free reconstruction of ten years of all-sky Fermi/LAT data provided by Platz et al. (2023) to carry out a pixel-by-pixel spectral analysis of the Bubbles. We recover the position-dependent spectral shape and normalization that would be required for parent proton or electron cosmic ray populations to produce the Bubbles' observed gamma-ray spectra. We find that models in which the gamma-ray emission is driven by either hadronic or leptonic processes can explain the data equally well. The cosmic ray population driving the emission must have either broken power-law or exponentially cut-off spectra, with break or cutoff energies that are almost constant with latitude but spectral indices below the break that harden towards the Bubbles' southern tip. For the leptonic channel, reproducing the observed position-dependent gamma-ray spectrum also requires a cosmic ray electron energy density that grows with distance from the Galactic plane and increases towards the edges of the Bubbles. This finding disfavors scenarios for the origin of the Bubbles where a population of cosmic ray electrons is accelerated near the Milky Way center and subsequently advected out to the extremities of the Bubbles.

Figures (11)

• Figure 1: Template-free reconstruction of the non-dust associated all-sky diffuse $\gamma$-ray photon flux, $I^\mathrm{nd}$, taken from the M2 model of platz_multi-component_2023; the data shown are for the energy bin centered at 133 GeV. The data are shown in Mollweide projection using Galactic coordinates. The colorbar is plotted on a log scale. The black line is a by-eye tracing of the FBs' apparent edges, and the white dashed line shows the region $-40^\circ < \ell < 40^\circ$, $-60^\circ < b < 60^\circ$ on which we focus in this work. The four white points labeled a, b, c, and d are chosen to show example results at different points in the sky.
• Figure 2: Our model of the specific energy density $u_\lambda$ of the ISRF (including stellar, dust, and CMB components) as a function of wavelength $\lambda$, drawn primarily from popescu_radiation_2017, at two sample positions: the Galactic center ($r=0$, $z=0$; blue) and at a point 8 kpc above the Galactic center ($r=0$, $z=8$ kpc; orange).
• Figure 3: The reduced $\chi^2$ for each model and source particle population (left panel for protons and right panel for electrons, PL, BPL, and EPL levels from top to bottom) for each pixel. The boundary of the FBs is marked by a yellow line, which is the same in each panel and the same as the black line in \ref{['fig:gamma_ray_data']}. We see that a power law model with a break or an exponential cutoff provides a better fit than a simple PL model in the FBs region.
• Figure 4: Best-fit model spectra versus energy-binned data for the six different models evaluated at the point $b=45^\circ$ and $l=0.38^\circ$ (labeled 'a' in \ref{['fig:gamma_ray_data']}) within the FBs. The blue points with error bars show the $\gamma$-ray data, and the computed SED is shown by solid black (PL), green (EPL), and red (BPL) lines. The EPL and BPL models perform better than the PL models for both hadronic (left) and leptonic (right) cases.
• Figure 5: Best-fit parameters for the EPL models (protons on the left and electrons on the right). The plots in the top panel shows the model amplitude and the middle panel shows the power-law index. We see a spectral hardening towards the Southern tip for both. The bottom panel shows the cut-off energy. It is of the order of a few TeV and does not vary much with latitude.