Fermi Bubbles: Galactic Gamma-Ray Lobes

• Fermi Bubbles are enormous, symmetric gamma-ray lobes extending roughly 10 kpc from the Galactic Center, characterized by hard spectral emission and sharp edges.
• Multiwavelength observations—from microwave to X-rays—reveal associated structures that shed light on cosmic-ray transport, magnetic fields, and nuclear wind dynamics.
• Diverse formation scenarios, including leptonic jet models and diffusive cosmic-ray injection, highlight their role as natural laboratories for investigating Galactic feedback and acceleration processes.

The Fermi Bubbles are two giant gamma-ray lobes centered on the Galactic Center, extending up to $55^\circ$ above and below the Galactic plane, or roughly $10\ \mathrm{kpc}$ into each Galactic hemisphere. They are characterized by spectrally hard gamma-ray emission, sharp edges, and approximate bilateral symmetry, and they are spatially associated with the microwave haze, polarized radio lobes, and large-scale X-ray structures. Their existence is firmly established, but both the dominant radiation mechanism and the driving engine—past activity of Sgr A$^*$, nuclear star formation, or repeated smaller outbursts—remain debated (Solares et al., 2015, Fox et al., 2019, Yang et al., 2018).

1. Discovery and defining observational properties

The modern picture of the bubbles is explicitly multiwavelength. The central review of spectroscopic work notes radio detection as early as Sofue & Handa (1984), X-ray and mid-IR detection by Bland-Hawthorn & Cohen (2003), the microwave “WMAP haze” identified by Finkbeiner (2004), and gamma-ray identification by Su et al. (2010) and Dobler et al. (2010) (Fox et al., 2019). In gamma rays, the structures appear as two large, homogeneous regions of spectrally hard emission, centered on the Galactic Center and extending to about $\pm 55^\circ$ in latitude (Solares et al., 2015).

The gross morphology is unusually simple for such a large Galactic structure. The bubbles are approximately symmetric about the Galactic plane, have smooth large-scale surfaces, and show sharp boundaries rather than diffuse fading into the halo. Their total solid angle is about $1\ \mathrm{sr}$, and their gamma-ray surface brightness is approximately uniform up to $|b| \approx 55^\circ$ (Yang et al., 2018, Malyshev, 2017). These properties immediately suggest a nuclear origin and make the bubbles a nearby laboratory for Galactic feedback, cosmic-ray transport, and magnetized outflows (Fox et al., 2019).

2. Multiwavelength counterparts and large-scale environment

A major empirical anchor of bubble studies is the close association between the gamma-ray lobes and the microwave haze. In a Fermi-LAT analysis of 50 months of data, the inverse-Compton interpretation was shown to be compatible with the WMAP and Planck haze if the same electron population produces synchrotron emission in a magnetic field $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$, with an allowed approximate range of $5$–$20\,\mu{\rm G}$ (Collaboration, 2014). This connection is central because the gamma rays mainly probe electrons above $\sim 100$ GeV, whereas the microwave haze probes mainly $10\ \mathrm{kpc}$0–$10\ \mathrm{kpc}$1 GeV electrons (Collaboration, 2014).

The radio environment is broader than the GeV morphology alone. Polarized radio lobes seen in S-PASS extend to approximately $10\ \mathrm{kpc}$2, and the associated structures include the North and South ridges, the Galactic Center spur, and limb-brightened spurs (Fox et al., 2019). Several MHD studies interpret this as evidence for a partly ordered field threading a larger bubble complex than the gamma-ray-defined lobes (Barkov et al., 2013, Ruszkowski et al., 2013).

At X-ray energies, the bubble context is again larger. The 2021 MeV study notes that the eROSITA X-ray bubbles are larger, more spherical, and edge-brightened relative to the gamma-ray bubbles, with average $10\ \mathrm{kpc}$3–$10\ \mathrm{kpc}$4 keV surface brightness $10\ \mathrm{kpc}$5, or roughly $10\ \mathrm{kpc}$6, nearly two orders of magnitude above the Fermi Bubble gamma-ray intensity (Negro et al., 2021). This disparity suggests that the X-ray, radio, microwave, and gamma-ray structures need not trace exactly the same plasma phase or acceleration epoch.

3. Gamma-ray spectrum, substructure, and high-energy constraints

A benchmark spectral analysis used 50 months of Fermi-LAT data between 100 MeV and 500 GeV above $10\ \mathrm{kpc}$7 and found that the bubble spectrum is not well described by a simple power law. It is well fit either by a log parabola or by a power law with exponential cutoff,

$10\ \mathrm{kpc}$8

with

$10\ \mathrm{kpc}$9

The abstract summarizes this as index $^*$0 and cutoff $^*$1 GeV, and a simple power law is excluded at more than $^*$2 significance (Collaboration, 2014).

The same study measured a total gamma-ray luminosity for $^*$3 and 100 MeV–500 GeV of

$^*$4

and estimated a median boundary width

$^*$5

reported in the abstract as $^*$6 (Collaboration, 2014). It also confirmed a south-eastern “cocoon,” with $^*$7 values between 95 and 975 depending on foreground model, but found no significant evidence for a jet (Collaboration, 2014).

Spatial-spectral uniformity remains partly unsettled. The 50-month analysis reported no significant spectral variation across the bubbles (Collaboration, 2014). By contrast, a later 60-month analysis found, robustly, a relative deficit of the flux at low energies toward the top of the South Bubble, and also found that at high energies the bubbles are more extended (Yang et al., 2014). The same work stressed that the detailed spectral behavior of each slice is somewhat dependent on the assumed background model, and that the North Bubble is much less secure because of Loop I confusion (Yang et al., 2014).

At TeV energies, non-detections have become an important constraint. The 2021 MeV study summarizes that HAWC found no significant northern-bubble emission above 1.2 TeV, and H.E.S.S. found no significant low-latitude TeV emission, constraining the cutoff to $^*$8 GeV–$^*$9 TeV (Negro et al., 2021). An earlier HAWC conference proceeding had already framed the logic explicitly: TeV observations are crucial because “a steeper cutoff will favor a leptonic model” (Solares et al., 2015).

4. Emission mechanisms and spectral diagnostics

In the leptonic inverse-Compton picture, the gamma rays are produced by relativistic electrons scattering the interstellar radiation field. A gamma-ray fit to the Fermi-LAT data implies an electron spectrum

$\pm 55^\circ$0

with

$\pm 55^\circ$1

and a total electron energy above 1 GeV

$\pm 55^\circ$2

This scenario fits the gamma-ray data very well and also naturally connects to the microwave haze (Collaboration, 2014). Its main physical difficulty is cooling: for 1 TeV electrons in a $\pm 55^\circ$3 field and the ISRF at 5 kpc, the cooling time is only about $\pm 55^\circ$4, or $\pm 55^\circ$5 Myr if only IC losses are considered, implying either very fast transport or in-situ reacceleration (Collaboration, 2014).

In the hadronic picture, gamma rays arise from inelastic proton-proton collisions and $\pm 55^\circ$6-decay. The corresponding proton spectrum is

$\pm 55^\circ$7

with

$\pm 55^\circ$8

Assuming $\pm 55^\circ$9, the required CR proton energy above 1 GeV is

$1\ \mathrm{sr}$0

with proton-proton collision timescale

$1\ \mathrm{sr}$1

Primary hadronic gamma rays alone fit the data less well than IC, with IC preferred over primary-only hadronic models at at least $1\ \mathrm{sr}$2, but hadronic models including IC emission from secondary leptons can also fit the gamma-ray data well (Collaboration, 2014).

The principal discriminant is no longer the GeV spectrum alone. The same 2014 analysis emphasized that a pure hadronic explanation of both gamma rays and the microwave haze does not work well, because synchrotron from secondary $1\ \mathrm{sr}$3 is too faint by a factor of roughly 3–4 and too soft spectrally (Collaboration, 2014). The 2018 origin review made the same point more generally: purely hadronic models fail to reproduce the microwave haze without an additional population of primary electrons (Yang et al., 2018).

The MeV band has therefore become strategically important. In the 2021 study, the benchmark leptonic model remains bright across roughly $1\ \mathrm{sr}$4 to $1\ \mathrm{sr}$5, whereas the hadronic benchmark is weaker and strongly magnetic-field dependent. COSI is projected to detect a hadronic component in the range $1\ \mathrm{sr}$6–$1\ \mathrm{sr}$7 MeV during the prime mission if the bubble magnetic field is $1\ \mathrm{sr}$8, while an AMEGO-X-like mission would detect a pure leptonic component through almost the entire MeV gap, roughly $1\ \mathrm{sr}$9–$|b| \approx 55^\circ$0 (Negro et al., 2021).

5. Gas kinematics, multiphase structure, and magnetic organization

Imaging establishes the bubbles’ shape, but spectroscopy has provided the strongest direct evidence that they are also a multiphase nuclear wind. Radio spectroscopy shows a deficiency of H I within 2 kpc of the Galactic Center and, inside this cavity, about 200 compact H I clouds. These clouds lie within roughly $|b| \approx 55^\circ$1 of the Galactic plane, have temperatures of about $|b| \approx 55^\circ$2, typical neutral hydrogen column densities of about $|b| \approx 55^\circ$3, and are interpreted as neutral clumps entrained in a biconical nuclear wind. Existing radio analyses give a neutral-gas mass outflow rate of about $|b| \approx 55^\circ$4 (Fox et al., 2019).

Ultraviolet spectroscopy reveals the ionized phases of the same flow. Warm-ionized tracers at $|b| \approx 55^\circ$5 include O I, N I, C II, Si II, Si III, S II, and Fe II, while C IV, Si IV, and N V trace gas at $|b| \approx 55^\circ$6. High-velocity absorption is more common along sightlines through the bubbles than outside them, and a foreground-background stellar pair demonstrates absorbing gas beyond $|b| \approx 55^\circ$7, with additional absorption between $|b| \approx 55^\circ$8 and $|b| \approx 55^\circ$9 in the background star alone. Current UV studies imply a kinematic age of roughly $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$0–$$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$1 and warm-ionized mass outflow rates of $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$2–$$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$3, but whether the wind accelerates or decelerates with distance remains unresolved (Fox et al., 2019).

Magnetic structure is a second major theme. A 3D MHD study showed that an explosion from a few million years ago can push and shear a surrounding magnetic loop anchored in the molecular torus around the Galactic Center, generating a spiral, partly ordered magnetic structure around the bubble region. In that framework, the bubble lifetime is estimated as $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$4 yr, with total injected energy $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$5 erg, and consecutive explosive events may match the morphology better than a single one (Barkov et al., 2013). A related AGN-jet MHD model found that a single population of cosmic-ray leptons can simultaneously explain the gamma-ray and microwave spectra, while a second, more recent pair of jets embedded in the bubbles helps reproduce the centrally peaked microwave emission; that model also predicts strongly polarized radio emission and a layer of enhanced rotation measure in the shock-compressed region (Ruszkowski et al., 2013).

6. Formation scenarios and unresolved questions

The formation literature spans a wide dynamic range in age, power source, and transport physics. One class of models treats the bubbles as long-lived hadronic reservoirs supplied by sustained star formation in the Galactic Center. In that picture, the relevant timescale is $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$6 Gyr, the characteristic bubble density is $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$7, the star formation rate is $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$8, and the GC injects $$B = 8.4 \pm 0.2\,{\rm [stat]}^{+11.2}_{-3.5}\,{\rm [syst]}\ \mu{\rm G}$$9 in cosmic rays; the bubbles then act as calorimeters for a hard proton population (Crocker et al., 2010).

A second class invokes repeated stellar capture or tidal disruption by Sgr A$5$0. Early versions adopted capture rates of $5$1 or $5$2, with $5$3 erg or $5$4 erg per event, heating Galactic Center gas to $5$5 keV, launching a wind with velocity $5$6, and generating many shocks that accelerate electrons to $5$7TeV or tens of TeV (Cheng et al., 2011, Dogiel et al., 2011). A 2024 synthesis retained the repeated-TDE engine, estimating ten to hundred events per million years and an average power of $5$8, then coupled it to Kompaneets-type expansion, Rayleigh-Taylor disruption of the shell, hydrodynamic turbulence in the envelope, Lighthill-generated MHD fluctuations, and stochastic re-acceleration of electrons; in that framework the preferred nonthermal emission is again leptonic (Dogiel et al., 2024).

A third class minimizes the need for a special accelerator by treating the bubbles as passive reservoirs illuminated by ordinary Galactic cosmic rays. In the diffusive-injection model, the gamma rays are produced by Galactic CR protons leaking into slowly expanding bubbles, with an energy-independent diffusion coefficient inside the bubbles

$5$9

best-fit $20\,\mu{\rm G}$0, and expansion speed constrained by the spectrum to

$20\,\mu{\rm G}$1

implying

$20\,\mu{\rm G}$2

That model reproduces the observed intensity profile and the integrated spectrum between $20\,\mu{\rm G}$3 and $20\,\mu{\rm G}$4 GeV without invoking additional particle production processes (Thoudam, 2013).

A fourth class places acceleration inside the bubbles. In the stochastic-acceleration model, electrons undergo volume-filling 2nd-order Fermi acceleration by plasma turbulence, with characteristic parameters $20\,\mu{\rm G}$5, $20\,\mu{\rm G}$6, and total energy in electrons above 100 MeV of $20\,\mu{\rm G}$7; the model predicts nearly constant surface brightness, sharp edges, and limb-brightening at $20\,\mu{\rm G}$8 GeV, while matching the microwave haze requires $20\,\mu{\rm G}$9 rather than the fiducial $\sim 100$0 (Mertsch et al., 2011). A related hadronic “scaled-up SNR” model treats the bubbles as forward shocks in the halo, with a favored bubble age of $\sim 100$1 yr and the specific prediction that the apparent bubble size should be larger in the TeV band than in the GeV band (Fujita et al., 2013).

Purely geometrical and shell-dynamical interpretations also exist. In a superbubble treatment, a thin-shell momentum-conserving model propagating in a medium with inverse-square density decline gave a stated reliability of $\sim 100$2 when compared with a digitized bubble section, outperforming the thermal and self-gravitating reference models (Zaninetti, 2018). This suggests that some large-scale aspects of the morphology can be captured without specifying the nonthermal particle physics in detail.

The field’s unresolved issues are therefore not the existence or basic phenomenology of the bubbles, but the coupling between dynamics, transport, and radiation. A 2018 review concluded that purely hadronic models are incomplete because they do not explain the microwave haze without extra primary electrons; that the simplest in-situ acceleration models struggle with the flat projected intensity profile; and that leptonic jet models currently reproduce the broad gamma-ray, microwave, X-ray, and radio-polarization phenomenology particularly well, but require past accretion rates $\sim 100$3 of Eddington (Yang et al., 2018). Future progress is expected from denser UV and radio spectroscopy, X-ray/UV metallicity and kinematic measurements, deeper TeV limits, neutrino searches, and especially MeV observations, where leptonic and hadronic spectra diverge most strongly (Fox et al., 2019, Negro et al., 2021).