The strange glow at the center of the Milky Way may be dark matter

A faint glow of gamma rays hangs over the center of the Milky Way galaxy, spreading across thousands of light years and refusing to let go of its source. For many years, one of the strongest arguments against the dark matter explanation was that the signal looked like a cluster of hidden stars. New analysis adds further complexity to the case.

This signal, known as the Galactic Center Excess (GCE), is a roughly spherical surplus of gamma rays observed near the center of the galaxy by NASA’s Fermi Global Telescope. Since it first appeared as data more than a decade ago, it has stimulated one of astrophysics’ most stubborn debates.

One possibility is that the glow comes from dark matter particles annihilating each other, a process long theorized to potentially produce gamma rays. The other is that the surplus comes from vast numbers of millisecond pulsars, rapidly rotating neutron stars that are too faint to pick out one by one.

“The signal is particularly difficult to interpret because the galactic center is a very bright and crowded region of the gamma-ray sky,” said study author Florian List, a researcher at the University of Vienna.

Previous research supporting the pulsar idea focused on how gamma rays are distributed across the sky. If the excess is from many unresolved point sources rather than a smooth signal, the map should show subtle deviations from the random pattern expected from diffuse radiation such as dark matter. These studies argued that the excess has point source properties.

But they came with trade-offs. To keep the calculations manageable, they typically ignored two things: the correlation between neighboring pixels and the energy carried by individual photons.

A new study sought to exceed both limits. The team trained a convolutional neural network on one million simulated Fermi gamma-ray maps. Of these, 790,000 were for training, 200,000 for validation, and 10,000 for testing. The simulated observations covered 10 logarithmically spaced energy bins from 2 to 20 GeV and were constructed to reflect the changing response of the telescope across energies.

This was important because galactic supercenters have a unique spectrum, and different possible sources can leave different energy signatures. By combining spatial and spectral information for the first time, networks will be able to ask tougher questions. That is, it’s not just about whether the glitter looks clumpy, but how bright that clumping should be.

The answer has been to make the light population much less bright than many previous studies had suggested.

Previous analyzes had pointed to unresolved sources just below Fermi’s detection threshold, but the new method found that point sources causing the exceedance must be so dim that they collectively become almost indistinguishable from the smooth emission.

“Our new analysis shows that the source must be so faint that it is almost indistinguishable from the emissions expected from dark matter annihilation,” said study author Nick Rodd, a scientist at Lawrence Berkeley National Laboratory.

This change has a major impact on the interpretation of pulsars. The paper says that to explain the signal, the median of the source count distribution requires about 200,000 sources in the galactic center region, but even the top 90 percent quantile still implies about 35,000 sources. This is far more than the hundreds to thousands of sources assumed in some previous scenarios. For comparison, one influential non-Poisson distribution template fit in 2016 implied approximately 200 sources.

The new model also found that the estimated source population is almost entirely below the one-photon threshold. This means that the light source is often so dim that it does not contribute even a single detected photon on average. At that level, it becomes very difficult to tell the difference between an unresolved point source and a truly smooth emission.

The team tested it firsthand. Another neural network trained to distinguish between point source radiation and Poisson-like radiation was able to exclude only 3% of the map as inconsistent with Poisson radiation with 95% confidence. In practical terms, this means that the signal appears to be much more compatible with the kind of smooth gamma-ray glow expected from dark matter than what previous analyzes had concluded.

However, this does not mean that dark matter has been discovered.

“The origin of the galactic supercenter is one of the longest-running debates in astrophysics,” List said. “Our study does not show that dark matter is responsible for the signal, but suggests that it is still too early to rule out this possibility.”

This sentence is important. Because the new paper is as much about removing objections as it is about proving a case. For many years, one of the strongest arguments against dark matter has been the claim that excess matter looks like a collection of statistically unresolved point sources. This analysis weakens that argument by showing that the supposed light source must be weak enough to blend into something very close to smooth radiation.

The study utilized 812 weeks of Fermi data collected from August 4, 2008 to February 23, 2024. The researchers looked at inner galaxies, defined as the region within 25 degrees of the galactic center, while masking bright galactic planes and some of the known cataloged sources. They modeled several types of gamma-ray emission, including the diffuse galactic background, isotropic emission, Fermi bubbles, disk populations of point sources, and the supercenter itself.

In particular, the addition of energetic information changed the estimated source number distribution of galactic center excess more than in the case of disk populations where point sources were already expected. This contrast lends weight to the idea that excess centers are a special case rather than a general failure of the method.

The authors are careful not to exaggerate their results. The biggest open issue remains background modeling.

The galactic center is a mess. Diffuse emission of gamma rays from normal astrophysical processes is bright, structured, and difficult to fully model. The researchers found that changing the diffuse background model could move the predicted median number of light sources from about 10,000 to about 100,000, even if the light sources were still far below the old bright light source image.

That sensitivity means the core debate is not over yet. This paper argues that further research is needed into improved diffusion models, wider energy ranges, and machine learning techniques that more directly incorporate smooth emission degeneracy into the analysis.

Still, the results are important because they reopen a space of possibilities that many researchers have increasingly treated as cornerstones. The glow at the Milky Way’s center may still come from a huge number of hidden millisecond pulsars, but their explanation looks more difficult than before. Meanwhile, dark matter is still alive and well in debate.

The immediate impact is not the detection of dark matter, but a reset of how we interpret galactic center signals. By incorporating photon energy into the analysis, this study weakens the leading statistical case against dark matter and raises the bar for pulsar-based explanations.

This provides a clearer goal for future exploration. Sophisticated machine learning tools that can improve background models, track millisecond pulsars in radio and other observations, and test whether the Milky Way’s central glow is really smooth or just looks that way.

In a field where dark matter remains frustratingly out of reach, narrowing the debate about one of the most debated signals in the sky is a meaningful advance.