Simulated galaxy clusters match X-ray outskirts—but their cores look too faint

Testing the average cluster

Galaxy clusters grow as matter falls in from their surroundings, making their outskirts important places to study how these enormous structures assemble. But the hot gas there emits much less X-ray radiation than the gas near their centres. Stacking images of multiple clusters can strengthen the faint average signal and smooth out individual peculiarities.

Aleksei Kruglov, Natalya Lyskova, Ildar Khabibullin, Veronica Biffi and Klaus Dolag used this approach to compare clusters in the Magneticum cosmological simulations with an existing observational stack from the SRG/eROSITA telescope. Their study is available as a revised arXiv preprint, and its landing record reports acceptance by the Journal of High Energy Astrophysics, or JHEAP.

The main result is a close match across much of the clusters’ radial extent—not a perfect reproduction everywhere, and not a secure detection of cluster emission arbitrarily far into their surroundings.

Making simulations resemble observations

The researchers selected 34 massive simulated clusters for their main comparison with the previously published stack of 38 observed clusters. Using the PHOX software package, they generated predictions for thermal gas emission in the soft X-ray band from 0.3 to 2.3 kiloelectronvolts.

They then applied eROSITA’s energy-dependent collecting response and image smoothing to make the simulated images more comparable to the observations. Images were rescaled using each cluster’s R500: the radius enclosing an average density 500 times the Universe’s critical density at that cluster’s redshift. Appropriate brightness corrections accompanied this rescaling.

The team masked neighbouring haloes and bright substructures, with additional manual filtering in some cases, before combining the images. X-rays emitted directly by active galactic nuclei were excluded, although the simulated effects of those nuclei on the gas remained part of the underlying model.

The simulated emission included surrounding structures within the cluster’s redshift slice, more than roughly 100 megaparsecs deep. It was therefore not simply an isolated cluster viewed against empty space.

A strong match outside the core

Between approximately 0.4 and three times R500, the simulated and observed surface-brightness profiles agree within twice the estimated uncertainty. The comparison spans more than two orders of magnitude in brightness, supporting the simulation’s description of the average gas distribution over this range.

Closer to the centre, however, the simulated clusters are systematically fainter. Around 0.1 R500, their brightness falls below the observed profile by as much as a factor of two.

The authors consider imperfect modelling of feedback from active galactic nuclei the most plausible explanation. Energy released around accreting central black holes can redistribute gas and reduce its central density; in the simulations, that redistribution may be excessive. This is an interpretation rather than a demonstrated cause. The paper also considers metallicity assumptions and differences in cluster populations, but argues that these are less likely to explain the discrepancy.

Where the outskirts blend into their surroundings

Beyond roughly three times R500, the stacked simulated profile becomes increasingly dominated by a nearly radially constant but spatially fluctuating component. This local background includes surrounding structures, diffuse gas and residual substructures; it should not be mistaken for emission belonging solely to the central cluster.

Its estimated brightness is about five millionths of a count per second per square arcminute, nearly two orders of magnitude below the measured eROSITA background. Unlike the real observational background, the simulated component excludes instrumental and particle backgrounds, Galactic foreground emission and the cosmic X-ray background.

Subtracting a constant estimate of this local component leaves a highly uncertain outer profile. Beyond about three R500, the paper’s signal-to-noise estimate drops below one. Crucially, its uncertainties are based on brightness differences among four angular sectors, capturing uneven emission around the stack.

A population average is not a typical object

Expanding the simulated sample from 34 to 84 clusters produced similar profiles but did not overcome the low outer signal-to-noise ratio. This does not establish that larger surveys cannot help: the authors discuss a separate, much larger observational study reporting more extended emission, while noting its different uncertainty method.

Another caution comes from comparing the stack with the median of individual sector profiles. Outside R200c, approximately 1.6 R500, the median falls below the stack, indicating that scattered bright regions increasingly influence the average.

The study therefore supports Magneticum’s average cluster-gas distribution outside the core while identifying two limits: central feedback remains imperfectly reproduced, and the far-out stacked glow is a population-level measurement that need not resemble any typical individual cluster.