As early as the 1930s, Fritz Zwicky noticed that galaxies in clusters move too fast to be held by the visible mass. Thus the mystery of dark matter was born. Today, galaxy clusters—giant reservoirs of dark matter and hot gas glowing in X-rays—serve as key laboratories for cosmology. But to extract precise cosmological parameters from them, we need accurate scales.
The authors turned radial gas profiles—temperature, density, pressure, and mass—into a data sequence, since each successive layer is physically linked to the previous one. A gated recurrent network (GRU), trained on thousands of hydrodynamic simulations from the TheThreeHundred project, "reads" this sequence from the center outward and predicts the total enclosed mass. An ensemble of 50 models also provides robust uncertainties.
The trained network achieves a prediction scatter of about 5% over most of the radius—by a factor of 2 better than traditional hydrostatic estimates. There is no bias, even for highly disturbed, merging systems. The model generalizes to simulations with different physics (active galactic nucleus feedback and others), as well as to sparse, "observational" profiles with just 7–14 data points. Application to real data from the REXCESS and X-COP surveys (spectroscopy at redshifts up to z~0.1) showed that classical hydrostatic masses are systematically underestimated by about 10% for the most massive clusters.
The method provides a physically motivated, unbiased way to weigh clusters without assuming equilibrium. This will strengthen cosmological tests, especially with new surveys like eROSITA.
The technique can be extended to the Sunyaev–Zeldovich effect and gravitational lensing, creating unified multi-wavelength scales, and applied to the vast CHEX-MATE data.
It will improve the precision of cosmological parameters and the calibration of mass–observable relations.
Apply to the full CHEX-MATE sample and incorporate additional observables.
It is directly linked to questions about the nature of dark matter and the distribution of baryons in large-scale structure.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors converted radial gas profiles (temperature, density, pressure, mass) into sequences and trained a GRU ensemble on thousands of hydrodynamic simulations from The Three Hundred to predict the total enclosed mass of clusters. The model achieves a scatter of ~5% — twice as good as hydrostatic estimates — and shows no bias even for disturbed systems. Applied to REXCESS and X-COP data, it revealed that classical masses are systematically underestimated for the most massive clusters.
Where its strength is
The strength of this work lies in turning radial profiles into sequences that a GRU reads from the center outward, capturing the physical connection between layers without assuming equilibrium. This provides a data-driven way to weigh clusters, directly linking simulations and observations.
Where it could go next
Worth trying
Why this may matter
This approach could become a standard tool for next-generation cosmological surveys (eROSITA, Athena), where accurate cluster masses are critical for measuring dark energy parameters and structure growth.
Nearby works in our archive
The closest work in the archive sits at 0.66 by meaning — the usual level for our corpus.
as written by the authors arXiv:2607.20721v1
Accurate cluster mass measurements are crucial for cosmology, yet conventional hydrostatic equilibrium (HSE) methods can suffer from systematic biases, particularly in dynamically disturbed systems. We present a gated recurrent unit (GRU) based deep learning framework for predicting three-dimensional mass profiles of galaxy clusters from spherically averaged intra-cluster medium (ICM) radial profiles. By treating ICM profiles as sequential data, the GRU captures radial dependencies and naturally handles profiles with different radial samplings. We train and validate the model using high-resolution hydrodynamical simulations from The Three Hundred Project, achieving unbiased mass predictions with a typical 1$$\sigma$$ scatter of $$\sim$$5% over most of the cluster region, significantly improving upon HSE estimates. The model provides radius-dependent uncertainty estimates and remains robust against variations in data quality and cluster morphology. When trained jointly on independent simulation suites (GIZMO-SIMBA and GADGET-X), it successfully generalises across both simulations. Feature importance analysis shows that enclosed gas mass is the dominant predictor, with pressure and temperature providing additional information on the radial mass distribution. We further apply the GRU model to X-ray observations of the REXCESS and X-COP cluster samples from XMM-Newton and compare the inferred mass profiles with HSE estimates. The HSE masses are systematically lower than the GRU predictions for the higher-mass X-COP sample, while the REXCESS sample shows mass differences that are close to zero on average. This work provides a data-driven framework for cluster mass inference that bridges simulations and observations and can be extended to multi-wavelength datasets, including Sunyaev-Zel'dovich and optical observations.
The nearest massive cluster, Coma, weighs as much as a million billion Suns, but 85% of that mass is still invisible.