Picasso: a machine learning model to paint intracluster gas on gravity-only simulations

Post-processing techniques emulating baryonic physics in gravity-only simulations are a cornerstone of modern cosmology. In particular, predicting the thermodynamic properties of the intracluster gas is necessary to exploit these simulations for galaxy cluster cosmology in the millimeter-wave and X-ray domains. In this talk, I will introduce picasso, a model to predict thermodynamic properties of the intracluster medium from halos in gravity-only simulations. Predictions are based on the combination of an analytical gas model, mapping gas properties to the gravitational potential, and of a machine learning model to predict the model parameters for individual halos based on their scalar properties such as mass and concentration. The model is trained using pairs of gravity-only and hydrodynamic simulations with identical initial conditions. When trained on non-radiative hydrodynamic simulations, picasso can make remarkably accurate and precise predictions of intracluster gas thermodynamics. Training the model on full-physics simulations yields robust predictions as well, albeit with slightly degraded performance. The model can be trained to make accurate predictions from minimal information, at the cost of modestly reduced precision. The picasso model is publicly available as a Python package, which includes trained models that can be used to make predictions easily and efficiently, in a fully auto-differentiable and hardware-accelerated framework.