About clusterps

clusterps is a Python package for the analysis of the velocity field in galaxy clusters. The code uses simulation-based inference (SBI) to forward model measurements of gas bulk velocities and velocity dispersions obtained by X-ray micro-calorimeter instruments such as XRISM/Resolve. Given a model for the emission measure distribution of a source and a pointing configuration, the code generates simulations of a Gaussian random field from any given power spectrum shape, and trains a neural network to learn the relation between power spectrum parameters and velocity measurements.

Motivation

The launch of the XRISM satellite has allowed for the first time to measure the velocity field of the gas in galaxy clusters. However, the relation between the measured bulk velocities and velocity dispersions, and the underlying velocity power spectrum, is not trivial to infer. This package attempts to fill this gap by forward modelling the measured velocities under the assumption that the velocity field can be described by a Gaussian random field with a parametric power spectrum. The code generates a large number of simulations spanning a pre-defined prior parameter range, and then trains a neural network to learn the relation between the model parameters and the mock velocity measurements. The trained model can then be applied to an observational dataset to retrieve the parameters of the power spectrum model.

Features

  • The code allows the user to easily define the emission measure model and the observational configuration

  • Inference can be performed for any user-defined power spectrum shape and prior range

  • Built-in coverage tests (TARP, 1D coverage) to test the accuracy of the model and its sensitivity to any given model parameter

  • Inference and posterior chain visualization

  • Easy-to-use model saving/reloading options

The current implementation has been developed in Python 3 and tested on Python 3.9+ under Linux and Mac OS.

Citation

If you find the software useful, please cite Eckert et al. 2022 . The associated paper also provides extensive validation tests of the provided software.