Beyond the Horizon: GPU-accelerated plasma simulations and machine-learning analysis of accretion flows

The OVERFLOW project aims to constrain the unknown microphysics of black hole (BH) accretion-disk coronae with state-of-the-art radiative plasma simulations and machine-learning-accelerated observational data analysis. Currently, no first-principles models exist for the plasma energization mechanism operating in the corona. Similarly, previous astrophysical observations have not been able to constrain the exact geometry of the inner accretion flow or pinpoint the origin of the observed variability.

Project brings together expertise in ray timing & polarization observations,  first-principles modeling of radiative plasmas, and high-performance computing solutions. The common goal is to combine the latest computational tools, observational methods, and synergistic expertise within the consortium to illuminate the physics of the inner accretion flows.

The first aim is to constrain the possible energy dissipation mechanisms by comparing radiative plasma simulations to the latest BH polarization and variability observations. The second aim is to constrain the accretion flow geometry by constructing a bottom-up, physics-first flow model that is directly validated against spectro-polarimetric observations. Thirdly, using the radiative output from long-duration plasma physics simulations and machine-learning-accelerated new data analysis techniques to constrain the origin of the observed (quasi-periodic) X-ray variability. Finally, all these tasks are made possible by simultaneously porting open-source plasma code to a machine-agnostic form by adopting a performance portability library.

Overall, OVERFLOW will open a new, first-of-its-kind possibility to validate plasma physical models against astrophysical BH observations. It will also support the national and global high-performance scientific software infrastructure by providing the community with highly optimized but user-friendly radiative plasma code. CSC plays a key technical and computational role in the OVERFLOW project. Its main responsibility is to enable and optimize the high-performance computing side of the research.

This project has received funding from the Research Council of Finland​​​​​​​ under funding decision No 372880.