Level of Education of Students Involved

Undergraduate

Faculty Sponsor

Stan Zygmunt

College

College of Arts & Sciences (CAS)

Discipline(s)

Astronomy

ORCID Identifier(s)

0009-0008-0839-627X

Presentation Type

Poster Presentation

Symposium Date

Spring 4-30-2026

Abstract

The Event Horizon Telescope (EHT), in its famous 2019 image, proved that directly imaging a supermassive black hole at the event horizon scale is possible. Understanding these observations requires effort on multiple fronts, including comparing them against synthetic images generated by post-processing general relativistic magnetohydrodynamics (GRMHD) simulations, which describe how magnetized plasma behaves in the strong-field limit, such as near a black hole. Bridging the gap between these simulations and observations involves general relativistic ray-tracing (GRRT), which tracks how light rays propagate through curved spacetime, producing synthetic images that can be directly compared to EHT data. The quantitative techniques that best compare simulation to observation, such as Bayesian parameter inference and forward modeling, require a large sample of these synthetic images, so rapidly and repeatedly generating these images across extensive parameter spaces is paramount. This task remains computationally intractable without dedicated supercomputing resources.

To address these limitations, we present PHOENIX, an open-source graphics processing unit (GPU) implementation of the RAPTOR GRRT code. By using NVIDIA's Compute Unified Device Architecture (CUDA) to port RAPTOR's core ray-tracing and polarized radiative transfer algorithms to run on consumer GPU hardware, PHOENIX achieves orders-of-magnitude speedups while maintaining convergence in integrated Stokes parameters, the quantities that describe the polarization state of light, to within 2% of RAPTOR. This efficiency makes large-scale GRMHD post-processing and Bayesian inference workflows accessible even without HPC infrastructure, enabling more of the EHT community to explore GRMHD data.

Download

Share

COinS