Researchers have developed neural operator surrogates to accelerate complex astrophysical simulations of black hole accretion. These models, including a Physics Informed Fourier Neural Operator (PINO) and an OFormer-style Transformer Neural Operator, are trained using data from the Black Hole Accretion Code (BHAC). The PINO model successfully learned to predict plasmoid formation in special-relativistic resistive MHD scenarios, a feat a data-only baseline could not achieve. The OFormer model was applied directly to adaptive mesh grids for simulating relativistic jets, marking a novel application of neural operators in this context. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Neural operator surrogates could significantly speed up complex astrophysical simulations, enabling broader parameter exploration in black hole research.
RANK_REASON This is a research paper detailing the application of neural operators to astrophysical simulations.