Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ASURA-FDPS-ML: Star-by-star Galaxy Simulations Accelerated by Surrogate Modeling for Supernova Feedback

Keiya Hirashima, Kana Moriwaki, Michiko S. Fujii, Yutaka Hirai, Takayuki R. Saitoh, Junnichiro Makino, Ulrich P. Steinwandel, Shirley Ho

TL;DR

The paper tackles the computational bottleneck of resolving supernova feedback at star-by-star precision in galaxy simulations. It introduces ASURA-FDPS-ML, a hybrid framework that couples direct simulations with a data-driven surrogate (enhanced by Gibbs sampling) to model SN feedback over a short time window, allowing larger timesteps while preserving essential dynamics. The surrogate, based on a U-Net, is trained on high-density SN environments and reproduces the evolution of density, temperature, and velocity fields, yielding star formation histories and outflow rates in agreement with fully resolved runs and achieving approximately $\sim 75\%$ cost savings. This approach enables multi-scale, star-by-star galaxy simulations that capture the inhomogeneous SN shell expansion within a turbulent ISM, bridging the gap between microphysical SN processes and galactic-scale evolution.

Abstract

We introduce new high-resolution galaxy simulations accelerated by a surrogate model that reduces the computation cost by approximately 75 percent. Massive stars with a Zero Age Main Sequence mass of more than about 10 $\mathrm{M_\odot}$ explode as core-collapse supernovae (CCSNe), which play a critical role in galaxy formation. The energy released by CCSNe is essential for regulating star formation and driving feedback processes in the interstellar medium (ISM). However, the short integration timesteps required for SNe feedback have presented significant bottlenecks in astrophysical simulations across various scales. Overcoming this challenge is crucial for enabling star-by-star galaxy simulations, which aim to capture the dynamics of individual stars and the inhomogeneous shell's expansion within the turbulent ISM. To address this, our new framework combines direct numerical simulations and surrogate modeling, including machine learning and Gibbs sampling. The star formation history and the time evolution of outflow rates in the galaxy match those obtained from resolved direct numerical simulations. Our new approach achieves high-resolution fidelity while reducing computational costs, effectively bridging the physical scale gap and enabling multi-scale simulations.

ASURA-FDPS-ML: Star-by-star Galaxy Simulations Accelerated by Surrogate Modeling for Supernova Feedback

TL;DR

The paper tackles the computational bottleneck of resolving supernova feedback at star-by-star precision in galaxy simulations. It introduces ASURA-FDPS-ML, a hybrid framework that couples direct simulations with a data-driven surrogate (enhanced by Gibbs sampling) to model SN feedback over a short time window, allowing larger timesteps while preserving essential dynamics. The surrogate, based on a U-Net, is trained on high-density SN environments and reproduces the evolution of density, temperature, and velocity fields, yielding star formation histories and outflow rates in agreement with fully resolved runs and achieving approximately cost savings. This approach enables multi-scale, star-by-star galaxy simulations that capture the inhomogeneous SN shell expansion within a turbulent ISM, bridging the gap between microphysical SN processes and galactic-scale evolution.

Abstract

We introduce new high-resolution galaxy simulations accelerated by a surrogate model that reduces the computation cost by approximately 75 percent. Massive stars with a Zero Age Main Sequence mass of more than about 10 explode as core-collapse supernovae (CCSNe), which play a critical role in galaxy formation. The energy released by CCSNe is essential for regulating star formation and driving feedback processes in the interstellar medium (ISM). However, the short integration timesteps required for SNe feedback have presented significant bottlenecks in astrophysical simulations across various scales. Overcoming this challenge is crucial for enabling star-by-star galaxy simulations, which aim to capture the dynamics of individual stars and the inhomogeneous shell's expansion within the turbulent ISM. To address this, our new framework combines direct numerical simulations and surrogate modeling, including machine learning and Gibbs sampling. The star formation history and the time evolution of outflow rates in the galaxy match those obtained from resolved direct numerical simulations. Our new approach achieves high-resolution fidelity while reducing computational costs, effectively bridging the physical scale gap and enabling multi-scale simulations.

Paper Structure

This paper contains 12 sections, 9 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Simulation code
  4. Initial Condition
  5. Star formation model
  6. Stellar feedback
  7. Required timestep for SN feedback
  8. Hybrid on-the-fly ML-driven simulations
  9. Dataset for Our Surrogate Modeling
  10. Test for our surrogate model of SN feedback
  11. Results for an isolated dwarf galaxy simulation
  12. Morphology Comparison

Figures (2)

  • Figure 1: Schematic diagram for our presenting hybrid on-the-fly ML-driven framework.
  • Figure 2: Comparison between simulations and ML for an isolated SN. Upper: The numerical simulation results. Lower: The numerical simulation results with our surrogate SN feedback model. Cross section of snapshots at $t=0, ~10^5, ~ \mathrm{and} ~4 \times 10^5$ years are listed up from left to right.