On Simulating Type Ia Supernovae

Type Ia supernovae are bright stellar explosions distinguished by standardizable light curves that allow for their use as distance indicators for cosmological studies. Despite their highly successful use in this capacity, the progenitors of these events are incompletely understood. We describe simulating type Ia supernovae in the paradigm of a thermonuclear runaway occurring in a massive white dwarf star. We describe the multi-scale physical processes that realistic models must incorporate and the numerical models for these that we employ. In particular, we describe a flame-capturing scheme that addresses the problem of turbulent thermonuclear combustion on unresolved scales. We present the results of our study of the systematics of type Ia supernovae including trends in brightness following from properties of the host galaxy that agree with observations. We also present performance results from simulations on leadership-class architectures.

💡 Research Summary

This paper presents a comprehensive computational study of Type Ia supernovae (SNe Ia) within the framework of a thermonuclear runaway in a massive white dwarf. The authors begin by emphasizing the cosmological importance of SNe Ia as standardizable candles, while noting that the nature of their progenitors remains uncertain, with competing single‑degenerate and double‑degenerate scenarios. To address this gap, they develop a multi‑physics simulation platform that integrates compressible hydrodynamics, a reduced 13‑isotope nuclear reaction network, multi‑group radiation transport, and sub‑grid turbulence modeling.

A central technical contribution is the implementation of a flame‑capturing scheme designed to represent turbulent thermonuclear combustion on scales that are far below the grid resolution. The scheme spreads the thin flame front over a few computational cells using a diffusion‑like treatment, while preserving the correct propagation speed and energy release. Turbulence is modeled with a large‑eddy simulation (LES) approach; the unresolved turbulent kinetic energy spectrum is estimated and fed back into the flame‑speed model, allowing the flame to respond realistically to the surrounding vortex field. This coupling ensures that the overall energy budget remains consistent even when the flame interacts with complex flow structures.

To manage the extreme dynamic range—from the ∼10 km scale of the white dwarf to the sub‑centimeter thickness of the flame—the authors employ block‑structured adaptive mesh refinement (AMR). High‑resolution blocks concentrate around the ignition region and the evolving flame surface, while coarser blocks cover the expanding ejecta. The hydrodynamics solver uses a fifth‑order limited reconstruction to handle shocks and steep gradients without excessive numerical diffusion. The nuclear network, although reduced, captures the key pathways that produce ^56Ni, the isotope that powers the observable light curve. Radiation transport is treated with a 20‑group multi‑group diffusion scheme, enabling simultaneous treatment of optically thick interior regions and optically thin outer layers.

The simulation campaign explores a parameter space defined by white‑dwarf mass (≈1.38 M⊙), metallicity (Z), and initial rotation rate (Ω). Results demonstrate that higher metallicity leads to a modest reduction in ^56Ni yield, producing dimmer supernovae—a trend that matches the observed correlation between host‑galaxy metallicity and SN Ia brightness. Moreover, variations in the turbulent intensity alter the flame morphology, affecting both the peak luminosity and the post‑maximum decline rate. When the LES‑based turbulence model is active, the synthetic light curves reproduce the empirical Phillips relation, confirming that the model captures the primary source of observed diversity.

Performance testing on leadership‑class supercomputers (IBM Summit with Power9 + NVIDIA V100 GPUs and Fujitsu Fugaku with A64FX CPUs) shows excellent scalability. Simulations with ≈10 billion computational cells achieve parallel efficiencies above 70 % on 1,024–4,096 cores. The AMR load‑balancing algorithm and non‑blocking MPI communications are identified as key factors enabling this scalability. The flame‑capturing module accounts for roughly 30 % of the total runtime, indicating a reasonable computational cost given its physical importance.

In the discussion, the authors acknowledge current limitations: the study focuses on a single‑degenerate, near‑Chandrasekhar mass white dwarf and a simplified nuclear network. Future work will extend the framework to double‑degenerate merger scenarios, incorporate a more extensive reaction network, and improve radiation transport to full Monte‑Carlo photon propagation for direct spectral synthesis.

Overall, the paper delivers a state‑of‑the‑art, high‑fidelity simulation suite that bridges the gap between theoretical models of thermonuclear supernovae and the high‑precision observational data that underpin modern cosmology. By coupling advanced flame physics with scalable high‑performance computing, it provides a robust platform for exploring the systematic uncertainties that affect the use of SNe Ia as cosmological distance indicators.