SuperSNEC: Fast and Accurate Light Curve Production for Large Hydrodynamic Model Grids Using Adaptive Gridding

TL;DR

Applied to SN 2011dh, SN~1993J, and SN 2020oi, SuperSNEC recovers light-curve parameters consistent with the literature; in particular, SN 2020oi is well reproduced by a purely radioactive model, with no clear evidence that an additional power source is required.

Abstract

We present SuperSNEC, an accelerated version of the SuperNova Explosion Code (SNEC) designed for rapid production of large radiation-hydrodynamic model grids using low-zone-count simulations ($\sim100$ zones). The main advance is adaptive gridding of the computational grid, which preserves light-curve fidelity relative to a high-resolution SNEC baseline ($\sim1000$ zones) while delivering a runtime improvement of ${\sim}420\times$. SuperSNEC also includes solver optimizations, optimized radioactive-energy deposition and ray-tracing, improved $^{56}$Ni mixing controls, and a smooth photosphere luminosity correction that suppresses low-resolution artifacts. We quantify the speed-accuracy trade-off for a 100-zone configuration against a 1000-zone reference and define baseline settings for efficient large-grid inference of stripped-envelope supernovae. Our optimized 100-zone setup achieves an RMS light-curve residual of $0.022$ mag relative to the 1000-zone reference, at a runtime of $<2$ seconds per model. Applied to SN 2011dh (Type IIb), SN~1993J (Type IIb), and SN 2020oi (Type Ic), SuperSNEC recovers light-curve parameters consistent with the literature; in particular, SN 2020oi is well reproduced by a purely radioactive model, with no clear evidence that an additional power source is required.

Figures (2)

• Figure 1: Bolometric light-curve comparison of five configurations against a 1000-zone SNEC reference (black dashed line). Top panels: luminosity; bottom panels: magnitude residuals $\Delta m = -2.5\log_{10}(L/L_\mathrm{ref})$, with the green band marking $\pm0.1$ mag. Left: full evolution (0--58 d); right: early-time zoom (0--3 d). The original SNEC code at $N=100$ (purple dotted; $\Delta m_\mathrm{all}=0.167$ mag) diverges strongly at $t\lesssim0.5$ d and $t\gtrsim10$ d. SuperSNEC with the same fixed computational grid (orange dot-dashed) improves via adaptive $\rm ^{56}Ni$ cadence and solver tuning. The adaptive-runtime grid brings a further factor-of-three improvement: the baseline ($N=100$, $n_\mathrm{quad}=70$; blue solid line) and fast ($N=60$, $n_\mathrm{quad}=50$; red solid line) configurations both remain well within 0.1 mag across all phases. The SuperSNEC fixed-grid model uses $f_\mathrm{Ni}=0.10$; the baseline uses $f_\mathrm{Ni}=0.20$, $n_\mathrm{quad}=70$; the fast model uses $f_\mathrm{Ni}=0.20$, $n_\mathrm{quad}=50$. All three share $\alpha_\mathrm{surf}=7$, $t_\mathrm{relax}=5$ d, $\Delta t_{\mathrm{Ni,min}}=5\times10^4$ s, $\Delta t_{\mathrm{Ni,max}}=5.456\times10^5$ s. The original SNEC uses $\Delta t_{\mathrm{Ni,min}}=5\times10^4$ s with no adaptive cadence.
• Figure 2: Bolometric light-curve (left) and photospheric velocity (right) comparison of $N$=100 SuperSNEC models (blue lines) to SN 2011dh (top), SN 1993J (middle), and SN 2020oi (bottom) data (black markers). Velocities for SN 2011dh and SN 1993J are Fe ii$\lambda5169$ absorption-minimum measurements; SN 2020oi uses Si i$\lambda10460$ as Fe ii$\lambda5169$ could not be reliably measured. Observational data: SN 2011dh bolometric LC from Ergon2014Ergon2015, velocities from Marion2014; SN 1993J bolometric LC constructed from Richmond1994 photometry with UV/IR blackbody corrections, velocities measured from UNLV Supernova Archive spectra Barbon1995; SN 2020oi bolometric LC constructed from the Rho2021 optical photometry using a blackbody bolometric correction, with Si i$\lambda10460$ velocities also from Rho2021.