AREPO-RSG: Aspherical Circumstellar Material and Winds from Pulsating Dusty Red Supergiants in Global 3D Radiation Hydrodynamic Simulations

TL;DR

This study addresses the origin of dense, confined CSM around hydrogen-rich SNe by performing global 3D radiation-hydrodynamic simulations of red supergiant envelopes with AREPO-IDORT, initialized from MESA models and augmented with an artificial core. The results show that large-amplitude radial pulsations lift surface material to radii up to ~$3\times 10^{14}$ cm, forming a bound inner CSM of ~0.01 $M_\odot$, and that dust formation leads to anisotropic, dust-driven outflows at rates ~$(1\!-\,10)\times 10^{-6}$–$10^{-5}$ $M_\odot\,\mathrm{yr^{-1}}$, with 3D convection imparting strong angular density variations and clumpy morphology. The simulations yield a two-zone CSM density profile—an inner bound atmosphere plus an outer wind—that broadly matches the CSM inferred for SN 2013fs and aligns with an analytical description; the CSM is highly aspherical due to surface convection, suggesting that CSM in many Type II SNe may be dominated by bound, pulsation-driven material rather than steady winds or binary interactions. The work implies that 3D effects must be incorporated in SN modeling (potentially via effective clumping) and provides a 1D analytical CSM model for use in SN light-curve and spectral modeling, with testable predictions that pulsating progenitors and confined CSM should be common among hydrogen-rich SNe and detectable with upcoming Rubin, ULTRASAT, and UVEX observations.

Abstract

Recent observations have revealed a surprisingly large fraction of hydrogen-rich supernovae (SNe) interacting with dense confined circumstellar material (CSM), whose origin is heavily debated. Exploiting our recent implementation of a sophisticated radiation transport scheme in the moving-mesh code AREPO, we perform full-sphere 3D radiation hydrodynamic simulations of red supergiant envelopes. For $10\, M_\odot$ and $20\, M_\odot$ core-carbon-burning stars, we find that large-amplitude radial pulsations lift the surface material of density $10^{-14}$-$10^{-12}\; \mathrm{g\; cm^{-3}}$ to the circumstellar environment up to $3\times10^{14}$ cm, consistent with the inferred density for the interacting SN 2013fs. There, radiation acts on dust to drive highly anisotropic outflows of $10^{-6}$-$10^{-5}\, M_\odot\, \mathrm{yr^{-1}}$. The total CSM masses for both simulations are $\sim 0.01\, M_\odot$. Due to convection, the CSM density structure has order-of-magnitude angular variations, dominated by large-scale asymmetries. We suggest that (1) the CSM around the progenitor is bound material instead of a widely-assumed steady wind, (2) highly aspherical CSM is common and can be created by surface convection rather than only from binary interactions, and (3) 3D effects need to be incorporated in 1D SN modeling, potentially via effective clumping. Based on our simulations, we propose a 1D analytical CSM model to be directly used for SN observable modeling. We predict that progenitor pulsations (seen in SN 2023ixf) and highly-confined CSM (seen in SN 2013fs) should be common among most hydrogen-rich SNe. This can be tested with progenitor monitoring using Rubin Observatory and near-future high-cadence surveys such as ULTRASAT and UVEX.

Figures (13)

• Figure 1: Locations of two 3D pre-SN AREPO-RSG simulations on the Hertzsprung-Russell diagram. The evolutionary tracks of MESA models are indicated in black solid lines. We mark the 1D MESA models (empty stars) as initial conditions for the 3D simulations (filled stars with errorbars indicating the $3\sigma$ variations due to temporal variability). The effective temperatures for the simulations are spherically-averaged (see Appendix \ref{['apx:method:analysis']}). Background gray scatter dots indicate observed Galactic RSG population obtained through TiO bands levesque2005ApJ or SED fitting gazak2014ApJ. The small gray stars mark the Type IIP/L supernova progenitors identified in pre-explosion images compiled by vandyk2025Galaxies, where we highlight the well-studied interacting SNe 2023ixf and 2024ggi. We also highlight the dusty progenitor of SN 2025pht detected by the James Webb Space Telescope kilpatrick2025arXive-prints. Both the absolute values and the uncertainties of the bolometric luminosity may be significantly underestimated beasor2025ApJ. Background gray lines show contours of constant radius.
• Figure 2: Dense CSM episodically lifted by semi-regular pulsation in the $10\, M_\odot$ simulation (left) and $20\, M_\odot$ simulation (right). From top to bottom, we show the bolometric luminosity, CSM mass, spherically-averaged density, and spherically-averaged radial velocity. In the top two rows, black solid lines show the general trends of variations filtering out the high-frequency variability. The bolometric luminosity from the MESA models used as initial conditions are indicated with gray dashed horizontal lines. In the bottom two rows, contour lines indicate the iso-density levels $[10^{-8},10^{-12},10^{-14},10^{-16}]$$\mathrm{g\, cm^{-3}}$. The cyan lines indicate the spherically-averaged Rosseland radius defined in Appendix \ref{['apx:method:analysis']}. The right-hand side of the vertical dotted lines indicate the relaxed phase as defined at the end of Section \ref{['sec:method']}.
• Figure 3: The CSM density structure from our 3D simulations is consistent with the range of densities inferred for SN 2013fs. Different colored lines indicate the spherically-averaged density profiles at different times, for the $10\, M_\odot$ simulation (orange) and the $20\, M_\odot$ simulation (red), which are much more extended than the initial conditions from MESA (black solid lines). The white-edged solid lines represent the density profiles averaged both in time and angles. The black dashed lines show the density profiles described by the analytical 'two-zone model' detailed in Section \ref{['sec:result:agb']}. We also highlight the inferred density profile from three well-studied interacting SNe, SN 2013fs yaron2017Nat.Phys., 2023ixf nayana2025ApJ and SN 2024ggi zhang2024ApJajacobson-galan2024ApJshrestha2024ApJertini2025AAchen2025ApJ, as labeled. The background gray solid lines indicate density profiles for constant mass-loss rates assuming a constant wind velocity $30\, \mathrm{km\, s^{-1}}$.
• Figure 4: Convection creates a clumpy stellar surface, and aspherical circumstellar material, leading to an anisotropic outflow. Left: A 2D density slice of the $10\, M_\odot$ simulation at 60.5 years. Right: Projected density maps along three spheres indicated by the dashed circles in the left plot (where the color at the edge of each density map corresponds to that of the appropriate circle). An interactive 3D visualization can be found here: https://jingzema.com/AREPO-RSG/arepo_rsg_csm_fast.html
• Figure 5: Time sequence of the convective envelope variations within one pulsation cycle for the $10\, M_\odot$ simulation. The top row shows the variability of the absolute bolometric magnitude $M_\mathrm{bol}$ in black and the spherically-averaged Rosseland radius $\langle R_\mathrm{ross}\rangle$ in orange. During this one pulsation cycle, we select 4 snapshots equally spaced in time (indicated by gray vertical lines in the top row), and plot the bolometric intensity looking from the x axis (second row), y-z mid-plane slice of the density (third row), and y-z mid-plane slice of the radial velocity (last row).