star formation, star cluster dynamics, prestellar cores, epoch of reionization, radiation-hydrodynamic code, n-body code, computational astrophysics.
Development of Next-generation AMR Radiation Hydrodynamics Code on GPUs
Link to heading
Quokka (Wibking & Krumholz 2022, He, Wibking & Krumholz 2024a, He, Wibking & Krumholz 2024b) is a two-moment AMR (Adaptive Mesh Refinement) radiation hydrodynamics code for astrophysical simulations. It integrates self-gravity, particles, chemistry modules, and magnetic fields (in progress) and is optimized for both CPUs and GPUs. Since 2023, I have been a core developer on the Quokka team. For additional details about this code, visit the Quokka GitHub page or the Quokka documentation page.
Massive Magnetically-critical Prestellar Cores and Formation of Large Turbulent Circumstellar Disks
Link to heading
In He & Ricotti 2023, we investigate the formation and collapse of prestellar cores at ~10 AU resolution in a set of radiation-magneto-hydrodynamic simulations of giant molecular clouds. We adopt, for the first time to our best knowledge, realistic initial/boundary conditions by zooming-in onto individual massive prestellar cores within the GMC. We identify two primary fragmentation modes: quasi-spherical and filamentary. In both modes the fragments eventually become embedded in a quasi-steady accretion disk or toroid with radii ∼ 500 − 5000 AU. Our simulations reveal that each core converts nearly 100 percent of the gas mass into a few massive stars forming near the disk center. The most massive cores, exceeding tens of solar masses, forms a cluster through competitive accretion, while smaller cores tend to align with the turbulent core model.
In a subsequent work (He & Ricotti 2024), we explore how do large Keplerian disks form in magnetically critical or near-critical cores. We discover that the magnetic field topology within these cores is highly turbulent and incoherent, which diminishes the effect of magnetic braking by roughly an order of magnitude. This substantial turbulence, driven by the non-axisymmetric gravitational collapse of the gas, primarily supports the vertical structure of the disks.
Escape of Lyman Continuum Photons from Resolved Stellar Populations
Link to heading
I published the first systematic study of the escape of ionizing photons from resolved stars in GMCs into the intercloud gas via a large set of radiation-MHD simulations (He, Ricotti & Geen 2020). We found that $f_{\rm esc}$ increases with decreasing mass and with increasing mean density of a GMC. GMCs with densities typical of local star formation regions have negligible $f_{\rm esc}$ (below 10%). This relation is explained by a simple model where two timescales are compared: the lifetime of the dominating UV sources (O/B stars) and the cloud destruction time. The former is nearly constant for massive clusters and slightly higher for less massive ones where O/B stars are lacking. The latter is found to be several cloud crossing times of the HII fronts at ~7 km/s, which is the characteristic timescale of photoionization feedback. We concluded that the sources of ionizing photons responsible for the epoch of reionization, one of the most important yet poorly understood stages in cosmic evolution, must have been very compact star clusters forming in dense environments different from today's galaxies.
Understanding Photoionization Feedback in Star Formation from Resolved Simulations
Link to heading
"Pillars of Creation" caused by photoionization feedback (He, Ricotti & Geen, 2019). The thermal pressure resulting from photoionization heating efficiently disperses the parent cloud. I show that photoionization feedback is efficient at dispersing dense molecular clouds before the onset of supernova explosions. Based on a set of simulations of collapsing molecular clouds extending a wide range of masses and densities, I find a strong linear correlation between the star formation timescale regulated by photoionization feedback, $t_{\rm SF}$, and the cloud crossing time of HII-fronts, $t_{\rm cr}$, typically at a speed of $c_s \sim 7$ km/s. I find the following empirical relation: the ratio of $t_{\rm SF}$ to the free-fall time, $t_{\rm ff} = (\pi/2) (R/v_{\rm esc})$, roughly equals $v_{\rm esc}/2.6$ km/s, where $v_{\rm esc}$ is the escape velocity at the cloud surface. We also find that the star formation efficiency per free-fall time scales with the density to the second power and approaches ~20 percent in very dense scenarios. Molecular clouds with escape velocities above ~7 km/s and high density (> 1000 cm$^{-3}$) deposit over 15 percent of the gas into stars and form gravitationally bound star clusters which appears to be globular cluster progenitors.
A simulation snapshot from He, Ricotti & Geen, 2019 that demonstrates the Adaptive Mesh Refinement technique of the RAMSES code. A large part of my research relies on running supercomputer simulations using RAMSES-RT, a radiation-magneto-hydrodynamic code for general purpose simulations of self-gravitating fluid dynamics. RAMSES-RT is one of few radiation-MHD codes available that is suitable for star formation simulations. The utilization of the Adaptive Mesh Refinement strategy enables the grid-based hydro solver to deal with extremely high dynamic range.
In He (2021), I present a new analytic method for calculating galaxy two-point correlation functions (TPCFs) that is accurate, efficient, and applicable to surveys with regular geometries. Closed-form formulas are derived for the normalized random-random pair counts RR for rectangular, cuboidal, circular and spherical survey volumes. Algorithms are also suggested to compute the normalized data-random pair counts DR analytically by fully accounting for edge effects. When tested on a galaxy catalog from the EAGLE simulation, this new analytic method computes RR and DR with perfect accuracy and zero variance, with speeds 3-6 orders of magnitude faster than Monte Carlo methods and 2.5 orders of magnitude faster than tree-based algorithms. For surveys with masks, irregular shapes or weighted patterns, the method is limited, but it provides significant speed improvements for basic TPCF calculations .
The emission from neutron stars and accretion disks in low-mass X-ray binaries is not isotropic. Previous predictions of the anisotropy factors for burst flux assumed a flat disk. However, recent observations showed that the reflection fraction can reach ~6 at the tail of a superburst, which is much higher than what a flat disk can account for. To investigate this discrepancy, I created numerical models to calculate the anisotropy factors for different disk shapes (He & Keep, 2016). One model includes a disk with a thick bulge in the inner region. The anisotropy factors of the direct and reflected burst flux, as well as the anisotropy of the persistent flux, are presented. The results show that reflection fractions larger than unity are produced when the inner accretion disk steeply increases in height, blocking part of the star from view. This geometry could be induced by the X-ray burst if X-ray heating causes the inner disk to puff up. These findings suggest that the anisotropy of radiation in low-mass X-ray binaries is strongly influenced by the shape and height of the inner accretion disk.
