Tidal Disruption Events

Repeating Partial TDEs: Evolution over Multiple Outbursts

Stars grazing SMBHs on bound orbits may survive tidal disruption. Using 3D hydrodynamic simulations, I show that Sun-like stars lose progressively more mass in each encounter and are doomed to complete disruption, producing numerous weak flares followed by exponentially brightening luminous flares.

Repeating Partial TDEs: Lessons Learned from ASASSN-14ko

Inspired by the discovery of the periodic nuclear transient ASASSN-14ko, I studied the disruption and fallback of stars on eccentric orbits around SMBHs. I show that eccentric TDEs can produce significantly shorter-lived flares than parabolic TDEs, and that the lack of amplitude evolution in ASASSN-14ko favors an evolved star on a grazing orbit.

Type Ia Supernovae

SN 2022joj: A Peculiar SN Ia from an Asymmetric Helium-shell Double Detonation

SN 2020jgb is a peculiar SN Ia showing dramatic red-blue color evolution on the rise. Its early-time and late-time observables independently point to an asymmetric double detonation of a sub-Chandrasekhar-mass white dwarf, with a viewing angle opposite the ignition point of the shell detonation.

SN 2020jgb: A Peculiar SN Ia Triggered by a Thick-Helium-shell Detonation

SN 2020jgb is a peculiar SN Ia whose unusually red colors, strong blue line blanketing, and prominent helium absorption in the NIR are best explained by a thick helium-shell (~0.1 solar mass) detonation atop a sub-Chandrasekhar-mass white dwarf.

ML for Transient Astronomy

LS-PSC: Morphological Star-Galaxy Classification for Legacy Surveys

A supervised machine learning approach using XGBoost to separate point sources from resolved galaxies in DESI Legacy Surveys (LS). The LS Point Source Catalog (LS-PSC) provides classification scores for ~3 billion sources, achieving superior accuracy compared to traditional morphological typing and enabling real-time extragalactic transient identification.

Adopted By

Open-source Software for Data Processing

HostSub_GP: Galaxy Background Removal in Transient Spectroscopy

A powerful new framework (powered by JAX) for precise host-galaxy subtraction in long-slit transient spectroscopy. It leverages archival imaging and Gaussian-process modeling to build high-fidelity models of galaxy light, enabling robust background removal in complex environments and substantially outperforming traditional techniques.

Publications

L. A. Kwok, C. Liu, S. W. Jha, et al., 2025, arXiv:2510.09760.
L. A. Kwok, S. Schulze, C. Liu, et al., 2025, Transient Name Server AstroNote, 129, 1.

BayeSpecFit: Fitting Blended Spectral Features in Supernovae

A Bayesian framework for fitting blended spectral features in SN spectra (powered by PyMC), using a combination of physical priors and flexible line profiles to robustly decompose complex absorption features and extract accurate velocity and equivalent width measurements.