RESEARCH

My research projects in reverse chronological order. My full publication list is available in the CV section.
Click on the images for more details.


Ongoing: Meso-scale Simulations of Multiphase Accretion Flows onto Supermassive Black Holes

yz-projections of number density at various scales. Panels a and b are taken from the fiducial simulation (8 levels of SMR) to show the galaxy-scale dynamics, while the other panels are taken from the meso-scale simulation (20 levels of SMR). The color scales are logarithmic, and are allowed to span dynamically. The blue streamlines trace the motion of gas. The dashed circle in panel g denotes the accretion zone of the meso-scale simulation.

Current state-of-the-art numerical simulations of galaxies are largely successful at recreating the accretion flows morphology observed, especially on the galaxy-scale (~0.1 to hundreds of kpc) and the event-horizon-scale (~10^-2 pc). However, our current theoretical understanding of the accretion dynamics between the Bondi radius and the outer edge of the accretion disk (the meso-scale) lags behind observations. My thesis project aims to provide theoretical explanations for the observed meso-scale accretion flows. To achieve this goal, I run galaxy-scale MHD simulations of the accretion flow and relativistic jet feedback motivated by observations of the elliptical galaxy M87 using the adaptive mesh refinement code Athena++. Then, through a nested zoom-in technique, I perform meso-scale simulations of the same system to study the detailed physics of the accretion flow. These simulations will ultimately provide boundary and initial conditions at the horizon-scale for future GRMHD simulations as well as more accurate subgrid models for cosmological simulations. Look out for our first paper coming soon to the arXiv in November!
Advisor: Yuan Li (UMass Amherst)

Segmentation of Current Sheets in Magnetized Plasma Turbulence with Computer Vision

Panel (a): volume rendering of the projection of the current density onto the B field, j∥. Panel (b): xy-slice of j_par. Streamlines of the in-plane magnetic field b_perp are in the foreground. Panel (c): fiducial SOM segmentation, color coded by the cluster identified via aweSOM. Green and orange correspond to regions in between double current sheets, purple to current sheets aligned with B, and blue to current sheets antialigned with B. Blue contours are regions where j_par > 2 j_par,rms, and red contours are regions where j_par < -2 j_par,rms.

Magnetohydrodynamical (MHD) turbulence is a ubiquitous phenomenon in the universe. The intermittency in plasma turbulence causes the formation of current sheets -- two-dimensional sheet-like structures of intense current flows. The magnetic field across these sheets usually experiences a reversal in polarity. Through tearing-mode instability, current sheets are disrupted, leading to magnetic reconnection, which is responsible for heating the plasma and accelerating particles. As a Pre-doctoral Fellow at the Center for Computational Astrophysics, I developed a machine learning framework, aweSOM, to automatically segment and track current sheets in 3D plasma simulations based on the self-organizing map technique. We find that in many test cases, the algorithm can effectively segment the current sheets from simulations.
We recently published both the science paper in ApJ Letters: 2025ApJ...985L..31H, and the package in the Journal of Open Source Software: 2025JOSS...10.7613H
Advisors: Joonas Nättilä, Jordy Davelaar, and Lorenzo Sironi (CCA / Columbia University)

Measuring Turbulence in the Interstellar Medium with Young Stars

Figure 1 of our recent paper (arXiv:2205.00012), showing four of the nearest star-forming region to the Solar system: Orion, Ophiuchus, Perseus, and Taurus. In the foreground are young stars, in the background is Hα intensity (top) and Hα line-of-sight velocity (bottom).

Stars are born out of the turbulent molecular gas of the interstellar medium. Once formed, they decouple from the surrounding gas, but still retain their pre-natal kinematics. Using Gaia astrometric measurements in conjunction with APOGEE-2, we calculate the velocity structure function of stars in four nearby star-forming regions and compare them to the structure functions found in Hα and CO gas.
In our 2022 paper, we found that the structure functions of the four star-forming regions in the Solar neighborhood shows a universal scaling of turbulence when traced by young stars. We also found evidence of local supernova energy injection in Orion and Ophiuchus, which is supported by observational studies.
This work is a follow-up to our 2021 paper, which investigates young stars in the Orion Complex specifically.
Read our press release here. Watch our interview for the American Astronomical Society's Journal Author Series here.
I also mentored a post-baccalaureate student, Benjamin Velguth, who recently published a follow-up work, examining the role of turbulence, gravity, supernovae, and magnetic fields in star formation through stellar kinematics.
Advisor: Yuan Li (UMass Amherst)

Weak Emission-Line Quasars (WLQs) in the Context of C Ⅳ Emission-line Properties


Preliminary result from our recently submitted paper. We found a strong correlation between the C IV rest-frame equivalent widths and the C IV || Distance parameter, which holds for all quasars in our sample.

WLQs are a peculiar subset of luminous active galactic nuclei whose emission spectra show extremely weak emission line profiles of Lyα+N V λ1240 and/or C IV λ1549. Much effort has been invested to explain the weakness in emission profiles of these quasars over the last two decades. My current project proposes a new way of looking at WLQs not as a disjoint subset of quasars but rather as quasars that lie preferentially towards the extreme end of the C IV || Distance parameter space.
Read our paper: 2023ApJ...950...97H.
Advisor: Ohad Shemmer (UNT)

Similarity Mapping of the Milky Way Using Neural Style Transfer

Background: Cartesian projection of the Milky Way H-alpha map along the Galactic equator, with the Galactic center at the middle of the frame. Foreground: each white rectangle is a region identified by the VGG-19 algorithm as 'similar' to the Orion region. We observe that all highlighted frames are regions with high star formation rates, indicating that the neural network picked up this information from the flux and velocity input by color channels.

As a graduate student advisor, I helped M.S. students in the A.I. summer program at UNT apply the pre-trained VGG-19 deep neural network architecture to the Milky Way map from the Wisconsin H-Alpha Mapper survey. Using the Orion Molecular Cloud Complex as a reference image, neural style transfer was able to identify various star-forming regions in the Milky Way. This result shows neural style transfer's promise as a quick targets identification pipeline for future sky surveys.
Read our project summary poster here.
Advisors: Mark Albert and Yuan Li (UNT - now at UMass)

Generating Initial Data for Binary Neutron Stars Merger Simulations


Isobaric contours plot of a binary neutron stars merger. The more compact star is 1.7 solar mass, while the puffier star is 1.14 solar mass. Each contour traces a curve of equal gas density, and brighter color indicates higher density.

I worked as an undergraduate research student at the Rochester Institute of Technology's Center for Computational Relativity and Gravitation from 2018 to 2020. There, I modified and documented the LORENE initial data code to generate physically motivated binary neutron stars pre-merger. Furthermore, I performed several GRMHD simulations of binary neutron stars merger using the Einstein Toolkit, and reported our findings at the 2019 Midwest Relativity Meeting in Grand Rapids, Michigan.
Read our conference presentation here.
Watch one of my simulations here.
Advisors: Joshua Faber (RIT) and Eric Blackman (UofR)