Research interests · portfolio

Models that expose the hidden dynamics.

A growing collection of work in kinetic and fluid plasma simulation, collisionless shocks, laboratory astrophysics, reacting flow, numerical methods, and machine learning for physical systems.

Featured · ALPINE electromagnetic PIC

Sustained magnetron plasma–RF dynamics

ALPINE—Arbitrary Lagrangian Particles Interpolated to Nonuniform Electromagnetics—is a plasma simulation suite built around unstructured finite-element fields, particle dynamics, and compact-support transfer methods. This full-2π six-vane calculation uses charge-conserving particle–field coupling to follow cathode emission, formation of a magnetically confined electron hub, and its interaction with the RF cavity.

At the operating-point window, a seeded π mode grows, organizes the electron cloud into three rotating spokes, and locks to a rotating layer of the hub. The synchronized views show the wave component of B_z, electron azimuthal density, and the π-mode envelope.

Normalized simulation units. Visualization from the ALPINE M4-phase3 sustained-operation model.

FLASH 4.8 cylindrical Z-pinch at 147.92 nanoseconds showing density, electron temperature, and azimuthal magnetic field with a strongly modulated liner contour

FLASH 4.8 · radiation MHD

Instability in a driven Z-pinch

An axisymmetric r–z FLASH 4.8 calculation follows a beryllium liner imploding around preheated deuterium in a one-centimeter periodic axial domain. Tabulated material properties and a prescribed pulsed-power current waveform drive the liner, while a reproducible 0.1% broadband density perturbation seeds instability.

By 147.92 ns and 23 MA, the saturated modes have produced a strongly nonlinear axial structure: approximately 266 µm RMS and 1.59 mm peak-to-peak liner modulation. The panels show mass density, electron temperature, and azimuthal magnetic-field magnitude; the white contour marks a liner material fraction of 0.5.

Open the full-resolution figure

Late-time state from the periodic Z-pinch run at step 173,399.

Three-dimensional hybrid simulation with magnetic field slices and ion populations

Featured · hybrid simulation

Super-Alfvénic expansion in magnetized plasma

A massively parallel three-dimensional hybrid model treats ion species kinetically with particle-in-cell methods while representing electrons as an inertialess, charge-neutralizing fluid. The work connects field structure, particle dynamics, and laboratory measurements across ion timescales.

Read the dissertation

Figure adapted from S. Eric Clark, Simulations of Super Alfvénic Laser Ablation Experiments in the Large Plasma Device, UCLA doctoral dissertation (2016).

Space-time magnetic field contour from a hybrid simulation of shock formation

Collisionless shocks

From piston formation to propagating structure

Hybrid simulations resolve magnetic compression, debris–ambient coupling, and the evolution of super-magnetosonic disturbances into collisionless shock structures.

Space-time magnetic-field contour from the 2016 UCLA dissertation.

Schematic of probes, laser path, and diagnostics on the Large Plasma Device

Experiment ↔ model

Benchmarking against laboratory plasma

Experiments in UCLA’s Large Plasma Device provided magnetic-field, potential, imaging, and probe data for direct comparison with two- and three-dimensional hybrid simulations.

Experimental schematic from the 2016 UCLA dissertation.

Simulated transverse magnetic field map with vector overlay

Field reconstruction

Model-to-data field comparison

Vector and scalar field products turn simulation state into observables that can be compared directly with diagnostic planes, lineouts, and time histories.

Magnetic-field comparison from the 2016 UCLA dissertation.

Portfolio in development

Machine learning for physical systems

Current local work explores scientific machine-learning workflows, surrogate models, and analysis methods for simulation data. Individual projects, validation results, and visual artifacts will be added here as they are prepared for public release.

Research map

Questions across scale and formulation.

The common thread is identifying the smallest model that retains the mechanism under study, then verifying where its assumptions break.

Hybrid kinetic–fluid PICKinetic ion populations coupled to fluid electron closures for laboratory, astrophysical, and fusion plasma research.
MHD & fluid plasma modelsEquilibria, macroscopic dynamics, closure behavior, and comparisons against kinetic descriptions.
Reacting & rarefied flowContinuum CFD, DSMC, chemistry, transport, thermal coupling, and vacuum-system dynamics.
Scientific machine learningData-driven models that respect physical structure, with validation against simulation and experiment.
GPU scientific computingAlgorithms, numerical performance, and scalable workflows for compute-intensive models.
Diagnostics & inferenceConnecting simulated fields and distributions to measurable quantities and uncertainty.

More simulations, images, and movies are on the way.

Explore the publication record