AI/ML for plasma physics
Learning collision operators with differentiable simulators
J. Plasma Phys. 92, E76 (2026)
Welcome to the UCLA Plasma Simulation Group
Our group develops and uses state-of-the-art kinetic plasma simulation software, increasingly combined with AI/ML, to make fundamental discoveries in plasma-based acceleration and light sources, plasma astrophysics, laser–plasma interactions, and the nonlinear optics of plasmas.
Our AI/ML research aims to accelerate simulation- and data-driven discovery, design better simulation algorithms, and develop new theoretical models of plasma phenomena.
The group is co-led by Professors W. B. Mori and E. Paulo Alves.
Recent highlights
All research →
AI/ML for plasma physics
Building physical symmetries into machine-learned plasma models
Phys. Rev. Research 8, 013200 (2026)
Plasma-based acceleration
Realigning beams in a plasma wakefield accelerator stage
Phys. Rev. Lett. 135, 195002 (2025)
Software
All software →We develop and share production and educational codes used by plasma physicists around the world.
OSIRIS
Fully relativistic, massively parallel electromagnetic PIC code, now open source with GPU support.
QuickPIC
Fully parallel, relativistic 3D quasi-static PIC code for plasma wakefield acceleration.
QPAD
Quasi-3D quasi-static PIC code combining the quasi-static approximation with azimuthal decomposition.
OSHUN
Parallel relativistic 2D3P Vlasov–Fokker–Planck code.
UPIC & skeleton codes
Trusted, bare-bones parallel PIC components and mini-apps for building new codes.
Educational codes
Jupyter notebooks and small PIC codes for teaching plasma physics.
Latest publications
All publications →- An explicit algorithm in the quasi-static approximate PIC program QPAD
R. Tang, W. An, H. Wang, et al. · Plasma Physics and Controlled Fusion (2026)
- Generalized multi-dimensional conservation laws for stimulated Raman and Brillouin scattering in a density gradient
V. Patel, S. Chase, F. S. Tsung, et al. · Accepted in Physical Review E (2026)
- Laser–plasma amplification of an ultrabroadband laser pulse to 0.3 TW
J. L. Shaw, M. V. Ambat, K. R. McMillen, et al. · Nature Photonics 20, 1070-1074 (2026)
- GPU-MPI parallelisation for QuickPIC, a three-dimensional quasi-static particle-in-cell algorithm
Y. Tian, Y. Wang, T. N. Dalichaouch, et al. · Journal of Plasma Physics 92, E103 (2026)
- Generating intense attosecond pulses and vectorizing polarization states from laser–plasma interactions
P. Geng, Y. Wu, Z. Fan, et al. · Optica 13, 1271 (2026)