A generative foundation model for 2D magnetized plasma turbulence
Abstract
We present a proof-of-principle study demonstrating a conditional generative foundation model for self-organizing magnetized plasma turbulence. Built on a multichannel flow-matching framework with a spectral decomposition strategy, the model is trained on direct numerical simulations of 2D modified Hasegawa-Wakatani (MHW) turbulence across the hydrodynamic to adiabatic parameter range. Without retraining, this \textit{single} model successfully performs three distinct downstream tasks: (i) unconditional generation of statistically faithful quasi-steady-state snapshots that recover the directional energy spectra and amplitude statistics, from which statistical transport quantities, including turbulent flux and effective diffusivity, alongside their uncertainties, are inferred directly; (ii) super-resolution from 4× and 8× downsampled observations to reconstruct fine-scale turbulence and large-scale zonal structures; and (iii) full-field inpainting from sparse spatial observations covering as little as ∼ 10% of the domain. The generated fields faithfully reproduce regime-dependent turbulence vs. zonal-flow states and energy partitioning. Under extreme sparsity the long-correlation zonal flows are pinned by only a handful of observations, while the residual uncertainty concentrates in the small-scale turbulent fluctuations. These results establish flow-matching generative models as powerful, unified surrogates for state generation, statistical inference, and sparse-data reconstruction, providing a scalable baseline for 3D plasma turbulence and experimental applications.
Related
- Scalar Dissipation Criticality in Compressible Magnetized Turbulence
- Photon Acceleration in Magnetized Plasma: A Mechanism for Fast Radio Bursts
- Counter-rotating density and current structures in a partially magnetized E×B plasma
- Virtual photon bremsstrahlung in a magnetized quark-gluon-plasma
- Interactions Between Two Magnetized Plasma Pressure Filaments: Drift-Alfvén Modes and Transport
BibTeX
@misc{guo2026generative,
title = {{A generative foundation model for 2D magnetized plasma turbulence}},
author = {Junyi Guo and Meet H. Parikh and Jian-Xun Wang and Ben Zhu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.36155},
eprint = {2610.36155},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.36155},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A generative foundation model for 2D magnetized plasma turbulence
AU - Junyi Guo
AU - Meet H. Parikh
AU - Jian-Xun Wang
AU - Ben Zhu
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2610.36155
UR - https://arxiv.org/abs/2610.36155
AB - We present a proof-of-principle study demonstrating a conditional generative foundation model for self-organizing magnetized plasma turbulence. Built on a multichannel flow-matching framework with a spectral decomposition strategy, the model is trained on direct numerical simulations of 2D modified Hasegawa-Wakatani (MHW) turbulence across the hydrodynamic to adiabatic parameter range. Without retraining, this \textit{single} model successfully performs three distinct downstream tasks: (i) unconditional generation of statistically faithful quasi-steady-state snapshots that recover the directional energy spectra and amplitude statistics, from which statistical transport quantities, including turbulent flux and effective diffusivity, alongside their uncertainties, are inferred directly; (ii) super-resolution from 4× and 8× downsampled observations to reconstruct fine-scale turbulence and large-scale zonal structures; and (iii) full-field inpainting from sparse spatial observations covering as little as ∼ 10% of the domain. The generated fields faithfully reproduce regime-dependent turbulence vs. zonal-flow states and energy partitioning. Under extreme sparsity the long-correlation zonal flows are pinned by only a handful of observations, while the residual uncertainty concentrates in the small-scale turbulent fluctuations. These results establish flow-matching generative models as powerful, unified surrogates for state generation, statistical inference, and sparse-data reconstruction, providing a scalable baseline for 3D plasma turbulence and experimental applications.
ER -