Data Assimilation of Plasma Dynamics and Chemistry

Low-temperature plasmas exhibit complex, multiscale dynamics involving electron kinetics, chemical reactions, and transport across timescales from nanoseconds to milliseconds and spatial scales from Debye lengths to device dimensions. This complexity requires modeling at multiple fidelities, from direct kinetic and particle-based methods to simplified fluid models, each relying on assumptions about electron energy distributions, reaction rates, and transport coefficients that are often difficult to determine.

Data assimilation (DA) offers a systematic way to address these challenges by combining physics-based models with sparse experimental data. While high-fidelity simulations are often too expensive for iterative design and optimization, DA leverages lower-complexity models informed by data to efficiently estimate critical parameters. This talk highlights the use of ensemble Kalman filtering (EnKF) and extended Kalman filtering (EKF) to estimate non-Maxwellian electron energy distributions and spatiotemporal plasma dynamics.

I will first present work using EnKF to estimate non-Maxwellian electron energy distributions and plasma chemistry in real-time from optical emission spectroscopy. This method achieves accuracy comparable to particle-in-cell simulations while reducing computational time from weeks to minutes. I will then show our work on DA of 1D partial differential equations using sparse measurements. Using EKF, we estimate spatially and temporally varying plasma states, enabling robust sensitivity analysis to identify key physical processes that influence measurements and guide high-fidelity modeling. The development of DA for more realistic simulations, incorporation of diverse measurements, and real-time control and optimization will be discussed.