Machine Learning-Based Prediction of Electron Energy Distribution Function in Low-Pressure Argon Plasmas

Low-pressure plasmas have been extensively utilized in various areas including surface processing applications due to its efficient molecular activation capability. The electron energy distribution function (EEDF), as a key parameter reflecting plasma states and internal energy transfer processes, plays a vital role in achieving precise plasma control and guiding the optimization of plasma sources and operating parameters. To enable effective diagnostics of EEDFs, this study proposes an approach based on machine learning combined with optical emission spectroscopy. An argon collisional radiative model was established to correlate spectral data with internal plasma physical properties. By integrating machine learning techniques into this model, predictions of the EEDF and electron density were achieved. Validation was performed by comparing preset reference data with machine learning predictions, demonstrating the method's accuracy in EEDF diagnostics. Furthermore, the approach was validated using experimental EEDF data obtained from argon discharges at pressures ranging from 10 to 30 mTorr and power levels from 200 to 800 W using Langmuir probes. Results show good agreement between the predicted EEDF and experimentally measured values, further confirming the effectiveness of the proposed method.