Machine Learning for Predicting Multipactor Susceptibility in Planar RF Structures

Multipactor discharge [1] is a persistent challenge in high-power microwave (HPM) and accelerator systems, where secondary electron avalanches can cause heating, vacuum degradation, and component failure. This work presents the first supervised machine learning (ML) framework for multipactor prediction, trained on high-fidelity 3D Particle-in-Cell (PIC) simulation data [1] in planar geometries. The model maps operational, geometric, and material-dependent secondary electron yield (SEY) [2] parameters to the time-averaged electron growth rate, enabling rapid reconstruction of susceptibility charts [1]. Among the models evaluated, tree-based ensemble methods such as Random Forest and Extra Trees demonstrate superior generalization to unseen materials compared to neural networks such as multilayer perceptron (MLP). MLPs were optimized via Bayesian hyperparameter tuning using a scalarized objective that jointly maximized Intersection over Union (IoU) and Structural Similarity Index (SSIM), improving spatial and structural fidelity. Performance metrics, including IoU, SSIM, and Pearson correlation, show close agreement with simulation benchmarks. Principal Component Analysis (PCA) attributes generalization limits to material feature-space disjointedness.

[1] A. Iqbal, D. Wen, J. Verboncoeur, and P. Zhang, Recent advances in multipactor physics and mitigation, High Voltage 8, 1095 (2023).

[2] R. M. Vaughan, Secondary emission formulas, IEEE Transactions on Electron Devices 40, 830 (1993).