Yulu Hu

To improve the throughput of electron-beam cross section measurement on existing testbeds, we propose a “ $5\times 5$ coarse sampling $+ 15\times 15$ machine-learning reconstruction” scheme. Paired $5\times 5$ / $15\times 15$ current-distribution matrices generated by the MTSS simulation software are used for training and validation. Methodologically, we integrate interpolation alignment and residual correction into a lightweight pipeline: interpolation first performs scale alignment and spatial interpolation to provide a stable prior; supervised learning then pixelwise corrects the systematic interpolation bias, yielding reliable reconstructions from few samples. Featuring low complexity and latency, this pipeline easily integrates into legacy platforms, enabling rapid $5\times 5$ – $15\times 15$ reconstruction on commodity hardware. Compared with the simulated $15\times 15$ Ground Truth, the residual model achieves pixel-level errors of mean absolute error (MAE) $= 1.287\times 10^{-4}$ and root mean squared error (RMSE) $= 1.804\times 10^{-4}$ . In engineering metrics, the beam-waist radius (r $99\cdot 5$ ) errors on the validation set are MAE $\approx ~16.86~\mu $ m and RMSE $\approx ~20.83~\mu $ m. Diagonal line-profile comparisons of the current maps further show samplewise MAE/RMSE at the $10^{-4}$ level, indicating controllable deviations that meet efficiency-improvement requirements and demonstrate practical feasibility for extension to experimental measurements.

This study presents an innovative structural design for pixelated electron guns, coupled with a dedicated structural screening algorithm. By constructing a convolutional neural network (CNN) model for the anode component, we achieve high-precision forward prediction of key electron gun parameters—emission current, operational range, and beam waist radius. Simulation-based validation demonstrates a minimal average relative error (ARE) of 0.54% between predicted and simulated results. Furthermore, leveraging the established CNN framework with parameter standardization methodology, we demonstrate rapid inverse synthesis of pixelated electron guns. The inverse design yields an ARE of 8.27% compared to the target specifications. This integrated approach significantly broadens design flexibility while enhancing developmental efficiency for electron gun systems.

A klystron is a microwave vacuum electronic device that converts the kinetic energy of an electron beam into microwave signal energy based on the principle of velocity modulation. Owing to its high peak power, high efficiency, and high gain, the klystron has been widely used in scientific research, national defense, and industrial applications. The interaction between the electron beam and the radio-frequency (rf) field within the resonant cavities, known as the beam-wave interaction, is the core physical process governing klystron operation. A detailed understanding of this process, together with accurate numerical simulation using computational solvers, is essential for optimizing device performance and shortening the development cycle. This paper reviews the large-signal beam-wave interaction solvers commonly used in klystron design and development in mainland China, along with several internationally recognized simulation codes. The discussed solvers cover a range of computational dimensions from one-dimensional (1D) to three-dimensional (3D). Finally, the challenges and future prospects for the development of klystron beam-wave interaction solvers are discussed.