arXiv subjects

Linlin Zhong

Publications and source records attributed to Linlin Zhong.

9 recordsLinked to original sources

arXiv2026

Latent Residual-Closure Fourier Neural Operator for Robust Multi-Field Solving in Particle-in-Cell Simulations

Particle-in-cell (PIC) simulations are widely used for kinetic plasma modeling in energy applications, but their efficiency is often limited by repeated field solves on dense meshes. This work proposes a Latent Residual-Closure Fourier Neural Operator (LRC-FNO) for robust surrogate multi-field solving in PIC simulations. Rather than treating field prediction as a purely data-driven regression task, LRC-FNO formulates PIC field solving as a two-level residual-closure problem involving source compression and source-to-field operator mapping. An autoencoder extracts compact representations of particle-deposited source fields, while a Latent Closure Refiner recovers unresolved residual structures lost during compression. A Coarse-FNO Solver captures the dominant field response, and a Residual-Closure FNO restores full-resolution corrections. The method is tested on three benchmarks with increasing complexity: 1D linear Landau damping (LLD), 2D two-stream instability (TSI), and a 2D scrape-off layer (SOL) fusion plasma model. In LLD and TSI, LRC-FNO better preserves charge-to-potential mapping, potential-mode evolution, residual charge structures, and particle-field energy exchange during closed-loop PIC integration. In the SOL case, LRC-FNO achieves relative L2 errors of 0.0447 for the self-consistent potential and 0.0251 for the magnetic vector potential in single-step prediction. More importantly, when used as a neural initial guess with 20 iterative corrections, LRC-FNO maintains strong physical consistency in extrapolated closed-loop simulations, preserving charge and current density structures over a time range close to twice the training horizon. These results demonstrate that LRC-FNO can serve as both a fast surrogate field solver and a high-quality initialization strategy for iterative PIC field solvers.

physics.comp-ph↗

arXiv2026

DeepPropNet: an operator learning-based predictor for thermal plasma properties

Thermal plasma properties play a critical role in plasma simulations and plasma-related applications. However, their strong nonlinear dependence on temperature, pressure, and gas composition makes accurate and efficient evaluation challenging. In this work, an operator learning-based model, termed DeepPropNet, is proposed for fast prediction of thermodynamic and transport properties of thermal plasmas. Two architectures are developed, including a single-property model (S-DeepPropNet) and a Mixture of Experts (MoE)-based multi-property model (MoE-DeepPropNet). The proposed models learn the nonlinear mapping from plasma operating conditions to physical properties based on high-fidelity datasets. The MoE architecture enables efficient multi-property prediction within a unified framework. Predictions are performed for binary SF6-N2 and ternary C4F7N-CO2-O2 mixtures. The results show that the proposed models achieve high accuracy, with relative L2 errors on the order of 10-3 to 10-2, while maintaining strong generalization capability under unseen conditions. The applicability of DeepPropNet is further demonstrated by coupling with finite volume method (FVM) and physics-informed neural networks (PINNs). The results indicate that DeepPropNet provides an efficient and scalable approach for plasma property prediction and plasma simulations.

physics.plasm-ph↗

arXiv2025

PaRO-DeepONet: a particle-informed reduced-order deep operator network for Poisson solver in PIC simulations

Particle-in-Cell (PIC) simulations are widely used for modeling plasma kinetics by tracking discrete particle dynamics. However, their computational cost remains prohibitively high, due to the need to simulate large numbers of particles to mitigate statistical noise and the inefficiency in handling complex geometries. To address these challenges, we propose PaRO-DeepONet, a particle-informed reduced-order surrogate framework that integrates Proper Orthogonal Decomposition (POD) with Deep Operator Network (DeepONet). By performing manifold sampling of particle evolution in PIC simulations and constructing snapshot matrices from various particle states, PaRO-DeepONet applies POD projection to extract latent charge density features that are shared across similar plasma scenarios. These reduced features are fed into the branch network of DeepONet, enabling efficient mapping from charge distributions to electrostatic potential fields. We validate PaRO-DeepONet on four representative benchmark cases. The model consistently achieves relative L2 errors below 3.5%, with the best case reaching as low as 0.82%. Notably, although the charge deposition becomes highly discrete due to particle sparsity, PaRO-DeepONet still reconstructs smooth and continuous potential fields, demonstrating strong robustness to sparse inputs and excellent generalization beyond traditional grid-based solvers. In terms of computational performance, PaRO-DeepONet reduces total PIC simulation time by 68%-85% and cuts the Poisson solver runtime by up to 99.6%. By significantly reducing computational cost while maintaining high accuracy, this work establishes machine learning-based operator surrogates as a promising new paradigm for next-generation kinetic plasma simulations.

physics.comp-ph↗

arXiv2025

NAS-PINNv2: Improved neural architecture search framework for physics-informed neural networks in low-temperature plasma simulation

Limited by the operation and measurement conditions, numerical simulation is often the only feasible approach for studying plasma behavior and mechanisms. Although artificial intelligence methods, especially physics-informed neural network (PINN), have been widely applied in plasma simulation, the design of the neural network structures still largely relies on the experience of researchers. Meanwhile, existing neural architecture search methods tailored for PINN have encountered failures when dealing with complex plasma governing equations characterized by variable coefficients and strong nonlinearity. Therefore, we propose an improved neural architecture search-guided method, namely NAS-PINNv2, to address the limitations of existing methods. By analyzing the causes of failure, the sigmoid function is applied to calculate the architecture-related weights, and a new loss term is introduced. The performance of NAS-PINNv2 is verified in several numerical experiments including the Elenbaas-Heller equation without and with radial velocity, the drift-diffusion-Poisson equation and the Boltzmann equation. The results again emphasize that larger neural networks do not necessarily perform better, and the discovered neural architecture with multiple neuron numbers in a single hidden layer imply a more flexible and sophisticated design rule for fully connected networks.

physics.comp-ph↗

arXiv2024

DeepCSNet: a deep learning method for predicting electron-impact doubly differential ionization cross sections

Electron-impact ionization cross sections of atoms and molecules are essential for plasma modelling. However, experimentally determining the absolute cross sections is not easy, and ab initio calculations become computationally prohibitive as molecular complexity increases. Existing AI-based prediction methods suffer from limited data availability and poor generalization. To address these issues, we propose DeepCSNet, a deep learning approach designed to predict electron-impact ionization cross sections using limited training data. We present two configurations of DeepCSNet: one tailored for specific molecules and another for various molecules. Both configurations can typically achieve a relative L2 error less than 5%. The present numerical results, focusing on electron-impact doubly differential ionization cross sections, demonstrate DeepCSNet's generalization ability, predicting cross sections across a wide range of energies and incident angles. Additionally, DeepCSNet shows promising results in predicting cross sections for molecules not included in the training set, even large molecules with more than 10 constituent atoms, highlighting its potential for practical applications.

physics.plasm-ph↗

arXiv2023

NAS-PINN: Neural architecture search-guided physics-informed neural network for solving PDEs

Physics-informed neural network (PINN) has been a prevalent framework for solving PDEs since proposed. By incorporating the physical information into the neural network through loss functions, it can predict solutions to PDEs in an unsupervised manner. However, the design of the neural network structure basically relies on prior knowledge and experience, which has caused great trouble and high computational overhead. Therefore, we propose a neural architecture search-guided method, namely NAS-PINN, to automatically search the optimum neural architecture for solving certain PDEs. By relaxing the search space into a continuous one and utilizing masks to realize the addition of tensors in different shapes, NAS-PINN can be trained through a bi-level optimization, where the inner loop optimizes the weights and bias of neural networks and the outer loop the architecture parameters. We verify the ability of NAS-PINN by several numerical experiments including Poisson, Burgers, and Advection equations. The characteristics of effective neural architectures for solving different PDEs are summarized, which can be used to guide the design of neural networks in PINN. It is found that more hidden layers do not necessarily mean better performance and sometimes can be harmful. Especially for Poisson and Advection, a shallow neural network with more neurons is more appropriate in PINNs. It is also indicated that for complex problems, neural networks with residual connection can improve the performance of PINNs.

physics.comp-ph↗

arXiv2022

Accelerating physics-informed neural network based 1D arc simulation by meta learning

Physics-Informed Neural Networks (PINNs) have a wide range of applications as an alternative to traditional numerical methods in plasma simulation. However, in some specific cases of PINN-based modeling, a well-trained PINN may require tens of thousands of optimizing iterations during training stage for complex modeling and huge neural networks, which is sometimes very time-consuming. In this work, we propose a meta-learning method, namely Meta-PINN, to reduce the training time of PINN-based 1-D arc simulation. In Meta-PINN, the meta network is first trained by a two-loop optimization on various training tasks of plasma modeling, and then used to initialize the PINN-based network for new tasks. We demonstrate the power of Meta-PINN by four cases corresponding to 1-D arc models at different boundary temperatures, arc radii, arc pressures, and gas mixtures. We found that a well-trained meta network can produce good initial weights for PINN-based arc models even at conditions slightly outside of training range. The speed-up in terms of relative L2 error by Meta-PINN ranges from 1.1x to 6.9x in the cases we studied. The results indicate that Meta-PINN is an effective method for accelerating the PINN-based 1-D arc simulation.

physics.plasm-ph↗

arXiv2022

Low-temperature plasma simulation based on physics-informed neural networks: frameworks and preliminary applications

Plasma simulation is an important and sometimes only approach to investigating plasma behavior. In this work, we propose two general AI-driven frameworks for low-temperature plasma simulation: Coefficient-Subnet Physics-Informed Neural Network (CS-PINN) and Runge-Kutta Physics-Informed Neural Network (RK-PINN). The CS-PINN uses either a neural network or an interpolation function (e.g. spline function) as the subnet to approximate solution-dependent coefficients (e.g. electron-impact cross sections, thermodynamic properties, transport coefficients, et al.) in plasma equations. On the basis of this, the RK-PINN incorporates the implicit Runge-Kutta formalism in neural networks to achieve a large-time-step prediction of transient plasmas. Both CS-PINN and RK-PINN learn the complex non-linear relationship mapping from spatio-temporal space to equation's solution. Based on these two frameworks, we demonstrate preliminary applications by four cases covering plasma kinetic and fluid modeling. The results verify that both CS-PINN and RK-PINN have good performance in solving plasma equations. Moreover, the RK-PINN has ability of yielding a good solution for transient plasma simulation with not only large time step but also limited noisy sensing data.

physics.plasm-ph↗

arXiv2020

Deep learning for thermal plasma simulation: solving 1-D arc model as an example

Numerical modelling is an essential approach to understanding the behavior of thermal plasmas in various industrial applications. We propose a deep learning method for solving the partial differential equations in thermal plasma models. In this method a deep feed-forward neural network is constructed to surrogate the solution of the model. A loss function is designed to measure the discrepancy between the neural network and the equations describing thermal plasmas. A good neural network is obtained by minimizing this loss function. We demonstrate the power of this deep learning method by solving a 1-D arc decaying model which is consist of three cases: stationary arc, transient arc without considering radial velocity, and transient arc with radial velocity respectively. The results show that the deep neural networks have excellent ability to express the differential equations describing thermal plasmas. This could bring us a new and prospective numerical tool for thermal plasma modelling.

physics.comp-ph↗