Anisotropic Maxwell neural operator for rapid parametric full-wave modelling of ion cyclotron resonance heating

Abstract

Full-wave calculations of ion cyclotron resonance heating (ICRH) under different plasma dielectric conditions require repeated assembly and solution of large-scale discretised systems, limiting parameter sweeps and multi-case response analysis. We therefore propose an anisotropic Maxwell neural operator (AMNO) for rapid parametric modelling of ICRH full-wave responses for the Experimental Advanced Superconducting Tokamak (EAST), which learns, within the one-parameter dielectric-field family generated by varying the hydrogen minority fraction X_H over 0.01-0.05 under otherwise fixed settings, a shared solution operator from the spatially varying complex anisotropic dielectric-tensor field to the three-component complex electric field under frequency-domain Maxwell constraints. It represents global spatial coupling through spectral operator layers and local fine-scale responses, and combines sparse reference-field supervision with the frequency-domain Maxwell-equation residual. Comparisons with COMSOL reference solutions for the same EAST frequency-domain Maxwell-dielectric model show that AMNO reconstructs the principal spatial and spectral features and maintains stable accuracy for unseen interpolation test cases. With reference-field points reduced to 7.5% of the dense full-wave set, AMNO reduces the relative L_2 error by 66.1%-89.9% compared with a sparsely supervised Fourier neural operator (FNO-Sparse) under the same supervision and requires about 0.25 s for single-case inference. AMNO thus reduces dependence on dense reference-field supervision while enabling subsecond parametric complex-field inference, providing a physics-constrained and data-efficient surrogate for rapid in-range X_H sweeps and cross-case response analysis within the modelled EAST configuration.