IEEE Transactions on Microwave Theory and Techniques
Quantum Computing Method for Fast Frequency Sweep of Electromagnetic Problems Based on Krylov Subspace and Finite Element Method
Xiaolong Li, J. Z. Han, Runqing Zhang, Feng Feng, Zhen Chen, Jingyi Feng, Jinyi Liu, Jiali Zhang, Haifeng Yu, Kaixue Ma, Qi‐Jun Zhang
Abstract
Recent advancements in hybrid quantum-classical algorithms (HQAs) have demonstrated significant potential for solving large-scale electromagnetic (EM) problems based on the finite element method (FEM). However, existing quantum EM solvers are fundamentally restricted to discrete, point-by-point frequency evaluations. For broadband microwave component design, executing iterative quantum circuit optimizations at each sampled frequency imposes a substantial computational overhead, limiting the practical applicability of broadband quantum EM simulation. To overcome this limitation, we propose a hybrid quantum-classical Krylov subspace projection (HQKSP) method for the fast frequency sweep of FEM-based EM simulations. Instead of repetitive quantum evaluations across the entire frequency band, the proposed method constructs a reduced-order model by generating a set of basis vectors that span a Krylov subspace at a single expansion frequency. To mitigate the quantum circuit depth limitations associated with strict second-order moment matching, we introduce a hardware-efficient recursive strategy to sequentially generate the basis vectors, bypassing the need for deep linear combinations of unitaries (LCUs). A quantum postprocessing mechanism is then developed to track and compensate for the accumulated relative phase offsets, restoring the phase coherence of the subspace. Subsequently, a direct quantum Galerkin projection is employed to compress the massive FEM system matrices into a low-dimensional space. Because the reduced matrix elements are evaluated only once at the expansion point, the quantum projection overhead is independent of the frequency sweep density. Therefore, the proposed HQKSP method enables the accurate extraction of complex-valued S-parameters while offering a potential reduction in frequency-dependent quantum computational overhead for broadband EM simulations. Numerical experiments on microwave components demonstrate that the proposed HQKSP method achieves the fast frequency sweep via quantum evaluations and classical postprocessing, maintaining close agreement with classical full-wave simulations.