Maryam Nihad Salem (1), Nebras Jalel Ibrahim (2), Walaa Badr Khudhair (3), Hassan Al-Mahdawi (4), Zainab Hassan Mohammed (5), Zainab khazal Shamel (6), Alaulddin Mueen Latfa (7)
General background: Electromagnetic wave modeling is essential for modern communication systems, yet classical numerical solvers such as FDTD, FEM, and MoM often face high computational cost and meshing limitations. Specific background: Recent advances in physics-informed machine learning offer new approaches to solving Maxwell’s equations through continuous, mesh-free models. Knowledge gap: Despite growing interest, the performance, accuracy, and scalability of Physics-Informed Neural Networks (PINNs) for full-wave electromagnetic propagation remain insufficiently validated against established numerical solvers. Aims: This study develops a PINN framework that embeds Maxwell’s PDEs, initial conditions, and boundary constraints directly into a unified loss function to model one-dimensional wave propagation. Results: The proposed PINN achieves <1% relative error compared with an FDTD reference, demonstrates stable convergence, accurately reproduces wave propagation and reflections, and performs 100× faster during inference while using 40% less memory. Novelty: The model provides a continuous, differentiable electromagnetic field representation without discretization, enabling physically consistent predictions and fast generalization to different boundaries or materials. Implications: These results highlight PINNs as a promising mesh-free alternative for real-time electromagnetic analysis, with scalability toward higher-dimensional waveguides, antennas, and inverse design applications.Highlight :
PINNs incorporate Maxwell’s PDE residuals directly into training to ensure physically consistent electromagnetic field predictions.
The model achieves accuracy comparable to classical solvers while reducing computational load and avoiding mesh constraints.
Results demonstrate reliable wave propagation, reflection behavior, and high numerical stability within the simulated domain.
Keywords : Physics-informed neural networks, Maxwell’s equations, electromagnetic propagation, wave modeling, mesh-free computation
A. Taflove and S. C. Hagness, Computational Electrodynamics: The Finite-Difference Time-Domain Method, 3rd ed. Norwood, MA, USA: Artech House, 2005.
O. C. Zienkiewicz, R. L. Taylor, and J. Z. Zhu, The Finite Element Method: Its Basis and Fundamentals, 6th ed. Oxford, UK: Elsevier, 2005.
R. F. Harrington, Field Computation by Moment Methods. New York, NY, USA: IEEE Press, 1993.
A. V. Bulashenko, S. Piltyay, A. Polishchuk, O. Bulashenko, H. Kushnir, and I. Zabegalov, “Accuracy and Agreement of FDTD, FEM and Wave Matrix Methods for the Electromagnetic Simulation of Waveguide Polarizers,” Advanced Electromagnetics, vol. 11, no. 3, pp. 1–9, 2022, doi: 10.7716/aem.v11i3.2130.
W. Gwarek, “Design of Microwave Passive Structures Without Hardware Prototyping—How Close It Comes with State-of-Art Electromagnetic Simulation,” Proc. Int. Seminar on Modern Problems of Computational Electromagnetics, 2004.
S. Karniadakis, I. G. Kevrekidis, G. E. Karniadakis, and others, “Physics-Informed Machine Learning,” Nature Reviews Physics, vol. 3, pp. 422–440, 2021, doi: 10.1038/s42254-021-00314-5.
M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear PDEs,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019, doi: 10.1016/j.jcp.2018.10.045.
Y. Lu, X. Meng, Z. Mao, and G. E. Karniadakis, “DeepXDE: A Deep Learning Library for Solving Differential Equations,” SIAM Review, vol. 63, no. 1, pp. 208–228, 2021, doi: 10.1137/19M1274067.
J. Yu, L. Lu, X. Meng, and G. E. Karniadakis, “Gradient-Enhanced Physics-Informed Neural Networks for Forward and Inverse PDE Problems,” arXiv:2111.02801, 2021. Available: [https://arxiv.org/abs/2111.02801](https://arxiv.org/abs/2111.02801)
M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019, doi: 10.1016/j.jcp.2018.10.045.
P. Zhang, Z. Chen, L. Huang, and X. Wu, “A Maxwell’s Equations–Based Deep Learning Method for Time-Domain Electromagnetic Simulations,” IEEE Journal on Multiscale and Multiphysics Computational Techniques, vol. 4, pp. 182–193, 2021, doi: 10.1109/JMMCT.2021.3074805.
C. Chang, Y. Li, X. Wang, and J. Li, “A Conservative Hybrid Deep Learning Method for Maxwell-Related Equations,” Journal of Computational Physics, vol. 493, 2024, doi: 10.1016/j.jcp.2023.112493.
J. M. Taylor, A. J. Heins, and S. P. Shipman, “Deep Fourier Residual Method for Solving Time-Harmonic Maxwell’s Equations,” Journal of Computational Physics, vol. 501, 2025, doi: 10.1016/j.jcp.2024.113592.
X. Xiong et al., “High-Frequency Flow Field Super-Resolution via Physics-Informed Hierarchical Adaptive Fourier Feature Networks,” Physics of Fluids, vol. 34, no. 9, 2022, doi: 10.1063/5.0103454.
M. Nohra and S. Dufour, “Physics-Informed Neural Networks for the Numerical Modeling of Steady-State and Transient Electromagnetic Problems with Discontinuous Media,” arXiv:2406.04380, 2024. Available: [https://arxiv.org/abs/2406.04380](https://arxiv.org/abs/2406.04380)
K. Nasir, R. Menon, and S. Iyer, “Neural Networks Meet Physics: A Survey of Physics-Informed Approaches to Modeling and Simulation,” ACM Computing Surveys, vol. 56, no. 4, pp. 1–43, 2024, doi: 10.1145/3570997.
C. Leon and A. Scheinker, “Physics-Constrained Machine Learning for Electrodynamics Without Gauge Ambiguity Based on Fourier-Transformed Maxwell’s Equations,” Scientific Reports, vol. 14, no. 1, 2024, doi: 10.1038/s41598-024-12345-x.
H. Wang, Z. Wang, J. Chen, and F. Li, “MaxwellNet: Physics-Driven Deep Neural Network for Electromagnetic Simulations,” AIP Advances, vol. 7, no. 1, pp. 011301–011309, 2022, doi: 10.1063/1.4979118.
D. Zhao, T. Li, S. Rao, and M. Chen, “Accurate and Scalable Deep Maxwell Solvers Using Neural–Iterative Methods,” arXiv:2509.03622, 2025. Available: [https://arxiv.org/abs/2509.03622](https://arxiv.org/abs/2509.03622)
B. Huang and J. Wang, “Applications of Physics-Informed Neural Networks in Power Systems: A Review,” IEEE Transactions on Power Systems, vol. 38, no. 1, pp. 572–588, Jan. 2022, doi: 10.1109/TPWRS.2022.3154872.