PIC2O-Sim: A Physics-Inspired Causality-Aware Dynamic Convolutional Neural Operator for Ultra-Fast Photonic Device FDTD Simulation

TL;DR

FDTD simulations in photonics are accurate but computationally expensive, hindering design feedback loops. The authors introduce PIC2O-Sim, a physics-inspired causality-aware neural operator that encodes permittivity and preserves spatial resolution, enabling autoregressive, time-bundled prediction of electromagnetic fields. Key innovations include a device encoder, Dilated Position-Adaptive Convolution (DPAConv) backbones, and a multi-stage time-bundling strategy to balance fidelity, scalability, and speed; collectively, these yield substantial improvements over baselines and large reductions in parameters, with orders-of-magnitude speedups over traditional FDTD. The work offers a practical surrogate for rapid photonic device design, though it notes remaining challenges with long-rollout error accumulation and points to future error-suppression or non-autoregressive approaches.

Abstract

The finite-difference time-domain (FDTD) method, which is important in photonic hardware design flow, is widely adopted to solve time-domain Maxwell equations. However, FDTD is known for its prohibitive runtime cost, taking minutes to hours to simulate a single device. Recently, AI has been applied to realize orders-of-magnitude speedup in partial differential equation (PDE) solving. However, AI-based FDTD solvers for photonic devices have not been clearly formulated. Directly applying off-the-shelf models to predict the optical field dynamics shows unsatisfying fidelity and efficiency since the model primitives are agnostic to the unique physical properties of Maxwell equations and lack algorithmic customization. In this work, we thoroughly investigate the synergy between neural operator designs and the physical property of Maxwell equations and introduce a physics-inspired AI-based FDTD prediction framework PIC2O-Sim which features a causality-aware dynamic convolutional neural operator as its backbone model that honors the space-time causality constraints via careful receptive field configuration and explicitly captures the permittivity-dependent light propagation behavior via an efficient dynamic convolution operator. Meanwhile, we explore the trade-offs among prediction scalability, fidelity, and efficiency via a multi-stage partitioned time-bundling technique in autoregressive prediction. Multiple key techniques have been introduced to mitigate iterative error accumulation while maintaining efficiency advantages during autoregressive field prediction. Extensive evaluations on three challenging photonic device simulation tasks have shown the superiority of our PIC2O-Sim method, showing 51.2% lower roll-out prediction error, 23.5 times fewer parameters than state-of-the-art neural operators, providing 300-600x higher simulation speed than an open-source FDTD numerical solver.

Figures (16)

• Figure 1: (a) Our ML-based model shows orders-of-magnitude speedup over FDTD solvers. (b) Global-view FNO li2021fourier with truncated modes cannot learn local convolution. (c) (Top) Due to light speed limitation, space causality implies local receptive fields; (Bottom) Wave propagation depends on local permittivity distributions. (d) Expanding convolutional kernel sizes for long-term prediction is not scalable in runtime and memory, especially for dynamic convolution PAConv su2019pixel.
• Figure 2: Illustration of the causality-constrained space-time locality and the theoretical receptive field.
• Figure 2: Compare different task partitioning when predicting 160 frames. Kernel sizes (KS) are adjusted to match the suitable receptive field for the predicted frames ($T$) per step. Partitioning into 2 stages gives the best results.
• Figure 3: PIC2O-Sim framework overview.
• Figure 4: Devices in a mini-batch are replicate-padded to the same size with a fixed spatial resolution.