Photonic computing for accelerating computational lithography: a review and outlook

Computational lithography has become indispensable for sustaining pattern fidelity and manufacturing yield as semiconductor technology advances toward full-chip inverse lithography technology (ILT). However, the increasing demands for high-fidelity physical modeling and large-scale simulation impose severe burdens on conventional electronic computing platforms in terms of energy consumption, memory bandwidth, and data movement. This review systematically examines the foundations of computational lithography and analyzes the current acceleration approaches, including Graphics Processing Unit (GPU)-based parallel computing and artificial intelligence (AI)-based surrogate models, along with their limitations. We further discuss photonic computing as a promising emerging paradigm, emphasizing its intrinsic parallelism, low latency, energy efficiency, and physical compatibility with linear operations. Photonic systems show strong potential for accelerating key computational kernels such as convolution, Fourier transform, and neural network inference. The optoelectronic hybrid architecture represents a viable near-term solution, in which photonic processors should be viewed as linear-kernel coprocessors rather than standalone replacements for electronic lithography engines, with particular attention to challenges in I/O bottlenecks, system stability, and numerical precision. This work provides a comprehensive framework for understanding the synergy between photonic computing and computational lithography and outlines future research directions.