Scalable unitary computing using time-parallelized photonic lattices

Exploiting alternative physical dimensions beyond the spatial domain has been intensively explored to improve the scalability in photonic computing. One approach leverages dynamical systems for time-domain computation, enabling universal and reconfigurable unitary operations. Although this method yields O(N) scaling in both device footprint and gate count, the required computation time increases by O(N²), which hinders practical implementation due to limitations in quality factors and modulation speeds of optical elements. Here, we propose time-parallelized photonic lattices that achieve O(N) time scalability while preserving the O(N) spatial scaling. We devise a pseudospinor buffer operation that temporally stores the optical information, thereby enabling parallel unitary computation. The proposed method not only mitigates the requirement for high-quality factors but also provides robustness against a broad range of defects, demonstrating the feasibility of time-domain photonic computation.