All-optical computing promises fast, energy-efficient processing and has regained momentum as artificial intelligence workloads strain electronic hardware. Among optical approaches, free-space diffractive optics enables rapid, multidimensional information processing with high parallelism, but advancing these systems requires a fundamental understanding of their computational capacity and how to exploit it. Whereas electronic computation relies on cascaded linear and nonlinear operations to implement digital circuits, neural networks, and machine vision, we show that diffractive optical systems can implement complex logic circuits by collapsing serial chains of nonlinear functions into a single optical stage. We first realize all basic logic gates and then demonstrate parallel half-/full adders and subtractors with one-stage readout. We further demonstrate an 8-bit ripple-carry adder by cascading eight one-shot optical full adders. Finally, we show scalability to hundreds of parallel inputs and direct two-dimensional image processing using spatially localized nonlinear optical logic functions. All-optical computing could enable faster, more energy-efficient information processing than electronics. Here, the authors show diffractive optical systems can perform multi-stage nonlinear logic circuits in a single optical stage, enabling basic gates, parallel adders/subtractors, ripple carry adders, and image processing.
Bhatt et al. (Fri,) studied this question.