Large-scale nonlinear optical computing with incoherent light via linear diffractive systems

Researchers from UCLA, led by Professor Aydogan Ozcan, have published a breakthrough paper demonstrating that linear optical systems can perform complex nonlinear computations using ordinary, incoherent light. The key innovation is an optimized diffractive processor—a stack of passive, patterned surfaces—that, when combined with a simple intensity-only encoding of the input data, can approximate a vast array of nonlinear mathematical functions. This directly addresses a major hurdle in optical computing: the weak nonlinearity of most materials and the high power required to trigger nonlinear optical effects. Their framework is specifically designed for real-world conditions where light is not perfectly coherent, like ambient illumination or light from an LCD screen.

Numerical simulations show the system's remarkable scale, capable of computing up to one million different nonlinear functions simultaneously in a single snapshot. The outputs are spatially multiplexed across a densely packed detector array. The team also provided an experimental proof-of-concept using incoherent light from an LCD, employing a model-free, in-situ learning strategy to optimize the diffractive surfaces and detector geometry directly on the physical hardware, accounting for imperfections. This work fundamentally redefines what's possible with passive optics, establishing diffractive processors as universal, massively parallel function approximators for incoherent and partially coherent light, moving beyond the limitations of traditional laser-based systems.