2025/01/23 by José C. A. Rocha, J. C. A. Rocha, Rocha, José C. A. +6 · 1 voice · 4 citations
Computer Science · Materials Science · Physics and Astronomy · #Neural Networks and Reservoir Computing #Metamaterials and Metasurfaces Applications #Random lasers and scattering media
paper · pdf · doi:10.1038/s41467-025-66798-2
Multi-plane light converters (MPLCs) - also known as diffractive neural networks - convert an orthogonal set of optical fields into any other orthogonal set via a unitary transformation. MPLC design typically involves optimising a digital model. However, inherently high levels of complexity mean that even a minor mismatch between this model and the physically realised MPLC leads to a severe reduction in performance. Here we create a self-configuring MPLC, converging in minutes while automatically absorbing unknown misalignments and aberrations into the design. To achieve this we introduce 'multi-plane wavefront shaping' - allowing multiple spatial light modes to be reshaped simultaneously. Convergence is accelerated via a high-speed MPLC platform incorporating a kHz-rate phase-only light modulator. Using this approach we demonstrate arbitrary optical transformations and universal mode sorters. Our work paves the way towards ultra-high-fidelity MPLCs with potential applications to optical communications, photonic computing and imaging.