Computing coherent light scattering on the millimetre scale using a recurrent neural network without training

Heterogeneous materials such as biological tissue scatter light in random, yet deterministic, ways. Wavefront shaping can reverse the effects of scattering to enable deep-tissue microscopy. Such methods require either invasive access to the internal field or the computational solving of an inverse problem. However, calculating the coherent field on a scale relevant to microscopy remains excessively demanding for consumer hardware. Here we show how a recurrent neural network can mirror Maxwell's equations without training. By harnessing public machine learning infrastructure, the light-field throughout a $6 \, \textrm{mm}^2$ area or $110^3 \, \mu\textrm{m}^3$ volume can be calculated in 16 minutes. The elimination of the training phase cuts the calculation time and, importantly, it ensures a fully deterministic solution, free of training bias. We integrated our method with an open-source electromagnetic solver. This enables any researcher with an internet connection to calculate complex light-scattering in volumes that are larger by two orders of magnitude.Comment: 7 pages, 3 figure