3D Computer Generated Holography by Nonconvex Optimization

Our algorithm solves for the 2D phase in Fourier space that best approximates a user-specified volumetric intensity pattern near the image plane. We introduce a customizable cost function that allows for application-specific solutions in neural photostimulation, machining or 3D lithography, where intensity responses may be nonlinear.

By using a thresholded multi-part cost function for binary target intensities, we demonstrate robustness and high-fidelity patterning across a large number of depth planes. Our algorithm out-performs existing techniques in both simulation and experiment.