Abstract

An iterative hybrid processor capable of studying physical processes involving 2-D second-order partial differential equations is implemented by combining an electronic microcomputer (DEC 11/73) and an analog optical processor in a closed-loop architecture. Such an architecture enables a division of computational labor: 2-D global operations can be allocated to the analog optical processor and neighborhood operations and control functions to the microcomputer. The analog optical processor is a confocal feedback system (CFS) with dynamic range enhancement provided by a photorefractive image amplifier. As an initial application, the hybrid processor is programmed to simulate the time evolution of a collisionless plasma using a standard particle simulation code. Poisson’s equation, relating the electric potential to the charge distribution, is optically solved in real time by choice of the filters inserted in the midplane of the CFS and by control of the temporal phase of the coherent feedback. The charge distribution for the next iteration is then calculated from the electric potential by the microcomputer using a set of linear motion equations. Analysis for the hybrid processor predicts a 100-fold time savings over commonly available mainframe computers for the simulation of plasmas containing in excess of 10⁶ particles. Theoretical principles, convergence criteria, and experimental results for the processor are discussed.

© 1986 Optical Society of America