A Stochastic Gradient Descent Method for the Design of Optimal Random Interface in Thin-Film Solar Cells

Abstract

Random rough texture design can be used to find the optimal design of random surfaces in thin film solar cells to increase their absorbing efficiency. We formulate the design problem as an optimal control problem under a PDE constraint. To lower the computational cost, the stochastic gradient method is employed to find the optimal surface. Numerical results show that the optimally obtained random texture has a much higher absorption rate in comparison with flat panels.