Yoh-Han Pao

Computational Intelligence Procedures for Monitoring and Control of MBE Growth of Arbitrary Film Structures

Abstract This paper describes the use of computational intelligence procedures for real-time interpretation of multiwavelength ellipsometry data and the use of the results in monitoring and control of Molecular Beam Epitaxy (MBE) growth of thin films. Data reduction is carried out with the Guided Evolutionary Simulated Annealing (GESA) approach, which is a genetic algorithm like optimization method, and a neural net model of the available multiwavelength spectroscopic materials data. The smaller noise level in the Ψ parameter was taken advantage of to improve the accuracy of the results. Control of the film growth is accomplished through adjustments of crucible temperatures. Due to the constraints in data acquisition and processing, such adjustments may be in either feedback mode or feedforward mode depending on the profile of the structure and the switching between the two modes are carried out automatically. Results from two growth examples of different target profiles are reported for the system of AlxGai_xAs film on GaAs substrate. In one case, there was a single step change in composition while linear and parabolic grades were included in the other.