Neural network based reverse-back procedure for photoacoustic electronic characterization of semiconductors

Abstract

In this paper, a procedure for determining the coefficient of ambipolar diffusion of semiconductors based on their photoacoustic response is developed. The procedure is based on the processing of the experimental response using previously designed neural network which accurately performs thermal characterization and determines the thickness of the semiconductor sample combined with a reverse-back procedure in which the theoretical model of the photoacoustic response is used, which depends on the coefficient of diffusion of charge carriers in nonlinear mode. With the parameters obtained this way, theoretical photoacoustic response is repeatedly generated and compared to the experimental signal until a satisfactory match is achieved. Experimental measurements were previously performed on Si n-type circular plates with thicknesses levels of 830 μm, 417 μm and 128 μm using a transmission minimum volume open-cell experimental set-up. The accuracy of the procedure is discussed. The coefficients obtained by this procedure show good agreement with letarture predictions.