Silicon wafer characterisation by laser ultrasonics and neural networks

Lamb waves propagating in plate like structures, have been investigated to characterise a silicon wafer in a non destructive way. A laser ultrasonic technique was first used for the generation and the detection of these guided acoustic waves. A two-dimensional Fourier transform was next performed on the whole experimental data to obtain the dispersion curves, revealing the propagation of symmetric and antisymmetric modes. The sample characterisation was carried out in an original way, directly comparing the experimental results to theoretical ones. A feedforward neural network, specifically trained to recognize the same propagation behaviours, achieved to fit the experimental dispersion curves with a theoretical anisotropic model. Knowing the wafer thickness and density, the elastic constants were then determined without any user intervention. This non contact method showed very good agreement with the results obtained in the literature, and was found to be prompt and easy to automate.