Determination of optical properties of Al 80Mn 20 quasi-crystalline alloy using neural networks

Abstract

The optical absorption of Al 80Mn 20 quasi-crystalline alloy is calculated from the reflectivity measurements at normal incidence in the energy range from 0.5 to 6.2 eV, using neural networks method. At the beginning, the sample is prepared in the alternate thin layers stacking form of Al and Mn (Multi-layer) selected thicknesses so as to have a total rate of Mn of 20% in alloy. The multilayer is treated thermically at 320 °C during one hour with the intention to obtain a sample of homogeneous composition by interdiffusion in solid state. Optical spectrum obtained by analysis of Kramers–Kronig of the reflectance data, presents an ambiguous behaviour at low energies. The calculations carried out by neural networks method, with a precision of calculation of about 10 −5, could dissipate this ambiguity. The behaviour at low energies of optical absorption spectrum calculated by the neural networks method is different and finds on the contrary a physical explanation. To argue the choice of the solutions provided by neural networks method, a preliminary test was applied to aluminium, by using literature data.