Identification of crystalline structures using Mössbauer parameters and artificial neural network

Abstract

Mossbauer spectroscopy is a useful technique for characterizing the valencies, electronic and magnetic states, coordination symmetries and site occupancies of Fe cations. The Mossbauer parameters of Isomer Shift (I.S.) and Quadrupole Splitting (Q.S.) are useful to distinguish paramagnetic ferrous and ferric ions in several substances, while the internal magnetic field provides information on the crystallinity. A correlation is being sought between Mossbauer parameters and several structure properties of some iron-containing minerals using Artificial Neural Networks (ANN). Distinct regions of crystalline structures are defined when any two parameters are plotted, but in several cases superposition of these regions leads to erroneous conclusions. We have tried to eliminate this difficulty by using convenient axes. These axes form n-dimensional vectors as input to our ANN. In recent years ANN has shown to be a powerful technique to solve problems as pattern recognition, optimization, preview ups and downs in stock market, automatic control and identification of a mineral from a Mossbauer spectrum or Mossbauer data bank. Using ANN we have been successful in identification of crystalline structures from plots of Mossbauer spectral parameters of I.S., Q.S., and structure properties of mean metal-oxygen distance in coordination site. Results using ANN in identification of crystalline structures using Mossbauer parameters of I.S., Q.S., and polyhedral volume of a coordination site are presented.