An approach based on neural computation to simulate transition metals using tight binding measurements

A theoretical study of neural networks modeling, based on the tight binding approach, is proposed in this study. The aim of the present contribution is to establish a network topology to compute the binding energy parameters of transition metals. However, because of the different types of crystallization fcc, bcc, hcp, and sc of transition metals, neural network topology determination cannot be easily established, i.e. it would not be able to collect the data to feed the neurocomputing model. Hence, in order to overcome this problem, it would be helpful to distinguish one common structure from fcc, bcc, hcp, and sc. We observe that the structures fcc, bcc, and sc on (111) sheet and hcp on (0001) sheet form one common structure that is a two-dimensional hexagonal lattice. As the first application, the applicability of this choice of two-dimensional hexagonal lattice has been demonstrated by the ability to build a neurocomputing model able to determine the energy band structures of transition metals. Once the architecture is established, a second investigation will show the capability of data development techniques, by using the proposed common structure in our model, to calculate the binding energy parameters for transition metal atoms.