Neurocomputing approach to predict 2D ferromagnetic lattices by spin dynamics

Abstract

An artificial neural networks modeling, based on spin dynamics, has been proposed to predict the 2D ferromagnetic crystals. The purpose is to establish an artificial neural network model (ANNM), which is able to predict the ordered magnetic surfaces. In crystal surfaces, Wood’s notation is used to describe different configurations. From the magnons dispersion curves and group velocity data of 2D spins crystallographic waveguide, we deduce the data used as an up input–output set to feed the ANNM. The input data are being given by the magnons modes over high symmetry points, or, by the group velocity of the magnon modes. However, the basis vectors of matrix notation and the rotation angle that aligns the unit cell of the reconstructed surface are giving the output data. The proposed neurocomputing model was capable of predicting the 2D crystals for the possible arrangements that the surface might take. In addition, the magnon excitations data were a useful approach for building a neural model which was able to predict and classify the magnetic surfaces of the crystals.