Designing Cu-Zr Glass Using Multiobjective Genetic Algorithm and Evolutionary Neural Network Metamodels–Based Classical Molecular Dynamics Simulation

Abstract

Shear deformation analysis and diffusion behavior of Copper-Zirconium Bulk Metallic Glasses (BMG) is studied for different combinations of processing parameters. Melt holding temperature, melt holding duration, cooling rate of melt and composition of BMG are varied to obtain BMGs of different structures. The as-quenched structures are characterized using Radial Distribution Function (RDF) and Self-Diffusion constant values (of each atom type). The objective of this study is to design a Cu-Zr BMG which can absorb maximum energy and still deform as little as possible while maximizing its diffusivity during shear deformation of the structure. For this, metamodels were constructed by feeding the Molecular Dynamics (MD) results to an Evolutionary Neural Network (EvoNN) so as to generate the desired objective functions which are then optimized through multiobjective genetic algorithm. This led to identification of some hitherto unknown structures, characterized by atomic coordinates, which have good resistance to shear deformation and at the same time possess high diffusivity.