The development of new materials and the improvement of the properties of conventional materials have been based on the results of many experiments conducted by changing the chemical composition and the manufacturing conditions; empirical rules and phenomenology form the basis for this development and improvement. Although these have mainly been adopted for material development, a quite different technological approach has recently been developed. This new technology involves forming the structure of a metallic material with desirable properties based on theoretical assumptions from the structure of atoms and crystals. The manufacturing conditions for a material with that structure are then theoretically investigated. This procedure is conducted by computer simulation, and is called computational metal physics or computational materials science. The figure shows the models that are mainly used at present, classified in accordance with the quantity of atoms of the relevant system. From the most microscopic standpoint, the energy level of the electrons of one atom or atoms within a unit lattice is first analyzed by quantum mechanics. Next, when dealing with the interaction or migration of a large number of atoms, it is analyzed by molecular dynamics, the Monte Carlo method, or some other method. These analyses enable such properties as the lattice parameters, elastic modulus, specific heat, and electrical conductivity to be presumed, as well as structural information such as lattice defects, the grain boundary and surface, and transport properties such as diffusivity and viscosity. In the macroscopic interpretation, the mean field theory based on thermodynamics and a finite-element method are applied. The present level of computational materials science is already sufficient to explain some of the phenomena that are empirically known, and great progress with this science can be expected in future.