| アブストラクト: |
Prediction of structures and properties of materials from the knowledge of their chemical composition has been a longstanding problem in materials science. So far, many inventions with associated improvements have been made to simulate materials structures and properties quantitatively from first principles, i.e. starting from the Coulomb interaction between atoms and electrons without using empirical parameters. Density functional theory (DFT) is one of the most popular bases of the first-principles simulations, for which W. Kohn got the Nobel Prize in chemistry in 1998. A uniqueness of DFT lies in the fact that it deals with one-electron density distribution but not the many-body wave function of the electronic system. With DFT, after some approximations for practical use, we can obtain the electron density distribution in the ground state and forces acting on atoms in materials. Sometimes coupled with molecular dynamics method for treating motion of atoms, the theory has been successfully applied to simulations of electronic band structures, crystal structures, surface reconstructions, the structure of liquids or amorphous states, and so on.
Although its successes are so impressive, however, limitation of DFT, at least within the present approximation, has also been clarified: cohesive energy estimates for solids are overestimated, van der Waals interaction is not reproducible, activation barrier of chemical reactions is underestimated, and electronic excitation spectrum is not correctly calculated. It is also well known that the present DFT cannot correctly predict metal-insulator transition governed by electron correlation. To overcome these difficulties, various attempts have been made so far, some of which are not based on DFT but on the wave function theory (WFT): the variational Monte Carlo method (VMC), the diffusion Monte Carlo method (DMC) and the trans-correlated method (TC) are such examples. In this talk, I will briefly review these methods and propose that their combination will be a powerful tool to go beyond the present limitation of the first-principles simulation of materials. |
