报告题目：Ab initio Quantum Monte Carlo method; Role in Materials Genome
报告人：Professor Ryo Maezono1
1 School of Information Science, JAIST, Asahidai 1-1, Nomi, Ishikawa 923-1292, JAPAN.
Dr. Ryo Maezono (PhD/Applied Physics) is a full Professor at JAIST (Japan Advanced Institute of Science and Technology), school of Information Science, working on Simulation Science Research area. He got his BSc (1995) and PhD (2000) in Applied Physics at Tokyo University, majoring condensed matter theory working on the phase diagrams of magnetic oxides. He was a JSPS fellow (Tokyo University/1999-2000), working on the magnetic properties of oxides. He got a postdoctoral position at Cavendish Laboratory, Cambridge University (EPSRC fellow/2000-2002), and moved to NIMS (National Institute of Materials Science, Japan), as a tenure researcher (2001-2007). In 2007, he moved to JAIST and now a tenure faculty there. Since his postdoc in Cambridge, he has worked on Diffusion Monte Carlo (DMC) electronic structure calculations using massive parallel computations. He has published several DMC works using world top class huge parallel calculations, exploring cutting-edge of numerical quantum many-body problem.
The ab intio Quantum Monte Carlo method is another universal framework to achieve electronic structure calculations to be compared with other approaches like DFT (density functional theory) or MO (molecular orbital methods). It solves the ab initio Schrodinger equation without taking one-body forms, by using numerical Monte Carlo samplings applied directly to many-body wave functions, not decomposed into orbital functions. Because of less assumptions required, the method achieves more reliable ab initio evaluations especially on the electronic correlations over DFT or MO. With high parallel efficiency achieving around 99%, the method is rapidly getting more practical to be used over such problems like magnetism, surfaces, inter-molecular interactions, and other fundamental problems, those forming generally difficult challenges to the conventional DFT and MO. Unlike DFT with the ambiguities due to the choice of exchange-correlation functionals, or MO with the complexities in the various correction schemes indispensable, the QMC realizes simpler and straightforward evaluations toward high accuracy. These advantages push the ab initio QMC method to be a promising 'killer application' running on flagship supercomputers in US for Materials Genome, with the highest efficiency to extract the power of massive parallel computations. In the talk, we present our group's efforts for decades using several implementations like CASINO, QMCPack, and TurboRVB.
Keywords: Quantum Monte Carlo; ab initio; Material Genome