• Home
• Outline

Outline

Background

For the past 50 years, materials science has provided the scientific basis for innovative devices as well as generated new concepts such as macroscopic quantum phenomena and revolutionized basic science.  This trend is expected to continue as the next generation of innovations in materials science will be through nanotechnology. Thus, the new manufacturing paradigm will be based on quantum theory. Currently desired structures can be constructed from selected elements using various material-construction and nanofabrication technologies. Hence, the current challenge is to determine how and what elements should be arranged. This challenge can be addressed using a computational scientific approach based on the first principle of quantum theory, which, owing to its scientific depth, essentially contributes to materials design.

Materials design has two important aspects: exploring whether a structure can be feasibly produced and understanding the properties of materials or structures. Materials science strives to elucidate the properties of materials and structures. However, common knowledge obtained in the twentieth century is becoming obsolete due to the emergence of the nanoworld in which the intrinsic properties of elements as well as their nanoscale configurations significantly influence the electron states and new characteristics hidden in bulk materials are revealed. Mutually competing factors such as covalency, ionicity, and electron correlation are intertwined with the nanoscale configuration of materials, yielding new complex correlation phenomena. Therefore, elucidating the correlations among multiple factors in real materials according to the first principle of quantum theory and establishing robust methodologies to design new materials are indispensable.

The producibility of materials or structures depends on the dynamics of material production and fabrication. Electron excitations on the attosecond scale induce atom transfer reactions on the picosecond scale, which produces new material phases. Materials and structures are fabricated through the dynamics of complex correlations on multiple time scales. Hence, the key to exploring new material functions is to understand and control the mechanisms of material production, while clarifying the non-equilibrium dynamics in material production processes on multiple time scales. Additionally, the discoveries of new material phases using recently developed fast spectroscopy have been frequently reported and elucidating non-equilibrium elementary processes has become increasingly important. Highly quantitative ab initio quantum simulations can reveal non-equilibrium dynamics by exploring the fundamental principles of elementary processes on space and time scales that are experimentally difficult to reproduce.

Advanced diffraction and spectroscopic experiments using next-generation neutron, muon, and X-ray sources are about to be launched at JPARC and SPring-8 in Japan. The computational scientific approach, based on the first principle of quantum theory adopted in this research project, should provide a robust theoretical backbone to explain the experimental results obtained from these progressive experimental techniques as well as reveal complementary knowledge. The computics approach combined with experiments should trigger a breakthrough in materials science and materials design.

In addition to theoretical and experimental approaches, a computational scientific approach has been steadily developed since the 1980s. In particular, numerous scientific and engineering fields have acknowledged the usefulness of ab initio calculations, which have been primarily based on density functional theory. However, to elucidate and predict complex correlations and non-equilibrium dynamics, existing techniques must be significantly improved and new computational methodology must be developed to explore numerous phenomena that cannot be described by existing methodology. Developments of supercomputers are certainly another factor to promote computational methodology for materials design.

A next-generation supercomputer Kei with a theoretical peak performance of 10 petaflops is scheduled for FY2012 in Kobe. However, this next-generation supercomputer requires advances in computational science. The performance of a single computing node is approaching its physical limit in fabrication. Thus, future high performance supercomputers will be inevitably based on a massively parallel architecture. In fact, Kei employs such architecture with a total of 640,000 computing cores. One drawback of massively parallel computers is that existing programs will not work effectively, and in many cases, only a few percent of the theoretical peak performance will be achieved. Based on a comprehensive understanding of computer architecture, new numerical algorithms must be implemented and parallel programs among nodes and cores must be developed. Unfortunately, the independent research and development in computational science and computer science in the last quarter century are obsolete from the perspective of effective use of such massively parallel high-performance computers. Additionally, specific computational engines for dedicated purposes using new hardware, including new generation of graphics processing units (GPUs) along with generic nodes, should be utilized in future next-generation exaflops supercomputers, which are expected to be available in 2018. Hence, high performance computing (HPC) will have a hierarchical multiplexing structure.

Computational science, which was originally initiated as a third approach, is entering a new phase.  Due to the qualitative changes in HPC, which represent a national project to develop the next generation of supercomputers, a new approach that integrates computational science with computer science and fully applies mathematical science is required. This approach is called "Computics". To reveal the mysteries of nature and to exploit new horizons in materials science, new academic fields that complement and collaborate with experimental studies must be established.

Objectives

Due to the reasons mentioned in the Background, we pursue the following five objectives.