Purpose outline

The strong force that binds quarks, the smallest constituent of matter, within a proton can be explained by the theory of quantum chromodynamics (QCD). Currently, the only method that can precisely calculate this theory is lattice QCD using supercomputers. Our research group led by the University of Tsukuba (PACS Collaboration) has been conducting large-scale lattice QCD simulations using various supercomputers, including CP-PACS, the world’s fastest supercomputer in 1996 developed at the University of Tsukuba. We are currently performing ultra-large-scale lattice QCD simulations using the world’s second fastest supercomputer, Fugaku. In this project, we will perform research toward further precise lattice QCD simulations based on our achievements.

On the other hand, through ultra-large-scale lattice QCD simulations using Fugaku, we have begun to understand bottlenecks in the calculations that are different from those in conventional calculations. It is important to fundamentally solve these problems for further large-scale simulations. We will develop methods to solve these problems by applying AI technology, which has been attracting a great deal of attention in recent years.

Research plan

In this project, two subprojects ① “New physics search beyond the Standard Model by using physics point ultra-large-scale lattice QCD” and ② “Development of AI technology for next-generation lattice QCD” is established. Those subprojects will promote research on ultra-large-scale lattice QCD precision calculations that take full advantage of the performance of the fastest supercomputer in Japan, Fugaku, and the development of new lattice QCD calculation techniques by applying AI technology.

In subproject ①, we will conduct world-leading ultra-large-scale lattice QCD simulations using an highly optimized lattice QCD code for Fugaku developed by our research group led by University of Tsukuba. Furthermore, using those calculation results, we will contribute to study exploring physical phenomena that cannot be explained by the Standard Model of elementary particles. In subproject ②, we will develop and research new calculation methods that incorporate AI technology, especially machine learning technology, which is now being applied to lattice QCD calculations, with the aim of speeding up lattice QCD simulations and establishing an AI R&D infrastructure for next-generation large-scale lattice QCD simulations.

These studies will be conducted jointly by two research groups. One is a research group led by University of Tsukuba, which has been studying large-scale lattice QCD simulations for many years. The other is studying the application of AI technology to lattice QCD calculations. We also aim to create a new research axis by the synergistic effects of the complementary research of the different groups.