The author has been conducting, since April 2022, a “Hosei University RCCMS (Research Center for Computing and Multimedia Studies)” Research Project, entitled “Numerical Computation and Deep Learning Libraries Development on lab2022 Computers”. During the year 2023-2024, the author has tried to build an environment for artificial intelligence (AI) computations, with the final goal to integrate AI into high performance computing (HPC). In this article, the author explains separately the current statuses of the HPC and AI areas, discusses the possible future convergence of these two different areas, and briefly explains an example of AI computations.

The author has been conducting, since April 2022, a “Hosei University RCCMS (Research Center for Computing and Multimedia Studies)” Research Project, entitled “Computational Science Research Collaboration”. The author has carried out research on quantum chemical computation methods for understanding the mechanism of radioactive cesium concentration in mushrooms, in collaboration with the Center for Computational Study and e-Systems (CCSE) from Japan Atomic Energy Agency (JAEA). The author has also successfully installed and set up the computer software, GRRM, MOPAC, gr2moc, used for this project on lab2022 computers. This article reports the present on-going research activities within this Research Project.

In this study, the influences of the incoming boundary layer thickness on the suppression effects on the corner separation by the bowed blade were examined numerically. The test cascades were linear compressor cascades composed of straight blades and of bowed ones. The positive dihedral angles of the bowed blades at the endwall were varied from 10 deg. to 40 deg. at intervals of 10 deg. At an inlet boundary plane, a uniform velocity distribution and a non-uniform one with a 99% boundary layer thickness of 8.1% span were applied as a boundary condition. The computed results clarified the following trends: Compared with the typical straight blade, the present bowed blades suppress the corner separation regardless of the incoming boundary layer thickness. Meanwhile, the increase of the incoming boundary layer thickness decreases the most effective dihedral angle, at which the mass-averaged total pressure loss coefficient at the cascade outlet becomes minimum.

Triple point and critical point of water were estimated by constant energy molecular dynamics. The basic cell was cubic, where the central part was lattice of the cubic ice at the initial stage. The number of molecules in the basic cell was 5832. The SPCE model was assumed. The pressure and potential energy were obtained as functions of temperature. The critical point was assigned by the ratio of temperature T and the initial temperature T₀, T/T₀. The estimated triple point and critical point temperature were reasonable compared with experimental data.

Critical point of argon was estimated by constant energy molecular dynamics. The basic cell was cubic, where the central part was lattice at the initial stage. The number of molecules in the basic cell was 6912. The Lennard-Jones potential function was assumed. The pressure and potential energy were obtained as functions of temperature. The critical point was assigned by fluctuation of temperature and the ratio of temperature T and the initial temperature T₀, T/T₀. The estimated critical point temperature and pressure were reasonable compared with experimental data.

Optical spectrum prediction based on first principles calculations is important for the development of optical materials. In particular, real-time Dependent Density Functional Theory (TDDFT) is one of the most widely used computational methods. The output of a real-time TDDFT calculation is the dipole moment, which is Fourier transformed (FT) and converted to polarizability. From this polarizability, an optical spectrum is obtained. However, if the time series data are not sufficient, the FT results are ambiguous. To solve this problem, we introduced Singular Spectral Analysis (SSA), which decomposes time-series data into several fundamental oscillations. Only the major components are extracted, and the essential time-series data are reproduced. During this process, high-frequency oscillations that are perceived as noise are removed. We applied this technique to ethylene TDDFT time-series data. By focusing on the fundamental vibrations that constitute the band edges, which are important for understanding their optical properties, we were able to clarify the signal. Furthermore, because the fundamental vibrations can be separated, it was found that the time series data can be extended and supplemented when the time series data is inadequate.

To reduce hypervelocity impact damages from micro-debris and meteoroids, some spacecraft have been equipped with multi-layered bumpers. However, due to rapidly growing number of micro-spacecraft in both Earth orbit and interplanetary space, their small size and lightweight are under severe resource constraints; they are mostly operated without dedicated bumpers against impact protection. If Multi-Layer Insulation (MLI) which covers outers of most spacecraft for thermal control can protect against such hypervelocity impacts, it may effectively ensure the safety of micro-spacecraft that are difficult to install dedicated bumpers. In this study, four existing MLI configurations were selected from previous studies, and the effects on each MLI configuration against impacts by particles simulating micrometeoroids were investigated by using numerical simulation in the mass and velocity ranges that cannot be tested on the ground.

The 2024 Symposium of Research Center for Computing and Multimedia Studies, Hosei University was held on 20 February 2024. This short article introduces outline of the 2024 symposium.