発表者名 
藤本 健 
指導教員名 
齊藤 英治 教授 
発表題目（英語） 
Physics Simulation with Graph Neural Network 
要旨（英語） 
Computation time is one of the hardest challenge many physicist have ever faced with. Especially in fluid simulation, solving Navier Stokes e.q. for a plenty number of particles is quite arduous procedure from the perspective of computational time. Recent research, however, succeeded in reducing computational time drastically by applying Graph Neural Network to calculation of multi particles dynamics.
Graph Neural Network(GNN) has been one of the most prospective method of deep learning algorithm for last 5 years, which takes graph structure data as input and output target value. By focusing on the connection of data, which is often ignored in simulation or theoretical analysis of Physics, it is shown that computationally reasonable dynamics calculation can be achieved and directly obtain the new types of Hamiltonian expression in complex systems.
In this presentation, starting from the basic explanation of GNN and its application, I’ll give some detailed contents, evaluation of GNN based simulator method, and some amazing results of simulation. 
発表言語 
日本語 

発表者名 
古山 昂樹 
指導教員名 
志村 努 教授 
発表題目（英語） 
1D computer generated holographic memory 
要旨（英語） 
Currently a holographic memory is attracting attention as a next generation optical memory. Holographic memories have large capacity and high data transfer rate. We have proposed a time series signal holographic memory using a volume hologram, and have investigated its write/read characteristics and theoretical recording density limit. The conventional volume holographic memory has a very large recording density, but it has some disadvantages. One is the influence of the thermal expansion/shrinkage of the hologram. The Bragg condition is not maintained when the temperature changes, and the signal disappears due to the thermal expansion of the hologram. Second is the issue of reproducibility. Volume holographic memories cannot be duplicated in the entire disc at once, such as by injection molding like ROM type optical disks.
Thus, we have proposed a holographic memory using a surface hologram [1]. Since it uses the RamanNath diffraction, the signal is reconstructed even if the hologram expands or shrinks. Though the recording density is almost the same as that of conventional optical discs, holographic memory discs can be produced at single process by injection molding, etc. and the data transfer rate is comparable to that of volume holographic memory.
Now, we propose a 1D surface holographic memory and are analyzing its write/read characteristics. In this system, strips of holograms are recorded on the hologram disc and read out sequentially. Because each holograms do not overlapped spatially, unlike conventional holographic memory systems, the crosstalk noise between pages will be reduced and it has a possibility to improve the recording density. The hologram to be recorded is obtained by using the Computer Generated Hologram method and we plan to make it by metasurfaces. In this presentation, I will show the simulation results of 1D surface holographic memory and talk fabrication process with metasurface.
[1] S. Hirayama et al., Photonics, 6(2), 70 (2019). 
発表言語 
日本語 

発表者名 
増木 亮太 
指導教員名 
有田 亮太郎 教授 
発表題目（英語） 
Validity and Range of Applicability of Quasiharmonic Approximation on Calculation of Thermal Expansion and Phonon Frequency Shift 
要旨（英語） 
The thermal expansion is one of the most fundamental properties of solids. Because the volumechange of materials cause problems in various situations, the way to control the thermal expansion is intensely studied. Therefore, it is of great importance to clarify the physical mechanism of thermal expansion in detail.
Quasiharmonic approximation(QHA), which neglects all the effect of the phonon anharmonicity except the volumedependence of the phonon frequencies, is the most widely used method for the firstprinciples calculation of the thermal expansion. It is empirically known that QHA gives accurate results for the thermal expansion coefficient. However, QHA often fails to correctly describe the temperaturedependent shift of the phonon frequencies even for simple materials like silicon. This is quite
paradoxical and casts doubt to the validity of QHA because the thermal expansion coefficient is closely related to the Gruneisen parameter, which is linked to the phonon frequency shift.
In this research, we formulate a theory of the thermal expansion based on the selfconsistent phonon(SCP) theory, which takes into account the anharmonic effect in a nonperturbative way, and resolve the paradox. We prove that the Gruneisen formula also holds within SCP theory. By using the perturbation expansion, we show that the phonon anharmonicity gives a small correction to the theramal expansion coefficient. On the other hand, we show that the phonon anharmonicity gives a dominant correction to the Tdependent shift of the phonon frequencies. We perform numerical calculations on some materials to show that our theory is correct. 
発表言語 
英語 

発表者名 
松井 彬 
指導教員名 
有田 亮太郎 教授 
発表題目（英語） 
Numerical study on the transport phenomena in the antiferromagnetic skyrmion system 
要旨（英語） 
The antiferromagnetic (AFM) skyrmion has been attracting attention owing to its promising application to memory devices; the AFM skyrmion is free from the stray fields and the skyrmion Hall effect, which has been preventing the application of the ferromagnetic skyrmion. [1] Despite the recent observation of the AFM skyrmion phase in a bulk system [2], the basic properties of the AFM skyrmion, such as its stability and the transport phenomena, are yet to elucidate.
In this study, we numerically calculate the topological charge and spin Hall conductivity for the AFM skyrmion system. Exploiting the stateoftheart realspace calculation method, we investigate the transport phenomena of the AFM skyrmion system for various skyrmion sizes and coupling strengths. We reveal that the transport phenomena in the AFM skyrmion system can significantly deviate from that in the conventional ferromagnetic skyrmion, which holds great promise in future applications to devices.
[1] B. Göbel et al., Physics Reports 895, 128 (2021)
[2] S.Gao et al., Nature 586, 3741 (2020) 
発表言語 
英語 
