HPCS News picks up researchers who conduct their advanced scientific research leveraging large-scale computer in the Cybermedia Center.
vol.15 Understanding the cleaning process in semiconductor manufacturing
Affiliation : Clean Technology Development Operations, SCREEN Semiconductor Solutions Co., Ltd.
In the semiconductor manufacturing process, single-wafer cleaning equipment that can treat rotating wafers with liquids containing various chemical substances is used to remove contaminant metals and nano-size particles from silicon wafers. In order to get high-performance and high-efficiency cleaning methods, it is important to know the liquid state such as the velocity and the thickness of the liquid film on the rotating wafer. However, the wafer rotates at high speed and the liquid film is extremely thin, making it extremely difficult to actually know the state of the chemically dangerous liquid. In this study, we attempted to clarify the liquid flow on a rotating wafer using CFD simulation using the LES-VOF method. Although the LES-VOF method can obtain detailed and highly accurate results, it takes too much time with normal computing power. By using the computing power of the supercomputer SQUID, we were able to clarify the detailed state of the liquid in a short period of time.
vol.14 An elucidation of the triggered activity formation mechanisms leading to lethal arrhythmias
Affiliation : Associate Professor, Faculty of Medicine, Kanazawa Medical University
The heart is the pump responsible for blood circulation, the work of life. Electricity (excitation) flowing accurately within the heart is required to function properly for the blood pump. An abnormality in the electrical flow (propagation of excitation waves caused by the excitation of myocardial cells) is an arrhythmia, and arrhythmias in the ventricles may be lethal. Although it has believed that excessive sustained excitation of myocardial cells causes abnormal electrical impulses, leading to the development of lethal arrhythmias, how ventricular arrhythmias are triggered by their electrical impulses is still not understood. In this study, we repeated simulations of excitation propagation on a myocardial sheet model consisting of human ventricular myocytes under various conditions using OCTOPUS in Cybermedia Center, Osaka University, and succeeded in elucidating the mechanism of the triggering phenomenon (triggered activity), resulting in the development of the lethal arrhythmia. These results will be useful for establishing methods to predict and prevent arrhythmia developments.
vol.13 First-principles calculations for single-molecule electronic transport
Affiliation : Assistant professor, Graduate School of Engineering Science, Osaka University
Single-molecule devices attract attentions to understand electron transport mechanisms in molecular scales, and find molecular design principles for promising organic devices. The exploring the relation between the bridging structure and electric conductance is important for this purpose. Since it is not possible to directly observe the bridging structure of the molecule between electrodes, we employed first-principles calculations to identify the molecular bridging structures. Electronic transport calculations are computationally expensive, since we need electrode atoms in addition to the molecule. Therefore, a large-scale computer like OCTOPUS, Cybermedia Center, is required. We have developed a method to unveil the junction structures by combining the unsupervised clustering and first-principles calculations and applied it for a molecule having complex anchors.
vol.12 Toward accurate prediction of cooling heat loss in reciprocating engines: Clarification of wall heat transfer mechanism in pulsating turbulent flows
Affiliation : Associate professor, Faculty of Engineering Science, Kansai University
In reciprocating internal combustion engines, such as automobile engines, 20 to 30% of the thermal energy generated by fuel combustion is lost as cooling heat loss. Cooling heat loss is the fraction of thermal energy transferred to the coolant through the wall, and must be minimized to improve engine thermal efficiency since it does not contribute to engine output. In this study, the complex flow and thermal fields in the engine cylinder caused by the reciprocating motion of the piston is modeled as “pulsating turbulent flow with strong repetitive acceleration/deceleration and wall heat transfer.” Highly accurate turbulence simulation method was applied to clarify the mechanism of wall heat transfer in the pulsating turbulent flow. We believe that, through such simulations, clarifying the limitations of heat transfer models widely used in current design processes will contribute to the development of advanced models that can predict heat loss with higher accuracy.
vol.11 Molecular simulation on water and ions in the self-assembled membranes - Prediction of molecular properties by using computer simulation -
Affiliation : Specially Appointed Lecturer, Graduate School of Information Science, University of Hyogo
Specially Appointed Lecturer, Graduate School of Information Science, University of Hyogo
Abstract: The water-treatment membrane is a key issue to keep the safe and secure environments for water supply toward the sustainable development goals. In order to understand the nature of molecular transport in the membrane, we’re challenging the computational observation of water and ion dynamics in the nanomaterial by using the supercomputer. In this study, the reason why water molecules move fast inside the nanoconfinement is clarified with the simulation methods of quantum chemistry and classical molecular dynamics.
vol.10 Field-induced chiral transition and paramagnetic current in superconductivity ～Chiral phenomena generated by paramagnetic coupling of orbital magnetization to magnetic field～
Affiliation : Assistant Professor, Department of Material Science, Graduate School of Engineering, University of Hyogo
Electrons are pairing in the superconducting state, and various superconducting phenomena occur as features of spin and orbital of electron pairs. One of the phenomena is the chiral state with intrinsic orbital magnetization which generates the spontaneous field. It is not clarified why the chiral superconducting state yields. We study the nature of chiral state by simulation based on the Ginzburg-Landau equation. The numerical analysis clarifies the field-induced chiral transition, paramagnetic supercurrent, and the inversion of chirality. These phenomena are caused by the paramagnetic coupling between the orbital magnetization and the external magnetic fields when the superconducting state varies depending on the distance. The vector parallelization is useful for analyzing the dependence of chiral superconducting state, and it is performed by the SX-Aurora TSUBASA at the Cybermedia Center.
vol.09 Numerical simulation of cavitating turbulent flow
Affiliation : Assistant Professor, Department of Mechanical Engineering, Graduate School of Engineering, Osaka University
When the pressure drops below the saturated vapor pressure in liquid flows, bubbles appear and disappear in a short time. This phenomenon is called cavitation, and it has various harmful effects on fluid machinery such as pumps and marine propellers. In Japan, numerical methods of cavitating flow have been developed, especially in response to the demand for flow analysis in liquid hydrogen turbopumps of rocket engines. However, the quantitative reproducibility of the lift coefficient is still not sufficient even flows around basic hydrofoils. In this study, the reproducibility of lift of cavitating flow around a hydrofoil was improved by resolving fine turbulent vortices. The computation to capture the fine-scale turbulence vortices tends to be large scale. Supercomputer SX-ACE can execute such computation in a high speed.
vol.08 Analysis of time-course single-cell data for identifying genes that control disease
Affiliation : Assistant Professor, Department of RNA Biology and Neuroscience, Graduate School of Medicine, Osaka University
Organs and tissues consist of cell populations, and genes expressed in these cells can be investigated at a single-cell level. A cell changes from a stem cell into a mature specialized cell, which is called cellular differentiation. It is important to trace changes of gene expression levels during cellular differentiation. For instance, comparing time-course single-cell data of a normal mouse with those of a diseased one by a computational method will reveal which genes can control the disease. In this study, we develop a computational tool for efficiently comparing differentiation trajectories derived from single-cell data for two related experimental conditions. https://github.com/ykat0/capital
vol.07 Towards high performance data analytics to provide practical solutions to real world problems
Affiliation : Specially Appointed Associate Professor, Advanced High-Performance Computing System Architecture Joint Research Division, Cybermedia Center, Osaka University
With the experience of Cybermedia Center accumulated through the operation and use of large-scale computer systems, we are conducting research to build computing platform that enables high-performance big data analysis, and to produce practical and beneficial solutions to solve real world problems from the user’s perspective. In this video, we introduce a wide variety of applications leveraging machine learning and AI technology as practical examples using a large-scale computer system, which include (i) periodontal disease diagnosis and cephalometric landmark detection in dental clinical practice, (ii) individual identification of coconut crab in ecological research, (iii) colorization of black and white images after the atomic bombing in Hiroshima. (https://www.aca.cmc.osaka-u.ac.jp/)
vol.06 Real-Scale Social Simulations
Affiliation : 1Professor, Department of Informatics, Kansai University, 2Graduate School of Informatics, Kansai University
Social simulations are attracting attentions since one of them is awarded the Novel Memorial Prize in Economic Sciences. However, many social simulations are regarded too abstract. We are developing nation-wide synthetic populations in order to implement real-scale social simulations. We synthesize populations using available statistics in finer districts in cities, towns or villages.