RIKEN Advanced Institute for Computational Science

Menu
Menu
Kcomputer 京について

Research Highlights

The K computer is being used in a broad range of fields including drug discovery, earthquake/tsunami research, weather forecasting, space science, manufacturing and material development.
Here are some examples of scientific results using the K computer.

You can find the most recent list of research achievements in "Press Release" and "Topics".

You can search some of the published achievements of researches using the HPCI system including the K computer on the HPCI Publication Database.

*HPCI(High Performance Computing Infrastructure)

2016.10.25 update

K computer and high-tech weather radar come together to predict sudden torrential rains

T. Miyoshi, M. Kunii, J. Ruiz, G.-Y. Lien, S. Satoh, T. Ushio, K. Bessho, H. Seko, H. Tomita, and Y. Ishikawa, "“Big Data Assimilation” Revolutionizing Severe Weather Prediction", Bulletin of the American Meteorological Society, doi: 10.1175/BAMS-D-15-00144.1

Related Links:

RIKEN (Press Release)
RIKEN (K computer Newsletter)

2016.10.25 update

Current atmospheric models underestimate the dirtiness of Arctic air

Journal:Sato, Y. et al. Unrealistically pristine air in the Arctic produced by current global scale models. Sci. Rep. 6, 26561; doi: 10.1038/srep26561 (2016).

Related Links

RIKEN (Press Release)

2016.10.03 update

The metal-insulator transition depends on the mass of the Dirac electrons

Yuichi Otsuka, Seiji Yunoki, and Sandro Sorella, "Universal quantum criticality in the metal-insulator transition of two-dimensional interacting Dirac electrons", Physical Review X, doi: 10.1103/PhysRevX.6.011029

Related Links:

RIKEN (Press Release)

2016.10.03 update

K computer provides new insights into fullerenes

Chan, B., Kawashima, Y., Katouda, M., Nakajima, T. & Hirao, K. From C60 to infinity: large-scale quantum chemistry calculations of the heats of formation of higher fullerenes. Journal of the American Chemical Society 138, 1420–1429 (2016). doi: 10.1021/jacs.5b12518

Related Links:

RIKEN (Press Release)
RIKEN (Riken Research Highlight)

2016.06.13 update

K computer takes first place in Graph500

The K computer reclaimed the top spot in Graph500, a new benchmark that seeks to gauge the ability of supercomputers to process data-intensive loads. The goal of Graph500 is to improve computing related to complex data problems in five key areas: cybersecurity, medical informatics, data enrichment, social networks, and symbolic networks.

In Graph500, the performance metric used is the speed of a breadth-first graph search, measured by the number of traversed edges per second (TEPS), with "edges" indicating the connection between two data points called “nodes”. Using 82,944 compute nodes of the K computer, the researchers solved a search of an extremely large graph (1 trillion nodes and 16 trillion edges) in 0.45 second. The K computer grabbed the top spot with a score of 38,621 giga TEPS.

This achievement demonstrates the flexibility of the K computer in a wide range of applications, including Big Data analysis.

This news was announced on July 14 (July 15 Japan time) at ISC15 held in Germany. Complete results of Graph 500 (July 2015):

Related Links:

RIKEN (Press Release)

2016.06.06 update

New Materials Development Technology achieved in 2015 -Enormous potential to produce a variety of tires that will provide new value-

Sumitomo Rubber Industries, Ltd. (SRI) has been working to develop the next generation of their New Materials Development Technology: "ADVANCED 4D NANO DESIGN". This technology will utilize the K computer and other cutting-edge equipment to perform highly advanced material simulations and analysis, thereby allowing SRI to achieve significant improvements in terms of the three major conflicting tire performance traits: grip performance, fuel efficiency and wear performance.

This technology will utilize the world-class SPring-8 and J-PARC research facilities to analyze and better understand the chemical structure of rubber and the behavior of the atoms and molecules that make up rubber materials. Using the K computer to build highly realistic models for more accurate simulations of rubber, this technology will allow us to see and understand the phenomena behind the atomic and molecular behavior of rubber materials through large-scale modeling and simulation.

The development is proceeding according to schedule and SRI planned to officially unveil the new technology at the Tokyo Motor Show in October. "ADVANCED 4D NANO DESIGN" will be incorporated in actual product development starting in 2016.

Figure: MD simulation of biological molecule

Related Links:

Sumitomo Rubber ( Press Release)

2016.06.06 update

New software allows simulation of molecular dynamics in large systems

Scientists have developed a software package called GENESIS (GENeralized Ensemble Simulation System), which is designed to perform molecular dynamics (MD) simulations for studying large biological systems containing 10 or even 100 million atoms. This software promises to usher in a new era in computational biophysics and biochemistry by allowing scientists to make connections between molecular and cellular-level understanding and to integrate experimental knowledge with theoretical and computational insights.

A key advantage of GENESIS is its superior computational efficiency on massive parallel supercomputers like the K computer. Using GENESIS, more than ten thousand CPUs can be used in parallel without any reduction in computational efficiency.

To investigate biomolecular dynamics and function within more realistic cellular environments than conventional simulations of biomolecules (e.g. proteins, DNA, membranes, and oligo-saccharides), much larger biological systems need to be simulated in milli- or microsecond time scales. GENESIS has the potential to be a good computational platform in this context, as it will help to overcome the current limitations, in terms of size and time, of biological MD simulations.

Jung J., Mori T., Kobayashi C., Matsunaga Y., Yoda T, Feig M., and Sugita Y. WIREs Comput Mol Sci. doi: 10.1002/wcms.1220.

Figure: MD simulation of biological molecule

Related Links:

RIKEN:
(GENESIS official site)
( Press Release)

2016.06.06 update

Largest dark matter simulation using K computer and supercomputer ATERUI

- Elucidating the formation and evolution of dark matter halos over 13.8 billions years, from the early universe to the present with the greatest accuracy ever -

It is said that 5 times as much dark matter exists compared to baryonic matter*, which is the ordinary substance we see as atoms and molecules. It is important to understand the evolution of dark matter in order to reveal the structural formation of the Universe. Through the effects of gravity, dark matter halo merges together to form larger structures. There is an extended dark matter halos around a galaxy whose mass is 10 times as large as the total mass of all the stars in the galaxy.

The research group calculated gravitational evolution of 550 billion dark matter particles from the early Universe, during a period of 13.8 billion years . The size of the volume for the simulations was 5.4 billion light years on each axis. The simulation in this huge space yields the world’s highest resolution ever, and is the world’s largest simulation of dark matter structural formation.

Mock catalogues of galaxies and active galactic nuclei will be prepared. These will be comparable to actual astronomical observations, and will be publicly available in the near future.

*Baryon: A baryon is a composite subatomic particle made up of three quarks. The most familiar baryons are the protons and neutrons that make up most of the mass of the visible matter in the universe.

Ishiyama T., Enoki M., Kobayashi M., Makiya R., Nagashima M.. and Oogi T., Publ Astron Soc Jpn, doi: 10.1093/pasj/psv021.

Figure: Dark matter distribution (Chiba University)

Related Links:

Chiba University (Japanese Press Release)
Center for Computational Astrophysics (Japanese Press Release)

2016.05.30 update

Successful real-time flood analysis using supercomputer-based high resolution tsunami modeling improves precision of tsunami and helps cities become more resilient to damage

Tsunami height calculations three minutes after the start of The Great East Japan Earthquake were underestimations, exposing an urgent need to improve methods of real-time estimation. The disaster also revealed the need for information other than wave height, such as the extent of flooding.

The joint research team of Tohoku University and Fujitsu Laboratories Ltd. developed a high-resolution (5 meter mesh) tsunami model, which runs on a supercomputer. This tsunami model quickly predicts the extent of flooding based on observation data at the time of earthquake. The researchers verified this model on the K computer, and they found that simulations of the general extent of flooding in the City of Sendai could be completed in two minutes. Performing the same computations on a workstation would take several days.

By generating high-resolution and real-time predictions of general tsunami-flooding conditions using this technology, the researchers will contribute to more effective disaster-response measures.

Oishi, Y., Imamura, F., and Sugawara, D. (2015) Geophys. Res. Lett. 42: 1083–1091, doi: 10.1002/2014GL062577.

Figure :Predicted tsunami arrival times (Tohoku University, Fujitsu Laboratories)

Related Links:

JAMSTEC (Press Release)
Fujitsu (Press Release)
Tohoku University (Japanese Press Release)

2016.05.30 update

K computer used to elucidate mechanisms of electron acceleration in a strong shock wave Novel theory for production of high-energy electrons

Astrophysical plasma exploding from supernovas interacts with interstellar gas, producing a strong shock wave. A shock wave shines brightly because high-energy electrons that are accelerated to nearly the speed of light (relativistic electrons) emit electromagnetic waves in various wavelengths (from radio to gamma rays).

The researchers performed calculations involving 10 billion plasma particles by using the K computer, and elucidated the structures of shock waves, which have never before been provable. At the shock front, part of the plasmas are reflected upstream (toward the direction of the astral body), creating filamentary structures, where magnetic islands are formed. Repetitive collisions of magnetic islands and electrons were found to generate high energy electrons.

This achievement will contribute to elucidation of the “existence of relativistic electrons”, which is one of the great questions of astrophysics.

Matsumoto Y, Amano T, Kato T. N., Hoshino M. (2015) Science 347:974-978. doi:10.1126/science.1260168.

Figure : Supercomputer simulations of a strong collisionless shock revealing spontaneous turbulent reconnection (Chiba University)

Related Links

Science (Abstract)
Chiba University (Japanese Press Release)
The University of Tokyo (Japanese Press Release)
Center for Computational Astrophysics (Japanese Press Release)