Cloud-based service simulates atomistic behavior of undiscovered materials up to 20 million times faster than conventional method

TOKYO--(BUSINESS WIRE)--Preferred Computational Chemistry (PFCC), a joint venture between Preferred Networks, Inc. (PFN) and ENEOS Corporation (ENEOS), has launched Matlantis™, a cloud-based, high-speed universal atomistic simulator for artificial intelligence (AI)-driven materials discovery, for companies and organizations in the United States.

Based on a proprietary AI technology featured in Editor’s Highlights in the scientific journal Nature Communications last year, Matlantis is currently helping over 50 Japan-based companies and organizations search for new materials on computers up to 20 million times faster than the conventional density functional theory (DFT) method.

Since PFCC launched Matlantis in Japan in July 2021, PFN and ENEOS have been continuously updating its simulation accuracy and universality while PFCC expanded the client base, incorporated user feedback and prepared for international launches.

The current version of Matlantis now available in the U.S. can simulate up to about 19,000 atoms at a time and supports any combinations of 72 elements which comprise 99.9969% of the total mass of all elements above the earth’s crust^[source]. The latest version 4 of the neural network potential named Preferred Potential (PFP), on which Matlantis is based, has been trained with a vast dataset of over 33 million virtual structures of molecules and crystals, enabling users to simulate behavior of undiscovered materials at an atomic level while searching for promising candidates. PFP’s training dataset was generated using supercomputers with computational resources equivalent to what a single graphic processing unit (GPU) would need 1,650 years of computation time for.

In three months between January and March 2023, Matlantis simulated over 2.6 trillion atoms for its clients in a range of domains including academia, automotive, chemical, electronics and energy industries. The use of target materials includes batteries and semiconductors as well as catalysts, absorbents and alloys for clean energy, environment and sustainable industry processes.

“The U.S. has been the epicenter of countless technological breakthroughs and we are truly excited to bring Matlantis to help computational scientists in the U.S. make even more of them,” said Daisuke Okanohara, CEO of PFCC. “We will also launch Matlantis in other markets to accelerate innovation by making it faster for researchers around the world to discover new materials for a sustainable future.”

To help researchers learn about the latest trend in materials informatics and application of neural network potential for materials discovery, PFCC plans to host a free webinar at 7:00 - 8:30 pm Eastern Time on Monday, May 29, 2023 with Ju Li, a professor at the Department of Nuclear Science and Engineering and the Department of Materials Science and Engineering at Massachusetts Institute of Technology. Registration is currently open at matlantis.com.

The latest version of Matlantis’s neural network potential was developed using National Institute of Advanced Industrial Science and Technology’s AI Bridging Cloud Infrastructure (ABCI) as well as PFN’s in-house supercomputers.

About Preferred Computational Chemistry

Preferred Computational Chemistry, Inc. (PFCC) was established in June 2021 in Tokyo as a joint venture between Preferred Networks, Inc. (PFN) and ENEOS Corporation (ENEOS) for the sale of Matlantis™, a cloud-based universal atomistic simulator jointly developed by PFN and ENEOS for high-speed materials discovery and development. PFCC's mission is to accelerate innovation and support companies and organizations to discover innovative materials for a sustainable future.

Matlantis™ is a trademark or registered trademark of Preferred Computational Chemistry, Inc. in Japan and elsewhere. Chainer™ and MN-Core™ are trademarks or registered trademarks of Preferred Networks, Inc. in Japan and elsewhere.

Contacts

Yumi Sakaguchi or Tomoyuki Akiyama

Preferred Computational Chemistry

[email protected]