Source | Matlantis Inc.
Matlantis Inc. (Cambridge, Mass., U.S.), the U.S. hub of the materials discovery arm of Japan’s AI company Preferred Networks Inc. (PFN, Tokyo) has announced an update to its Matlantis universal atomistic simulator, and the opening of its office in Cambridge, Massachusetts, both of which aim to accelerate the adoption of AI‑driven materials research across North America.
Matlantis supports large-scale material discovery by reproducing new materials’ behavior at an atomic level on the computer. Launched in July 2021 as a cloud-based software-as-a-service (Saas), Matlantis is used by more than 100 companies and organizations for discovering various materials including catalysts, batteries, semiconductors, alloys, lubricants, ceramics and chemicals. Its delivery of speed, elemental versatility and quantum-level accuracy also make it a powerful, flexible tool for composites manufacturers (details below).
The recent update introduces Version 8 of PFN’s AI technology named Preferred Potential (PFP), which the company says enables researchers across industries to accelerate discovery, improve predictive performance and unlock new frontiers in materials science with high levels of simulation accuracy.
According to Matlantis, PFP 8.0 marks a significant milestone as the “first” universal machine learning interatomic potential (MLIP) to be trained with datasets developed with a method called r2SCAN (restored-regularized strongly constrained and appropriately normed) functional. PFP versions up to Version 7 previously relied on datasets generated with a method called Perdew-Burke-Ernzerhof (PBE) functional, which has also been widely adopted by MLIPs other than PFP. It is known, however, that PBE has certain limitations in simulation accuracy — how closely computer-based simulations of materials’ behavior align with real-world experimental results.
The introduction of the r2SCAN method is the culmination of PFN’s continuous efforts over the past couple of years to overcome the accuracy limitations of the PBE-based approach. Developing training datasets with the r2SCAN method is reported to be more computationally intensive, requiring three to five times the computing time compared to the PBE method. However, because PFP 8.0 is now trained with the datasets built with r2SCAN as well as PBE, Matlantis users can achieve up to doubled the simulation accuracy in the same timeframe as the previous version.
“We believe this will further pave the way for the era of computer-based materials discovery,” says Daisuke Okanohara, CEO of Matlantis.
Matlantis enables research teams to:
Perform simulations from the first day of use. Provided as an Saas, users can access Matlantis via a browser and start searching for new materials. Because Matlantis’ MLIP has already been trained with massive datasets, users can immediately focus on material discovery without spending time building machine learning models.
Search a wide variety of undiscovered materials. As a universal atomistic simulator, Matlantis covers a wide variety of materials for batteries, semiconductors, catalysts and more, without the need of changing AI models depending on their types.
Accelerate materials discovery. Researchers can complete simulations in just a few hours that would otherwise require years of conventional DFT calculations. This speed-up transforms iterative design in materials discovery, reshaping the R&D process so that computational insights lead experiments, instead of merely validating them afterward.
Achieve higher simulation accuracy. With the new training datasets built with the r2SCAN method, Matlantis can simulate material properties with higher accuracy than common MLIPs in the same timeframe, further narrowing the gap between simulations and experiments.
How can Matlantis support composites manufacturing?
Matlantis PFP 8.0 supports everything from tailored fiber resin interfacial design to durability assessment and hybrid structure optimization, which can slash R&D timelines, prototype cost and accelerate innovation in fiber-reinforced and honeycomb composites.
More specifically, Matlantis can simulate complex architectures (such as fiber-matrix interfaces, damage progression and honeycomb tessellations) containing tens of thousands of atoms in seconds rather than months. This capability enables researchers to do virtual exploration of composite microstructures and interactions before committing to expensive experimental prototypes.
The simulator can also model nearly any element, including those in carbon, glass and aramid fibers, metallic coatings, resin matrices and even core materials like aluminum honeycomb. Crucial interfacial phenomena like fiber sizing, resin adhesion or moisture uptake, can be accurately modeled at the atomic level, guiding surface treatments and coatings for better bonding.
Finally, Matlantis accelerates hybrid R&D workflows, evaluating wide ranges of fiber matrix combinations (such as carbon/epoxy, glass/vinyl ester, aramid with toughened resin and sandwich cores) computationally before physical testing. By narrowing down the most promising systems virtually, Matlantis reduces prototyping cycles and material waste — ideal for composites manufacturers aiming for faster time-to-market.
“With PFP 8.0 we finally have an MLIP that keeps the best DFT‑level fidelity while spanning the most of periodic table,” says Matlantis technical advisor professor Ju Li, Ph.D., widely recognized for his work on atomistic modeling and materials research. “That accuracy‑plus‑speed combination lets engineers generate phase diagrams or screen multicomponent systems in hours or several days rather than weeks or months — work that directly informs alloy design, battery materials, and other high‑value applications. Establishing a U.S. office means we can collaborate even more closely with industrial and academic partners here, shorten feedback loops and bring new Matlantis capabilities to market faster.”
PFP 8.0 is developed using PFN’s supercomputer and AI Bridge Cloud Infrastructure (ABCI) 2.0 and 3.0 provided by Japan’s National Institute of Advanced Industrial Science and Technology (AIST) and AIST Solutions Co., Ltd. The use of ABCI 3.0 is supported by the ABCI 3.0 Development Acceleration Program.
Lu public network security: 37140202000173
