Leading materials discovery firm unveils Crystal Structure Prediction tool, with Honda as an early adopter.
BOSTON and TOKYO, Jan. 28, 2026 — Matlantis today announced the launch of Matlantis CSP (Crystal Structure Prediction), a new feature in its universal atomistic simulator that quickly identifies previously unknown stable crystal structures from the vast search space of atomic configurations and compositions in a given elemental system.
Novel materials are critical for addressing challenges like decarbonization and next-gen energy, but materials research and development has long relied on repeated synthesis experiments—even when success rates are low. Matlantis CSP adds a computational screening phase earlier in the workflow, enabling teams to eliminate physically impossible options upfront and concentrate on the most promising candidates.
Honda R&D is implementing Matlantis CSP to boost exploration efficiency in materials development—covering multi-component systems and metastable structures that were previously hard to assess because of high computational costs.
Surpassing the boundaries of traditional CSP methods
Previously, crystal structure prediction faced several persistent obstacles: DFT-based evaluations could take hours per structure, search processes often leaned toward specific compositions when exploring variable composition spaces, and large-scale runs might need complex environment setup and specialized knowledge. Matlantis CSP is built to overcome these limitations by integrating Matlantis’ core technology—its universal machine-learning interatomic potential PFP (Preferred Potential)—with proprietary algorithms and a parallel processing framework optimized for large-scale CSP. This provides:
1) High-throughput structure assessment: Leveraging PFP, Matlantis CSP can calculate energies in seconds to minutes per structure while preserving accuracy. It also includes safeguards to ensure robust calculation completion without stopping due to anomalous atomic configurations that often occur during searches.
2) Thorough and efficient search across composition space: Matlantis CSP features a proprietary algorithm built to explore the entire composition space while maintaining diversity in sampled structures. Compared to random search, it boosts efficiency by roughly 3–6 times, allowing comprehensive exploration without gaps across any composition.
3) Parallel processing framework tailored for the Matlantis ecosystem: To handle tens of thousands of trials quickly, Matlantis CSP optimizes memory usage and parallel execution for Matlantis. Users can start large-scale searches right away, with no need for complex environment configuration.
“We have high hopes for CSP as a technology that will drastically enhance exploration efficiency in materials development,” stated Mitsumoto Kawai, Chief Engineer of Device Process at Innovative Research Excellence, Honda R&D Co., Ltd. “With CSP, crystal structure searches—including multi-component systems and metastable structures that were once unfeasible—are now possible. Narrowing down promising crystal structures and compositions with high confidence before conducting experiments will not only raise the likelihood of creating next-gen materials but also cut down development timelines.”
Here is the link to the Honda case study.
Matlantis CSP has already yielded early results across various systems—oxides, alloys, and phosphides—uncovering over 10 previously unknown stable crystals. In the Ga–Au–Ca system, it found 13 new crystals, which significantly updates the phase diagram compared to existing databases.
“With Matlantis CSP, we’re making crystal structure prediction feasible at actual research scale—allowing teams to explore wider composition spaces, spot promising candidates earlier, and cut down time spent on low-probability experiments,” said Daisuke Okanohara, CEO of Matlantis. “We’re pleased to see Honda R&D acknowledge the impact CSP can make, and we’re eager to speed up the journey from simulation to synthesis with partners across industries.”
About Matlantis
Co-developed by PFN and ENEOS, Matlantis is a universal atomistic simulator that facilitates large-scale material discovery by replicating new materials’ atomic-level behavior on computers. PFN and ENEOS integrated a deep learning model into a traditional physical simulator to boost simulation speed by tens of thousands of times and support a broad range of materials. Matlantis was launched in July 2021 as a cloud-based software-as-a-service (SaaS) by Matlantis Corp. (formerly Preferred Computational Chemistry)—a firm jointly backed by PFN, ENEOS, and Mitsubishi Corporation.
Matlantis is utilized by more than 150 companies and organizations to discover diverse materials, such as catalysts, batteries, semiconductors, alloys, lubricants, ceramics, and chemicals. For additional details, please visit: .
Media Contact:
Emily Townsend
Scratch Marketing + Media on behalf of Matlantis
