A Composition-Only Descriptor for Predicting Single-Phase High-Entropy Perovskite Oxides
High-entropy perovskite oxides (HEPOs) offer a vast compositional space through multicomponent substitution across the A and B sites of the ABO3 framework, but identifying which compositions form single-phase perovskites still relies on trial-and-error synthesis or per-composition computation, slowing materials discovery. Here, we propose a composition-only phase-formation descriptor, η, developed for equimolar five-component A-site systems, (A1A2A3A4A5)TiO3, by combining machine-learned-potential stability evaluations with synthesis and X-ray diffraction (XRD) labels. We show that phase-formation tendency can be captured by two compositional mismatch factors: ionic radius dispersion at the multicomponent site and average valence deviation from the charge-neutrality ideal. Using composition information alone, η outperformed the Goldschmidt tolerance factor and other composition-only descriptors in classifying the 40 XRD-labeled compositions and matched or exceeded the explicitly computed energy above hull, Ehull. In prospective tests, the synthesis outcomes of 16 previously untested compositions spanning the low- to high-η range were all consistent with the η ranking. Literature-reported B-site HEPOs were enriched on the low-η side, with 71.4% falling within the top 10% low-η region of a virtual candidate space, suggesting transferability across sites. This descriptor provides a practical pre-synthesis metric for prioritizing promising HEPO candidates.
Authors
- Atsunori Matsuda (ORCID: https://orcid.org/0000-0002-6493-1205)
- Kenta Hongo (ORCID: https://orcid.org/0000-0002-2580-0907)
- Ryo Maezono (ORCID: https://orcid.org/0000-0002-5875-971X)
- Hiroyuki Muto (ORCID: https://orcid.org/0000-0003-2816-1415)
- Go Kawamura (ORCID: https://orcid.org/0000-0001-6585-7636)
- Liangzhi Kou (ORCID: https://orcid.org/0000-0002-3978-117X)
- Wai Kian Tan (ORCID: https://orcid.org/0000-0002-0014-5475)
- Nazmul Shahadath
- Hanzo Tsubota (ORCID: https://orcid.org/0009-0002-3746-8056)
- Tom Ichibha
Institutions
- Toyohashi University of Technology (JP)
- Queensland University of Technology (AU)
- Japan Advanced Institute of Science and Technology (JP)
- Institute of Science Tokyo (JP)
Publication Details
- Journal
- Chemistry of Materials
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1021/acs.chemmater.6c02470
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Asahi Glass Foundation
- Japan Society for the Promotion of Science