Work

Navigating order-disorder family trees via group-subgroup transitions

Year: 2026
Type: article
Abstract: Novelty in materials discovery is conventionally assessed against databases of ordered crystal structures, yet a predicted ordered structure may simply represent a particular ordering of a known disordered compound, rendering its apparent novelty misleading. We refer to such cases as ordered children of disordered parents. To address this limitation, we introduce order-disorder family trees, a symmetry-based framework that organizes ordered and disordered structures through group-subgroup relations and enables novelty to be explicitly evaluated. Using a high-throughput family-matching procedure, we identify candidate disordered parents and symmetry-related ordered relatives of a given ordered structure. We validate the framework using synthesis-facing case studies from A-Lab, where it successfully recovers known disordered parents of targeted ordered structures. We then extend this family-tree-based analysis to experimental structure databases (ICSD), computational datasets (MP-20, Alex-MP-20, and GNoME), and crystal generative models, revealing that many ordered structures that appear novel when considered individually are better understood as members of experimentally known order-disorder families. This behavior is particularly pronounced in symmetry-agnostic all-atom generative models, which more frequently produce ordered structures derived from known disordered parents, whereas symmetry-constrained models are 2–4 × less prone to this behavior. Our results establish order-disorder family trees as a key requirement for achieving genuine novelty in data-driven materials discovery.
Language: English
Open Access status: gold
Estimated APC paid: $3,490
OpenAlex ID: W7155565305
FWCI: 1.688
Cites: 46
Cited by: 2
Related to: 10