split MLIP for layered materials greatly improves moiré-structure verification accuracy and energy and force prediction

AI editorPublished 2026.07.31Posted 2026.09.09 10:38 KST

A split MLIP that separates intralayer and interlayer interactions in layered van der Waals materials, together with a data curation approach, was presented. In large-scale moiré-structure prediction and verification, energy and force prediction improved by about 10 times over existing models, and a physical verification metric using the distribution of stacking configuration was also proposed. [1]

1/ 원문 최초 공개: 2026년 7월 31일.
2/ 최종 출판본(Version of record): 2026년 8월 31일.
3/ AI 활용 방법: split machine-learned interatomic potential
4/ 적용 과제: 층상 van der Waals 재료의 moiré 구조 대규모 예측·검증
5/ 보고된 결과: 에너지·힘 예측 정확도가 기존 모델 대비 약 10배 향상되고, stacking configuration 분포 기반 검증 지표를 제안
6/ For layered van der Waals materials, the researchers presented a split MLIP and data curation approach that separates intralayer and interlayer interactions.
7/ This method was used for large-scale structural prediction and verification of moiré structures, and energy and force prediction accuracy was reported to have improved by about 10 times compared with existing models.
8/ From the perspective of materials research, this result shows that more stable potential learning is possible by separating interactions in layered materials, and it can be interpreted as a useful computational framework for handling complex structural spaces such as moiré structures.
9/ It is also notable that, rather than relying only on force and energy errors, a physical verification metric based on the distribution of stacking configuration was proposed.
10/ However, the reported performance improvement depends on the layered-material data and validation setting.
11/ Therefore, whether the same level of improvement is reproducible in other material classes or levels of structural complexity requires separate verification. [1]

How to apply this

In layered van der Waals materials, one can try energy and force prediction and structural verification of moiré structures with a split MLIP that separates intralayer and interlayer interactions.

Organize DFT structure, energy, and force data so that intralayer and interlayer interactions are separated, and curate them by stacking type.
Train the split MLIP to predict energy and force simultaneously, and compare it with a baseline model such as a general MLIP or a single potential using the same partition.
Generate moiré-structure candidates at scale, then calculate the predicted stacking configuration distribution to see whether physically plausible stacking statistics are obtained.
Evaluate error and distribution together on an independent test set and in extrapolation regions to confirm that not only numerical improvement but also structural verification is preserved.

Data needed: DFT calculation data for layered van der Waals materials, training samples with intralayer and interlayer configurations separated, and moiré-structure samples containing various stacking configurations are needed.
Checking the result: On an independent test set, compare energy and force errors with baseline models, and confirm whether the stacking configuration distribution is physically reasonable and reproduces known structural trends.
What can go wrong: If the training data are biased toward specific materials or stacking patterns, the performance improvement and verification metrics may not hold outside that range.

이 가이드는 원문의 방법을 바탕으로 AI가 제안한 적용 방향입니다. 원문에 적힌 내용이 아니며 같은 결과가 재현된다는 보장도 아닙니다.

웹 자료를 바탕으로 AI가 자동 요약한 글입니다. 연구의 세부 조건과 결과는 연결된 원문을 확인해주세요.

split MLIP for layered materials greatly improves moiré-structure verification accuracy and energy and force prediction

How to apply this

Related concepts and practice

Sources