标题： 基于通用机器学习原子间势的表面稳定性建模：全面的解理能基准研究 显示英文标题

标题： Surface Stability Modeling with Universal Machine Learning Interatomic Potentials: A Comprehensive Cleavage Energy Benchmarking Study

摘要： 机器学习原子间势（MLIPs）通过弥合量子力学精度与经典模拟效率之间的差距，彻底改变了计算材料科学，使对周期表中材料特性的前所未有的探索成为可能。 尽管它们在预测体材料性质方面取得了显著成功，但尚未有系统评估来确定这些通用MLIPs（uMLIPs）在预测断裂能方面的表现，断裂能是决定断裂、催化、表面稳定性和界面现象的关键性质。 在此，我们使用我们之前建立的包含36,718个层状结构的密度泛函理论（DFT）数据库，对19种最先进的uMLIPs进行了全面基准测试，该数据库涵盖了元素、二元和三元金属化合物。 我们评估了多种架构范式，并分析了它们在化学组成、晶体系统、厚度和表面取向方面的性能。 我们的结果表明，训练数据的组成比架构的复杂性更为重要：在Open Materials 2024（OMat24）数据集上训练的模型，该数据集强调非平衡构型，在87%的情况下能够正确识别热力学最稳定的表面终止，且没有任何显式的表面能训练，其平均绝对百分比误差低于6%。 相比之下，在仅基于平衡数据集训练的架构相同模型表现出五倍高的误差，而在表面吸附物数据上训练的模型则灾难性地失败，误差增加了17倍。 值得注意的是，使用适当数据训练的简单架构在准确性上可以与复杂的变换器相媲美，同时提供10-100倍的计算速度提升。 这些发现表明，社区应关注战略性的训练数据生成，以捕捉相关的物理现象。 显示英文摘要

摘要： Machine learning interatomic potentials (MLIPs) have revolutionized computational materials science by bridging the gap between quantum mechanical accuracy and classical simulation efficiency, enabling unprecedented exploration of materials properties across the periodic table. Despite their remarkable success in predicting bulk properties, no systematic evaluation has assessed how well these universal MLIPs (uMLIPs) can predict cleavage energies, a critical property governing fracture, catalysis, surface stability, and interfacial phenomena. Here, we present a comprehensive benchmark of 19 state-of-the-art uMLIPs for cleavage energy prediction using our previously established density functional theory (DFT) database of 36,718 slab structures spanning elemental, binary, and ternary metallic compounds. We evaluate diverse architectural paradigms, analyzing their performance across chemical compositions, crystal systems, thickness, and surface orientations. Our results reveal that training data composition dominates architectural sophistication: models trained on the Open Materials 2024 (OMat24) dataset, which emphasizes non-equilibrium configurations, achieve mean absolute percentage errors below 6% and correctly identify the thermodynamically most stable surface terminations in 87% of cases, without any explicit surface energy training. In contrast, architecturally identical models trained on equilibrium-only datasets show five-fold higher errors, while models trained on surface-adsorbate data fail catastrophically with a 17-fold degradation. Remarkably, simpler architectures trained on appropriate data achieve comparable accuracy to complex transformers while offering 10-100x computational speedup. These findings show that the community should focus on strategic training data generation that captures the relevant physical phenomena.