Department of Electrical and Computer Engineering, Northwestern University, Evanston, IL, 60208, USA.
Sci Rep. 2022 Jul 13;12(1):11953. doi: 10.1038/s41598-022-15816-0.
While experiments and DFT-computations have been the primary means for understanding the chemical and physical properties of crystalline materials, experiments are expensive and DFT-computations are time-consuming and have significant discrepancies against experiments. Currently, predictive modeling based on DFT-computations have provided a rapid screening method for materials candidates for further DFT-computations and experiments; however, such models inherit the large discrepancies from the DFT-based training data. Here, we demonstrate how AI can be leveraged together with DFT to compute materials properties more accurately than DFT itself by focusing on the critical materials science task of predicting "formation energy of a material given its structure and composition". On an experimental hold-out test set containing 137 entries, AI can predict formation energy from materials structure and composition with a mean absolute error (MAE) of 0.064 eV/atom; comparing this against DFT-computations, we find that AI can significantly outperform DFT computations for the same task (discrepancies of [Formula: see text] eV/atom) for the first time.
虽然实验和密度泛函理论（DFT）计算一直是理解晶体材料化学和物理性质的主要手段，但实验成本高昂，DFT 计算耗时且与实验存在显著差异。目前，基于 DFT 计算的预测模型为进一步的 DFT 计算和实验提供了一种快速筛选候选材料的方法；然而，这些模型继承了基于 DFT 的训练数据中的大差异。在这里，我们展示了如何通过将人工智能与 DFT 结合起来，通过专注于预测“给定材料结构和组成的材料形成能”这一关键材料科学任务，比 DFT 本身更准确地计算材料性质。在一个包含 137 个条目的实验保留测试集中，人工智能可以根据材料结构和组成预测形成能，平均绝对误差（MAE）为 0.064 eV/原子；与 DFT 计算相比，我们发现人工智能在同一任务上的表现明显优于 DFT 计算（差异为[公式：见文本] eV/原子），这是首次实现。
