Towards understanding structure–property relations in materials with interpretable deep learning

Tien-Sinh Vu, Minh-Quyet Ha, Duong-Nguyen Nguyen, Viet-Cuong Nguyen, Yukihiro Abe, Truyen Tran, Huan Tran, Hiori Kino, Takashi Miyake, Koji Tsuda & Hieu-Chi Dam

npj Computational Materials9: 215 (2023)

A central challenge in the field of materials science involves the use of both experience and theory to explore the compositions and structures of materials with specific properties and subsequently validating them through experimentation. Traditionally, materials have been characterized based on their elemental compositions and structures. Researchers have primarily relied on their knowledge and experience to predict certain properties of hypothetical materials with specific compositions and structures. Traditional material calculations can provide accurate information on the physical properties of hypothetical materials, but they still have certain limitations. Unlike traditional approaches, materials informatics (MI) approaches initially involve the conversion of primitive data descriptions into appropriate representations that can be used for mathematical reasoning and inference. Recently, various deep learning (DL)-based MI approaches have been developed to address the challenges associated with material representation and to predict physical properties. However, DL models currently employed in materials research exhibit certain limitations in delivering meaningful information for interpreting predictions and comprehending the relationships between structure and material properties. In this work, Tien-Sinh Vu et al. from Japan Advanced Institute of Science and Technology, proposed an interpretable DL architecture that incorporates the attention mechanism to predict material properties and gained insights into their structure–property relationships. The proposed architecture was evaluated using two well-known datasets (the QM9 and the Materials Project datasets), and three in-house-developed computational materials datasets. Train–test–split validations confirmed that the models derived using the proposed DL architecture exhibit strong predictive capabilities, which are comparable to those of current state-of-the-art models. Furthermore, comparative validations, based on first-principles calculations, indicated that the degree of attention of the atoms’ local structures to the representation of the material structure is critical when interpreting structure–property relationships with respect to physical properties. These properties encompass molecular orbital energies and the formation energies of crystals. The proposed architecture shows great potential in accelerating material design by predicting material properties and explicitly identifying crucial features within the corresponding structures.

编辑概述

材料科学领域中的一个核心挑战是使用经验和理论来探索具有特定性质材料的组成和结构，并通过实验进行验证。传统上，材料可由其元素组成和结构进行表征。研究人员主要依靠知识和经验来预测具有特定组成和结构材料的某些性质。传统的材料计算尽管能够提供关于假设材料物理性质的准确信息，但仍具有一定的局限性。与传统方法不同，材料信息学（MI）方法首先将原始数据描述转换为可用于数学推理和推断的适当表示。近年来，许多基于深度学习（DL）的MI方法被开发出来，以应对材料表示方面的挑战并预测其物理特性。然而，目前在材料研究中使用的DL模型在提供用来解释预测和理解材料构效关系的有效信息方面表现不足。在本工作中，来自日本科学与技术高级研究所的Tien-Sinh Vu等人，提出了一种可诠释DL架构，该架构结合了注意力机制来预测材料特性并深入理解其构效关系。作者使用两个著名的数据集（QM9和Materials Project数据集），以及三个内部开发的计算材料数据集，对所提出的架构进行了评估。训练-测试-分割验证证实了使用DL架构导出的模型具有强大的预测能力，可与当前最先进的模型相媲美。此外，基于第一性原理计算的比较验证表明，在解释与物理性质有关的构效关系时，原子的局部结构对材料结构表示的注意程度至关重要。这些性质包括分子轨道能量和晶体形成能。通过预测材料性质并明确识别相应结构中的关键特征，本工作所提出的架构在加速材料设计方面显示出了巨大的潜力。