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

Deep learning(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.To address these limitations,we propose an interpretable DL architecture that incorporates the attention mechanism to predict material properties and gain insights into their structure–property relationships.The proposed architecture is 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 confirm 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,indicate 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.

Role of atomic-scale thermal fluctuations in the coercivity

The microscopic mechanism of coercivity at finite temperature is a crucial issue for permanent magnets.Here we present the temperature dependence of the coercivity of an atomistic spin model for the highest-performance magnet Nd_(2)Fe_(14)B.For quantitative analysis of the magnetization reversal with thermal fluctuations,we focus on the free energy landscape as a function of the magnetization.The free energy is calculated by the replica-exchange Wang–Landau method.This approach allows us to address a slow nucleation problem,i.e.,thermal activation effects,in the magnetization reversal.We concretely observed that the thermal fluctuations lead to a downward convexity in the coercivity concerning the temperature.Additionally,through analyzing the microscopic process of the thermal activation(nucleation),we discover the activation volume is insensitive to a magnetic field around the coercivity.The insensitivity explains the linear reduction of the free energy barrier by the magnetic field in the nucleation process.