Generative Design of Inorganic Compounds Using Deep Diffusion Language Models.

Due to the vast chemical space, discovering materials with a specific function is challenging. Chemical formulas are obligated to conform to a set of exacting criteria, such as charge neutrality, balanced electronegativity, synthesizability, and mechanical stability. In response to this formidable task, we introduce a deep-learning-based generative model for material composition and structure design by learning and exploiting explicit and implicit chemical knowledge. Our pipeline first uses deep diffusion language models as the generator of compositions and then applies a template-based crystal structure prediction algorithm to predict their corresponding structures, which is then followed by structure relaxation using a universal graph neural network-based potential. Density functional theory (DFT) calculations of the formation energies and energy-above-the-hull analysis are used to validate new structures generated through our pipeline. Based on the DFT calculation results, six new materials, including Ti2HfO5, TaNbP, YMoN2, TaReO4, HfTiO2, and HfMnO2, with formation energy less than zero have been found. Remarkably, among these, four materials, namely, Ti2HfO5, TaNbP, YMoN2, and TaReO4, exhibit an e-above-hull energy of less than 0.3 eV. These findings have proved the effectiveness of our approach.

TCSP: a Template-Based Crystal Structure Prediction Algorithm for Materials Discovery.

Fast and accurate crystal structure prediction (CSP) algorithms and web servers are highly desirable for the exploration and discovery of new materials out of the infinite chemical design space. However, currently, the computationally expensive first-principles calculation-based CSP algorithms are applicable to relatively small systems and are out of reach of most materials researchers. Several teams have used an element substitution approach for generating or predicting new structures, but usually in an ad hoc way. Here we develop a template-based crystal structure prediction (TCSP) algorithm and its companion web server, which makes this tool accessible to all materials researchers. Our algorithm uses elemental/chemical similarity and oxidation states to guide the selection of template structures and then rank them based on the substitution compatibility and can return multiple predictions with ranking scores in a few minutes. A benchmark study on the 98290 formulas of the Materials Project database using leave-one-out evaluation shows that our algorithm can achieve high accuracy (for 13145 target structures, TCSP predicted their structures with root-mean-square deviation < 0.1) for a large portion of the formulas. We have also used TCSP to discover new materials of the Ga-B-N system, showing its potential for high-throughput materials discovery. Our user-friendly web app TCSP can be accessed freely at www.materialsatlas.org/crystalstructure on our MaterialsAtlas.org web app platform.

Crystal Composition Transformer: Self-Learning Neural Language Model for Generative and Tinkering Design of Materials.

Self-supervised neural language models have recently achieved unprecedented success from natural language processing to learning the languages of biological sequences and organic molecules. These models have demonstrated superior performance in the generation, structure classification, and functional predictions for proteins and molecules with learned representations. However, most of the masking-based pre-trained language models are not designed for generative design, and their black-box nature makes it difficult to interpret their design logic. Here a Blank-filling Language Model for Materials (BLMM) Crystal Transformer is proposed, a neural network-based probabilistic generative model for generative and tinkering design of inorganic materials. The model is built on the blank-filling language model for text generation and has demonstrated unique advantages in learning the "materials grammars" together with high-quality generation, interpretability, and data efficiency. It can generate chemically valid materials compositions with as high as 89.7% charge neutrality and 84.8% balanced electronegativity, which are more than four and eight times higher compared to a pseudo-random sampling baseline. The probabilistic generation process of BLMM allows it to recommend materials tinkering operations based on learned materials chemistry, which makes it useful for materials doping. The model is applied to discover a set of new materials as validated using the Density Functional Theory (DFT) calculations. This work thus brings the unsupervised transformer language models based generative artificial intelligence to inorganic materials. A user-friendly web app for tinkering materials design has been developed and can be accessed freely at www.materialsatlas.org/blmtinker.

Deep Learning-Based Prediction of Contact Maps and Crystal Structures of Inorganic Materials.

Crystal structure prediction is one of the major unsolved problems in materials science. Traditionally, this problem is formulated as a global optimization problem for which global search algorithms are combined with first-principles free energy calculations to predict the ground-state crystal structure of a given material composition. These ab initio algorithms are currently too slow for predicting complex material structures. Inspired by the AlphaFold algorithm for protein structure prediction, herein, we propose AlphaCrystal, a crystal structure prediction algorithm that combines a deep residual neural network model for predicting the atomic contact map of a target material followed by three-dimensional (3D) structure reconstruction using genetic algorithms. Extensive experiments on 20 benchmark structures showed that our AlphaCrystal algorithm can predict structures close to the ground truth structures, which can significantly speed up the crystal structure prediction and handle relatively large systems.

High-Throughput Discovery of Novel Cubic Crystal Materials Using Deep Generative Neural Networks.

High-throughput screening has become one of the major strategies for the discovery of novel functional materials. However, its effectiveness is severely limited by the lack of sufficient and diverse materials in current materials repositories such as the open quantum materials database (OQMD). Recent progress in deep learning have enabled generative strategies that learn implicit chemical rules for creating hypothetical materials with new compositions and structures. However, current materials generative models have difficulty in generating structurally diverse, chemically valid, and stable materials. Here we propose CubicGAN, a generative adversarial network (GAN) based deep neural network model for large scale generative design of novel cubic materials. When trained on 375 749 ternary materials from the OQMD database, the authors show that the model is able to not only rediscover most of the currently known cubic materials but also generate hypothetical materials of new structure prototypes. A total of 506 such materials have been verified by phonon dispersion calculation. Considering the importance of cubic materials in wide applications such as solar panels, the GAN model provides a promising approach to significantly expand existing materials repositories, enabling the discovery of new functional materials via screening. The new crystal structures discovered are freely accessible at www.carolinamatdb.org.

Engineering magnetic anisotropy and exchange couplings in double transition metal MXenes via surface defects.

Two-dimensional (2D) materials have been experimentally proven to manifest almost all types of material properties observed in bulk materials. However, 2D magnetism was elusive until recently. In this work, we used an approach that synergistically uses density functional theory, and Monte Carlo methods to investigate the magnetic and electronic properties of magnetic double transition metal MXene alloys (Hf2MnC2O2and Hf2VC2O2) by exploiting realistic surface terminations via creating surface defects including oxygen vacancies and H adatoms. We found that introducing surface oxygen vacancies or hydrogen adatoms is able to modify the electronic structures, magnetic anisotropies, and exchange couplings. Depending on the defect concentration, a ferromagnetic half-metallic state can be realized for both Hf2VC2O2and Hf2MnC2O2. Bare Hf2VC2O2exhibits easy-axis anisotropy, whereas bare Hf2MnC2O2exhibits easy-plane anisotropy; however, defects can change the latter to easy-axis anisotropy, which is preferable for spintronics applications. The considered defects were found to modify the magnetic anisotropy by as much as 300%. Defects also produce an inhomogeneous pattern of exchange couplings, which can further enhance the Curie temperature. In particular, Hf2MnC2O2H0.22was predicted to have a Curie temperature of about 171 K due to a combination of easy-axis anisotropy and a connected network of enhanced exchange couplings. Our calculations suggest a route toward engineering exchange couplings and magnetic anisotropy to improve magnetic properties.

Revealing the Formation Energy-Exfoliation Energy-Structure Correlation of MAB Phases Using Machine Learning and DFT.

MAB phases became popular as ultrahigh-temperature materials with high damage tolerance and excellent electrical conductivity. MAB is used to exfoliate two-dimensional (2D) transition-metal borides (MBenes), which are promising materials for developing next-generation nanodevices. In this report, we explore the correlation between the formation energy, exfoliation energy, and structural factors of MAB phases with orthorhombic and hexagonal crystal symmetries using density functional theory (DFT) and machine learning. For this, we developed three different machine learning models based on the support vector machine, deep neural network, and random forest regressor to study the stability of the MAB phases by calculating their formation energies. Our support vector machine and deep neural network models are capable of predicting the formation energies with mean absolute errors less than 0.1 eV/atom. MAB phases with the chemical formulas, MAB, M2AB2, and M3AB4, where M = Nb, Mn, Ti, W, V, Sc, Cr, Hf, Mo, Zr, Ta, and Fe, and A = group III-A elements (Al, Ga, In and Tl), were investigated to find out the formation energy and their structure correlation. We demonstrated that the stability of a MAB phase for a given transition-metal decreases when the A element changes from Al to Tl. DFT revealed that M-A and B-A bond strength strongly correlates with the stability of MAB phases. In addition, the exfoliation possibility of 2D MBenes becomes higher when the A element changes from Al to Tl because of weakening of M-A and B-A bonds.