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

Abstract

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 on our MaterialsAtlas.org web app platform.

Crystal Transformer: Self-learning neural language model for Generative and Tinkering Design of Materials

“…While prediction models have been proposed for synthesizability prediction [53], formation energy prediction [54], and e-above-hull calculation, these models and algorithms usually require the availability of the crystal structures. Fortunately, recent progress in template based [55,52], deep learning based [56], and global optimization based crystal structure prediction tools [57,58] have made it possible to guess the crystal structures for increasing families of materials, which can be combined with our composition generators to explore and discover new materials.…”

“…Next, we calculate their total energy and predict their e-above-hull energies to rank these candidates. We then pick the top 100 formulas with the lowest predicted e-above-hull energy and apply our TCSP, a template based crystal structure prediction algorithm [52] to obtain the structures. For the predicted structures with the best quality scores, we run DFT relaxation to get the final structures and calculate their formation energy and e-above-hull using DFT method (see Method).…”

Applications of Artificial Intelligence and Machine Learning Algorithms to Crystallization

“…The crystal structure is defined by its lattice and motif (size of the unit cell and space group), and a description of the cell contents, which together provide the pattern by which molecules tesselate by translation to create a (perfect, pure) crystal. The approaches to CSP are based on (i) ab initio global optimization based on first-principles, (ii) machine-learning techniques to speed up, support the decision making, and cluster outcomes in intermediate steps in the workflow, and (iii) template-based elemental substitution . The steps in the workflow of CSP are outlined in Figure .…”

“…The steps in the workflow of CSP are outlined in Figure . Typically, the process to obtain such a stability-ranked list of crystal structures starts with the specification of the chemical system (contents of the unit cell, given as 2D input molecules), a reference temperature (0 K), pressure (0 Pa), and a metric corresponding to the number of entities in the unit cell (formula unit) and the space groups , considered, all in order to constrain the search space enough for computations to be feasible. ,, Quantum mechanics (QM)-based energy evaluations, , such as Density Functional Theory (DFT) calculations, can be used directly or train molecular mechanics force fields to optimize the crystal lattice energy. The lattice energy is given by the summation of intermolecular (electrostatic and repulsion-dispersion) and intramolecular forces (difference of the energy of the molecule in gas to crystal state) .…”

Machine-Learning-Assisted Acceleration on High-Symmetry Materials Search: Space Group Predictions from Band Structures

“…Inverse design − has been an important direction in material chemistry because of rapid progress and growing demands in new material developments. There are two major approaches: (i) search of wanted material systems from existing materials − and (ii) fabrications of new materials − with desired properties. For the first one, a direct query of material systems at existing material databases − is efficient.…”