The development of modern civil industry, energy and information technology is inseparable from the rapid explorations of new materials. However, only a small fraction of materials being experimentally/computationally studied in a vast chemical space. Artificial intelligence (AI) is promising to address this gap, but faces many challenges, such as data scarcity and inaccurate material descriptors. Here, we develop an AI platform, AlphaMat, that can complete data preprocessing and downstream AI models. With high efficiency and accuracy, AlphaMat exhibits strong powers to model typical 12 material attributes (formation energy, band gap, ionic conductivity, magnetism, bulk modulus, etc.). AlphaMat’s capabilities are further demonstrated to discover thousands of new materials for use in specific domains. AlphaMat does not require users to have strong programming experience, and its effective use will facilitate the development of materials informatics, which is of great significance for the implementation of AI for Science (AI4S).
30 December 2021 in ACS Applied Materials & Interfaces
The physics of solid carbon dioxide and its different polymorphs are not only of great practical and fundamental interest but also of considerable importance to terrestrial and planetary chemistry. Despite decades of computer simulations, the atomic-level structures of solid carbon dioxide polymorphs are still far from well understood and the phase diagrams of solid carbon dioxide predicted by traditional empirical force fields or density-functional theory are still challenged by their accuracies in describing the hydrogen bonding and van-der-Waals interactions. Especially the “intermediate state” solid carbon dioxide phase II, separating the most stable molecular phases from the intermediate forms, has not been demonstrated accurately and is the matter of a long standing debate. Here, we introduce a general ab initio electron-correlated method that can predict the Gibbs free energies and thus the phase diagrams of carbon dioxide phases I, II and III, using the high-level second-order Møller-Plesset perturbation (MP2) theory at high pressures and finite temperatures. The predicted crystal structures, phase transitions, and Raman spectra are in excellent agreement with the experiments. The proposed model not only reestablishes the position of solid carbon dioxide in phase diagram but also holds exceptional promise in assisting experimental studies of exploring new phases of molecular crystals with potentially important applications.