TapTechNews reported on May 15th that Microsoft Research AI for Science recently launched the MatterSim model, which can accurately and efficiently simulate materials and predict performance in a wide range of elements, temperatures, and pressures, helping to digitize material design. Exploring new materials is crucial for technological advancements in nanoelectronics, energy storage, and healthcare. One of the key challenges in material design is predicting material properties without actual synthesis and testing. The MatterSim model designed by Microsoft combines deep learning technology to simulate a variety of materials such as metals, oxides, sulfides, and halides in different states (crystals, amorphous solids, and liquids) within a wide range of synthesis and working temperatures, pressures. The training process of MatterSim uses large-scale synthetic data, obtained through active learning, molecular dynamics simulations, and generative models. This data generation strategy ensures the model's wide coverage of material space, allowing it to predict energy, forces, and stresses at the atomic level with accuracy comparable to first-principles predictions. MatterSim reduces data requirements by 90%-97% for fine material simulations and performance predictions. Through customization, MatterSim optimizes the task with only 3% of the original data, achieving the expected experimental precision simulation.