Accelerating materials design with AI emulators and generators

Materials design is a challenging and time-consuming process that requires exploring a vast and complex chemical space. To accelerate this process, we present MatterSim and MatterGen, two novel models that can emulate and propose novel materials with desired properties. MatterSim is a machine learning model actively trained from large-scale first-principles computations for efficient atomistic simulations at first-principles level and accurate prediction of materials’ properties across the periodic table and across a wide range of temperatures and pressures. MatterGen is an atomistic generative model that is able to propose novel and stable materials across the periodic table. Furthermore, the model can be fine-tuned to conditionally generate stable, novel materials with desired chemistry, symmetry, as well as mechanical, electronic and magnetic properties. These models unlock the large-scale discovery, exploration, and simulation of novel crystalline materials under a wide range of thermodynamic conditions, and open new possibilities for computational materials design.