AI helps design new materials that work in the real world
The “CrysVCD” tool developed at MIT could cut the huge amounts of time and money spent on screening out chemically unstable designs.
Artificial intelligence (AI) models can now generate millions of new material designs in mere minutes. However, this advancement has not led to a significant increase in the utilization of these materials for enhancing product performance, such as computer chips and rockets. A primary reason for this disparity is that current models fail to effectively account for the chemical stability of the generated materials, rendering them useless in practical applications.
Consequently, industries must dedicate substantial computational resources to filtering out unstable materials, often leaving only a small fraction of viable options.
MIT researchers have devised a framework designed to be implemented at the onset of the material generation process, thereby markedly improving the stability rate while preserving targeted material properties. Known as "crystal generator with valence-constrained design" or CrysVCD, this approach ensures that every design adheres to fundamental chemical principles concerning electron arrangements around atomic materials before initiating the resource-intensive generation phase.
The team, led by associate professor Mingda Li, published their findings in Nature Computational Science, demonstrating that CrysVCD can enhance the valence shell rules of several prevalent material models and achieve high lattice-dynamics stability in nearly 70 percent of computational material generations. The researchers also showcased the potential of their method in creating materials with desired properties, such as high thermal conductivity or high dielectric constant, essential for computer chips and data centers.
Li likened material-generating models to DVDs and CrysVCD to the DVD player, emphasizing the compatibility of their system with any existing model, not just diffusion models. Co-authors include Mouyang Cheng, Weiliang Luo, Hao Tang, Bowen Yu, Yongqiang Cheng, Weiwei Xie, Ju Li, and Heather Kulik, all affiliated with MIT's departments of Materials Science and Engineering, Chemistry, Chemical Engineering, Physics, and Nuclear Science and Engineering.
The researchers utilized AI diffusion models in conjunction with a language model, initiating the process by generating chemically valid formulas followed by the diffusion model, which generates atomic structures of crystal materials in synchronization with the material generation model. This technique significantly accelerates the screening process, allowing the generation of higher-quality materials.
Written by urgent.news from MIT News Research's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
