AivexaNewsSearch
AI news for builders and product teamsChecked every hour

AI helps design new materials that work in the real world

Collected Sep 30, 2026

MIT researchers have developed a framework called crystal generator with valence-constrained design, or CrysVCD, that applies key chemistry rules about electrons around atoms before the expensive materials generation step. The work was published in Nature Computational Science.

The approach addresses a longstanding problem: current AI models can generate millions of material designs but do not reliably factor in chemical stability. Screening out unstable materials consumes about 90 percent of the computational cost for creating usable materials and can take weeks or months, according to doctoral student Mouyang Cheng. Large companies with big computing budgets can afford such screening, but many small companies and research labs cannot.

CrysVCD combines a language model with diffusion models. The language model first produces chemically valid formulas; the diffusion model then generates the corresponding atomic crystal structure. Associate professor Mingda Li compared material-generating models to DVDs and CrysVCD to a DVD player, saying it can be plugged into any model, not only existing diffusion models but also future ones.

The researchers reported that CrysVCD allowed several commonly used material models to meet valence shell rules more often and achieved high lattice-dynamics stability, described as a stringent stability test, in nearly 70 percent of computational material generations. When fine-tuned on stability metrics, the approach produced crystalline materials with 68 percent mechanical stability and 85 percent metastability. The approach created more stable materials an order of magnitude more efficiently than methods that screen after generation, the researchers said.

The team also used the approach to generate material candidates with high thermal conductivity and easy polarization in an electric field. Ju Li, MIT's Carl Richard Soderberg Professor in Power Engineering, noted such materials are useful for the semiconductor industry and data center cooling.

The approach works best with solid structures with highly ordered internal arrangements and does not work with every kind of material.

The paper's authors include Mouyang Cheng, Weiliang Luo, Hao Tang, Bowen Yu, Yongqiang Cheng of Oak Ridge National Laboratory, Weiwei Xie of Michigan State University, Ju Li, and Heather Kulik. The work was supported in part by the U.S. Department of Energy, a Mathworks Engineering Fellowship, the National Science Foundation, and the U.S. Defense Threat Reduction Agency.

Read at MIT News · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

The “CrysVCD” tool developed at MIT could cut the huge amounts of time and money spent on screening out chemically unstable designs.