MIT researchers create CrysVCD AI to speed materials discovery
MIT researchers have developed an AI tool called CrysVCD that helps scientists discover new materials much more efficiently.
Instead of wasting time on unstable designs like older models, CrysVCD uses chemical constraints early to reduce unstable candidates and improve stability, saving tons of time and money.
CrysVCD yields 70% stable candidates
Typical diffusion-based material generation can take about 1,000 steps to create one material, and in testing, about 70% of its suggested materials passed tough stability checks.
The team used the framework to search for materials with high thermal conductivity and strong dielectric properties, key for better cooling in data centers and the semiconductor industry.
Plus, because it's efficient, smaller labs can now join the discovery game too.