Artificial Intelligence Powers Breakthrough In Large-Scale Atom Simulations
The integration of artificial intelligence techniques has allowed scientists to achieve the most accurate simulation of objects made up of tens or millions of atoms. Due to the enormous computational power needed, previous simulations of the behavior and interactions of atoms have been limited to small molecules. There are ways to simulate larger numbers of atoms over time. However, these methods heavily rely upon approximations that do not provide detailed molecular information.
The team of Boris Kozinsky, at Harvard University, has developed Allegro. It uses artificial intelligence to simulate systems with tens or millions of atoms. Kozinsky’s team used Perlmutter – the eighth most powerful supercomputer in the world – to simulate the complex interaction of 44,000,000 atoms that make up the HIV protein shell. They also successfully simulated vital biological molecules like cellulose, haemophilia-associated protein, and a tobacco plant virus.
Kozinsky stresses that this method can accurately simulate any object based on atoms with exceptional precision and scaleability. The system can be used to solve a variety of materials science issues, such as batteries, catalysis and semiconductors.
Source:
Artificial Intelligence Powers Breakthrough in Large-Scale Atom Simulations
