The most accurate simulation of objects made from tens of millions of atoms has been run on one of the world’s best supercomputers with the help of artificial intelligence.
Existing simulations that describe in detail how atoms behave, interact and evolve are limited to small molecules, due to the computational power required. There are techniques to simulate a much larger number of atoms over time, but these rely on approximations and are not precise enough to extract many detailed features of the molecule in question.
NOW, Boris Kozinsky at Harvard University and his colleagues have developed a tool, called Allegro, that can accurately simulate systems with tens of millions of atoms using artificial intelligence.
Kozinsky and his team used the world’s 8th most powerful supercomputer, Perlmutter, to simulate the 44 million atoms involved in the protein shell of HIV. They also simulated other common biological molecules such as cellulose, a protein missing in people with hemophilia and a common tobacco plant virus.
“Anything that is essentially composed of atoms, you can simulate with these methods with extremely high precision, and now also on a large scale,” says Kozinsky. “It’s a demonstration, but by no means limited to this area.” The system could also be used for many problems in materials science, such as the study of batteries, catalysis and semiconductors, he says.
To be able to simulate such a large number of particles, the researchers used a kind of AI called a neural network to calculate the interactions between atoms that were symmetrical at all angles, a principle called equivariance.
“When you develop networks that very fundamentally include these symmetries…you get these big improvements in accuracy and other properties that we’re interested in, such as the stability of the simulations or the speed at which the machine learning model learns as that you teach it with more data,” says a team member Albert Musaelianalso at Harvard.
“It’s a tour de force in programming and demonstrating that these machine learning potentials are now scalable,” says Gabor Csanyi at the University of Cambridge.
However, simulating biological molecules like these is more a demonstration that the tool works for large systems than a handy boost for researchers, since biochemists already have sufficiently precise tools that can be run much more quickly, he said. Where this could be useful is for materials with lots of atoms that experience extreme shocks and forces on very short timescales, such as in planetary cores, Csányi says.
Topics: