Sign In

Communications of the ACM

ACM TechNews

Scientists ­se Machine Learning to Translate 'hidden' Information That Reveals Chemistry in Action

A sketch of the new method that enables fast determination of the three-dimensional structure of nanocatalysts.

A team from Brookhaven National Laboratory used neural networks and machine learning to teach computers to decode previously inaccessible information from x-ray data.

Credit: Brookhaven National Laboratory

Researchers at Brookhaven National Laboratory and Stony Brook University say they have used neural networks and machine learning to teach computers to decipher previously inaccessible information from x-ray data and apply it to decoding three-dimensional nanoscale structures.

They say their on-the-fly technique could help improve the ability of catalysts to quickly drive reactions toward desired products.

The team worked out a way to analyze an aspect of the x-ray absorption spectrum associated with low-energy waves that are less vulnerable to heat and disorder.

The key to decoding the spectrum was developed by training computers to find the connections between hidden features of the spectrum and the structural details of the catalysts. The researchers employed theoretical modeling to generate simulated spectra of several hundred thousand model structures, using them to teach the computer to identify the spectral features and their correlation with the structure.

Brookhaven's Anatoly Frenkel says a neural network was built to handle spectrum-to-structure conversion.

From Brookhaven National Laboratory
View Full Article


Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account