Next-Level Screening: Predicting Material Properties


A new paper from the Clean Energy Project team reveals that they can now use multi-layer artificial neural networks to predict the electrical properties of novel molecules without actually simulating the entire molecule. This advance was made possible by the enormous number of simulations done for the Clean Energy Project, and promises to enable screening of many more molecules than the team was able to address in their previous work.



Paper Title:
"Learning from the Harvard Clean Energy Project: The Use of Neural Networks to Accelerate Materials Discovery"

Published in journal:
Advanced Functional Materials

Authors:
Edward O. Pyzer-Knapp, Kewei Li and Alan Aspuru-Guzik

Layman Abstract:

In this paper Edward, from Harvard's Clean Energy Project, shows how techniques from the field of machine learning can be used to speed up materials screening. By using a special class of neural networks, known as multi-layer perceptrons, he is able to predict the properties of a molecule to a high degree of accuracy before any calculations are performed. They show how using this method one can eliminate almost 99% of a screening library or molecules without having to calculate it. By eliminating the molecules unlikely to be useful, they greatly increasing the range of molecules that can be considered by the Clean Energy Project.

Technical Abstract:

Here, the employment of multilayer perceptrons, a type of artificial neural network, is proposed as part of a computational funneling procedure for high-throughput organic materials design. Through the use of state of the art algorithms and a large amount of data extracted from the Harvard Clean Energy Project, it is demonstrated that these methods allow a great reduction in the fraction of the screening library that is actually calculated. Neural networks can reproduce the results of quantum-chemical calculations with a large level of accuracy. The proposed approach allows to carry out large-scale molecular screening projects with less computational time. This, in turn, allows for the exploration of increasingly large and diverse libraries.

Access to Paper:

To view the paper, please click here.

Related Articles