6 Oct 2017
Autonomously driven cars may be the most significant breakthrough in the field of artificial intelligence (AI) so far, but less glamorous applications in search engines and spam filters illustrate the versatility of innovations AI has already enabled. Now, researchers have found a way to use AI to greatly accelerate chemical analysis.
Infrared spectroscopy, one of the most valuable experimental methods for examining molecules, is used to investigate chemical ‘fingerprints’ of substances and materials to provide information on their composition and properties. Due to the complexity of these spectra, computer-aided simulation is indispensable – but while quantum chemical calculations enable extremely precise prediction of infrared spectra, in practice, these are limited to relatively small chemical systems, due to the high computational effort required.
An international group of researchers led by Philipp Marquetand from the Faculty of Chemistry at the University of Vienna, Austria, has now found a way to accelerate these simulations using artificial neural networks – mathematical models of the human brain.
These networks can learn the complex quantum mechanical relationships necessary for the modelling of infrared spectra by using only a few examples. This allows scientists to carry out simulations that would take modern supercomputers thousands of years to process within a few minutes, without sacrificing reliability. ‘We can now finally simulate chemical problems that could not be overcome with the simulation techniques used up to now,’ said Michael Gastegger, the first author of the study.
The researchers are confident that their method of spectra prediction will become widely used in the analysis of experimental infrared spectra.
- Public Evening Lecture: Are Machines Better than Humans?
- Materials Chemistry Group
- Fine Chemicals Group
- Young Chemists Panel