University of Luxembourg, University of Warwick and TU Berlin develop a new deep machine learning algorithm

An interdisciplinary team of chemists, physicists, and computer scientists from the University of Luxembourg, the University of Warwick and the Technical University of Berlin have developed a deep machine learning algorithm that can predict the quantum states of molecules, so-called wave functions, which determine all properties of molecules. 

Artificial intelligence and machine learning algorithms are routinely used to predict our purchasing behaviour and to recognise our faces or handwriting. In scientific research, artificial intelligence is establishing itself as a crucial tool for scientific discovery. In chemistry, AI has become instrumental in predicting the outcomes of experiments or simulations of quantum systems. Artificial intelligence achieves this by learning to solve fundamental equations of quantum mechanics.

Solving these equations in the conventional way requires massive high-performance computing resources (months of computing time) which is typically the bottleneck to the computational design of new purpose-built molecules for medical and industrial applications. The newly developed AI algorithm can supply accurate predictions within seconds on a laptop or mobile phone.

Dr. Reinhard Maurer from the Department of Chemistry at the University of Warwick comments: “This has been a joint three year effort, which required computer science know-how to develop an artificial intelligence algorithm flexible enough to capture the shape and behaviour of wave functions, but also chemistry and physics know-how to process and represent quantum chemical data in a form that is manageable for the algorithm.”

The teams have been brought together during an interdisciplinary 3-month fellowship program at the Institute for Pure and Applied Mathematics (IPAM) at the University of California (UCLA) on the subject of machine learning in quantum physics.

Prof. Klaus Robert-Muller from the Institute of Software Engineering and Theoretical Computer Science at the Technical University of Berlin adds: “This interdisciplinary work is an important progress as it shows that, AI methods can efficiently perform the most difficult aspects of quantum molecular simulations. Within the next few years, AI methods will establish themselves as essential part of the discovery process in computational chemistry and molecular physics.”

Prof. Alexandre Tkatchenko from the Department of Physics and Materials Research at the University of Luxembourg concludes: “This work enables a new level of compound design where both electronic and structural properties of a molecule can be tuned simultaneously to achieve desired application criteria.”

The article "Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions" was published on 15 November 2019 in Nature Communications.

This communication was first published 18 November 2019 by the University of Luxembourg.