Adaptive machine learning framework to accelerate ab initio molecular dynamics. Issue 16 (23rd December 2014)

• Record Type: Journal Article
• Authors: Botu, Venkatesh; Ramprasad, Rampi; Rupp, Matthias
• Is Part Of: International journal of quantum chemistry. Volume 115:Issue 16(2015:Aug. 19)
• Page Start: 1074
• Page End: 1083
• Publication Date: 2014-12-23
• Subjects: Quantum chemistry -- Periodicals; 541.28
• DOI: 10.1002/qua.24836
• Languages: English
• ISSNs: 0020-7608

Abstract

<abstract abstract-type="main"> <title> <x xml:space="preserve">Abstract</x> </title> <p>Quantum mechanics‐based <italic>ab initio</italic> molecular dynamics (MD) simulation schemes offer an accurate and direct means to monitor the time evolution of materials. Nevertheless, the expensive and repetitive energy and force computations required in such simulations lead to significant bottlenecks. Here, we lay the foundations for an accelerated <italic>ab initio</italic> MD approach integrated with a machine learning framework. The proposed algorithm learns from previously visited configurations in a continuous and adaptive manner on‐the‐fly, and predicts (with chemical accuracy) the energies and atomic forces of a new configuration at a minuscule fraction of the time taken by conventional <italic>ab initio</italic> methods. Key elements of this new accelerated <italic>ab initio</italic> MD paradigm include representations of atomic configurations by numerical fingerprints, a learning algorithm to map the fingerprints to the properties, a decision engine that guides the choice of the prediction scheme, and requisite amount of <italic>ab initio</italic> data. The performance of each aspect of the proposed scheme is critically evaluated for Al in several different chemical environments. This work has enormous implications beyond <italic>ab initio</italic> MD acceleration. It can also lead to accelerated structure and property prediction schemes, and accurate force fields. © 2014 Wiley Periodicals, Inc.</p> </abstract>