Many technologically important phenomena coupling complex chemistry with high-gradient elastic fields, such as fracture by stress corrosion or thermal activation in covalent materials or embrittled metals, are beyond the reach of First-Principles Molecular Dynamics (FPMD) techniques. This is because the required model system sizes are too large, and/or the required simulation times are too long, for these traditional high-accuracy approaches. In most (but not all) situations, using classical MD is not a viable alternative, as suitably general and accurate (e.g., “reactive”) force fields are not available, nor is it clear how to produce fitting databases a priori guaranteed to contain the information necessary to describe all the chemical processes which might be encountered along the dynamics. This is notably true also for any force fields based on Machine Learning e.g., based on neural networks or Gaussian Processes (GP). Finally, “QM/MM” techniques combining quantum and classical zones in a single calculation also pose significant difficulties, especially when the target processes involve sustained mass transport into and out of the (e.g., fast moving) QM zone. I will argue that this situation requires novel MD techniques capable of either locating and using QM-accurate information pre-stored in massive databases or promptly generating this information if no such information is available. In practical realisations, these could be DFT-cured configurations databases coupled with online-learning force fields, where the computational workload of a MD run is minimised by allowing QM subroutine calls only when “chemically novel” configurations are encountered along the system’s trajectory. I will describe the implementation one such “Learn On the Fly” technique, effectively unifying First-Principles Molecular Dynamics and Machine Learning into a single, information efficient simulation scheme capable of learning/predicting atomic forces through Bayesian inference. Recently developed “covariant” kernels specifically designed for direct learning of atomic forces by GP regression are at the core of this technique, due to their symmetry properties and optimal energy conservation. Interestingly, QM-zone partitioning approaches followed by execution using any of the existing O(N3) QM engines is predicted to be a better option than using O(N) QM methods when dealing with large QM zones in learn on the fly calculations running on high-end parallel platforms.

Inference-boosted first-principles molecular dynamics for chemo-mechanical modelling

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.