(572e) Machine Learned Disposable Force Fields for Fluid-Solid Simulations: Application to Water Transport in Carbon and Boron-Nitride Nanotubes

Simulation techniques based on accurate and efficient representations of potential energy surfaces are crucial for the understanding of adsorption at solid–liquid interfaces. Here we present a machine learning framework that enables the efficient development and validation of force fields with quantum-level accuracy with the aid of data-driven machine learning algorithms. Instead of trying to deliver a globally optimal machine learning potential, we propose to develop models applicable to specific thermodynamic state points in a simple and user-friendly process: a “disposable” potential which can be sourced with minimal effort.

Essentially, after an initial ab-initio simulation, a machine learning potential is constructed with minimum human effort through a data-driven active learning protocol. Such models can afterward be applied in exhaustive simulations to provide reliable answers for the scientific question at hand. The accuracy of this approach is benchmarked with respect to the underlying ab-initio reference and the resulting models are evaluated in detail with an automated validation protocol that includes structural and dynamical properties and the precision of the force prediction of the models.

We showcase this methodology by shedding light onto the fascinating and puzzling behaviour of water in carbon nanotubes where experimental measurements have reported ultra-fast and radius-dependent water transport which are absent in isostructural boron nitride nanotubes. Our simulations using disposable models reveal a large, radius-dependent hydrodynamic slippage on both materials with water experiencing indeed a 5 times lower friction on carbon surfaces compared to boron nitride. Analysis of the diffusion mechanisms across the two materials reveals that the fast water transport on carbon is governed by facile oxygen motion, whereas the higher friction on boron nitride arises from specific hydrogen-nitrogen interactions.