Significance: Understanding complex materials, in particular those with solid–liquid interfaces, such as water on surfaces or under confinement, is a key challenge for technological and scientific progress. Although established simulation approaches have been able to provide important atomistic insight, ab initio techniques struggle with the required time and length scales, while force field methods can often be limited in terms of their accuracy. Here we show how these limitations can be overcome in a simple and automated machine learning procedure to provide accurate models of interactions at the ab initio level, as illustrated for a variety of complex aqueous systems. These developments open up the prospect of the straightforward exploration of many technologically relevant systems by molecular simulations.

We report classical atomistic molecular dynamics simulations of four structurally diverse model asphaltenes, a model resin, and their respective mixtures in toluene or heptane under ambient conditions. Relatively large systems (∼50 000 atoms) and long time scales (>80 ns) are explored. Wherever possible, comparisons are made to available experimental observations asserting the validity of the models. When the asphaltenes are dissolved in toluene, a continuous distribution of cluster sizes is observed with average aggregation number ranging between 3.6 and 5.6, monomers and dimers being the predominant species. As expected for mixtures in heptane, the asphaltene molecules tend to aggregate to form a segregated phase. There is no evidence of the distinct formation of nanoaggregates, and the distribution of clusters is found to be continuous in character. Analysis of the shape of the clusters of asphaltenes suggests that they are generally spherical in character, with the archipelago models favoring longer prolate structures and the continental model tending toward oblate structures. The aggregates are seen to be diffuse in nature, containing at least 50% solvent on average, being denser in heptane than in toluene. Mixtures of asphaltenes with different architecture are found to have cluster properties that are intermediate between those of the individual components. The presence of resins in the mixture does not appear to alter the shape of the asphaltene aggregates or their size or density when toluene is the solvent; on the other hand, the resins lead to an increase in the density of the resulting aggregates in heptane. Quantification of these observations is made from histograms of the cluster distributions, potential of mean force calculations, and an analysis of the shape factors. We illustrate how the time scales for complete aggregation of molecules in heptane are longer than the longest of the simulations reported in the open literature and as an example report a long simulation (0.5 μs) that fails to reach an equilibrium state, suggesting that acceleration techniques (e.g., using coarse-grained models) are needed to appropriately explore these phenomena.

A description of fluid systems with molecular-based algebraic equations of state (EoSs) and by direct molecular simulation is common practice in chemical engineering and the physical sciences, but the two approaches are rarely closely coupled. The key for an integrated representation is through a well-defined force field and Hamiltonian at the molecular level. In developing coarse-grained intermolecular potential functions for the fluid state, one typically starts with a detailed, bottom-up quantum-mechanical or atomic-level description and then integrates out the unwanted degrees of freedom using a variety of techniques; an iterative heuristic simulation procedure is then used to refine the parameters of the model. By contrast, with a top-down technique, one can use an accurate EoS to link the macroscopic properties of the fluid and the force-field parameters. We discuss the latest developments in a top-down representation of fluids, with a particular focus on a group-contribution formulation of the statistical associating fluid theory (SAFT-γ). The accurate SAFT-γ EoS is used to estimate the parameters of the Mie force field, which can then be used with confidence in direct molecular simulations to obtain thermodynamic, structural, interfacial, and dynamical properties that are otherwise inaccessible from the EoS. This is exemplified for several prototypical fluids and mixtures, including carbon dioxide, hydrocarbons, perfluorohydrocarbons, and aqueous surfactants.

We present a corresponding states correlation based on the description of fluid phase properties by means of an Mie intermolecular potential applied to tangentially bonded spheres. The macroscopic properties of this model fluid are very accurately represented by a recently proposed version of the Statistical Associating Fluid Theory (the SAFT-γ version). The Mie potential can be expressed in a conformal manner in terms of three parameters that relate to a length scale, σ, an energy scale, ε, and the range or functional form of the potential, λ, while the nonsphericity or elongation of a molecule can be appropriately described by the chain length, m. For a given chain length, correlations are given to scale the SAFT equation of state in terms of three experimental parameters: the acentric factor, the critical temperature, and the saturated liquid density at a reduced temperature of 0.7. The molecular nature of the equation of state is exploited to make a direct link between the macroscopic thermodynamic parameters used to characterize the equation of state and the parameters of the underlying Mie potential. This direct link between macroscopic properties and molecular parameters is the basis to propose a top-down method to parametrize force fields that can be used in the coarse grained molecular modeling (Monte Carlo or molecular dynamics) of fluids. The resulting correlation is of quantitative accuracy and examples of the prediction of simulations of vapor–liquid equilibria and interfacial tensions are given. In essence, we present a recipe that allows one to obtain intermolecular potentials for use in the molecular simulation of simple and chain fluids from macroscopic experimentally determined constants, namely the acentric factor, the critical temperature, and a value of the liquid density at a reduced temperature of 0.7.

Experimental measurements have reported ultrafast and radius-dependent water transport in carbon nanotubes which are absent in boron nitride nanotubes. Despite considerable effort, the origin of this contrasting (and fascinating) behavior is not understood. Here, with the aid of machine learning-based molecular dynamics simulations that deliver first-principles accuracy, we investigate water transport in single-wall carbon and boron nitride nanotubes. Our simulations 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. This work not only delivers a clear reference of quantum mechanical accuracy for water flow in single-wall nanotubes but also provides detailed mechanistic insight into its radius and material dependence for future technological application.