Using Computationally-Determined Properties for Machine Learning Prediction of Self-Diffusion Coefficients in Pure Liquids

Abstract

The ability to predict transport properties of liquids quickly and accurately will greatly improve our understanding of fluid properties both in bulk and complex mixtures, as well as in confined environments. Such information could then be used in the design of materials and processes for applications ranging from energy production and storage to manufacturing processes. As a first step, we consider the use of machine learning (ML) methods to predict the diffusion properties of pure liquids. Recent results have shown that Artificial Neural Networks (ANNs) can effectively predict the diffusion of pure compounds based on the use of experimental properties as the model inputs. In the current study, a similar ANN approach is applied to modeling diffusion of pure liquids using fluid properties obtained exclusively from molecular simulations. A diverse set of 102 pure liquids is considered, ranging from small polar molecules (e.g., water) to large nonpolar molecules (e.g., octane). Self-diffusion coefficients were obtained from classical molecular dynamics (MD) simulations. Since nearly all the molecules are organic compounds, a general set of force field parameters for organic molecules was used. The MD methods are validated by comparing physical and thermodynamic properties with experiment. Computational input features for the ANN include physical properties obtained from the MD simulations as well as molecular properties from quantum calculations of individual molecules. Fluid properties describing the local liquid structure were obtained from center of mass radial distribution functions (COM-RDFs). Feature sensitivity analysis revealed that isothermal compressibility, heat of vaporization, and the thermal expansion coefficient were the most impactful properties used as input for the ANN model to predict the MD simulated self-diffusion coefficients. The MD-based ANN successfully predicts the MD self-diffusion coefficients with only a subset (2 to 3) of the available computationally determined input features required. A separate ANN model was developed using literature experimental self-diffusion coefficients as model targets. Although this second ML model was not as successful due to a limited number of data points, a good correlation is still observed between experimental and ML predicted self-diffusion coefficients.

Predicting the Self-Diffusion Coefficient of Liquids Based on Backpropagation Artificial Neural Network: A Quantitative Structure–Property Relationship Study

“…Their liquid-phase ANN model contained 4698 data points with an R 2 of 0.9977 and an AARD of 12.5%; however, its input feature species involved critical properties such as the critical volume, pressure, and temperature of substances, and the ANN model was no longer applicable for substances whose critical properties were difficult to obtain. In addition, Allers et al 46 also developed the MD-simulated self-diffusion coefficient ANN model (MD-ANN) and the experimental self-diffusion coefficient ANN model (EXP-ANN) for predicting 102 pure liquids. Although only two to three descriptors were used as input features, the R 2 of EXP-ANN was only 0.89, and its MD-ANN, although higher than EXP-ANN, had an R 2 of 0.93.…”

“…44 Allers et al 45 developed artificial neural network (ANN) models for liquid, supercritical, and gaseous states to predict the self-diffusion coefficients of 118 pure compounds using 19 molecular descriptors, such as density, critical temperature, critical pressure, and critical volume. In another study, Allers et al 46 also established artificial neural network models to predict the self-diffusion coefficients of 102 pure liquids obtained by MD simulations and experiments. The descriptors consisted of physical properties obtained by MD simulations and molecular properties from quantum calculations, and only two to three descriptors were needed to predict the self-diffusion coefficients.…”

Prediction of Self-Diffusion in Binary Fluid Mixtures Using Artificial Neural Networks

“…Machine learning (ML) has proven to be a promising technique for the prediction of thermodynamic and transport properties. Example ML efforts for modeling of transport include the prediction of diffusion for organic compounds in air, binary gas mixtures, − organic compounds in water, − supercritical CO 2 , supercritical water mixtures, binary ionic mixtures, and mixtures of binary solvents and hydrocarbons. − Previous work from our group includes the use of artificial neural networks (ANNs) and random forest methods to predict self-diffusion coefficients of both Lennard-Jones (LJ) fluids and pure single component fluids. , We have also shown that ANNs can model diffusion of LJ fluids in pores, correct finite-size effects in MD simulations of self-diffusion and MS diffusion in binary LJ fluids, and predict diffusion using entirely computational-derived descriptors …”

“…42,43 We have also shown that ANNs can model diffusion of LJ fluids in pores, 44 correct finite-size effects in MD simulations of selfdiffusion and MS diffusion in binary LJ fluids, 45 and predict diffusion using entirely computational-derived descriptors. 46 In this paper, we develop ANN models to predict the selfdiffusion constants of each individual component in binary mixtures over the entire compositional range by developing a database from reported literature diffusion measurements. We also explore the role of molecular clustering and intermolecular interactions by incorporating self-and binary association cluster energies as input features for the ANNs along with other more common fluid properties and experimental conditions.…”

Prediction of transport property via machine learning molecular movements

“…Transport properties, which describe the movements of mass and heat, have critical roles in materials of various products, thus being widely investigated using MD and ML methods. 25,26 For example, the diffusion coefficient have been predicted for gases molecules, 11 ions 27,28 and other chemical compounds 29 to discover and design better materials for artificial membrane and batteries. Similarly, the viscosity measures the ability of lubricants to reduce friction and wear in machinery.…”