Managing Expectations and Imbalanced Training Data in Reactive Force Field Development: An Application to Water Adsorption on Alumina.

ReaxFF is a computationally efficient model for reactive molecular dynamics simulations that has been applied to a wide variety of chemical systems. When ReaxFF parameters are not yet available for a chemistry of interest, they must be (re)optimized, for which one defines a set of training data that the new ReaxFF parameters should reproduce. ReaxFF training sets typically contain diverse properties with different units, some of which are more abundant (by orders of magnitude) than others. To find the best parameters, one conventionally minimizes a weighted sum of squared errors over all of the data in the training set. One of the challenges in such numerical optimizations is to assign weights so that the optimized parameters represent a good compromise among all the requirements defined in the training set. This work introduces a new loss function, called Balanced Loss, and a workflow that replaces weight assignment with a more manageable procedure. The training data are divided into categories with corresponding "tolerances", i.e., acceptable root-mean-square errors for the categories, which define the expectations for the optimized ReaxFF parameters. Through the Log-Sum-Exp form of Balanced Loss, the parameter optimization is also a validation of one's expectations, providing meaningful feedback that can be used to reconfigure the tolerances if needed. The new methodology is demonstrated with a nontrivial parametrization of ReaxFF for water adsorption on alumina. This results in a new force field that reproduces both the rare and frequent properties of a validation set not used for training. We also demonstrate the robustness of the new force field with a molecular dynamics simulation of water desorption from a γ-Al2O3 slab model.

Comparisons of Molecular Structure Generation Methods Based on Fragment Assemblies and Genetic Graphs.

The use of quantitative structure-property relationships (QSPRs) helps in predicting molecular properties for several decades, while the automatic design of new molecular structures is still emerging. The choice of algorithms to generate molecules is not obvious and is related to several factors such as the desired chemical diversity (according to an initial dataset's content) and the level of construction (the use of atoms, fragments, pattern-based methods). In this paper, we address the problem of molecular structure generation by revisiting two approaches: fragment-based methods (FMs) and genetic-based methods (GMs). We define a set of indices to compare generation methods on a specific task. New indices inform about the explored data space (coverage), compare how the data space is explored (representativeness), and quantifies the ratio of molecules satisfying requirements (generation specificity) without the use of a database composed of real chemicals as a reference. These indices were employed to compare generations of molecules fulfilling the desired property criterion, evaluated by QSPR.

Inverse-QSPR for de novo Design: A Review.

The use of computer tools to solve chemistry-related problems has given rise to a large and increasing number of publications these last decades. This new field of science is now well recognized and labelled Chemoinformatics. Among all chemoinformatics techniques, the use of statistical based approaches for property predictions has been the subject of numerous research reflecting both new developments and many cases of applications. The so obtained predictive models relating a property to molecular features - descriptors - are gathered under the acronym QSPR, for Quantitative Structure Property Relationships. Apart from the obvious use of such models to predict property values for new compounds, their use to virtually synthesize new molecules - de novo design - is currently a high-interest subject. Inverse-QSPR (i-QSPR) methods have hence been developed to accelerate the discovery of new materials that meet a set of specifications. In the proposed manuscript, we review existing i-QSPR methodologies published in the open literature in a way to highlight developments, applications, improvements and limitations of each.

Simulations of Interfacial Tension of Liquid-Liquid Ternary Mixtures Using Optimized Parametrization for Coarse-Grained Models.

In this work, liquid-liquid systems are studied by means of coarse-grained Monte Carlo simulations (CG-MC) and Dissipative Particle Dynamics (DPD). A methodology is proposed to reproduce liquid-liquid equilibrium (LLE) and to provide variation of interfacial tension (IFT), as a function of the solute concentration. A key step is the parametrization method based on the use of the Flory-Huggins parameter between DPD beads to calculate solute/solvent interactions. Parameters are determined using a set of experimental compositional data of LLE, following four different approaches. These approaches are evaluated, and the results obtained are compared to analyze advantages/disadvantages of each one. These methodologies have been compared through their application on six systems: water/benzene/1,4-dioxane,water/chloroform/acetone, water/benzene/acetic acid, water/benzene/2-propanol, water/hexane/acetone, and water/hexane/2-propanol. CG-MC simulations in the Gibbs (NVT) ensemble have been used to check the validity of parametrization approaches for LLE reproduction. Then, CG-MC simulations in the osmotic (µsoluteNsolventP zzT) ensemble were carried out considering the two liquid phases with an explicit interface. This step allows one to work at the same bulk concentrations as the experimental data by imposing the precise bulk phase compositions and predicting the interface composition. Finally, DPD simulations were used to predict IFT values for different solute concentrations. Our results on variation of IFT with solute concentration in bulk phases are in good agreement with experimental data, but some deviations can be observed for systems containing hexane molecules.

Chemoinformatics at IFP Energies Nouvelles: Applications in the Fields of Energy, Transport, and Environment.

The objective of the present paper is to summarize chemoinformatics based research, and more precisely, the development of quantitative structure property relationships performed at IFP Energies nouvelles (IFPEN) during the last decade. A special focus is proposed on research activities performed in the "Thermodynamics and Molecular Simulation" department, i. e. the use of multiscale molecular simulation methods in responses to projects. Molecular simulation techniques can be envisaged to supplement dataset when experimental information lacks, thus the review includes a section dedicated to molecular simulation codes, development of intermolecular potentials, and some of their possible applications. Know-how and feedback from our experiences in terms of machine learning application for thermophysical property predictions are included in a section dealing with methodological aspects. The generic character of chemoinformatics is emphasized through applications in the fields of energy, transport, and environment, with illustrations for three IFPEN business units: "Transports", "Energy Resources", and "Processes". More precisely, the review focus on different challenges such as the prediction of properties for alternative fuels, the prediction of fuel compatibility with polymeric materials, the prediction of properties for surfactants usable in chemical enhanced oil recovery, and the prediction of guest-host interactions between gases and nanoporous materials in the frame of carbon dioxide capture or gas separation activities.

Prediction of Alternative Gasoline Sorption in a Semicrystalline Poly(ethylene).

In this work, we first report the acquisition of new experimental data and then the development of quantitative structure-property relationships on the basis of sorption values for neat compounds and up to quinary mixtures of some hydrocarbons, alcohols, and ethers, in a semicrystalline poly(ethylene). Two machine learning methods (i.e., genetic function approximation and support vector machines) and two families of descriptors (i.e., functional group counts and substructural molecular fragments) were used to derive predictive models. Models were then used to predict sorption variations when increasing the number of carbon atoms in a series of hydrocarbons and for n-alkan-1-ols. In addition to the performed internal/external validations, the model was further tested for surrogate gasolines containing ca. 300 compounds, and predicted sorption values were in excellent agreement with experimental data (R(2) = 0.940).

Surface photografting of acrylic acid on poly(dimethylsiloxane). Experimental and dissipative particle dynamics studies.

This work includes both experimental and theoretical studies of the wetting property changes of water on a surface of poly(dimethylsiloxane) (PDMS) modified with different amounts of acrylic acid (AA). The default surface properties of PDMS were changed from hydrophobic to hydrophilic behavior which was characterized with contact angle measurements by two approaches: (i) experimental tests of samples subjected to a photografting polymerization procedure to obtain a functionalized surface and (ii) DPD (dissipative particle dynamics) simulations which also involve the calculation of sets of repulsive parameters determined following two methods: the use of the "Blends" module in the Materials Studio software and the calculation of cohesive energy density with molecular simulations. Changes of contact angle values observed from both experimental and numerical simulation results provide qualitative and quantitative information on the wetting behavior of photografted surfaces.

Extension of a charged anisotropic united atoms model to polycyclic aromatic compounds.

A new potential model for polycyclic aromatic hydrocarbons has been developed on the basis of a charged anisotropic united atoms (AUA) potential with six AUA force centers and three electrostatic point charges per aromatic ring. Using quantum mechanical calculations, quadrupolar moments of several aromatic molecules were computed and a correlation has been observed that links the magnitude of the point charges with respect to the number of aromatic rings. The Lennard-Jones parameters of quaternary carbon atoms bridging two aromatic rings have been optimized with the minimization of a dimensionless error criterion incorporating various thermodynamic data of naphthalene. The new potential model, called ch-AUA, was then evaluated on its abilities to predict thermodynamic and transport properties for a series of polycyclic aromatic compounds in a wide range of temperatures. Although the relative errors with respect to the experimental density, vaporization enthalpy, and vapor pressure data are similar to those computed with the noncharged AUA potential, the new ch-AUA potential noticeably improves the prediction of the shear viscosities of polycyclic aromatic compounds. Comparisons between experimental viscosities of 1-methylnaphthalene at different pressures and those computed using the new ch-AUA and the noncharged AUA potentials show that the new potential improves the prediction of viscosities at high pressures.

Molecular dynamics study of hydrated imogolite. 2. Structure and dynamics of confined water.

The behaviour of water confined in an imogolite nanotube was studied by means of molecular dynamics simulations. The results of the study show an important difference between the interaction of water molecules with the internal and external surfaces of the nanotube. The analysis of the density profiles of confined molecules, of their spatial organisation, of the size of molecular clusters, of the lifetime of H-bonds in the system and of dynamical characteristics of molecules permits us to qualify the external imogolite surface as hydrophobic, whereas the internal surface reveals a hydrophilic character.