Flexible Polarizable Water Model Parameterized via Gaussian Process Regression

Water is one of the most common components in molecular dynamics (MD) simulations. Using Gaussian process regression for predicting the properties of a water model without the need of running a simulation whenever the parameters are changed, we obtained a flexible polarizable water model, named SWM4/Fw, that is able to reproduce many reference water properties. The added flexibility is critical for modeling chemical reactions in which chemical bonds can be stretched or even broken and for directly calculating vibrational spectra. In addition to being one of the few water models that are both flexible and polarizable, SWM4/Fw is also efficient thanks to the extended Lagrangian scheme with Drude oscillators. The overall accuracy is on par with or better than the related SWM4-NDP model.

Orcinol and resorcinol induce local ordering of water molecules near the liquid-vapor interface

Resorcinol and orcinol are simple members of the family of phenolic compounds present in particulate matter in the atmosphere; they are amphiphilic in nature and thus surface active in aqueous solution. Here, we used X-ray photoelectron spectroscopy to probe the concentration of resorcinol (benzene-1,3-diol) and orcinol (5-methylbenzene-1,3-diol) at the liquid-vapor interface of aqueous solutions. Qualitatively consistent surface propensity and preferential orientation was obtained by molecular dynamics simulations. Auger electron yield near-edge X-ray absorption fine structure (NEXAFS) spectroscopy was used to probe the hydrogen bonding (HB) structure, indicating that the local structure of water molecules near the surface of the resorcinol and orcinol solutions tends towards a larger fraction of tetrahedrally coordinated molecules than observed at the liquid-vapor interface of pure water. The order parameter obtained from the molecular dynamics simulations confirm these observations. This effect is being discussed in terms of the formation of an ordered structure of these molecules at the surface leading to patterns of hydrated OH groups with distances among them that are relatively close to those in ice. These results suggest that the self-assembly of phenolic species at the aqueous solution-air interface could induce freezing similar to the case of fatty alcohol monolayers and, thus, be of relevance for ice nucleation in the atmosphere. We also attempted at looking at the changes of the O 1b₁, 3a₂ and 1b₂ molecular orbitals of liquid water, which are known to be sensitive to the HB structure as well, in response to the presence of resorcinol and orcinol. However, these changes remained negligible within uncertainty for both experimentally obtained valence spectra and theoretically calculated density of states.

Salt Enrichment and Dynamics in the Interface of Supercooled Aqueous Droplets

The interconversion reaction of NaCl between the contact-ion pair (CIP) and the solvent-separated ion pair (SSIP) as well as the free-ion state in cold droplets has not yet been investigated. We report direct computational evidence that the lower is the temperature, the closer to the surface the ion interconversion reaction takes place. In supercooled droplets the enrichment of the subsurface in salt becomes more evident. The stability of the SSIP relative to the CIP increases as the ion-pairing is transferred toward the droplet's outer layers. In the free-ion state, where the ions diffuse independently in the solution, the number density of Cl^- shows a broad maximum in the interior in addition to the well-known maximum in the surface. In the study of the reaction dynamics, we find a weak coupling between the interionic NaCl distance reaction coordinate and the solvent degrees of freedom, which contrasts with the diffusive crossing of the free energy barrier found in bulk solution modeling. The H₂O self-diffusion coefficient is found to be at least an order of magnitude larger than that in the bulk solution. We propose to exploit the enhanced surface ion concentration at low temperature to eliminate salts from droplets in native mass spectrometry ionization methods.

Low Density Interior in Supercooled Aqueous Nanodroplets Expels Ions to the Subsurface

The interaction between water and ions within droplets plays a key role in the chemical reactivity of atmospheric and man-made aerosols. Here we report direct computational evidence that in supercooled aqueous nanodroplets a lower density core of tetrahedrally coordinated water expels the cosmotropic ions to the denser and more disordered subsurface. In contrast, at room temperature, depending on the nature of the ion, the radial distribution in the droplet core is nearly uniform or elevated toward the center. We analyze the spatial distribution of a single ion in terms of a reference electrostatic model. The energy of the system in the analytical model is expressed as the sum of the electrostatic and surface energy of a deformable droplet. The model predicts that the ion is subject to a harmonic potential centered at the droplet's center of mass. We name this effect "electrostatic confinement". The model's predictions are consistent with the simulation findings for a single ion at room temperature but not at supercooling. We anticipate this study to be the starting point for investigating the structure of supercooled (electro)sprayed droplets that are used to preserve the conformations of macromolecules originating from the bulk solution.

Machine learning for reparameterization of four-site water models: TIP4P-BG and TIP4P-BGT

Parameterizing an effective water model is a challenging issue because of the difficulty in maintaining a comprehensive balance among the diverse physical properties of water with a limited number of parameters. The advancement in machine learning provides a promising path to search for a reliable set of parameters. Based on the TIP4P water model, hence, about 6000 molecular dynamics (MD) simulations for pure water at 1 atm and in the range of 273-373 K are conducted here as the training data. The back-propagation (BP) neural network is then utilized to construct an efficient mapping between the model parameters and four crucial physical properties of water, including the density, vaporization enthalpy, self-diffusion coefficient and viscosity. Without additional time-consuming MD simulations, this mapping operation could result in sufficient and accurate data for high-population genetic algorithm (GA) to optimize the model parameters as much as possible. Based on the proposed parameterizing strategy, TIP4P-BG (a conventional four-site water model) and TIP4P-BGT (an advanced model with temperature-dependent parameters) are established. Both the water models exhibit excellent performance with a reasonable balance among the four crucial physical properties. The relevant mean absolute percentage errors are 3.53% and 3.08%, respectively. Further calculations on the temperature of maximum density, isothermal compressibility, thermal expansion coefficient, radial distribution function and surface tension are also performed and the resulting values are in good agreement with the experimental values. Through this water modeling example, the potential of the proposed data-driven machine learning procedure has been demonstrated for parameterizing a MD-based material model.

Rapid Water Transport through Organic Layers on Ice

Processes involving atmospheric aerosol and cloud particles are affected by condensation of organic compounds that are omnipresent in the atmosphere. On ice particles, organic compounds with hydrophilic functional groups form hydrogen bonds with the ice and orient their hydrophobic groups away from the surface. The organic layer has been expected to constitute a barrier to gas uptake, but recent experimental studies suggest that the accommodation of water molecules on ice is only weakly affected by condensed short-chain alcohol layers. Here, we employ molecular dynamics simulations to study the water interactions with n-butanol covered ice at 200 K and show that the small effect of the condensed layer is due to efficient diffusion of water molecules along the surface plane while seeking appropriate sites to penetrate, followed by penetration driven by the combined attractive forces from butanol OH groups and water molecules within the ice. The water molecules that penetrate through the n-butanol layer become strongly bonded by approximately three hydrogen bonds at the butanol-ice interface. The obtained accommodation coefficient (0.81 ± 0.03) is in excellent agreement with results from previous environmental molecular beam experiments, leading to a picture where an adsorbed n-butanol layer does not alter the apparent accommodation coefficient but dramatically changes the detailed molecular dynamics and kinetics.

Mechanism of anisotropic surface self-diffusivity at the prismatic ice–vapor interface

Models of mesoscopic ice roughening require improved understanding of surface attachment kinetics.

Structure, dynamics and thermodynamics of single-file water under confinement: effects of polarizability of water molecules

Various structural, dynamic and thermodynamic properties of water molecules confined in single-wall carbon nanotubes are investigated using both polarizable and non-polarizable water models.

Polarizable six-point water models from computational and empirical optimization

Tröster et al. (J. Phys. Chem B 2013, 117, 9486-9500) recently suggested a mixed computational and empirical approach to the optimization of polarizable molecular mechanics (PMM) water models. In the empirical part the parameters of Buckingham potentials are optimized by PMM molecular dynamics (MD) simulations. The computational part applies hybrid calculations, which combine the quantum mechanical description of a H2O molecule by density functional theory (DFT) with a PMM model of its liquid phase environment generated by MD. While the static dipole moments and polarizabilities of the PMM water models are fixed at the experimental gas phase values, the DFT/PMM calculations are employed to optimize the remaining electrostatic properties. These properties cover the width of a Gaussian inducible dipole positioned at the oxygen and the locations of massless negative charge points within the molecule (the positive charges are attached to the hydrogens). The authors considered the cases of one and two negative charges rendering the PMM four- and five-point models TL4P and TL5P. Here we extend their approach to three negative charges, thus suggesting the PMM six-point model TL6P. As compared to the predecessors and to other PMM models, which also exhibit partial charges at fixed positions, TL6P turned out to predict all studied properties of liquid water at p0 = 1 bar and T0 = 300 K with a remarkable accuracy. These properties cover, for instance, the diffusion constant, viscosity, isobaric heat capacity, isothermal compressibility, dielectric constant, density, and the isobaric thermal expansion coefficient. This success concurrently provides a microscopic physical explanation of corresponding shortcomings of previous models. It uniquely assigns the failures of previous models to substantial inaccuracies in the description of the higher electrostatic multipole moments of liquid phase water molecules. Resulting favorable properties concerning the transferability to other temperatures and conditions like the melting of ice are also discussed.

Comparison of select polarizable and non-polarizable water models in predicting solvation dynamics of water confined between MgO slabs

We present a molecular dynamics simulation study in which we compare and contrast the performance of a polarizable shell water potential model and non-polarizable water force field-extended simple point charge (SPC/EF) model in predicting the solvation dynamics of confined water molecules sandwiched between MgO(100) slabs. Structural features based on radial distribution functions, atomic density profiles, adsorption patterns, orientational ordering and dynamical correlations such as diffusional characteristics, hydrogen bonding lifetimes and residence probabilities are used as metrics for comparison. The simulations yield significant ordering of water molecules in the two layers adjacent to the oxide interface and the extent of ordering decreases with increasing distance from the oxide-water interface. These results elucidate that the dependence of local ordering and solvation dynamics on the molecular geometry and charge distribution, observed for typical three- and four-site water models, is generally lost for confined water if polarization is explicitly included. While the interfacial water structure predicted by the polarizable and non-polarizable models are similar, the confinement and interface proximity effects on the solvation dynamics are seen to be more pronounced for polarizable water models in comparison to non-polarizable ones. The study also shows that the polarizable water model over predicts the orientational order and under predicts the transport properties of confined water. In addition, analysis of the orientational preferences and hydrogen bonding characteristics of water near oxide interfaces suggests a higher degree of tetrahedral disorder in the polarizable shell compared to the non-polarizable SPC/E flexible model. The origin of the differences in solvation behavior of confined water between oxide slabs is analyzed based on the energetic contributions of the dispersive and electrostatic terms in the two force fields. Our findings suggest some new considerations regarding the role of polarization terms in predicting confinement and interface proximity effects that may guide future development of reliable polarizable water models for confined liquids.

Matching of additive and polarizable force fields for multiscale condensed phase simulations

Inclusion of electronic polarization effects is one of the key aspects in which the accuracy of current biomolecular force fields may be improved. The principal drawback of such approaches is the computational cost, which typically ranges from 3 - 10 times that of the equivalent additive model, and may be greater for more sophisticated treatments of polarization or other many-body effects. Here, we present a multiscale approach which may be used to enhance the sampling in simulations with polarizable models, by using the additive model as a tool to explore configuration space. We use a method based on information theory to determine the charges for an additive model that has optimal overlap with the polarizable one, and we demonstrate the feasibility of enhancing sampling via a hybrid replica exchange scheme for several model systems. An additional advantage is that, in the process, we obtain a systematic method for deriving charges for an additive model that will be the natural complement to its polarizable parent. The additive charges are found by an effective coarse-graining of the polarizable force field, rather than by ad hoc procedures.