A new method is presented to identify the truly interfacial molecules at fluid/fluid interfaces seen at molecular resolution, a situation that regularly occurs in computer simulations. In the new method, the surface is scanned by moving a probe sphere of a given radius along a large set of test lines that are perpendicular to the plane of the interface. The molecules that are hit by the probe spheres are regarded as interfacial ones, and the position of the test spheres when they are in contact with the interfacial molecules give an estimate of the surface. The dependence of the method on various parameters, in particular, on the size of the probe sphere is discussed in detail. Based on the list of molecules identified as truly interfacial ones, two measures of the molecular scale roughness of the surface are proposed. The bivariate distribution of the lateral and normal distances of two points of the interface provides a full description of the molecular scale morphology of the surface in a statistical sense. For practical purposes two parameters related to the dependence of the average normal distance of two surface points on their lateral distance can be used. These two parameters correspond to the frequency and amplitude of the surface roughness, respectively. The new method is applied for the analysis of the molecular level structure of the liquid-vapor interface of water. As an immediate result of the application of the new method it is shown that the orientational preferences of the interfacial water molecules depend only on the local curvature of the interface, and hence the molecules located at wells of concave curvature of the rippled surface prefer the same orientations as waters located at the surface of small apolar solutes. The vast majority of the truly interfacial molecules are found to form a strongly percolating two-dimensional hydrogen bonded network at the surface, whereas no percolation is observed within the second molecular layer beyond the surface.

The preferential orientation of the water molecules near the water/1,2-dichloroethane interface is analyzed in detail at different distances from the interface on the basis of a grand canonical ensemble Monte Carlo simulation. The orientation of the individual water molecules is described by the angular polar coordinates of the interface normal vector in a local coordinate frame fixed to the particular water molecule, and the bivariate joint distribution of the two polar angles is calculated. It is found that water molecules have two distinct orientational preferences, and these two preferences exist simultaneously among the water molecules penetrating farthest into the organic phase. In the first preferred orientation the plane of the molecule is parallel to the interface, whereas in the second the molecular plane is aligned perpendicularly to the interface and the molecular dipole vector declines from the plane parallel to the interface by about 30° pointing toward the organic phase. The first of the two preferred orientations is found to be present in the entire interfacial region and also, to a smaller extent, in the subsurface water layer adjacent to the interface. The second orientational preference is only present among the water molecules penetrating farthest into the organic phase. The two orientations correspond to the alignment of a hydrogen bonded pair of water molecules, in which the molecule located toward the aqueous phase has the first, whereas the one on the organic side the second of the two preferred orientations. The obtained picture is in a clear contrast with the findings of previous studies, in which the orientation of the water molecules was described by monovariate distributions of the alignment of one or more selected molecule-fixed vectors. In order to understand the origin of the difference between the present results and earlier findings we also calculate the monovariate distributions of the direction of three of such molecular vectors, i.e., the dipole vector of the water molecule, the vector joining the two H atoms, and the vector perpendicular to the molecular plane. The comparison of the obtained monovariate distributions with the bivariate joint distribution of the two polar angles reveals that the averaging of the bivariate distribution over any of its two angles completely obscures the dual orientational preference. The present study clearly points out the importance of choosing appropriate statistical distributions in the analysis of simulation results and demonstrates the pitfalls of averaging over too many variables.

No abstract

The adsorption isotherm of methanol on ice at 200 K has been determined both experimentally and by using the Grand Canonical Monte Carlo computer simulation method. The experimental and simulated isotherms agree well with each other; their deviations can be explained by a small (about 5 K) temperature shift in the simulation data and, possibly, by the non-ideality of the ice surface in the experimental situation. The analysis of the results has revealed that the saturated adsorption layer is monomolecular. At low surface coverage, the adsorption is driven by the methanol-ice interaction; however, at full coverage, methanol-methanol interactions become equally important. Under these conditions, about half of the adsorbed methanol molecules have one hydrogen-bonded water neighbor, and the other half have two hydrogen-bonded water neighbors. The vast majority of the methanols have a hydrogen-bonded methanol neighbor, as well.

Substantial progress in our understanding of interfacial structure and dynamics has stemmed from the recent development of algorithms that allow for an intrinsic analysis of fluid interfaces. These work by identifying the instantaneous location of the interface, at the atomic level, for each molecular configuration and then computing properties relative to this location. Such a procedure eliminates the broadening of the interface caused by capillary waves and reveals the underlying features of the system. However, a precise definition of which molecules actually belong to the interfacial layer is difficult to achieve in practice. Furthermore, it is not known if the different intrinsic analysis methods are consistent with each other and yield similar results for the interfacial properties. In this paper, we carry out a systematic and detailed comparison of the available methods for intrinsic analysis of fluid interfaces, based on a molecular dynamics simulation of the interface between liquid water and carbon tetrachloride. We critically assess the advantages and shortcomings of each method, based on reliability, robustness and speed of computation, and establish consistent criteria for determining which molecules belong to the surface layer. We believe this will significantly contribute to make intrinsic analysis methods widely and routinely applicable to interfacial systems.

Molecular dynamics simulations of the vapor-liquid interface of water-methanol mixtures of five different compositions were performed on the canonical (N,V,T) ensemble at 298 K. In addition, the vapor-liquid interface of the two neat systems was simulated, as well. The obtained configurations were analyzed by means of the novel identification of truly interfacial molecules method, which provides a full list of the molecules that are right at the surface (i.e., at the boundary of the two phases). The molecular level roughness of the surface, the adsorption of the methanol molecules at the surface layer, the orientation of the surface molecules, the residence time of the molecules at the surface layer, as well as the surface aggregation of the molecules were analyzed in detail. Both the frequency and the amplitude of the surface roughness were found to become larger with an increasing methanol content. This effect was found to be stronger for the amplitude, which falls in the range of 2-4 A, depending on the composition of the system. Methanol was found to be adsorbed at the surface layer, being preferentially at the humps of the molecularly rough surface. Surface methanol prefers to orient in such a way that the O-CH(3) bond remains perpendicular to the macroscopic plane of the surface, pointing the methyl group to the vapor phase. The main orientational preference of the water molecules is to lie parallel to the surface. Methanol was found to remain considerably longer at the surface layer of the mixed systems than water. Thus, contrary to the fact that the residence times of the two molecules were found to be rather similar to each other at the surface of their neat liquids, the residence time of the methanol molecules was an order of magnitude larger than that of water molecules at the surface of their mixtures. A strong lateral microscopic segregation of the molecules was observed at the surface layer; the minor component of the system (irrespective of whether it was water or methanol) was found to form two-dimensional aggregates, leaving the rest of the surface empty for the major component. The effect of the vicinity of the vapor phase on the properties of the molecules was found to vanish very quickly: the composition of the second layer as well as the properties of the molecules of this layer (e.g., dynamics and orientation) did not differ considerably from those in the bulk liquid phase.

The structure of water has been analyzed at eight different thermodynamic states from ambient to supercritical conditions both by molecular dynamics (MD) and Reverse Monte Carlo (RMC) simulation. MD simulations have been carried out with two different potential models, a polarizable potential and one of the most successful nonpolarizable models, i.e., the well known Simple Point Charge potential in its revised version labeled by E (SPC/E). It has been found that, although the polarizable model can reproduce the experimental partial pair correlation functions at the high temperature states better than the nonpolarizable one, it still cannot account for all the features of the measured functions. The experimental partial pair correlation functions have been well reproduced by the RMC simulations at every state point. The resulting structures have been analyzed in detail. It has been found that the tetrahedral orientation of the hydrogen bonded neighbors is already lost at 423k, whereas the hydrogen bonds themselves remain preferentially linear even above the critical point. In investigating the properties of the hydrogen-bonded clusters of the molecules it has been found that the space-filling percolating network, present under ambient conditions, collapses around the critical point.

Substantial improvements in the molecular level understanding of fluid interfaces have recently been achieved by recognizing the importance of detecting the intrinsic surface