We propose a novel interpretation of the water liquid-vapor interface vibrational sum-frequency (VSF) spectrum in terms of hydrogen-bonding classes. Unlike an absorption spectrum, the VSF signal can be considered as a sum of signed contributions from different hydrogen-bonded species in the sample. We show that the recently observed positive feature at low frequency, in the imaginary part of the signal, is a result of cancellation between the positive contributions from four-hydrogen-bonded molecules and negative contributions from those molecules with one or two broken hydrogen bonds. Spectral densities for each of these subgroups span the entire relevant spectral range. Three-body interactions within our newly developed E3B water simulation model prove to be critical in describing the proper balance between different hydrogen-bonded species, as (two-body) SPC/E, TIP4P, and TIP4P/2005 models fail to reproduce the positive feature. The results clarify the molecular origin of the VSF signal, and highlight the importance of many-body interactions for water in heterogeneous situations.

Phase-sensitive vibrational sum-frequency experiments on the water surface, using isotopic mixtures of water and heavy water, have recently been performed. The experiments show a positive feature at low frequency in the imaginary part of the susceptibility, which has been difficult to interpret, and impossible to reproduce using two-body (pairwise-additive) water simulation models. We have reparameterized a new three-body simulation model for liquid water, and with this model we calculate the imaginary part of the sum-frequency susceptibility, finding good agreement with experiment for dilute HOD in D(2)O. Theoretical analysis provides a molecular-level structural interpretation of these new and exciting experiments. In particular, we do not find evidence of any special ice-like ordering at the surface of liquid water.

The most common potentials used in classical simulations of liquid water assume a pairwise additive form. Although these models have been very successful in reproducing many properties of liquid water at ambient conditions, none is able to describe accurately water throughout its complicated phase diagram. The primary reason for this is the neglect of many-body interactions. To this end, a simulation model with explicit three-body interactions was introduced recently [R. Kumar and J. L. Skinner, J. Phys. Chem. B 112, 8311 (2008)]. This model was parameterized to fit the experimental O-O radial distribution function and diffusion constant. Herein we reparameterize the model, fitting to a wider range of experimental properties (diffusion constant, rotational correlation time, density for the liquid, liquid/vapor surface tension, melting point, and the ice Ih density). The robustness of the model is then verified by comparing simulation to experiment for a number of other quantities (enthalpy of vaporization, dielectric constant, Debye relaxation time, temperature of maximum density, and the temperature-dependent second and third virial coefficients), with good agreement.

Infrared spectroscopy of the water OH stretch provides a sensitive probe of the local hydrogen-bonding structure and dynamics of water molecules. Previously, we have utilized a mixed quantum/classical model to calculate vibrational spectroscopic observables for bulk water, ice, the liquid/vapor interface, and small water clusters, as well as water interacting with ions and biological molecules. These studies rely on spectroscopic maps that relate the OH stretching frequency and transition dipole to the local environment around a water molecule. Our spectroscopic maps were parametrized based on water clusters taken from bulk water simulations; in this article, we test the robustness of these maps for water in nonbulk-liquid environments. We find that the frequency, transition dipole, and coupling maps work as well for the water surface, ice Ih, and the water hexamer as they do for liquid water. This suggests that these maps may be generally applied to study the vibrational spectroscopy of water in diverse, potentially heterogeneous environments.

Using a newly developed and recently parameterized classical empirical simulation model for water that involves explicit three-body interactions, we determine the eleven most stable isomers of the water hexamer. We find that the lowest energy isomer is one of the cage structures, in agreement with far-IR and microwave experiments. The energy ordering for the binding energies is cage > glove > book > bag > chair > boat > chaise, and energies relative to the cage are in good agreement with CCSD(T) calculations. The three-body contributions to the cage, book, and chair are also in reasonable agreement with CCSD(T) results. The energy of each isomer results from a delicate balance involving the number of hydrogen bonds, the strain of these hydrogen bonds, and cooperative and anti-cooperative three-body interactions, whose contribution we can understand simply from the form of the three-body interactions in the simulation model. Oxygen-oxygen distances in the cage and book isomers are in good agreement with microwave experiments. Hydrogen-bond distances depend on both donor and acceptor, which can again be understood from the three-body model. Fully anharmonic OH-stretch spectra are calculated for these low-energy structures, and compared with shifted harmonic results from ab initio and density functional theory calculations. Replica-exchange molecular dynamics simulations were performed from 40 to 194 K, which show that the cage isomer has the lowest free energy from 0 to 70 K, and the book isomer has the lowest free energy from 70 to 194 K. OH-stretch spectra were calculated between 40 and 194 K, and results at 40, 63, and 79 K were compared to recent experiments, leading to re-assignment of the peaks in the experimental spectra. We calculate local OH-stretch cumulative spectral densities for different donor-acceptor types and compare to analogous results for liquid water.