Simulating liquid water near

Simulating Liquid Water near Mineral Surfaces Current Methods and Limitations... 

The strength of the water-metal interaction together with the surface corrugation gives rise to much more drastic changes in water structure than the ones observed in computer simulations of water near smooth nonmetallic surfaces. Structure in the liquid state is usually characterized by pair correlation functions (PCFs). Because of the homogeneity and isotropy of the bulk liquid phase, they become simple radial distribution functions (RDFs), which do only depend on the distance between two atoms. Near an interface, the PCF depends not only on the interatomic distance but also on the position of, say the first, atom relative to the interface and the direction of the interatomic distance vector. Hence, considerable changes in the atom-atom PCFs can be expected close to the surface. 

Numerous simulation studies of liquid water near various smooth and structured surfaces were reported (see Refs. [28, 32, 206] for more details). Near the surface, water molecules show orientational... 

The simulated free surface of liquid water is relatively stable for several nanoseconds [68-72] because of the strong hydrogen bonds formed by liquid water. The density decrease near the interface is smooth it is possible to describe it by a hyperbolic tangent function [70]. The width of the interface, measured by the distance between the positions where the density equals 90% and 10% of the bulk density, is about 5 A at room temperature [70,71]. The left side of Fig. 3 shows a typical density profile of the free interface for the TIP4P water model [73]. 

Recently, many experiments have been performed on the structure and dynamics of liquids in porous glasses [175-190]. These studies are difficult to interpret because of the inhomogeneity of the sample. Simulations of water in a cylindrical cavity inside a block of hydrophilic Vycor glass have recently been performed [24,191,192] to facilitate the analysis of experimental results. Water molecules interact with Vycor atoms, using an empirical potential model which consists of (12-6) Lennard-Jones and Coulomb interactions. All atoms in the Vycor block are immobile. For details see Ref. 191. We have simulated samples at room temperature, which are filled with water to between 19 and 96 percent of the maximum possible amount. Because of the hydrophilicity of the glass, water molecules cover the surface already in nearly empty pores no molecules are found in the pore center in this case, although the density distribution is rather wide. When the amount of water increases, the center of the pore fills. Only in the case of 96 percent filling, a continuous aqueous phase without a cavity in the center of the pore is observed. 

The xi, X2,... are displacements of the atoms, and this is a gaussian convolution of the mechanical potential energy I/(TV) with the variances depending on the masses of the atoms, on the temperature, and on h. If /(TV) is weakly dependent on the displacements xi,X2,..., then this FH model reduces to the QFH model, Eq. (3.67). Near minima of I/(TV) this gaussian convolution means that /(TV) > /(TV) this is an approximate description of zero point motion. Near maxima of /(TV) this gaussian convolution means that /(TV) < /(TV) this is an approximate description of barrier tunneling. Simulations of liquid water have been conducted... 

The properties of liquid water between electrodes are obviously of considerable importance. MD simulations of the system have been reported, with and without an applied electric field [47]. It is concluded that the presence of the walls is structurebreaking for the H-bond networic and that this has significant consequences for the self-diffusion of water molecules near the walls. 

In the previous section, we have shown that switching the picture from the nearly integrable Hamiltonian to the Hamiltonian with internal structures may make it possible to solve several controversial issues listed in Section IV. In this section we shall examine the validity of an alternative scenario by reconsidering the analyses done in MD simulations of liquid water. As mentioned in Section III, since a water molecule is modeled by a rigid rotor, and has both translational and rotational degrees of freedom. So, the equation of motion involves the Euler equation for the rigid body, coupled with ordinary Hamiltonian equations describing the translational motions. The precise Hamiltonian is therefore different from that of the Hamiltonian in Eq. (1), but they are common in that the systems have internal structures, and the separation of the time scale between subsystems appears if system parameters are appropriately set. 

With local information given by INM analysis in mind, we next see the character of rotational relaxation in liquid water. The most familiar way to see this, not only for numerical simulations [76-78] but for laboratory experiments, is to measure dielectric relaxation, by means of which total or individual dipole moments can be probed [79,80]. Figure 10 gives power spectra of the total dipole moment fluctuation of liquid water, together with the case of water cluster, (H20)io8- The spectral profile for liquid water is nearly fitted to the Lorentzian, which is consistent with a direct calculation of the correlation function of rotational motions. The exponential decaying behavior of dielectric relaxation was actually verified in laboratory experiments [79,80]. On the other hand, the profile for water cluster deviates from the Lorentzian function. As stated in Section III, the dynamics of finite systems may be more difficult to be understood. 

The simulations started from a condition in which no gas molecules were dissolved into the liquid water. During the simulations, gas molecules in the gas were gradually dissolved one by one into the liquid water, and the growth of the hydrate from the dilute solution was successfully observed on the interface between the hydrate and the liquid water (hereafter, the interface is referred to as the hydrate interface) for both systems. Then, we analyzed the mechanism of cage formation on the hydrate interface for both systems. In Figure 1(b), snapshots of gas molecules and the hydrogen-bonded networks (HB-networks) near the hydrate interface at the end of the simulation for the CH4 system are shown. 

Benjamin and coworkers studied various aspects of ion dynamics near the liquid/ vapor and the liquid/liquid interface. Time-dependent probability distributions of the ion position were studied near the interface between immiscible apolar and polar liquids [198], The simulation results were in almost quantitative agreement with a one-dimensional diffusion model. Small differences were attributed to the solvent reorganization dynamics. Later this work was extended to the ion and solvent dynamics following charge transfer [199] near the same polar/apolar interface and to the liquid water/vapor interface [200],... 

---