Computational Investigation of the First Solvation Shell Structure of

Nov 14, 2008 - Department of Chemistry, Louisiana Tech UniVersity, Ruston, Louisiana 71270 and Chemical & Materials. Sciences DiVision, Pacific Northw...
0 downloads 0 Views 550KB Size
Download PDF
J. Phys. Chem. B 2009, 113, 4141–4146

4141

Computational Investigation of the First Solvation Shell Structure of Interfacial and Bulk Aqueous Chloride and Iodide Ions† Collin D. Wick‡ and Sotiris S. Xantheas*,§ Department of Chemistry, Louisiana Tech UniVersity, Ruston, Louisiana 71270 and Chemical & Materials Sciences DiVision, Pacific Northwest National Laboratory, 902 Battelle BouleVard, P.O. Box 999, MS K1-83, Richland, Washington 99352, [email protected], [email protected] ReceiVed: July 30, 2008; ReVised Manuscript ReceiVed: September 23, 2008

Molecular dynamics simulations with polarizable interaction potentials were carried out to understand the solvation structure of chloride and iodide anions in bulk and interfacial water, showing qualitative similarities between the first solvation shell structures at the interface and bulk. For the more polarizable iodide, its solvation structure was found to be more anisotropic than chloride, and this trend persisted at both the interface and in the bulk. The anisotropy of the solvation structure correlated with polarizability, but it was also found to inversely correlate with anion size. When polarizability was reduced to near zero, a very small anisotropy in the water solvation structure around the ion still persisted. Polarizable anions were found to have on average an induced dipole in the bulk that was significantly larger than zero. This induced dipole resulted in the water hydrogen atoms having stronger interactions with the anions on one side of them, in which the dipole was pointing. In contrast, the other side of the anions, in which the induced dipole was pointing away from, had fewer water molecules present and, for the case of iodide, was rather devoid of water molecules all together at both the interface and in the bulk. This region formed a small cavity in the bulk, whereas at the air-water interface it was simply part of the air interface. In the bulk, this small cavity may be viewed as somewhat hydrophobic, and the need for the extinction of this cavity may be one of the major driving forces for the more polarizable anions to reside at the air-water interface. I. Introduction The understanding of electrolyte solutions, including their surface properties, has been a long-sought goal dating back nearly 100 years.1,2 Until recently, the common notion was that the air-water interface of simple aqueous electrolyte solutions was free of alkali halide ions,3 a view that was reinforced by molecular dynamics simulations with nonpolarizable potentials.4 This picture has changed with mounting evidence of anions being present at the interface from molecular dynamics simulations with polarizable potentials.5-7 In addition, recent experimental studies have presented evidence for interfacial anions.8-10 A general trend has emerged, namely that the larger, more polarizable halidessbromide and iodideshave a greater affinity for the interface than the smaller, less polarizable fluoride and chloride anions. The degree of polarizability, along with anion size,11 for these anions was found to be one of the primary origins for the greater interfacial enhancement of the former two.12 A typical scenario given to explain this picture is as follows. A polarizable anion in the bulk liquid has water molecules arranged rather symmetrically around it, but at the interface this symmetry is broken, resulting in the anion dipole being induced further at the interface, a fact that, in turn, induces stronger interactions with the water molecules in that region.13 Another factor that plays a critical role in the interfacial enhancement of ions was found to be the size of the anion,13 † Part of the special section “Aqueous Solutions and Their Interfaces”. * To whom correspondence should be addressed. E-mail: [email protected]. ‡ Louisiana Tech University. § Pacific Northwest National Laboratory.

with larger anions having greater interfacial affinity even if no polarizability is present.11 There have been many studies of the bulk solvation of ions, including significant experimental studies investigating the influence of alkali-halide salts on water vibrational spectroscopy14-17 and the influence of ions on water dynamics.18,19 However, understanding the bulk solvation structure around anions still remains an important issue and, to this end, computational studies can offer additional insight. Molecular dynamics simulations of nonpolarizable anions suggested that due to the water charge distributions, anisotropic solvation of anions is possible.20 When polarizability was taken into account, it was found that it induced a larger degree of anisotropic solvation.21-23 Car-Parrinello molecular dynamics simulations have previously found some evidence for anisotropic solvation for bromide in bulk water.24 It is clear that an anion at the air-water interface will be anisotropically solvated, and previous studies point to the likelihood that anions have anisotropic bulk solvation as well.21-24 What is less clear is how the bulk and interfacial solvation compare, and if any insights into the driving forces for anion interfacial activity can be garnered from this comparison. This paper describes the investigation and comparison of interfacial and bulk aqueous solvation of chloride and iodide anions. Chloride has a positive and iodide has a negative bulkto-interface free energy difference,12 and one goal of this study is to try to bring some insight into the factors that govern higher iodide interfacial activity than chloride. II. Computational and Simulation Methodology Molecular dynamics simulations with many-body potentials were carried out to study the structure of water in the first

10.1021/jp806782r CCC: $40.75  2009 American Chemical Society Published on Web 11/14/2008

4142 J. Phys. Chem. B, Vol. 113, No. 13, 2009 solvation shell around chloride and iodide, and how it is affected with respect to the environment. The rigid 4-site polarizable Dang-Chang model was used, which has a single Lennard-Jones (LJ) site located at the oxygen position, two positive partial charges located on the hydrogen atoms, and a negative partial charge and point polarizability located on an M-site, which is located along the bisector of the oxygen-hydrogen bonds.25 Sodium, chloride, and iodide were modeled by single-site potentials with LJ, their formal charge, and a point polarizability.12 An iterative self-consistent field technique was used, during which induced dipoles from point polarizabilities were modified until deviations in sequential iterations were less than 0.00001 D. The water molecule geometry was kept rigid with the SHAKE algorithm.26 Four system compositions were simulated in this study, including both 550 water molecules and a single sodium cation. In addition, one system contained a single chloride anion, another a single iodide anion, another an anion with the same LJ parameters as iodide, but with a polarizability of 0.01 Å3 instead of 6.9 Å3 (the polarizability of iodide), and finally a system contained an anion with the same LJ parameters as chloride, but with a higher polarizability (4.5 vs 3.69 Å3). For the modeling of the solvation structure, we used a slab geometry in which the z-axis was elongated, providing two air-water interfaces. The total z-length was approximately 3 times that of the liquid, which was 26 Å, and the xy dimensions were both 25.5 Å. Simulations were performed in the NVT ensemble, with the temperature set with the Berendsen thermostat.27 Four independent, well-equilibrated simulations were initiated for each system, with the anion located in a different location for each (two with the anion at the GDS, and two with the anion located in the liquid center), and special care was taken during the simulations to ensure that the anion did not approach the cation within 10 Å throughout the simulations. If a cation did approach the anion closer than 10 Å, then the simulation run was restarted with different starting conditions (with the cation only in a different location). Because of this procedure, no constrains were placed on the position of any of the atoms. The total simulation time for each independent run was 200 ps, totaling 800 ps for each system. To ensure that this time was adequate, all independent runs were inspected to verify that all values and observations presented were consistent between them. Also, in addition to the NVT simulations for the slab geometries, four independent 200 ps calculations of the same compositions, but in periodic bulk water, were performed in the NVE ensemble, after extensive equilibration in the NVT ensemble. These were used to calculate the relaxation times of the anions in bulk water. A potential truncation of 9 Å was enforced for LJ interactions with analytic tail corrections, and the particle mesh Ewald summation technique was used to handle the longrange electrostatics. III. Results Radial Distribution Functions. The anion-hydrogen and anion-oxygen radial distribution functions (RDFs) calculated at the interface and the bulk for both systems are shown in Figure 1. The interface and bulk were defined in order to represent the regions less than 3.5 Å and greater than 5 Å from the Gibbs dividing surface (GDS), respectively. The RDFs were both normalized by the bulk liquid density, which is why the interfacial RDFs approach a value near 0.5 and the bulk RDFs approach a value around 1. What is most noticeable between the bulk and interface for both anions is that their first RDF peaks are nearly identical. This is in contrast to a previous study

Wick and Xantheas

Figure 1. Radial distribution functions (RDFs) at the interface and bulk between chloride or iodide (g(r) is offset by -3) and the water hydrogen or oxygen atoms.

of ion solvation with nonpolarizable chloride anions that found that chloride had a lower first RDF peak at the interface when compared to the bulk when they were both normalized to approach the value of 1.28 Because the two studies were normalized by different densities, the current study gives evidence that the inclusion of polarizability increases the number of water molecules in the first solvation shell of interfacial water molecules. Also, it can be observed that the first peaks in the RDFs of Figure 1 are at slightly shorter distances at the interface than in the bulk. Integrating the RDFs and multiplying by the volume element and average bulk density can yield the average distance that a hydrogen atom can be found from an anion. This distance was calculated to be 2.17 and 2.19 Å for interfacial and bulk chloride anions, respectively, and 2.56 and 2.58 Å for interfacial and bulk iodide anions, both giving shorter distances at the interface. These factors, along with that the first RDF peak remains unchanged in the interfacial region, point to possibly stronger binding between the anion and water at the interface when compared to the bulk. This would be consistent with the picture of an interfacial ion having a significant induced dipole that points toward the bulk, and the water hydrogen atoms interacting with the anion in the region having somewhat stronger interactions than if with an anion with a lower induced dipole. However, it is interesting to note that, in general, the RDFs are qualitatively similar at the interface and at the bulk for the two anions. Radial-Angular Distribution Functions. Figure 2 shows the radial-angular distribution functions (RADFs) between the two anions and the water hydrogen atoms. The definition of the distance is the same as in the RDFs or between the anion and the water hydrogen, and the angle is defined between the anion-hydrogen vector and the induced dipole vector of the anion. A value of θ ) 0° (which is the region in the top middle of each diagram) defines the region where the induced dipole is pointing. The diagrams are in polar coordinates and cover all cases where the hydrogen-anion distance is less than 4.3 Å, thus accounting for the first two hydrogen-anion solvation shells. It should be noted that these diagrams equally cover cases where the dipole is both very small and large and is averaged over the total number of molecular dynamics steps sampled in the simulations. The diagrams present both the interfacial and bulk regions, as defined previously, for the chloride and iodide anions. The interfacial diagrams on the top in Figure 2 show that the water hydrogen atoms are most often found toward the side of the anions in which their dipoles are pointing. Furthermore, this effect is clearly stronger for iodide, which has a greater polarizability. This is in agreement with previous assessments

First Solvation Shell Structure

Figure 2. Radial-angular distribution function (RADF) between chloride (left) or iodide (right) anions with the water hydrogen atoms at the interface (top) and in the bulk (bottom). The angle is defined between the vector from the anion to the water hydrogen atom and the anion-induced dipole vector.

Figure 3. Radial-angular distribution functions (RADFs) between the anions and the water hydrogen atoms. The left panel represents an ion with the same LJ parameters as iodide, but with a polarizability of 0.01 Å3, whereas the right panel represents an ion with the same LJ parameters as chloride, but with a greater polarizability (4.5 vs 3.69 Å3).

of interfacial anion solvation.7,13,29 However, the arrangement of the first solvation shell of water molecules around the anions in the bulk (shown in the bottom panels), shares many qualitative similarities to the arrangement at the interface, even though the anisotropic nature is somewhat stronger at the interface. For both anions, the first solvation shell is clearly anisotropic around the anion, with a very low probability to find anions near the region in which the induced dipoles are pointing away, although it is very probable to find hydrogen atoms on the other side. The degree of the similarity between the interface and bulk first shell solvation structure will be discussed below. A question may arise as to the quality of sampling in the simulations, since only a single anion is present, and if these diagrams represent local minimums. It should be stated that these diagrams are from four independent simulations, in which the anion partitions between the bulk and interface for all of them. Also, the same diagram has been constructed for a concentrated 1 M CsI solution in an air-water system (not shown), and it was found to be very similar to the diagrams shown here. To better understand the role of polarizability, the RADFs for two additional “anions” were calculated in bulk water and are given in Figure 3. The left panel shows the results for an iodide-like anion, having the same LJ parameters as iodide, but with a polarizability of 0.01 Å3. This should allow for the polarizability to have a negligible effect on the water interactions, but at the same time, the small induced dipole will still respond to local electric fields. Because of this, if an anisotropic

J. Phys. Chem. B, Vol. 113, No. 13, 2009 4143

Figure 4. RDFs between iodide and the water oxygen atoms (displaced by 5), between iodide and the water hydrogen atoms (no displacement), and between an iodide-like atom with a polarizability of 0.01 Å3 and the water hydrogen atoms (displaced by -4). The solid lines represent the full RDFs, the dotted lines represent water molecules located near the direction that the induced dipoles are pointing (D60), and the dashed lines represent atoms located away from the direction of the anion induced dipole (D120).

solvation structure forms due to fluctuations in water structure, the induced dipole will point toward the region with more water molecules (since their hydrogen atoms will be pointing toward the anion). It should be noted that a short test calculation with a higher polarizability of 0.1 Å3 was carried out, and the same structure as in the left panel of Figure 3 was observed. There is a somewhat higher probability to find water molecules in the region where the induced dipole is pointing due to random fluctuations of water molecules surrounding the anion. However, comparing this figure with all of those in Figure 3 shows that the polarizability results in an anisotropic solvation environment that is greater than that of an anion with essentially no dipole. This view is consistent with previous investigations.24 The right panel in Figure 3 represents a chloride-like anion with a higher polarizability in bulk water. This shows that with increased polarizability, but the same size, a significantly more anisotropic solvation structure is found, which is also consistent with previous work.21 However, this solvation structure appears to be even more anisotropic than iodide, even though its polarizability is significantly lower (4.5 vs 6.9 Å3). The smaller size of the chloride-like anion appears to further promote the degree of anisotropy in comparison with iodide. The effect of a smaller size may be explained due to the fact that dipolar interactions are shorter range than ion interactions. Since dipolar interactions are shorter range, they will be less effective for larger anions such as iodide. However, since iodide has a much higher polarizability than regular chloride (3.69 Å3), it still has a more anisotropic solvation structure than chloride. Directed Radial Distribution Functions. To provide further analysis of the opposing regions in which the iodide induced dipoles are pointing toward and away, RDFs confined to these regions are plotted in Figure 4 for bulk anions. The confined regions were defined as those with an angle θ < 60° (D60 region) and θ > 120° (D120 region), from the angles taken from Figures 2 and 3 (the angle between the anion-induced dipole and the anion-hydrogen vector). Both the water hydrogen and oxygen atoms have their first RDF peaks in similar positions in both the D120 region and the overall RDF, but the D120 region is nearly double in magnitude as the overall RDF. For the anion with low polarizability, the ion-hydrogen RDF is moderately higher in the D120 region as with the overall RDF. The most pronounced difference between iodide and the low polarizable

4144 J. Phys. Chem. B, Vol. 113, No. 13, 2009

Figure 5. Distance at which the radial-angular distribution function first reaches the value of 1 as a function of the angle of the vector defined in the RADF between the water hydrogen atoms and the chloride anion (black lines), the iodide anion (green lines), and the chloride-like anion with a higher polarizability (red line).

iodide-like anion is in the comparisons in the RDFs for the D60 region. The low polarizable iodide-like anion has its first D60 RDF peak at a slightly longer distance, and around half of the height of the first D120 RDF peak. However, for iodide, the D60 region does not really show a significant first RDF peak, and slowly increases until it eventually plateaus around 4-4.5 Å. The iodide-oxygen D60 RDF structure has a somewhat pronounced peak, but its position is quite similar to that of where the iodide-hydrogen peak plateaus. This finding suggests that for the few water molecules that are present in the D60 region around a distance of 4.25 Å there is minimal preference for their orientation whereas, in contrast, there is strong orientational order in the D120 region. Excluded Volume. Figure 5 depicts the distance at which the RADF first reaches the value of 1 as a function of θ (as defined previously). This function is used to represent the region surrounding an anion in which there is a low probability to find another water oxygen or hydrogen atom, a concept similar to a cavity region for the anion. What can be observed is that at high values of (cos θ) (or at the region near where the induced dipole is pointing), the distance to reach the value of 1 between anions and hydrogen atoms is relatively short. In addition, at very negative values of (cos θ) this distance is significantly larger for the bulk and is nonexistent at the interface (since the induced dipole is pointing away from the vapor phase). For the bulk, this points to two distinct regions: one with short anion-hydrogen distances (region AD) and one with larger distances (region BD). For chloride, the distance in region BD is around 3.6 Å, in comparison to iodide for which this distance is around 4.0 Å. Thus, the distances in the BD region differ by 0.4 Å between the two, whereas the σ values vary by 0.8 Å (or 0.4 Å in ion radius), which is consistent with a view that atomic size most strongly affects the BD distance. To better gauge the effect of polarizability alone, the simulations from the chloridelike anion with an increased polarizability gave a distance of 3.7 Å, only slightly larger than that of regular chloride, suggesting a rather minor effect due to polarizability. The mean angular positions between regions AD and BD for chloride and iodide correspond to a (cosθ) value of -0.54 and -0.20, respectively. For the chloride-like anion with increased polarizability, this value between regions AD and BD was -0.11, showing the greatest BD region surface area. This is consistent with what was observed in Figure 3. The surface area exposed to area BD is increased when the polarizability is larger, but a larger anion size somewhat counteracts this effect as the smaller size lowers the angular size of the BD region.

Wick and Xantheas

Figure 6. Distributions of the average induced dipole (top) and the induced dipole vectors (bottom). The average induced dipole is divided by µ2.

Region AD is broader in the bulk than at the interface for both chloride and iodide anions by a similar amount (around ∆(cos θ) ) 0.25). From this analysis, the higher average induced dipole at the interface (which will be discussed below) results in a more-compact AD region for strong hydrogen-anion interactions. Although these interactions are also likely stronger at the interface for region AD, they only result in a slight shift of the first anion-hydrogen RDF peak, as seen in Figure 1. Dipole Distributions. The average dipole and the dipole vector distributions at the interface and bulk for chloride and iodide are given in Figure 6. It should be noted that the dipole vector distribution is essentially the dipole distribution (P(|µ|)), but with each point scaled by 1/µ2 to account for the fact that there are three dimensions. Because of this, P(µ) for the vector distribution at lower µ (or very large 1/µ2) will often have a higher uncertainty than at larger µ. For chloride in the bulk, the distributions are the smallest, with the vector distribution being most probable at zero, while still being fairly broad. For iodide, in contrast, the dipole distributions are at higher µ values, with the vector distribution peaking at values greater than µ ) 0. What is clear is that both chloride and iodide have, on average, significant dipoles in the bulk. In addition, the distributions show that dipoles with very low values exist, albeit with very low probabilities. At the interface, both iodide and chloride dipole distributions shift to larger values, and for iodide, no dipoles with values less than around 1.5 D were observed. These increased dipoles will favor stronger interactions with the water hydrogen atoms in the AD region of the interface. As to whether the increased dipole favors iodide over chloride at the interface is not absolutely clear. The described shifts in the dipole distributions appear slightly larger for iodide than for chloride. However, since the iodide anion is larger, its interactions with water are at a greater distance, whereas the dipole interactions are shorter range than the charge interactions. This issue will be discussed later. Dipole Autocorrelation Function. The dipole autocorrelation function was calculated for the anion-induced dipole. When the function reaches the value of exp(-1), the dipole was considered decorrelated. The values for iodide and the iodide-like anion (with a polarizability of 0.01 Å3) were 2.5 ( 0.5 ps and 0.17 ( 0.01 ps, respectively. The decorrelation of the low polarizable anion is around an order of magnitude faster than for iodide, likely occurring by a different mechanism. Regular fluctuations in water hydrogen interactions with the low polarizable anion may be responsible for its decorrelation time. For iodide, it will more likely decorrelate via a rotational mechanism, as it is quite rare for the induced dipole for iodide to approach zero. For comparison, the rotational decorrelation time for the dipole of

First Solvation Shell Structure water with the Dang-Chang potential was calculated to be 1.3 ps,30 of similar magnitude as the iodide dipole decorrelation time. IV. Discussion The anisotropic nature of the interactions between water and polarizable anions suggests that there are, in general, strong interactions on one side of the anions but relatively weak interactions on the other. If the BD region shown in Figure 5 was treated as an excluded volume, then it is apparent that as an anion transfers from the interface to the bulk, this region represents a cavity formed in bulk water. In addition, the similar RDFs shown in Figure 4 for D60 of oxygen and hydrogen atoms surrounding the iodide anion allude that there is little preferential water orientation in the BD region. Because of this, the BD cavity may be treated as somewhat hydrophobic. Using results from a previous study of the free energies of hard spheres of varying radius as a function of surface area31 and defining the radius and surface area for this cavity as equal to the BD region radius and angular size, respectively, a solvation free energy for this cavity in water can be estimated. For the chloride and iodide anions this would represent the surface area of the region of (cos θ) less than -0.4 and -0.77 kcal/mol, respectively, giving values of 1.2 and 3.9 kcal/mol as obtained from the data of the hard-sphere study.31 This leads to a difference between chloride and iodide in this “cavity free energy” of 2.7 kcal/ mol. This exercise is a very rough comparison but gives an idea of a ballpark estimate for how much free energy may be gained solely by region BD for the transfer of the anion from the interface to the bulk. A previous study utilizing the potential of mean force technique to determine the free energy of chloride and iodide across a vapor-liquid interface found the free energy for the transfer from the bulk to the interface for chloride and iodide to be around 1.8 and -1.5 kcal/mol, respectively,12 leading to a difference of 3.3 kcal/mol, a value that is very close to the rough estimate of the cavity free energy. It should be emphasized that the bulk solvation structure of iodide is rather dynamical, and there will be instances with water molecules occupying the described cavity, however this is not the case for a majority of configurations (as observed by the RADFs). Another difference between the interface and bulk is that the region of shorter anion-hydrogen atom distances is broader in the bulk than in the interface. The first anion-hydrogen RDF peak for the interface and bulk for both anions are very similar. The average number of hydrogen atoms within 2.8 Å of chloride or 3.1 Å of iodide is 5.1 and 5.5 for chloride in the bulk and interface and 4.7 and 5.2 for iodide, respectively. This shows that, on average, more water hydrogen atoms are present in the first solvation shell of the anions in the interface than in the bulk, despite the fact that the interactions cover a smaller total surface area at the interface. This is likely a result of the stronger anion-water interactions present at the interface. For a rough test of the effect that a larger induced anion dipole has on the interactions with water in the AD region, clusters containing a single anion and five water molecules, a number near the one within the anions’ first solvation shell, had their minimum energies calculated. The cluster energies included the average over multiple configurations (5 or more) initiated from different initial conditions in which at least one hydrogen from each water was bound with the anion. The water molecules were always located near the region in which the dipole was pointing. The anion dipole was fixed (with no fluctuation in ion dipoles allowed) for each of the calculations to correspond with the average value in the bulk and interface to determine the effect

J. Phys. Chem. B, Vol. 113, No. 13, 2009 4145 of the increased dipole on these interactions. The difference between cases for iodide with a set dipole of 3.5 D (interface average) and 2.6 D (bulk average) were 1.66 kcal/mol, and the differences for chloride with a dipole of 2.35 D (interface average) and 1.65 D (bulk average) were 2.06 kcal/mol. This shows that interactions of the anions with water molecules located in region AD are stronger at the interface at the average anion induced dipole, but the increases in energy are quite similar between chloride and iodide. V. Conclusions Detailed molecular dynamics simulations with polarizable potentials were carried out to study the aqueous solvation structures of chloride and iodide both in the bulk and at the air-water interface. The results show a qualitatively similar firstshell solvation structure at the interface and in the bulk, with anisotropic water solvation surrounding both anions. For the interface this resulted in water molecules with strong interactions with iodide in the region its dipole was pointing, and air on the other side. For bulk anions the anisotropic solvation persisted, but instead of air, a small cavity was found to form in the region opposite the direction of their induced dipoles. The anions at the interface had a somewhat more compact region of stronger interactions with water than the bulk, and had on average a higher induced dipole than when in the bulk. There are likely an array of factors that drive anions to the interface, mostly dependent on anion size and polarizability.13 However, one major factor proposed in this study that, to the best knowledge of the authors, has not been previously suggested, is that the extinction of a somewhat hydrophobic cavity in the bulk, due to the anisotropic solvation environment surrounding anions due to polarizability, is a driving force for anion interfacial activity. Further studies with more accurate descriptions of the underlying intermolecular interactions are warranted in order to validate this result. Acknowledgment. Part of this work was supported by the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences, US Department of Energy. Battelle operates the Pacific Northwest National Laboratory for the US Department of Energy. In addition, some of the research was funded by the Louisiana Board of Regents Research Competitiveness Subprogram contract No. 3LEQSF(2008-11)RD-A-21. The calculations were carried out using the resources from the Louisiana Optical Network Initiative (LONI). Additional computer resources were provided by the Office of Basic Energy Sciences, US Department of Energy. References and Notes (1) Heydweiller, A. Ann. Physik. 1910, 33, 145. (2) Onsager, L.; Samaras, N. N. T. J. Chem. Phys. 1934, 2, 528. (3) Randles, J. E. B. Phys. Chem. Liq. 1977, 7, 107–179. (4) Bhatt, D.; Chee, R.; Newman, J.; Radke, C. J. Curr. Opinion Colloid Interface Sci. 2004, 9, 145–148. (5) Tobias, D. J.; Hemminger, J. C. Science 2008, 319, 1197. (6) Jungwirth, P.; Tobias, D. J. J. Phys. Chem. B 2002, 106, 6361– 6373. (7) Dang, L. X.; Chang, T. M. J. Phys. Chem. B 2002, 106, 235–238. (8) Ghosal, S.; Hemminger, J. C.; Bluhm, H.; Mun, B. S.; Hebenstreit, E. L. D.; Ketteler, G.; Ogletree, D. F.; Requejo, F. G.; Salmeron, M. Science 2005, 307, 563–566. (9) Petersen, P. B.; Saykally, R. J. Chem. Phys. Lett. 2004, 397, 51– 55. (10) Liu, D. F.; Ma, G.; Levering, L. M.; Allen, H. C. J. Phys. Chem. B 2004, 108, 2252–2260. (11) Eggimann, B. L.; Siepmann, J. I. J. Phys. Chem. C 2008, 112, 210. (12) Dang, L. X. J. Phys. Chem. B 2002, 106, 10388–10394.

4146 J. Phys. Chem. B, Vol. 113, No. 13, 2009 (13) Vrbka, L.; Mucha, M.; Minofar, B.; Jungwirth, P.; Brown, E. C.; Tobias, D. J. Curr. Opinion Colloid Interface Sci. 2004, 9, 67–73. (14) Smith, J. D.; Saykally, R. J.; Geissler, P. L. J. Am. Chem. Soc. 2007, 129, 13847. (15) Terpstra, P.; Combes, D.; Zwick, A. J. Chem. Phys. 1990, 92, 65. (16) Schultz, J. W.; Hornig, D. F. J. Phys. Chem. 1961, 65, 2131. (17) Nickolov, Z. S.; Miller, J. D. J. Colloid Interface Sci. 2005, 287, 572. (18) Omta, A. W.; Kropman, M. F.; Woutersen, S.; Bakker, H. J. Science 2003, 301, 347. (19) Kropman, M. F.; Bakker, H. J. Science 2001, 291, 2118. (20) Rajamani, S.; Ghosh, T.; Garde, S. J. Chem. Phys. 2004, 120, 4457. (21) Carignano, M. A.; Karlstro?m, G.; Linse, P. J. Phys. Chem. B 1997, 101, 1142. (22) Dang, L. X.; Garrett, B. C. J. Chem. Phys. 1993, 99, 2972. (23) Stuart, S. J.; Berne, B. J. J. Phys. Chem. B 1999, 103, 10300.

Wick and Xantheas (24) Raugei, S.; Klein, M. L. J. Chem. Phys. 2002, 116, 196. (25) Dang, L. X.; Chang, T. M. J. Chem. Phys. 1997, 106, 8149–8159. (26) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comp. Phys. 1977, 23, 327–341. (27) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684–3690. (28) Benjamin, I. J. Chem. Phys. 1991, 95, 3698–3709. (29) Wick, C. D.; Dang, L. X. J. Chem. Phys. 2007, 126, 1347021-– 1347024. (30) Wick, C. D.; Dang, L. X. J. Phys. Chem. B, 2006, 110, 68246831. (31) Huang, D. M.; Geissler, P. L.; Chandler, D. J. Phys. Chem. B 2001, 105, 6704–6709.

JP806782R