Sie sind auf Seite 1von 11

Structure of the metal-aqueous electrolyte solution interface

rard,a) M. Kinoshita,b) N. M. Cann, and G. N. Patey D. R. Be


Department of Chemistry, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z1

Received 24 February 1997; accepted 23 June 1997 Theoretical results are given for aqueous electrolyte solutions in contact with uncharged metallic surfaces. The metal is modeled as a jellium slab and is treated using local density functional theory. The solution structure is obtained using the reference hypernetted-chain theory. The two phases interact electrostatically and the coupled theories are iterated to obtain fully self-consistent results for the electron density of the metal and surface-particle correlation functions. The metal-induced structure of pure water and aqueous electrolyte solutions as well as the electrostatic potential drop across the interface are discussed in detail. The results are compared with those for ions in simple dipolar solvents. It is found that the water molecules are ordered by the metal eld and that the surface-induced solvent structure strongly inuences the ion distributions. 1997 American Institute of Physics. S0021-96069751336-5

I. INTRODUCTION

The structure of water and electrolyte solutions at metal surfaces is of fundamental importance in electrochemistry and surface science. However, despite considerable effort, our microscopic understanding of the organization and behavior of particles at these interfaces has developed slowly and many basic questions remain. Further theoretical and experimental studies which focus on the microscopic structure and dynamics are needed. For bulk solutions in contact with metals, progress has been hampered by the lack of detailed experimental results for the interfacial structure. Experimental methods which probe interfaces must inevitably extract surface data amidst an overwhelming number of bulk particles. Hence, it is not surprising that the development of techniques which probe the structure of liquids near metals has lagged behind the corresponding bulk methods. However, experimental methods which probe the microscopic structure at metal-water interfaces are now becoming available.1,2 In recent years there have been many computer simulation studies of metal-water interfaces.318 In this work, the liquid is usually treated employing one of the standard water models and the surface-water interaction is modeled in different ways. Many models for the metal-water potential are based on calculations of the interaction of a water molecule with a metal cluster. For example, the models employed in Refs. 57,10, and 11 are based on the interaction potential between a water molecule and a Pt5 cluster calculated from ckel-MO theory.3,4 Results for the interaction of extended Hu water with a mercury cluster are also available16 and have been used in simulations.17,18 These potentials provide some idea of the interaction strength and of its dependence on separation and orientation. Thus they represent a signicant advance. It should be noted, however, that interaction potena

Present address: Department of Chemistry, University of Utah, Salt Lake City, Utah, 84112. b Permanent address: Research Section of Nuclear Chemical Engineering, Institute of Atomic Energy, Kyoto University, Uji, Kyoto 611, Japan. J. Chem. Phys. 107 (12), 22 September 1997

tials obtained with metal clusters are strongly dependent on cluster size.19 Surface-water interactions have also been modeled assuming a planar surface with a uniform distribution of Lennard-Jones particles for the metal atoms,9 or, sitesite potentials acting between the metal and water particles.8,12,13 In several studies9,12,14,15 ions have also been considered with the ion-metal potentials modeled in various ways. However, we note that at present simulations including solvent molecules and ions at nite concentration are problematic for several reasons, not the least of which is very slow convergence. Also, such simulations are usually performed by conning a xed number of ions and solvent particles between two plates, and there is no reliable way of estimating the bulk equilibrium concentration. In principle, Grand Canonical Monte Carlo calculations would solve this problem but such calculations would be extremely difcult in practice.20 The integral equation approach used in the present work is approximate but does not suffer from these difculties. In this paper, we examine the structure of water and aqueous electrolyte solutions at an isolated metal surface by extending a method recently developed for ions in a simple dipolar solvent.2123 In this approach, the metal is assumed to be a hard jellium slab and the electron density is obtained via density functional theory DFT with the mean electrostatic eld of the electrolyte solution included. The structure of the uid is obtained using the reference hypernetted-chain RHNC theory in the electrostatic eld of the metal. The coupled DFT/RHNC equations are iterated until selfconsistent electron and molecular distribution functions are obtained. As in earlier work with simple dipolar solvents,22 we nd that the metal-electrolyte-solution interface is a rather narrow region characterized by an inhomogeneous distribution of electrons in the metal and of solvent particles and ions in the uid. The water molecules are ordered at the interface and, as in the dipole solvent case, the solvent structure strongly inuences the ion distributions. However, the ion distributions obtained for the aqueous model are qualitatively different from those obtained for the simpler solvent. The remainder of this paper is divided into three sec 1997 American Institute of Physics 4719

0021-9606/97/107(12)/4719/10/$10.00

4720

rard et al.: Metal-aqueous electrolyte interface Be

TABLE I. The reduced multipole moments used in the water-like solvent model. The moments are given in spherical tensor notation Ref. 24 and the solvent diameter is taken to be 2.8 . Dipole moment Quadrupole moment Octupole moment 0 Q 1* 0 Q 2* 2 2 Q 2* Q 2 * 0 Q 3* 2 2 Q 3* Q 3 * 2.787 0.05 0.787 92 0.18 0.240 998

the elementary charge and e i( z ) and e s ( z ) denote the interaction energy of an electron with the average potential of the ions, i( z ), and solvent molecules, s ( z ). The quantities u jel and u xc are dened in Ref. 21. The potential at a distance z from the center of the jellium slab is approximated by the potential at a radius r d m /2 z z w from the macrosphere center i.e., z z w from the surface of the macrosphere. In this way, the ion potential is given by

tions. The model and theory for aqueous electrolyte solutions are described in Section II, the results are discussed in Section III and our conclusions are summarized in Section IV.
II. MODEL AND THEORY

i z

4e r

t 2 g m t g m t dt

4e

t g m t g m t dt , 2

As in our earlier work,21,22 the interfacial system consists of an uncharged innite metallic slab ( w ) of width 2 z w 64 immersed in a molecular uid. The metal is represented by a uniform positive background of number density n representing the atomic cores of the metal and an electron gas. The molecular uid consists of a solution of charged hard sphere cations ( ) and anions ( ) with bulk number densities and , respectively in a model water-like solvent ( s ) at a bulk density s . The ions are characterized by the valences and hard sphere diameters d . The water-like solvent is represented as a hard sphere of diameter d s 2.8 with embedded point dipole, quadrupole and octupole moments characteristic of a water molecule in the liquid phase. The reduced multipole moments of the water model are listed in Table I and are consistent with values used in earlier work.24 A complete description of the theory and method of solution has previously been reported2123 for the simpler dipolar solvent model and for charged hard spheres in a dipolar solvent. The theory consists of solving self-consistently the RHNC theory for the uid25 and the Hohenberg-Kohn-Sham local density functional theory for the metal.26 In the present article, the uid structure at the planar metal surface is approximated by the result obtained using the RHNC theory at a spherical surface. Clearly, the structure at a spherical surface approaches the at-plate limit as the diameter of the sphere is increased. In previous studies,2729 it has been found that a macrosphere of diameter d m 30d s 84 yields essentially at-plate results. The use of this approximation and the introduction of the more detailed water-like solvent requires that the interaction potentials spanning the interface be explicitly dened. In density functional theory, an effective potential is used to describe the interaction of an electron with its environment. In the present model, the effective potential is given by u eff z u z u xc z e i z e s z ,
jel

where g m ( r ) denotes the macrosphere-ion pair distribution functions. In the case of the solvent, the pair distribution function is expanded in the form21,30 g ms 01
m 0 mm m F 0 mm g 0 ; ms r R 0 0, , , m

where R is a Wigner generalized spherical harmonic, F 0 mm ( ) m f 0 mm / 2 m 1, f 0 mm is a nonzero constant, is the angle between the dipole vector and the surface normal and is the angle denoting the orientation of the solvent about the dipole axis. Using this expansion, the potential due to the solvent is given by

s z 4 s

m 0,

0 mm m f Q

2m1

3/2

0 mm t 1 m g 00; ms t dt ,

m where Q denotes the solvent multipole moments in spherical tensor notation see Table I and Ref. 24. The interactions of the ions and solvent molecules with the metal surface are given by

u w 01

u ws 01

z z w d /2, z z w d /2,

u jel z ,

z z w d s /2, z z w d s /2,

u el ws 01 ,

where u el ws is the electrostatic contribution to the jelliumsolvent potential. If the wall-solvent potential is expanded in the form21,30 u ws 01
0 mm m F 0 mm u 0 ; ws z 1 R 0 0, , , m

where z is the distance from the center of the jellium slab the origin z 0 is chosen at the center of the metal slab, u jel is the interaction energy of an electron with the positive jellium background, u xc is the exchange-correlation energy, e is

0 mm the projections u 0 ; ws ( z ) are given by

0 mm ;el u0 ; ws z

m Q

d m u jel z edz m

m ! F 0 mm

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

rard et al.: Metal-aqueous electrolyte interface Be

4721

for z z w d s /2. In the remainder of this article, we follow Blum31 and make the convenient choice, f mnl (2 m 1)(2 n 1), in the expansions of the pair interaction and correlation functions. The numerical methods employed in this work are based on an extension of the algorithms developed by Kinoshita and Berard23 for the simpler case of a purely dipolar hard sphere solvent model in the bulk and near surfaces. The stability, efciency and robustness of these methods extends to the more sophisticated water-like model used in this article. In addition, the convergence properties do not deteriorate signicantly for the present systems.
III. RESULTS

In this section, we consider results for pure water and aqueous electrolyte solutions in contact with a jellium surface. As in earlier studies for the simpler dipolar solvent model, we consider a mercury-like jellium metal having a Wigner-Seitz radius r s (4 n /3) 1/3 2.65a 0 , where a 0 is the Bohr radius. The density functional equations are evaluated on a grid of 0.02 extending 8 beyond the edges of the jellium slab. The RHNC equations are solved on a grid of 0.01d s over 2 12 points using the HendersonPlischke bridge functions32 as described in an earlier publication.22 All results given are at 298 K and the relevant ion and solvent densities are obtained using experimental values as in Ref. 24.
A. Pure solvent at a jellium surface

FIG. 1. RHNC projections of the wall-solvent pair distribution function for pure water in contact with inert - - - - and jellium metal ( ) surfaces. The solvent parameters are given in Table I and the Wigner-Seitz radius was taken as r s 2.65a o . Note that z * ( z z w )/ d s .

We begin by considering the pure water-like solvent at the model metal surface. The angle-averaged, wall-solvent 000 pair distribution function, g 00; ms , is shown in Fig. 1a and is compared with the result for water at an inert surface. The gure clearly shows that the metal has a signicant inuence on the solvent distribution. At the inert surface,28,29 the solvent contact density is relatively low and oscillations in the distribution function decay rapidly to bulk-like behavior 000 within a few solvent diameters. At the metal surface, g 00; ms also decays rapidly to bulk-like behavior. However, in this case the probability density shows a high surface contact value and pronounced oscillations. These differences result directly from the short-ranged attraction of the solvent to the jellium metal. For a macrosphere, the average surface area occupied by water molecules may be dened as A s , r 000 4 s 0minr 2 g 00; ms r dr A 9

where A is the surface area of a sphere with radius ( d m d s )/2 and r min is the separation corresponding to the 000 rst minimum in g 00; ms . In moving from an inert to a jellium metal surface, the average surface area per solvent molecule decreases from A s 13.1 2 at the inert surface to 10.2 2 at the jellium surface. When these values are compared with the area per molecule for a close-packed monolayer of spheres of diameter 2.8 , A s 6.8 2 , one can see that the inuence of the metal surface on the contact layer is signicant. The

average area per solvent molecule in the second layer can be obtained by integrating as in Eq. 9, but now with the limits ranging from the rst to the second minimum in the radial distribution function. One nds that the second layer is only slightly more dense due to the metal surface (12.0 2 per solvent molecule as opposed to 12.1 2 at the inert surface although in the metal case the probability density is much more strongly peaked at z * 1.5. In Fig. 1b, the rst angle-dependent projection of the 011 pair distribution function, g 00; ms , is shown. This projection is the dominant component of g ms describing the orientational polarization of the solvent. Again, the inuence of the metal surface on the solvent polarization is evident. At the inert 011 surface, g 00; ms only deviates signicantly from zero near contact. At the jellium surface, the solvent is strongly polarized from contact to z * 2.0. The negative contact value of 011 g 00; ms indicates that the positive ends of the solvent dipoles have a tendency to point outward consistent with the electric 011 eld at the metal surface. The change in sign of g 00; ms near z * 1.5 indicates a counter polarization of solvent not found in the simple dipolar models previously considered.21 In order to more fully elucidate the inuence of the metal on the solvent structure, it is useful to consider slices of the full pair distribution function, g ms ( , , z * ), for three wall-solvent separations see Fig. 2. Again, is the angle between the dipole and surface normal vectors and describes the orientation of the HOH plane about the dipole. These plots have the advantage that the full angle-space of

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

4722

rard et al.: Metal-aqueous electrolyte interface Be

FIG. 3. Water molecule congurations for various orientations ( , ). The black and gray spheres represent oxygen and hydrogen atoms, respectively. The thin sticks represent the lone pairs of electrons. The surface is located to the left of the solvent particles. The congurations labeled 15 are referenced in the text.

FIG. 2. Slices g ms ( , , z * ) for z * 1.5, and d

of the full jellium-solvent pair distribution function wall-solvent separations of a z * 0.5; b z * 0.75; c z * 1.7. The system parameters are as in Fig. 1.

the probability distribution functions is shown and both the dipole and OH bond distributions can be observed simultaneously. In order to help map the and angles to solvent orientations, Fig. 3 is also included and shows the conguration of a water molecule for several pairs of ( , ) coordinates. For example, when 0, Fig. 3 shows that rotates a water molecule from a surface normal orientation (cos 1) through the HOH plane. In this case, when cos 0.6, one of the OH bonds is directed outward approximately normal to the surface; when cos 0.6, the other OH bond is directed inward. When 90, rotates the dipole through the plane bisecting the water molecule; a vector directed from the oxygen atom to a lone pair of electrons O: points outward normal to the surface when cos 0.6 and inward when cos 0.6. The wall-solvent pair distribution function for water in contact with the jellium surface is shown in Fig. 2a. We see that at contact the most probable orientation of a water molecule has the dipole and the HOH plane nearly at against the surface cos 0, 90 conguration 1 in Fig. 3. This leads to strong interactions with other water molecules in the contact layer. On close inspection, one can see that the peak of the distribution is not exactly at cos 1 but is shifted to a small positive value indicating that orientations with the dipole directed away from the surface have a preference. This is consistent with the electric eld of the jellium which tends to point the positive end of the dipole outward. We also observe that the orientational distribution at contact is rather broad and that practically any orientation resulting from a moderate perturbation of conguration 1 in Fig. 3 is signicantly probable. This range of orientations includes cases with the O: vector directed both outward conguration 2 and inward conguration 3 with respect to the surface. Also, when cos 0, there is considerable exibility in

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

rard et al.: Metal-aqueous electrolyte interface Be

4723

leading to congurations with the HOH plane partially rotated (45 135) such that one OH bond and one O: vector are near the surface and the other set are tilted towards the bulk. If we look at a slice of the distribution function a little further from the surface Fig. 2b, z * 0.75] but still within the contact layer, we see that orientations with an OH vector roughly normal to the surface conguration 4 become more highly probable. At this separation orientations with an OH bond normal to the surface and orientations with the HOH plane at against the surface are nearly equally probable. At larger separations in the vicinity of the second peak in the solvent density, the orientational distribution changes rather abruptly Fig. 2c, z * 1.5]. Here the solvent strongly prefers orientations with an OH bond directed toward the metal surface conguration 5. This change in orientational preference is consistent with the change in sign of 011 the g 00; ms projection see Fig. 1b. These orientations allow tetrahedral contacts between water molecules in the rst and second layers and this helps stablize contact congurations where the OH bonds are canted into the surface conguration 2. A little further from the surface Fig. 2d, z * 1.7], one sees that orientations with lone pairs directed inward become signicantly probable. These orientations also allow tetrahedral contacts between water molecules in the rst and second layers and help stablize orientations in the contact layer where OH bonds are canted away from the surface conguration 3. In brief, the metal-water interactions tend to pull solvent particles towards the surface and induce a signicant amount of orientational polarization. The orientations of the water molecules in the contact layer seem to largely favor interactions among particles within that relatively dense layer. However, hydrogen-bonding type interactions between the rst and second layers also appear to be of some importance. We note that the structure of water at a metal surface differs signicantly from the structure at inert surfaces. At an inert hydrophobic surface the contact layer is less dense and the water molecules tend to orient such as to interact favorably with other molecules further from the surface.29,33,34 Hence, in the inert case one nds that molecular congurations with only a single OH bond or O: vector directed towards the surface are highly probable in the contact layer. We note that the water-mercury interface has been simulated17,18 using potentials which give a more realistic representation of the strong short-range interaction between water molecules and the metal. As one might expect, the contact layer obtained in these calculations is considerably denser than that found for the present model. Nevertheless, many features of the metal-water orientational correlations are similar to those discussed above.18
B. Aqueous electrolyte solutions

TABLE II. A summary of the aqueous electrolyte solutions considered in the present paper ( d s 2.8 ). Salt NaCl KF KCl KCl CsI d /ds 0.84 1.08 1.08 1.08 1.28 d /ds 1.16 0.84 1.16 1.16 1.44 Molarity 0.25 0.25 0.25 3.00 0.25

two projections of the wall-solvent pair distribution functions for 0.25 M and 3.0 M KCl solutions see Fig. 4. At the lower concentration solid curves, the projections are nearly indistinguishable from the projections of the pure solvent and this is also true for the higher order projections not shown in the gure. At 3.0 M, small variations in the projections are apparent. However, the solvent structure at the surface is only weakly disturbed and remains essentially as for the pure solvent case described above. In the RHNC theory, the reference system provides an approximation for the unknown bridge functions.25 In the present model for a metal surface, small variations in the reference system have only a minor inuence on the distribution functions obtained. This is illustrated in Fig. 4 where the hypernetted-chain HNC results for 0.25 M KCl calculated neglecting only the wall-solvent bridge functions are included. The HNC results dotted curves show some deviation from the RHNC curve for the angle-averaged distribu000 tion function, g 00; ms . In particular, the HNC closure gives a

In this section we consider electrolyte solutions at neutral jellium metal surfaces and examine the metal-ion distribution functions. The solutions considered are summarized in Table II with the ion diameters included. We rst consider

FIG. 4. Projections of the wall-solvent pair distribution function for KCl solutions in contact with a jellium metal surface. RHNC result are shown for 0.25 M ( ) and 3.0 M - - - solutions. The HNC results for the 0.25 M solution - - - - are also included.

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

4724

rard et al.: Metal-aqueous electrolyte interface Be

narrower distribution of particles in the contact layer and 011 fewer structural features appear. However, for g 00; ms and higher order projections not plotted the HNC and RHNC results are in very good agreement. Hence, although there are some differences in packing near the surface the orientational solvent structure is very similar in both case. The generally good agreement between HNC and RHNC results suggests that there is not a strong dependence on the approximation used for the bridge functions. It is shown below Figs. 5 and 7 that this is also true of the metal-ion distribution functions. This observation differs from the inert surface case where the results obtained were found to be sensitive to the closure approximation applied.29 It appears that, for the present model, the direct interactions with the jellium eld tend to dominate the closure and hence the relative importance of the bridge diagrams is reduced. In the remainder of this paper only RHNC results are presented unless otherwise indicated. The wall-ion distribution functions near the metal surface are plotted in Fig. 5 for four 0.25 M salt solutions. When the ions are smaller or comparable in size to a water molecule, as in the 0.25 M KF case, the distribution functions show a high concentration of anions near the surface. In the second and third layers, the anion concentration is lower than the bulk value. In contrast, for small cations a maximum in the density prole occurs at approximately one solvent diameter from contact and cations are very sparse in the contact layer. This arrangement of cations and anions is clearly evident for 0.25 M KF. As the anion diameter is increased as in the cases of Cl and I ), the wall-anion distribution functions become less structured with slight preferences shown for separations between the rst, second and third solvent layers. Apart from these weak structural features, for the larger anions the density proles are bulk-like except near contact where the density falls below the bulk value. As the cation diameter is increased (Na K Cs ), the wallcation distribution functions remain structured. However, the strong second-layer peak observed for small cations partially shifts to the rst layer, and for Cs the rst- and secondlayer peaks are roughly equal in height and area. If we assume that the ions do not greatly perturb the solvent packing near the surface perhaps true for smaller ions, then we can use the solvent orientational distributions discussed above to speculate about possible congurations near the surface and rationalize the observed surface-ion distributions. Possible congurations which would stablize the ion distributions and which are consistent with the solvent structure are shown in Fig. 6. Small anions are attracted to the surface by the jellium potential and can be efciently solvated by water molecules in the rst and second layers. This could account for the fact that uoride ions have a strong preference to be near the surface essentially located in the rst layer of solvent see Fig. 5. Larger anions experience a weaker attraction to the surface and have a very uniform distribution exhibiting little preference for any particular separation. Cations may be strongly solvated both near the surface or in the second solvent layer. For small cations, the strong water-ion attractions and the jellium-cation repul-

FIG. 5. The metal-cation ( ) and metal-anion - - - - pair distribution functions for several 0.25 M electrolyte solutions.

sion appear to combine such that the cations prefer the second solvent layer. The cation-water attraction and jelliumcation repulsion are both weakened for larger cations. As a result, the probability density in the contact layer increases. Metal-ion distribution functions using the jellium metal model have also been obtained for a simple dipolar solvent.22 The results found, particularly for smaller ions, differ signi-

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

rard et al.: Metal-aqueous electrolyte interface Be

4725

FIG. 6. An illustration of possible solvent and ion congurations near the jellium metal surface. Here the ions are assumed to be comparable in size to a water molecule, or smaller.

cantly from those for the water-like solvent. In the simple dipolar case, the solvent is strongly polarized by the jellium eld and the dipoles in the contact layer have a high probability of being oriented normal to the surface. In this case, the cations tend to be near the surface and the anions one solvent layer away such that they interact favorably with the negative and positive ends of the dipoles, respectively. For the present water-like model, the orientational structure of the solvent at the surface is not so simple and the preferred locations of the positive and negative ions are reversed. In both cases the solvent structure at the surface appears to strongly inuence the ion distributions. However, there is a difference in that for the simple dipole case the orientational order at the surface is largely determined by the direct jellium-dipole interactions, whereas, for the water-like model the waterwater interactions in the contact layer are clearly very important. Further insight into the nature and origin of the wall-ion distribution functions can be obtained by comparing results for different wall-ion interactions. This is done in Fig. 7 for 0.25 M KCl solution. The ion distributions at an inert surface are shown in Fig. 7a. In this case the jellium eld is absent and the ions prefer the bulk solution where they are better solvated.22 The jellium model creates a strong but shortranged eld which is attractive to anions and repulsive to cations.21,22 The ion-solvent interaction is also strong and long-ranged. Therefore, it is of interest to investigate the relative importance of the ion-metal and ion-solvent interactions in determining the metal-ion distribution functions. This can be done by calculating wall-ion distribution functions for a ctitious system where the jellium-ion interaction is removed from the metal-solution Hamiltonian. That is, Eq. 5 is replaced by u w 01

FIG. 7. Wall-ion pair distribution functions for 0.25 M KCl solution. RHNC results for an inert wall are shown in a, and for the ctitious case where the electrostatic metal-ion interaction is turned off in b. HNC results for the fully interacting metal surface are shown in c. The curves are as in Fig. 5.

, 0,

z z w d /2, z z w d /2,

10

but the jellium-solvent interaction Eq. 6 is retained in the Hamiltonian. The ion proles obtained are shown in Fig. 7b. We note that the K curve is signicantly altered particularly near contact and in fact it now closely resembles the true result for Cs compare with Fig. 5. This shows that for smaller cations the distribution functions are signicantly inuenced by the direct jellium-cation repulsion. However, the repulsive force is very short-ranged and for larger cations e.g., Cs ) the distribution function is altered only very near contact. The Cl prole remains fairly smooth in the ctitious case, but does differ a little from the true result see Fig. 5. These calculations show that for the water-like solvent the ion distributions are somewhat sensitive to the direct interaction with the jellium. This differs from the simple dipolar case where the ion proles are basically dictated by the solvent order and show little dependence on the direct jellium-ion interaction for both small and large ions.

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

4726

rard et al.: Metal-aqueous electrolyte interface Be

fers visibly from the net 0.25 M solution contribution. At 3.0 M, the potentials decay very quickly to zero as a result of ion screening. Again the individual solvent and ion contributions are large and signicant cancellation occurs. However, in this case the net solution contribution differs noticeably from the pure solvent result.
IV. SUMMARY AND CONCLUSIONS

FIG. 8. The metal-ion pair distribution functions for 3.0 M KCl solution. The curves are as in Fig. 5.

The ion distribution functions for 3.0 M KCl solution are plotted in Fig. 8. It can be seen that even at this concentration the ion proles are structurally very similar to the lower concentration results shown in Fig. 5. We do note that the peak at z * 1.5 in g w has increased signicantly in height showing an increased tendency for the positive ions to sit just outside the rst solvent layer. Using the various wall-particle distribution functions it is possible to calculate the potential drop, , across the wall-solution interface. This is dened as z z0 , 11

In this paper we have considered model aqueous electrolyte solutions in contact with a metal surface. The metal was taken to be a jellium slab and the electron density was obtained using density functional theory. The model electrolyte solution consisted of charged hard spheres in a hard multipolar water-like solvent. The properties of the solution where obtained using the RHNC approximation. Both phases were allowed to interact electrostatically and the DFT/RHNC equations were iterated until the electron density, the metalion and the metal-water distribution functions were fully self-consistent. The self-consistency of the present theory is a signicant improvement upon earlier work.

and results for both inert and jellium surfaces are given in Table III. The contributions from the solvent, s , ions, i , and, where appropriate, the jellium metal, j u jel/ e , are included in the table. We note that for the inert case the value obtained for the pure solvent is in good agreement with earlier work.28 For the inert surface it is also clear that although the solvent and ion contributions vary considerably, the net potential drop across the interface is virtually unchanged by the presence of ions. This observation persists even for 3.0 M KCl. At the jellium interface, the net solution contribution i.e., solvent plus ion to the potential increases by 50 mV with the addition of salt to the system. Again, this increase is nearly independent of ion size and salt concentration despite the fact that both the solvent and ion contributions show a strong dependence on these variables. The jellium contribution shows only small variations with ion size for 0.25 M electrolyte solutions. For 3.0 M KCl and for pure water, j increases by about 20 mV. The mean electrostatic potentials for 0.25 M and 3.0 M KCl solutions are plotted in Fig. 9 together with the individual solvent, ion and jellium contributions. The jellium potential decays rapidly in the uid region and is negligible for z * 0.6. At 0.25 M, the potentials due to the ions and solvent molecules are large but cancel in such as way that the net solution contribution is virtually indistinguishable from the pure solvent curve except at contact for clarity, the pure solvent result is plotted only inside of contact, where it dif-

FIG. 9. The self-consistent mean-eld electrostatic potential ( ) for a 0.25 M KCl, and b 3.0 M KCl in contact with a metal surface ( r s 2.65a o ). The gures include the separate contributions from the jellium - -, solvent and ions as well as the sum of the solvent and ion terms - - - -. The solvent potential in the absence of ions - - - is also included. In a, the dotted and dashed curves are indistinguishable outside of contact and only the dotted curve is shown.

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

rard et al.: Metal-aqueous electrolyte interface Be TABLE III. The potential drop, , in volts across jellium and inert surface-water interfaces including contributions from the solvent, s , ions, i , and jellium metal, j . All results are for solutions at 298 K. Unless indicated otherwise, all results were calculated using the RHNC theory. The dielectric constants, , of the bulk solution are included. Model pure water KF NaCl KCl KCl CsI Molarity 0.00 0.25 0.25 0.25 3.00 0.25

4727

85.92 73.65 75.45 75.48 44.18

s i 0.246 0.241 0.250 0.246 0.244 0.248

0.246 0.241 0.250 0.246 0.244 0.248

Inert surface 0.246 0.455 0.214 0.105 0.145 0.238 0.008 0.255 0.011 0.056 0.192 Jellium surface

vacuum pure water KF NaCl KCl KCl HNC KCl CsI

0.00 0.25 0.25 0.25 0.25 3.00 0.25

1.00 85.92 73.65 75.45 75.48 75.48 44.18

0.679 0.263 0.530 0.553 0.537 0.163 0.759

0.470 0.206 0.182 0.196 0.572 0.026

0.679 0.733 0.736 0.735 0.733 0.735 0.733

3.275 3.339 3.348 3.356 3.355 3.356 3.329 3.362

3.275 2.660 2.615 2.620 2.620 2.623 2.594 2.629

The structure of pure water at inert surfaces has been obtained using both simulation33 and the RHNC approximation28,34 employed in the present work. In the inert case, the average surface-water correlations have been interpreted with reference to ice-like structures. However, careful consideration of recent simulation results18 suggests that this description is perhaps not very appropriate. While the average correlations can be interpreted in this way, if one examines actual congurations near the surface one is struck by the fact that there is considerable disorder and well dened ice-like structures are not observed. In any case, for the metal model considered here the average distribution of water molecules near the surface is signicantly different from that found for inert surfaces. At contact, the most probable orientation of a water molecule has the HOH plane nearly at against the surface with the positive end of the dipole canted slightly outward. These structures appear to come mainly from interactions among water molecules in the relatively dense surface layer with the jellium eld acting to cant the dipole slightly away from the surface. However, the structure is not rigid and other congurations exist with signicant probability. There is also considerable hydrogen bonding between the rst and second water layers. A weakness of the present model is that it lacks the strong shortrange attractions found in quantum calculations involving water on mercury clusters.16 These interactions have been included in simulation studies17,18 and they result in a contact layer that is denser than that found here. However, the angular structure at the interface is similar to that obtained in the present calculations.18 The ion distribution functions are largely determined by the solvent ordering. However, unlike the simple dipolar case, the inuence of the direct metal-ion interaction is not negligible in the aqueous system and its effect depends on the ion diameter. The ion distribution functions indicate a loosely layered structure. For relatively small ion diameters,

the anions tend to reside adjacent to the surface and the cations prefer the second solvent layer. This contrasts with the dipolar solvent case where the opposite order i.e., anions in the second layer prevails and can be rationalized by considering the solvent structure near the surface. As the anion diameter is increased, the anion distributions tend to lose their structure and suggest a relatively smooth bulk-like density of anions across the interface. As the cation diameter is increased the probability of nding a cation near contact increases at the expense of the cation density in the second solvent layer. Calculations for a ctitious model where the direct metal-ion interactions were turned off but the metalwater interactions retained, indicated that this occurs because the metal-cation repulsion becomes less important with increasing cation size. Thus both the solvent structure and direct interactions with the jellium play a part in determining the metal-ion distribution functions. We have also calculated the potential drop across the metal-solution interface and have examined the various contributions. It was found that the jellium contribution is not greatly inuenced by the presence of ions and remains near the value for the jellium-pure-solvent interface. Further, while the separate contributions from the ions and solvent are sensitive to ion size and concentration, there are cancellations such that the net solution contribution is virtually independent of these parameters. This observation is consistent with the dipolar solvent case, and emphasizes again that the ion and solvent potentials are very strongly coupled and should never be treated as independent quantities in double layer theories. Finally, we note that the DFT/RHNC approach to metalaqueous electrolyte solution interfaces is not limited to the present model. It can be extended to include more realistic metal-water interactions and also more realistic metal-ion interactions if and when they are available. Work in this direction is planned.

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

4728

rard et al.: Metal-aqueous electrolyte interface Be cker, Gurskii, and K. Heinzinger, J. Phys. Chem. 100, 14 969 1996; J. Bo R. R. Nazmutdinov, E. Spohr, and K. Heinzinger, Surf. Sci. 335, 372 1995. 18 J. C. Shelley, G. M. Torrie, D. R. Berard, and G. N. Patey, J. Chem. Phys. to be published. 19 J. D. Head and S. J. Silva, J. Chem. Phys. 104, 3244 1996. 20 J.C. Shelley and G.N. Patey, J. Chem. Phys. 100, 8265 1994. 21 D. R. Berard, M. Kinoshita, X. Ye, and G. N. Patey, J. Chem. Phys. 101, 6271 1994. 22 D. R. Berard, M. Kinoshita, X. Ye, and G. N. Patey, J. Chem. Phys. 102, 1024 1995. 23 M. Kinoshita and D. R. Berard, J. Comput. Phys. 124, 230 1996. 24 P. G. Kusalik and G. N. Patey, J. Chem. Phys. 88, 7715 1988. In a coordinate frame where the dipole of the water molecule is directed along the z axis and OH bonds are in the xz plane, the relationships between the nonzero multipole moments of water in spherical and Cartesian tensor 0 0 2 form are as follows: Q 1 * a z , Q 2 * a zz / d s , Q 2 * a ( xx 0 2 2 2 3 * a zzz / d s and Q 3 * 3 a ( xxz y yz )/ 30d s y y )/ 6 d s , Q , where z is the dipole moment and and are the Cartesian components of the quadrupole and octupole tensors, respectively. Also, 3 a 1/d s k BT and k BT is the Boltzmann constant times the temperature. 25 J. Hansen and I. McDonald, Theory of Simple Liquids, 2nd ed. Academic, London, 1986. 26 J. A. Alonso and N. H. March, Electrons in Metals and Alloys Academic, New York, 1989. 27 G. M. Torrie, P. G. Kusalik, and G. N. Patey, J. Chem. Phys. 88, 7826 1988; 89, 3285 1988; 90, 4513 1989; 91, 6367 1989. 28 G. M. Torrie and G. N. Patey, J. Phys. Chem. 97, 12909 1993. 29 M. Kinoshita and M. Harada, Mol. Phys. 81, 1473 1994. 30 D. R. Berard and G. N. Patey, J. Chem. Phys. 95, 5281 1991. 31 L. Blum and A. J. Torruella, J. Chem. Phys. 56, 303 1972. 32 D. Henderson and M. Plischke, Proc. R. Soc. London Ser. A 400, 163 1985. 33 C. Y. Lee, J. A. McCammon, and P. J. Rossky, J. Chem. Phys. 80, 4448 1984. 34 G.N. Patey and G.M. Torrie, Chem. Script. 29A, 39 1989.

ACKNOWLEDGMENTS

The authors thank John Shelley for many useful discussions. The nancial support of the National Science and Engineering Research Council of Canada is gratefully acknowledged.
J. D. Porter and A. S. Zinn, J. Phys. Chem. 97, 1190 1993. M. F. Toney, J. N. Howard, J. Richer, G. L. Borges, J. G. Gordon, O. R. Melroy, D. G. Wiesler, D. Yee, and L. B. Sorensen, Nature London 368, 444 1994. 3 E. Spohr and K. Heinzinger, Ber. Bunsenges. Phys. Chem. 92, 1358 1988. 4 E. Spohr, J. Phys. Chem. 93, 6171 1989. 5 K. J. Schweighofer, X. Xia, and M. L. Berkowitz, Langmuir 12, 3747 1996. 6 X. Xia and M. L. Berkowitz, Phys. Rev. Lett. 74, 3193 1995. 7 K. Forster, K. Raghavan, and M. Berkowitz, Chem. Phys. Lett. 162, 32 1989; K. Raghavan, K. Forster, and M. Berkowitz, Chem. Phys. Lett. 177, 426 1991; K. Raghavan, K. Forster, K. Motakabbir, and M. Berkowitz, J. Chem. Phys. 94, 2110 1991. 8 S. Zhu and M. R. Philpott, J. Chem. Phys. 100, 6961 1994. 9 J. N. Glosli and M. R. Philpott, J. Chem. Phys. 96, 6962 1992; 98, 9995 1993. 10 J. I. Siepmann and M. Sprik, Surf. Sci. 279, L185 1992. 11 G. Nagy and K. Heinzinger, J. Electroanal. Chem. 296, 549 1990. 12 M. R. Philpott, J. N. Glosli, and S. Zhu, Surf. Sci. 335, 422 1995. 13 J. N. Glosli and M. R. Philpott, Electrochim. Acta 41, 2145 1996. 14 M. R. Philpott and J. N. Glosli, Chem. Phys. 198, 53 1995. 15 th, and K. Heinzinger, Electrochim. Acta 41, 2131 1996. E. Spohr, G. To 16 R. R. Nazmutdinov, M. Probst, and K. Heinzinger J. Electroanal. Chem. 369, 227 1994. 17 E. Spohr and K. Heinzinger, Chem. Phys. Lett. 123, 218 1986; E. Spohr, ibid. 207, 214 1993; K. Heinzinger, Fluid Equilibria 104, 277 1995; O. Pecina, W. Schmickler, and E. Spohr, J. Electroanal. Chem. 394, 29 th, E. Spohr, and K. Heinzinger, Chem. Phys. 200, 347 1995; G. To th and K. Heinzinger, ibid. 245, 48 1995; J. Bo cker, Z. 1995; G. To
1 2

J. Chem. Phys., Vol. 107, No. 12, 22 September 1997

Das könnte Ihnen auch gefallen