You are on page 1of 7

pubs.acs.

org/Langmuir
© 2010 American Chemical Society

Charge Reversal of Surfaces in Divalent Electrolytes:


The Role of Ionic Dispersion Interactions
Drew F. Parsons* and Barry W. Ninham
Research School of Physical Sciences and Engineering, Australian National University,
Canberra, ACT 0200, Australia

Received October 30, 2009. Revised Manuscript Received January 7, 2010

Surface potentials of alkali earth nitrates at a mica surface are calculated using a modified Poisson-Boltzmann
approach that includes nonelectrostatic ion-surface dispersion interactions. New ab initio dynamic polarizabilities are
used to determine dispersion interactions. A hydration model describing the hydration shell of cations is presented.
Excellent agreement with experiment is achieved, including charge reversal at high electrolyte concentration without the
need for site binding models. This suggests that specific ionic dispersion forces provide the mechanism for ion surface
binding. An asymptotic surface potential is found in the limit of very high concentration. A Hofmeister series is
predicted according to the strength of charge reversal, with Mg > Ca > Sr > Ba. The ion-surface dispersion
adsorption energies of hydrated ions appear to explain the apparent repulsive secondary hydration forces observed
experimentally between mica surfaces when taken with a surface hydration layer.

Introduction especially when complications due to surfactants are involved.8


The measurement of a surface potential provides the standard But more interesting is the appearance of charge reversal as a
characterization of the surface properties of colloids. The inter- function of electrolyte concentration alone. In contradiction to
pretation of surface potentials via double layer theory allows the simple electrostatic theory, at sufficiently high concentration,
relationship between surface charge and surface potential to be some electrolytes can induce a reversal in the surface polarity.
determined directly. Usually, the connection is made by the This can occur in 1:1 electrolytes,9-11 but it is more commonly
electrostatic Poisson-Boltzmann model. The relationship is observed with multivalent ions, both cations8,12,13 and anions.7
“direct” in the sense that if the surface charge is positive then This kind of charge reversal has been interpreted theoretically
the surface potential is positive. Conversely, if the surface charge by invoking site binding models.14-18 In these models, in an
is negative then the surface potential is negative. Screening by an extension of the original charge regulation model,19 the intrinsic
electrolyte at higher concentrations implies that that the magni- surface charge σ0 is modified by a surface layer of adsorbed
tude of the surface potential is reduced. In a simple electrostatic (“bound”) counterion σc. The amount of σc is established by a
theory, counterions will be attracted to the surface, effectively postulated equilibrium between the bulk counterion C in solution
neutralizing the surface charge so that the overall electrostatic and SC, bound to a surface site S,
potential decreases. However, the counterion charge at the sur-
face may never exceed the original bare surface charge. It can do S þ C h SC ð1Þ
no more than neutralize it. As the electrolyte concentration
The equilibrium constant Kc for this system is determined by a
increases, the magnitude of the surface potential will reduce
fitted surface binding energy. The counterion surface charge σc
asymptotically to zero with a purely electrostatic interpretation.
is then obtained from the adsorbed ion concentration [SC].
In the absence of additional effects such as chemisorption or
The total surface charge, σ = σ0 þ σc, may change polarity if
physisorption of ions to the surface, or ion-ion spatial correla-
|σc| > |σ0|. Essentially, the site binding model enables a larger
tions,1,2 the surface potential may never cross zero. Physisorption
via ion-surface dispersion interactions is addressed in this paper,
(8) Alkan, M.; Karadas-, M.; Dogan, M.; Demirbas- , O. € Colloids Surf., A 2005,
while ion-ion correlations are not.
259, 155–166.
In real systems, reversal of the surface polarity is indeed found. (9) Mietta, F.; Chassagne, C.; Winterwerp, J. J. Colloid Interface Sci. 2009, 336,
Charge reversal is obtained trivially by varying pH in biological 134–141.
(10) Parsons, D. F.; Ninham, B. W. J. Phys. Chem. A 2009, 113, 1141–1150.
systems due to protonation of acidic carboxylate and basic amine (11) Koelsch, P.; Viswanath, P.; Motschmann, H.; Shapovalov, V.; Brezesinski,
groups.3,4 At mineral surfaces, charge reversal is more subtle. pH G.; M€ohwald, H.; Horinek, D.; Netz, R. R.; Giewekemeyer, K.; Salditt, T.;
charge reversal effects may appear depending on the mineral,5-7 Schollmeyer, H.; von Klitzing, R.; Daillant, J.; Guenoun, P. Colloids Surf., A 2007,
303, 110–136.
(12) Lyons, J. S.; Furlong, D. N.; Healy, T. W. Aust. J. Chem. 1981, 34, 1177–
*To whom correspondence should be addressed. 1187.
(1) Greberg, H.; Kjellander, R. J. Chem. Phys. 1998, 108, 2940–2953. (13) Chakir, A.; Bessiere, J.; Kacemi, K. E.; Marouf, B. J. Hazard. Mater. 2002,

(2) Jimenez-Angeles, F.; Lozada-Cassou, M. J. Chem. Phys. 2008, 128, 174701. 95, 29–46.
(3) Molina-Bolı́var, J. A.; Galisteo-Gonzalez, F.; Hidalgo-Alvarez, R. Colloids (14) Scales, P. J.; Grieser, F.; Healy, T. W. Langmuir 1990, 6, 582–589.
Surf., B 2001, 21, 125–135. (15) Kekicheff, P.; Marcelja, S.; Senden, T. J.; Shubin, V. E. J. Chem. Phys.
(4) Fisher, D. J.; Richmond, D. V. J. Gen. Microbiol. 1969, 57, 51–60. 1993, 99, 6098–6113.
(5) Vane, L. M.; Zang, G. M. J. Hazard. Mater. 1997, 55, 1–22. (16) Pashley, R. M. J. Colloid Interface Sci. 1981, 83, 531–546.
(6) Alkan, M.; Demirbas, O.;€ Dogan, M. Microporous Mesoporous Mater. 2005, (17) Pashley, R. M.; Israelachvili, J. N. J. Colloid Interface Sci. 1984, 97, 446–
83, 51–59. 455.
(7) Marouf, R.; Marouf-Khelifa, K.; Schott, J.; Khelifa, A. Microporous (18) Tatulian, S. A. J. Colloid Interface Sci. 1995, 175, 131–137.
Mesoporous Mater. 2009, 122, 99–104. (19) Ninham, B. W.; Parsegian, V. A. J. Theor. Biol. 1971, 31, 405–428.

6430 DOI: 10.1021/la9041265 Published on Web 01/29/2010 Langmuir 2010, 26(9), 6430–6436
Parsons and Ninham Article

concentration of counterion to be found at the surface than would In previous papers, the dynamic polarizabilities underpinning a
be the case due to simple electrostatic interactions alone. This theory that includes dispersion forces have applied a single mode
results in charge reversal rather than charge neutralization. The model of the polarizabilities. The modal frequencies were inferred
success of the site binding model hinges on the reality of the from ionization potentials. Compared to ab initio polarizabilities,
surface equilibrium constant Kc. In some cases, it is simply treated the single mode model badly underestimates polarizabilities,
as a fitting parameter to match experiment.16,17 In other works, it especially for halide anions and at the optical/UV frequencies
is simply identified with the counterion concentration at the relevant to dispersion forces.10,21 We have recently reported
surface.15 ab initio polarizabilities for monovalent monatomic ions.35 We
The role of nonelectrostatic ionic forces, that is, quantum here report the ab initio polarizabilities for multivalent mona-
mechanically based van der Waals or dispersion forces, in tomic cations and for the polyatomic water molecule and nitrate
explaining ion specific effects has become apparent over the past anion. Water and nitrate are in reality nonspherical; however, for
decade or so.20 These ion specific forces are ignored in classical simplicity, we apply a mean spherical approximation to them.
electrostatic theory, whether or not the theory is extended to
include “hydration”. Their inclusion goes some way toward
Theoretical Model
explaining deviations from pure electrostatic theory both in bulk
solution, for example, self-energies and activity coefficients,21-23 The surface potential may be estimated in a Poisson-
and at surfaces, for example, ion specific direct force measure- Boltzmann approach by calculating the electrostatic potential
ments.24-29 Ion specificity (Hofmeister effects) in this theory ψ(z) due to the combination of surface charge and electrolyte
occur due to dispersion forces. The specificity arises due to the charge in solution. The surface potential is the value of ψ at z = 0.
different dynamic polarizabilities R(iω) of the ions that determine ψ(z) is obtained by solving the Poisson equation
their interactions. At the same time, the importance of ionic and
molecular polarizabilities has also become clear in molecular d2 ψ X
dynamics simulations. Polarizable water models such as q- ε0 ε 2
¼ - qi ci ðzÞ ð2Þ
dz
TIP4P/F30 or TIP4P-FQ31,32 reproduce bulk properties such as i

water dipole, temperature-dependent density, and dielectric con- where ci(z) is the concentration profile of ion i and qi is its charge. ε
stant. Likewise, polarizable potential models for ions show a is the dielectric constant of the solvent (water, ε = 78.36), and ε0 is
tremendous influence on the surface affinities of ions at an the permittivity of a vacuum. We have assumed a planar surface
interface predicted by computer simulation,32-34 with ionic here. The boundary condition for the Poisson equation is related
polarizability enabling positive adsorption of ions to the air- to the surface charge. We adopt a boundary condition of constant
water interface. surface charge σ0 at z = 0,
In this paper, we explore how ion-surface dispersion forces
may explain the phenomenon of surface charge reversal (or more dψ σ0
precisely, reversal in the polarity of the surface potential). The ¼ - ð3Þ
dz ε0 ε
nonelectrostatic dispersion potentials appear naturally alongside
electrostatic interactions in a modified Poisson-Boltzmann de- Surface charge will vary with pH due to protonation of surface
scription of the electrolyte. We obtain charge reversal without charge sites. By adopting constant surface charge, we imply that
having to invoke a site binding model of unknown origin. We our calculations are conducted at fixed pH. The magnitude of our
therefore argue that dispersion forces provide the underlying surface charge is calibrated against experimental surface poten-
mechanism by which the “site binding model” operates. A tials measured15 at pH 5.9. The fitted surface charge is taken to lie
correspondence may be established between our theory and the at z = 0, and ions are permitted to occupy the volume where z g 0,
site binding model by identifying the ion dispersion interaction at right up to the surface. Some implications of finite size effects,
the surface with the surface binding energy, rather than treating it with the ion radius providing a distance of closest approach to the
as a fitting parameter. The surface binding model may be thought surface, are discussed at the end of this paper.
of as a special case of the general theory of nonelectrostatic Ion concentrations are given as Boltzmann factors determined
interactions which considers nonelectrostatic interactions at the by the free energy of the ion,
interface alone. The general theory also includes nonelectrostatic
ionic forces in the bulk solution away from the surface. ci ðzÞ ¼ ci0 exp½ - Fi ðzÞ=kT ð4Þ

(20) Ninham, B.; Yaminsky, V. Langmuir 1997, 13, 2097–2108.


ci0 is the bulk concentration of ion i. We take the free energy to be
(21) Parsons, D. F.; Deniz, V.; Ninham, B. W. Colloids Surf., A 2009, 343, 57–63. the sum of the electrostatic interaction energy qiψ of the ion inside
(22) Bostr€om, M.; Ninham, B. W. Langmuir 2004, 20, 7569–7574. the electrostatic potential, plus the nonelectrostatic ion-surface
(23) Bostrom, M.; Ninham, B. J. Phys. Chem. B 2004, 108, 12593–12595.
(24) Bostrom, M.; Williams, D.; Ninham, B. Langmuir 2002, 18, 6010–6014. dispersion interactions Ui. That is,
(25) Tavares, F.; Bratko, D.; Blanch, H.; Prausnitz, J. J. Phys. Chem. B 2004,
108, 9228–9235.
(26) Bostr€om, M.; Ninham, B. W. Biophys. Chem. 2005, 114, 95–101.
Fi ðzÞ ¼ qi ψðzÞ þ Ui ðzÞ ð5Þ
(27) Bostrom, M.; Tavares, F. W.; Finet, S.; Skouri-Panet, F.; Tardieu, A.;
Ninham, B. W. Biophys. Chem. 2005, 117, 217–224. The conventional application of the Poisson-Boltzmann
(28) Bostr€om, M.; Deniz, V.; Franks, G. V.; Ninham, B. W. Adv. Colloid model at a surface, as used here, does not include ion-ion
Interface Sci. 2006, 123-126, 5–15.
(29) Bostr€om, M.; Lima, E. R. A.; Biscaia, E. C.; Tavares, F. W.; Nostro, P. L.; interactions. These could be modeled within the same PB frame-
Parsons, D. F.; Deniz, V.; Ninham, B. W. J. Phys. Chem. B 2009, 113, 8124–8127. work by adding an additional term to eq 5, which would integrate
(30) Habershon, S.; Markland, T. E.; Manolopoulos, D. E. J. Chem. Phys. 2009, the interactions of ion i with other ions j in the volume between
131, 024501.
(31) Rick, S. W.; Stuart, S. J.; Berne, B. J. J. Chem. Phys. 1994, 101, 6141–6156. surfaces.36 Alternatively ion-ion interactions and ion-ion
(32) Stuart, S. J.; Berne, B. J. J. Phys. Chem. 1996, 100, 11934–11943.
(33) Vrbka, L.; Mucha, M.; Minofar, B.; Jungwirth, P.; Brown, E. C.; Tobias,
D. J. Curr. Opin. Colloid Interface Sci. 2004, 9, 67–73. (35) Parsons, D. F.; Ninham, B. W. Langmuir 2010, 26, 1816-1823.
(34) Warren, G. L.; Patel, S. J. Phys. Chem. C 2008, 112, 7455–7467. (36) Burak, Y.; Andelman, D. J. Chem. Phys. 2001, 114, 3271–3283.

Langmuir 2010, 26(9), 6430–6436 DOI: 10.1021/la9041265 6431


Article Parsons and Ninham

correlations (including finite ion size effects) could be included via mined by Dagastine et al.43 and take mica as our surface, using the
integral equation methods such the hypernetted chain (HNC) dielectric model of Chan and Richmond.44 The excess polariz-
model.37,38 Ion-ion correlation effects become particularly sig- ability Ri* refers to the effective polarizability of the ion inside the
nificant at very high concentrations, but at the same considerably depolarizing medium of the solvent, and it is explained in more
increase the computational cost of the calculation. At the same detail below.
time, ion-surface dispersion interactions, which we do include The water dielectric data were obtained in numerical form from
via Ui, should be more significant than ion-ion dispersion Dagastine et al.,43 with the static dielectric constant45 taken to be
interactions, since the former decay with 1/z3 (see eq 6 below) ε = 78.36. The dielectric function for mica follows the model44
while the latter decay more quickly,21 with 1/r6. Hence, while a
superior model would be constructed by including ion-ion size εuv - 1
correlations, any discrepancy due to the neglect of ion-ion εðiωÞ ¼ 1 þ ð10Þ
1 þ ðω=ω0 Þ2 þ ω=g
dispersion interactions should be relatively mild.
In classical electrostatic Derjaguin-Landau-Verwey- with εuv = 2.45, ω0 = 2.38  1016 rad s-1, and g = 4.89  1016
Overbeek (DLVO) theory, Ui = 0; that is, ionic dispersion forces rad s-1. At zero frequency, the static dielectric constant of mica
are neglected. In fact, the classical DLVO theory, that handles was set to ε = 5.4.
electrostatic forces in a nonlinear theory and electrodynamic Ab Initio Polarizabilities. Dynamic polarizabilities were
fluctuation (dispersion) forces not at all or separately in a linear calculated using quantum chemical software Molpro.46 Electron
theory, can be proven to be physically inconsistent.20 However, correlation is generally significant, especially for anions35 where
ionic dispersion forces lie at the heart of our model, they are the correction due to electron correlation may be as much as 20-
treated nonlinearly via eq 4, and they enable charge reversal to 50%. Electron correlation was handled using the coupled-cluster
occur. single and double (CCSD) model. A total of 2100 frequency
Ion-Surface Dispersion Interactions. Ionic dispersion in- points were calculated for each ion, from 0 to 1017 Hz, corre-
teractions are driven by the dynamic polarizability Ri(iω) of an sponding to far UV/soft X-ray frequencies (around 5 nm in
ion, with the bulk of the interaction coming from the polariz- wavelength). Dunning’s augmented correlation consistent basis
ability at optical/UV frequencies. We adopt a semiclassical model sets47 were used for smaller ions, nitrate, and water, namely, aug-
of polarizability which includes finite ion size.10,35,39 The ion cc-pCVQZ for Be2þ and Mg2þ, aug-cc-pwCVQZ for nitrate, and
polarizability is taken to be spatially spread, following the ion’s aug-cc-pCV5Z for water. Effective core potentials (pseudopoten-
electron density, in a spherical Gaussian cloud exp(-r2/ai2)with tials) were applied to the heavier alkali earth ions Ca2þ through
Gaussian radius ai. The resulting ion-surface dispersion interac- Ra2þ, using the Stuttgart ECPnMDF family of basis sets.48
tion is given by35,39-42 Although we use the full ab initio data in this paper, it may be
useful in other applications to have a parametrized analytic model
Bi hi ðzÞ of the dynamic polarizability,
Ui ðzÞ ¼ ð6Þ
z3
X fj
where RðiωÞ ¼ R0 ð11Þ
j 1 þ ðω=ωj Þ2
" # ! " #  
2z 2z2 z2 4z4 z R0 is the static polarizability, and fj and ωj are the weight and
hi ðzÞ ¼ 1 þ pffiffiffi - 1 exp - - 1 þ erfc
πai ai 2 ai 2 ai 4 ai modal frequency, respectively, of each mode contributing to the
total dynamic polarizability. We have shown elsewhere35 that a
ð7Þ
single mode model provides a poor estimate of the polarizability,
The surface dispersion coefficient Bi is calculated from the excess particularly when the single modal frequency is derived from the
dynamic polarizability Ri*(iω) of the ion, multiplied by the ionization potential of an ion. In the latter case, the modal
difference in the dielectric responses of surface and solvent and frequency is too low such that the frequency dependence of the
summed over imaginary frequencies iωn, where ωn = 2πnkT/p: polarizability decays too rapidly. The strength of dispersion
energies is therefore underestimated. For anions, the error in
 the single mode ionization potential model is particularly stark,
kT X0 Ri ðiωn Þ Δðiωn Þ
B ¼ ð8Þ around 70%. It is less disastrous for metal cations at 20% error. A
2 n εw ðiωn Þ multimode fit, on the other hand, is accurate to within 0.03% over
with the entire frequency range. The multimode parametrization of the
ab initio polarizabilities is given in Table 1 for the alkali earth
εw ðiωÞ - εs ðiωÞ cations together with the nitrate anion and the neutral water
ΔðiωÞ ¼ ð9Þ
εw ðiωÞ þ εs ðiωÞ molecule. In the case of the latter polyatomic species, their
geometry was first optimized.
The prime (0 ) next to the summation symbol (Σ0 ) in eq 8 means Ab initio calculations do provide the full anisotropic polariz-
that the n = 0 term is taken with a factor 1/2. εw(iω) and εs(iω) are ability tensor. Anisotropic effects are likely to be interesting for
the dielectric functions of solvent (water) and surface, respec-
tively. We use the water dielectric susceptibility function deter- (43) Dagastine, R. R.; Prieve, D. C.; White, L. R. J. Colloid Interface Sci. 2000,
231, 351–358.
(44) Chan, D.; Richmond, P. Proc. R. Soc. London, Ser. A 1977, 353, 163–176.
(37) Kjellander, R. J. Chem. Phys. 1988, 88, 7129–7137. (45) CRC Handbook of Chemistry and Physics, 84th ed.; Lide, D. R., Ed.; CRC
(38) Wernersson, E.; Kjellander, R. J. Phys. Chem. B 2007, 111, 14279–14284. Press: Boca Raton, FL, 2004.
(39) Mahanty, J.; Ninham, B. W. Faraday Discuss. Chem. Soc. 1975, 59, 13–21. (46) Werner, H.-J., et al. MOLPRO, a package of ab initio programs, version
(40) Mahanty, J.; Ninham, B. W. Dispersion Forces; Academic Press: London, 2008.1; University College Cardiff Consultants Limited: Wales, UK, 2008; see http://
1976. www.molpro.net.
(41) Mahanty, J.; Ninham, B. W. J. Chem. Phys. 1973, 59, 6157–6162. (47) Woon, D. E.; Thom, H.; Dunning, J. J. Chem. Phys. 1995, 103, 4572–4585.
(42) Mahanty, J.; Ninham, B. J. Chem. Soc., Faraday Trans. 2 1974, 70, 637–650. (48) Lim, I. S.; Stoll, H.; Schwerdtfeger, P. J. Chem. Phys. 2006, 124, 034107.

6432 DOI: 10.1021/la9041265 Langmuir 2010, 26(9), 6430–6436


Parsons and Ninham Article

Table 1. Static Polarizabilities, Weights, and Characteristic Frequencies for Multimode Decompositions of the Dynamic Polarizabilities of
Multivalent Cations, Nitrate Anion, and Neutral Water (See eq 11)a
mode 1 mode 2 mode 3 mode 4 mode 5
ion R0 (Å3) f1 ω1(a.u.) f2 ω2 (a.u.) f3 ω3 (a.u.) f4 ω4 (a.u.) f5 ω5 (a.u.)

Be2þ 0.0077 0.8291 4.8511 0.1710 9.9219


Mg2þ 0.0704 0.4259 4.1734 0.4881 2.3315 0.0860 8.9125

Ca 0.4692 0.5426 1.2095 0.4327 1.5398 0.0201 3.1135 0.0047 7.7725

Sr 0.8684 0.5510 0.9755 0.3575 1.1576 0.0003 1.4005 0.0912 3.3075
Ba2þ 1.5957 0.6963 0.7937 0.2718 1.0529 0.0306 1.8000 0.0012 5.5701
Ra2þ 2.0192 0.6052 0.7127 0.3531 0.9763 0.0408 1.7664 0.0010 5.8336
nitrate 5.0215 0.3418 0.2536 0.3863 0.5638 0.2270 1.0631 0.0429 2.4249 0.0018 9.3252
water 1.4255 0.0204 0.1906 0.5479 0.4669 0.4008 1.0286 0.0305 2.9532 0.0005 19.6298
a
1 a.u. = 6.57968  1015 Hz.

nonspherical ions such as the flat nitrate ion, especially in regard Table 2. Hydration Numbers56 nw, Gaussian Radii10 a, and Dispersion
to interactions with a flat interface. However, the model of B Coefficients of Hydrated Cations and Nitrate at Mica Surface
dispersion forces applied here (eqs 6 and 7) has assumed that ion nw a (Å) B (10-50Jm3)
ions are spherical. The dispersion self-energies of anisotropic
ions21 and the ion-ion interactions of point charges with aniso- Mg2þ 10 1.54 -3.69
tropic polarizabilities49,50 have been described, but a model of Ca2þ 7 1.83 -3.74
Sr 2þ
6a
2.00 -3.49
ion-surface dispersion energies for nonspherical ions with finite Ba2þ 5 2.25 -2.91
size has not yet been developed. We therefore apply a spherical nitrate n/a 2.01 -1.08
approximation to water and nitrate using mean isotropic polar- a
The hydration number of Sr2þ is the average of the Marcus56 values
izabilities. for Ca2þ and Ba2þ.
Polarizabilities in Aqueous Solution. The effect of the water
microwave frequencies, are largely negligible. Thus, to a reason-
medium on ion polarizabilities is handled in two ways. We
able approximation, the polarizability of the hydration shell is
distinguish between the general effect of the bulk water medium
determined simply by the density effect. That is, the polarizability
surrounding the ion and the specific effect of the water molecules
of the hydrated ion is taken to be the sum of the polarizability Ri of
in the tightly held hydration layer.
the ion itself plus the polarizability Rw of the water molecules in
Hydrated Ion. In the case of a tightly bound hydration layer,
the hydration shell (nw in number). The hydrated polarizability of
the hydration water molecules are not quickly exchanged with the
a cosmotropic ion is then
bulk medium but are instead carried with the ion. In this case, the
hydration water molecules behave differently from water in the Rhyd ðiωÞ ¼ Ri ðiωÞ þ nw Rw ðiωÞ ð12Þ
bulk medium, partly due to their specific orientation around the
ion and partly due to the density of hydration shell, which will be In this paper, we take the number of water molecules nw to be
different from that of bulk water. This situation applies to the so- given by experimental hydration numbers.56 The hydration
called cosmotropic ions51-53 such as Liþ and multivalent ions number of Sr2þ is not reported by Marcus,56 whose hydration
Mg2þ, Ca2þ, which are characterized by strong electric fields due numbers we used for other ions. Experimental estimates of
to either high charge or small size. hydration numbers vary57 but generally place the hydration
Effectively, the hydration water molecules become part of the number of Sr2þ between that of Ca2þ and Ba2þ (but closer to
ion in such a way that the cosmotropic ion must be explicitly the former). For consistency with the Marcus hydration numbers
treated as a hydrated ion. This means that the polarizability of the of Ca2þ and Ba2þ (7 and 5, respectively), we therefore assign an
hydration water molecules must be added to the polarizability of hydration number of 6 to Sr2þ.
the ion and the size of the ion must be increased by the width of the Hydrated ion radii are taken from Parsons and Ninham,10 with
hydration shell. These two hydration effects have competing the width of the hydration shell being added to the intrinsic hard
implications for the overall dispersion interactions of the hydrated sphere radius of the ion. The hydration shell width is taken to be
ion. The greater total polarizability tends to strengthen dispersion one water radius, 1.14 Å (hard sphere radius). Hydration numbers
energies, while the larger hydrated ion size tends to weaken them. and Gaussian radii of the hydrated ions are given in Table 2. We
Applied to monovalent lithium and sodium ions, this hydration also list in Table 2 the dispersion B coefficients of the ions at mica
model has found success in explaining ion specificity at silica surface.
surfaces54 and in modeling the reversal of the Hofmeister series In accordance with Collins’ rule of matching water affinities,51,53
observed experimentally at silica compared to alumina surfaces.55 the procedure for the tightly bound hydration layer described here
The dominant component of dispersion energies arises from is applied only to cosmotropic ions.21 Chaotropic ions such as
the polarizability at optical/UV frequencies. Therefore, orienta- nitrate are considered to have a weakly bound hydration layer. It
tional effects of the hydration water molecules, corresponding to seems reasonable to propose that hydration water molecules of
chaotropic ions are readily exchanged with the bulk medium. In
this case, the hydration shell does not need to be considered
(49) Imura, H.; Okano, K. J. Chem. Phys. 1973, 58, 2763–2776.
(50) Barash, Y. S.; Ginzburg, V. L. Sov. Phys. Usp. 1984, 27, 467–491. explicitly. Rather, the hydration water molecules of chaotropic
(51) Collins, K. Biophys. Chem. 2006, 119, 271–281. ions are included within the normal bulk dielectric response of the
(52) Collins, K. D. Methods 2004, 34, 300–311. medium.
(53) Collins, K. Biophys. J. 1997, 72, 65–76.
(54) Salis, A.; Parsons, D. F.; Bostr€om, M.; Medda, L.; Barse, B.; Ninham, B. Bulk Depolarization: Excess Polarizability. When an ex-
W.; Monduzzi, M. Langmuir, published online October 15, 2009, http://pubs.acs.org/ ternal electric field polarizes an ion in solution, it also polarizes the
doi/abs/10.1021/la902721a.
(55) Parsons, D. F.; Bostr€om, M.; Maceina, T. J.; Salis, A.; Ninham, B. W.
Langmuir, published online November 4 2009, http://pubs.acs.org/doi/abs/10.1021/ (56) Marcus, Y. Pure Appl. Chem. 1987, 59, 1093–1101.
la903061h. (57) Hinton, J. F.; Amis, E. S. Chem. Rev. 1971, 71, 627–674.

Langmuir 2010, 26(9), 6430–6436 DOI: 10.1021/la9041265 6433


Article Parsons and Ninham

medium surrounding that ion. The polarized medium exerts a


depolarization field so that the local field experienced by the ion
differs from the external field. Treating the polarizable ion as a
dielectric sphere with permittivity εi, the polarization response of
the ion in solution may be described by an effective polariz-
ability35,58,59 R*,
i

 3Vi ½εi ðiωÞ - εw ðiωÞ


Ri ðiωÞ ¼ ð13Þ
4π ½εi ðiωÞ þ 2εw ðiωÞ

where εw is the dielectric function of the medium (water).


We describe the ionic permittivity via the simple relation εi =
1 þ 4πniRi = 1 þ 4πRi(iω)/Vi, where Vi is the ionic volume.10 This
simple formula is more robust to errors in the magnitude of the
ionic volume than the Clausius-Mossotti relation suggested Figure 1. Surface potential of calcium nitrate solution at mica
elsewhere.59,60 The latter may yield unphysical negative ionic surface as a function of electrolyte concentration. Experimental
dielectric values if the ionic volume has been underestimated. data15 (points) are compared against the theoretical curve (solid
The effective polarizability R*i obtained in this way is com- line).
monly referred to as the “excess polarizability”, since it is related
to the difference in the polarizability responses of ion and medium
(the “εi - εw” term in eq 13).
Note that the solvent has two effects on cosmotropic ions. First
the polarizability Rhyd of the hydrated ion is determined, that is,
the polarizability (in vacuum) of ion plus tightly held hydration
shell, eq 12. Then that hydrated ion polarizability is used to
determine the excess polarizability of the hydrated ion (eq 13).
The excess polarizability of a cosmotropic ion is not taken directly
from the ion’s intrinsic polarizability. In the case of chaotropic
ions, on the other hand, the excess polarizability is derived directly
from the intrinsic polarizability of the ion.

Results
Surface Potentials at a Mica Surface. The surface potential Figure 2. Relative surface concentration of adsorbed Ca2þ from
of calcium nitrate solution at a mica surface is shown in Figure 1 calcium nitrate at mica surface on a log scale as a function of bulk
as a function of electrolyte concentration. Experimental points15 concentration. The surface concentration [Ca2þ(surf)] is shown
are shown for comparison. The theoretical surface charge is taken relative to bulk concentration [Ca2þ(bulk)].
to be fixed at -0.012 C m-2, calibrated to fit to the experimental The surface concentration asymptotes to about 10 times the bulk
data point at 10-3 M, measured at pH 5.9. The acidic conditions concentration. This asymptotic behavior is not observed experi-
at pH 5.9 have reduced the surface charge from the fully mentally, since the highest reliably measured concentration is
dissociated surface charge of -0.33 C m-2 (1 site per 48 Å2) 0.1 M, with experimental measurements having poor reproduci-
due to protonation of the negatively charged surface sites. Our bility at higher concentrations.15 The asymptotic behavior is
fitted value of -0.012 C m-2 is consistent with the surface charge found at higher concentrations between 0.1 and 10 M. The charge
of -0.022 C m-2 predicted by a charge regulation model19,61 at reversal effect is stronger with trivalent cations; in these systems,
pH 5.9 with a mica equilibrium constant pKH = 5.85, with the the asymptotic surface potential at high concentrations can be
surface potential taken in dilute solution at -70 mV. deduced experimentally from force measurements.62
Figure 1 shows excellent agreement between our theoretical Ion Specificity: Shift of Isoelectric Point. The Hofmeister
model and the experimental data at all concentrations. Notably, series in the alkali earth cations may be explored by considering
the theoretical model is able to achieve charge reversal to the same the surface potentials of nitrate solutions of these cations as a
magnitude (þ12 mV at 0.1 M), and charge neutralization occurs function of salt concentration (Figure 3). In terms of the strength
at the same concentration, 0.03 M, as observed experimentally. of the charge reversal effect, we have Mg > Ca > Sr > Ba. The
Where the experimental data suggest a linear relationship isoelectric point (concentration of zero potential, or cvp) of each
between the surface potential and log-concentration, it is curious solution is shifted accordingly, with Mg showing a zero potential
to observe a flattening in the theoretical curve at very high at the lowest salt concentration (0.003 M) and Ba2þ reaching zero
concentrations. That is, the surface potential reaches an upper potential at the highest concentration (0.27 M). Ion specificity is
limit at high concentration, corresponding to an asymptotic value therefore strong, with the cvp’s changing by almost an order of
for the concentration of adsorbed Ca2þ at the surface. We show magnitude between cations, the full predicted set of cvp’s being
the ratio of the concentration of Ca2þ at the surface to the bulk 0.003, 0.024, 0.062, and 0.27 M for Mg, Ca, Sr, and Ba,
concentration, as a function of bulk concentration, in Figure 2. respectively.
The Hofmeister series seen in Figure 3 is consistent in three
(58) Landau, L. D.; Lifshitz, E. M. Electrodynamics of Continuous Media; ways. First, at any given concentration, the surface potentials
Pergamon Press: Oxford, 1960; Vol. 8.
(59) Netz, R. R. Curr. Opin. Colloid Interface Sci. 2004, 9, 192–197. follow the series Mg > Ca > Sr > Ba. Second, the same series is
(60) Boroudjerdi, H.; Kim, Y.-W.; Naji, A.; Netz, R.; Schlagberger, X.; Serr, A.
Phys. Rep. 2005, 416, 129–199.
(61) Miklavic, S. J.; Ninham, B. W. J. Colloid Interface Sci. 1990, 134, 305–311. (62) Pashley, R. M. J. Colloid Interface Sci. 1984, 102, 23–35.

6434 DOI: 10.1021/la9041265 Langmuir 2010, 26(9), 6430–6436


Parsons and Ninham Article

Figure 3. Surface potential of alkali earth nitrate solutions at mica


surface as a function of electrolyte concentration.
found in the isoelectric points (concentrations of zero potential). Figure 4. Surface hydration. There is overlap of the hydrated
Third, the series is maintained in the asymptotic surface potentials layers of the surface and hydrated ions, while unhydrated ions
found at very high concentration. It is curious that the most are excluded from the surface hydration layer.
polarizable ion, Ba2þ, has the smallest charge reversal effect. This
Short range repulsion is weaker in Kþ solutions than in Naþ or
is chiefly due to ion hydration, with Ba2þ having the smallest
Ca2þ solutions, and Pashley cites the surface free energy of Kþ at
number of hydration waters and therefore the smallest hydration
below 1kT, in contrast to that of Naþ. The surface dispersion free
enhancement to the total polarizability of the hydrated ion. Mg2þ,
energy of an unhydrated Kþ ion using our B coefficient at z = 0,
conversely, is the least polarizable of the cations but exhibits the
however, is 2.5kT. That is, the surface energy of Kþ at contact is
largest charge reversal effect due to its greater number of hydra-
about the same as that of hydrated Naþ rather than being
tion waters.
significantly smaller.
Ion Adsorption and Secondary Hydration Forces. Pashley,
This contradiction may be resolved by applying the concept of
Israelachvili, and co-workers have measured surface forces be-
the hydration layer to the surface itself,67 complementary to the
tween two mica surfaces in solution.16,17,63,64 They found the
ionic hydration layer modeled in this paper. The hydrated ion
measured forces at close separations (10-40 Å) to be larger than
model has been applied to tightly hydrated cosmotropic ions such
expected from DLVO theory alone. Pashley interpreted these
as the divalent cations or Naþ. Kþ on the other is classified10,21 as
repulsive forces to be derived from the adsorption of hydrated
chaotropic with a weakly held hydration layer. We have here
ions and therefore labeled them “secondary hydration forces”.65,66
applied Collin’s criteria for distinguishing between cosmotropic
The repulsive force was attributed to the adsorption free energy of
and chaotropic ions, using Jones-Dole viscosity coefficients.51-53
hydrated ions to the surface. In the case of Ca2þ, for instance, the
Tightly hydrated cosmotropic ions are those with a positive
free energy of adsorption was estimated empirically17 at 4kT.
Jones-Dole viscosity coefficient, and weakly hydrated chaotro-
We have shown how ion-surface dispersion interactions
pic ions are those with a negative Jones-Dole coefficient. Among
provide an explanation for the phenomenon of surface charge
the group I alkali cations, the boundary is found between Naþ
reversal. The surface dispersion interaction provides additional
and Kþ, with Naþ having only a slightly positive Jones-Dole
adsorption beyond that expected from electrostatic interactions.
coefficient. The notion of surface-induced ion dehydration, which
It is instructive to consider the value of the ion-surface
√ dispersion
may for instance occur at alumina surfaces,55 may be relevant to
interaction at contact (z = 0), Ud0 = 16B/(3 πa3). This quantity
borderline cases such as Naþ. Applying Collin’s law of matching
may be interpreted as the adsorption energy of the ion. In the case
water affinities51,53,68 to the ion-surface interaction, there will be
of hydrated, Ca2þ its value is -10.98 kJ mol-1, or -4.4kT at room
a greater affinity for hydrated ions to approach the hydrated
temperature (T = 298 K). The closeness of the magnitude of this
surface. The hydration layers of ion and surface may overlap,
quantity to Pashley’s free energy, at 4kT, is certainly not coin-
bringing the center of the ion close to the edge of the surface
cidental. It seems apparent that ion-surface dispersion interac-
hydration layer, as illustrated by hydrated Ca2þ in Figure 4.
tions provide, or at the very least contribute substantially to, the
This brings insight into the physical meaning of ion-surface
mechanism for ion adsorption to the mica surface which leads to
contact at z = 0. Rather than suggesting by z = 0 that the ion is
Pashley’s short-range repulsive forces.
embedding into the mica surface itself, we may interpret z = 0 to
This interpretation is supported also by the ion-surface
correspond to the outer edge of the surface hydration layer, that
dispersion energies for monovalent cations.35 The surface disper-
is, to a distance of closest approach of the hydrated ion. Contact
sion energies of hydrated Liþ and Naþ may be calculated at
at z = 0 is defined by the surface hydration layer. This model also
-1.6kT and -2.6kT, respectively. These magnitudes are consis-
partly explains why the mica surface charge has been calibrated to
tent with Pashley’s adsorption energies16 which are likewise
-0.012 C m-2, slightly lower than the -0.022 C m-2 expected
greater than 1kT.
from mica at pH 5.9 under a charge regulation model (see above);
The argument that ion adsorption driven by dispersion forces
it is an effective charge accounting for the displacement of the
leads to Pashley’s short-range repulsion may be strengthened
surface hydration layer.
further by considering the repulsive force in Kþ electrolyte.16
As for Kþ, as a chaotropic, weakly hydrated ion, it cannot
penetrate into the surface hydration layer, also illustrated in
(63) Israelachvili, J. N. Adv. Colloid Interface Sci. 1982, 16, 31–47.
(64) Ducker, W. A.; Pashley, R. M. J. Colloid Interface Sci. 1989, 131, 433–439.
(65) Pashley, R. J. Colloid Interface Sci. 1981, 80, 153–162. (67) Johnson, S. B.; Franks, G. V.; Scales, P. J.; Healy, T. W. Langmuir 1999, 15,
(66) Pashley, R. M.; Israelachvili, J. N. J. Colloid Interface Sci. 1984, 101, 511– 2844–2853.
523. (68) Collins, K. D.; Washabaugh, M. W. Q. Rev. Biophys. 1985, 18, 323–422.

Langmuir 2010, 26(9), 6430–6436 DOI: 10.1021/la9041265 6435


Article Parsons and Ninham

Figure 4. Its surface dispersion energy should therefore be taken on charge reversal effects in a HNC model are forthcoming.71 The
at the distance of closest approach, one ionic radius out from z = ab initio polarizabilities reported here and previously35 can help
0, where the Kþ surface dispersion energy drops to - 0.5kT. In these more sophisticated HNC models achieve quantitative
this way, the behavior of Kþ is brought in line with Pashley’s agreement between theory and experiment, and come closer by
observations. The link between our ion-surface dispersion en- refining the B values used, just as they have improved agreement
ergy and Pashley’s short-range repulsive forces appears to be with simple Poisson-Boltzmann theory.
consistent. The question of the effective surface charge and
location of z = 0 may be made more robust by explicitly adding Conclusion
the surface hydration layer to the calculation. The relationship By applying ab initio ionic polarizabilities with an hydration
between ionic dispersion forces and short-range repulsion be- model for alkali earth cations to a modified Poisson-Boltzmann
tween surfaces will be explored more closely in subsequent work. description of an electrolyte, we have been successful in reprodu-
Our model here does not allow for the extremely short-range cing charge reversal observed experimentally in surface potentials
oscillatory forces reported by Pashley and Israelachvili.66 A at mica surfaces. Charge reversal appears due to increased ion
multilayer model of the surface hydration layers or spatial surface affinity driven by nonelectrostatic ion-surface dispersion
structure in the solvent69,70 would be required to theoretically interactions (alongside ion correlation forces, not considered
model oscillatory forces. here). This same mechanism appears to be responsible for
The short-range repulsive forces have been termed as “hydra- short-range repulsive forces (hydration forces) observed experi-
tion forces” by Pashley, which seems somewhat mislabeled mentally.
insofar as the repulsion is driven by adsorption due to dispersion We suggest that the idea of a hydration layer, applied here to
forces (“adsorption forces” might be more apt). However, given hydration of ions, should also be considered at surfaces. It seems
the behavior of Kþ and the interpretation of its relationship to the likely that a tightly bound surface hydration layer will be relevant
hydration layer of the surface, “hydration forces” is perhaps not to some surfaces but not others, in the same way that cosmotropic
invalid. However, it should be understood to refer to hydration of ions in solutions have a tightly held hydration layer while
the surface as much as to hydration of the ions. chaotropic ions do not.
The agreement between Pashley’s calcium ion adsorption Ions were treated as spherical in this paper. We expect inter-
energy and our surface dispersion energy, at around 4kT, is esting anisotropic effects may appear if a theory of nonspherical
interesting because both Pashley’s and our modeling are based ions is developed. This will introduce ion shape as a third element
on a Poisson-Boltzmann description of the electrolyte which of ion specificity, following ion size and ion polarizability, that
neglects ion-ion correlations. The theoretical modeling of the will contribute to the ion specific effects of planar ions such as
calcium nitrate experiments motivating this paper used the nitrate, formate, or carbonate, and linear ions such as the cyanate
anisotropic hypernetted chain (HNC) model by Kjellander and series. Shape effects will likewise explain the specific behavior of
Marcelja et al.,15,37 which does include ion-ion correlations but anisotropically hydrated ions such as hydroxide. Hydrated hy-
neglects ionic dispersion interactions. The site binding model in droxide forms a pyramid with water at the base and the ion at the
these HNC calculations took a slightly lower Ca2þ adsorption apex,72,73 which will impact on the affinity of hydroxide to a
energy, at 2.5kT. The effect of ion-ion correlations, aside from surface. This kind of anisotropic hydrated system is best modeled
other effects such as inducing spatial ordering between ions, is to with explicit hydration waters, moving beyond the simple hydra-
provide a limit to the ion concentration adsorbed at the surface by tion model described in this paper.
preventing ions from crowding on top of each other. They The success found here in modeling charge reversal suggests
introduce a distance of closest approach to the ionic interactions. that the model of hydrated ions with dispersion forces may be
As a consequence, Marcelja’s lower adsorption energy indicates applicable to the exploration of ion surface effects at surfaces of
that ion-ion correlations prevent calcium ions, on average, from biological or medical interest. For instance, the lithium ion is used
approaching the surface hydration layer at z = 0. The calcium ion in treating mood and bipolar disorders74,75 while cerium(III)
surface dispersion energy falls back to 2.5kT at a distance of 0.73 nitrate can be used in medication for burns.76 Is the bioeffective-
Å from the surface, which, perhaps coincidentally, is close to the ness of these ions determined by their physical properties, through
radius of a bare (unhydrated) calcium ion. Our model of ionic the ion specific shift that they induce on the surface potentials of
dispersion forces therefore again seems to be consistent with HNC cell membranes?
calculations that include site binding. That is, site binding in a
HNC model can explained by ionic dispersion interaction taken in
Acknowledgment. The authors would like to thank Stjepan
conjunction with the distances of closest approach that form part
Marcelja for helpful and thought provoking discussions. We
of the ion-ion correlations of a HNC model. This approach has
acknowledge the assistance of the staff at the Australian National
been shown using model dispersion B coefficients to have an
Computational Infrastructure National Facility.
impact on forces between surfaces,38 increasing or reducing them
depending on the value of B and surface charge density. Reports (72) Tuckerman, M. E.; Marx, D.; Parrinello, M. Nature 2002, 417, 925–929.
(73) Wick, C. D.; Dang, L. X. J. Phys. Chem. A 2009, 113, 6356–6364.
(69) Basilevsky, M. V.; Parsons, D. F. J. Chem. Phys. 1998, 108, 9114–9123. (74) Schou, M.; Juel-Nielsen, N.; Str€omgren, E.; Voldby, H. J. Neurol.,
(70) Cherepanov, D. A. Phys. Rev. Lett. 2004, 93, 266104. Neurosurg. Psychiatry 1954, 17, 250.
(71) Wernersson, E. Theoretical Investigations of the Role of Ion-Ion Correla- (75) Burgess, S. S.; Geddes, J.; Hawton, K. K.; Taylor, M. J.; Townsend, E.;
tions and Ion-Specific Interactions in Electric Double Layers. Ph.D. Thesis, Jamison, K.; Goodwin, G. Cochrane Database Syst. Rev. 2001, 3, CD003013.
University of Gothenburg, 2009, http://hdl.handle.net/2077/21040. (76) Garner, J.; Heppell, P. Burns 2005, 31, 539–547.

6436 DOI: 10.1021/la9041265 Langmuir 2010, 26(9), 6430–6436

You might also like