9 1988 by The Humana Press Inc.

All rights of any nature whatsoeverreserved.

0163-4992/88/1200-0255503.00

Theory of Preferential Solvation of

Nonelectrolytes
A. BEN-NA[M
Department of Physical Chemistry, The Hebrew University,
Jerusalem, Israel
Received January 20, 1987; Accepted June 1, 1987

ABSTRACT
The concepts of local compositions around a solute and prefer-
ential solvation of a solute are defined in terms of the Kirkwood-Buff
integrals. The difference between the local and the bulk composition
is a measure of the preferential solvation of a solute with respect to
the various components of the solvent.
A statistical mechanical theory is developed that leads to simple
relationships between local compositions and experimentally measur-
able quantities. Some preliminary results on preferential solvation of
methane in mixtures of water-ethanol and water-p-dioxane are pre-
sented.
Index Entries: Solvation, of nonelectrolytes; nonelectrolytes, sol-
vation of; statistical mechanics, of solvation; solvation, statistical me-
chanics of; Kirkwood-Buff integrals; methane, solvation in water-
ethanol and in water-p-dioxane; water-ethanol, solvation of methane
in; water-p-dioxane, solvation of methane in.

INTRODUCTION
The problem of preferential solvation (PS) arises almost in any phys-
ical chemical study of solutes in mixed solvents. The study could be ther-
modynamic, spectroscopic, or kinetic (1). The behavior of the solutes
(e.g., chemical shift, diffusion, reactivity, and so on) depends on the
composition of the solvent. However, in order to understand how the
solvent composition affects the solute behavior, we need to know the

Cefl Biophysics 255 Vot 12, 1988

2.56 Ben-Maim

composition that the solute "sees," i.e., the composition in its immediate
vicinity. This is, in general, different from the bulk composition of the
mixed solvent.
There are essentially two questions that are of central importance in
the study of PS. First, how do we measure the PS of a given solute in a
given mixture? Second, what are the molecular reasons that cause a so-
lute to prefer one component over the other and, hence, alter the compo-
sition in its local environment?
The simplest approach to a n s w e r the first question is to follow some
property of a solute in a mixed solvent. For example, if ~a is the NMR
chemical shift (or other spectroscopic quantity) characteristic of the sol-
ute S in pure solvent A, and ~B the corresponding quantity for pure sol-
vent B, then one might relate the observed chemical shift of S in a mix-
ture of A and B, ~A,8, to 5A and 5B by the equation
~A,B = xA(local)~A + [1 -- XA(local)]58 [1.1]
where XA(Iocal) defined in Eq. [1.1] is a measure of the local composition
of the solution near the solute. This may or m a y not be different from the
bulk composition XA of the solvent mixture, XA being the mole fraction of
the c o m p o n e n t A in the mixture.
Although Eq. [1.1] can serve as an operational definition of XA(1Ocal),
it does not really tell us what is the local composition in the vicinity of the
solute S. We should not be surprised to find that different properties of S,
used in Eq. [1.1], will result in different values of xA(local). The reason is
that there is no theoretical support to the assumption that 5a,s is an aver-
age of 8A and 5B, as implied in Eq. [1.1]. Therefore, the approximation
involved in using Eq. [1.1] will, in general, be different for different prop-
erties of S in mixtures of A and B.
What we need is an u n a m b i g u o u s definition of, and a m e t h o d of
measuring, the local composition of the solvent, that is i n d e p e n d e n t of a
specific measurable property of S.
Perhaps the first thermodynamic treatment of the problem of PS was
presented by Grunwald et al. (2), w h o were interested in the solvation of
ions in mixtures of water and dioxane. This approach was further devel-
oped by Covington and N e w m a n (3). However, the ambiguity in the
very definition of the local composition has not been removed. Coving-
ton and N e w m a n defined solute species SAiBj that contains i molecules
of A and j molecules of B in their solvation shell. But they failed to define
the solvation shell, and therefore there remains the ambiguity of the as-
signment of which A or B molecules belong or do not belong to this shell
(besides, they based their theory on an unjustifiable assumption that i +
j = n, n being the coordination n u m b e r of S in pure A or B).
An attempt is made here to define the concept of PS unambiguously
and i n d e p e n d e n t l y of any modelistic assumptions on the system. The
definition of the local composition is presented in the next section. Later
we provide a method of measuring the PS of a solute S in a two-

Cell Biophysics VoL 12, 1988

 Theory of Solvation of Nonelectrolytes 257

component solvent mixture. Some preliminary experimental examples

are also presented. There is almost no relevant experimental data on
three-component systems we could have used. We hope the publication
of this paper will encourage experimentalists to undertake measure-
ments that eventually should lead to the answer to the second question
posed above; namely, why a solute will prefer one solvent molecule over
others.

THE FORMULATION OF THE PROBLEM OF PS

Consider a mixture of two components, NA molecules of A and N B

molecules of B, at some temperature T and pressure P. In such a system
the composition measured by the mol fraction XA = NA/(NA -4- NB) will be
the same at any point Ro within the system (except, of course, for a small
region near the boundaries of the system. We shall not be interested in
this region. However, it should be noted that the type of questions we
shall ask about the composition around a solute particle could also be
asked about this region.)
We shall refer to XA as the bulk composition of the system. Next con-
sider a very dilute solution of a solute S in our two-component mixture.
(In principle, we could have discussed also high concentrations of S in
the system. However, this will require consideration of three compo-
nents, S, A, and B in the surroundings of S. For simplicity, we treat only
the case of very dilute solutions, so that S "sees" practically only As and
Bs.)
Qualitatively, the question we would like to ask is quite simple.
What is the composition of the liquid in the immediate vicinity around
the solute S. Clearly, since the affinity of S toward A might be different
from its affinity toward B, we should expect that the composition near
the solute S will differ from the bulk composition XA.
The main question that will concern us in this section is h o w to
define the local region in the vicinity of S in which the composition is ex-
pected to be affected by the presence of S. Once we have defined the re-
gion, we shall provide a m e t h o d of measuring the composition in this
region.
Consider first a simple spherical solute, say argon, in a two-compon-
ent solvent, say water and ethanol. Let dR' = dx'dy'dz' be an element of
volume at a distance R' from the center of S. The average n u m b e r densi-
ties of A and B in this element of volume will be
pA(R') = pA(bulk)gAs(R') [2.1]
pB(R') = pB(bulk)gBs(R') [2.2]
w h e r e pA(bulk) and pB(bulk) are the bulk densities of A and B, respec-
tively, and gAs(R') and gBs(R') are the radial distribution functions for the
pair of species A,S and B,S, respectively.

Cell Biophysics VoL 12, 1988

258 Ben-Naim
Clearly, if we had the full information on these two radial distribu-
tion functions, we could have defined the local composition at any dis-
tance R' from the center of S by
XA(R') = pA(R')I[pA(R') + pB(R')] [2.3]
Furthermore, this local composition will be different at different dis-
tances, say R' and R" (see Fig. 1). If the solute is not spherical, e.g., S is a
protein molecule, then the local composition w o u l d be d e p e n d e n t on
both distance and relative orientation of the point at which we choose
our element of volume (see Fig. 2).
We k n o w from the general theory of liquids that the radial distribu-
tion functions normally will tend to unity at distances of the order of
m a g n i t u d e of a few molecular diameters. Thus, at these distances from
the center of S, all local densities will be identical to the bulk densities.
At short distances, however, large deviation from the bulk densities
are expected. A typical form of the radial distribution function for a one-
c o m p o n e n t simple liquid is depicted in Fig. 3. We observe that at very
short distances the local density will oscillate above and below the bulk
density. Clearly, similar behavior is expected in a multicomponent sys-
tem.
Unfortunately, there is no experimental data on the separate radial
distribution functions in two or more c o m p o n e n t systems. Even if we
had such information, it would have been too detailed to be useful for
practical purposes. Instead, we are interested in the overall composition
in the local neighborhood of the solute, which roughly coincides with the
region in which g(R) is significantly different from unity (Fig. 3). Fortu-
nately, this information may be obtained from thermodynamic quanti-
fies. The relevant quantities are the so-called Kirkwood-Buff integrals
(4,5). These are defined as follows.

GaB = f [gaB(R) - 1]4"rrR2dR [2.4]

0

Fig. 1. The average local density of the solvent around a spherical solute
S depends only on the distances ~R'! and iR"] from the center of S.
Cell Biophysics VoL 12, 1988
Theory o f Solvation o f Nonelectrolytes 2.59

Fig. 2. The average local density of the solvent around a protein S de-
pends on both distance and orientation of the element of volume relative to the
center of S.

w h e r e gAB(R) is the radial distribution function for the pair of species A

a n d B. If the molecules are not spherically symmetrical, then gAB(R) is the
orientationally averaged pair-correlation function. The integration is ex-
t e n d e d from zero to infinity. H o w e v e r , in m o s t practical cases gAB(R) dif-
fers from unity only at distances of the order of m a g n i t u d e of a few mo-
lecular diameters. Therefore, practically the m a i n contribution to the
integral comes from the region in which gAB differs considerably from
unity. This region can be c o n v e n i e n t l y referred to as the correlation re-
gion a r o u n d A (or B, d e p e n d i n g on the vicinity of which molecule we are
interested in).

Fig. 3. Schematic form of the pair correlation function g(R) for pure liquid
of simple spherical particles. Note that g(R) is practically unity at distances of a
few molecular diameters or.
Cell Biophysics VoL 12, ! 988
260 Ben-hlaim

The significance of the quantity GAB with respect to the question of

preferential solvation is the following. PAgAs(R)4~R2dR is the average
n u m b e r of A particles in the element of v o l u m e 4 wR2dR at the distance R
from the center of the solute S. On the other hand, pA4"rrR2dR is the aver-
age n u m b e r of A molecules in the same element of volume but taken rela-
tive to an arbitrary center in the liquid. Therefore, PA[XAs(R) -- 1]4~rR2dR
measures the excess, or deficiency, of A molecules in the spherical shell
4~rR2dR around S relative to the same spherical shell but at an arbitrary
location in the liquid. The quantity pAGAs, according to the definition in
Eq. [2.4], is simply the overall excess or deficiency of A molecules in the
entire volume a r o u n d S.
As we noted above, although the integration in Eq. [2.4] extends
from zero to infinity, in practice it can be cutoff at a distance of a few
molecular diameters.
We shall use the Kirkwood-Buff theory of solution (4,5) to relate GAB
to t h e r m o d y n a m i c quantifies in the next section. It should be noted that
these relationships are derived in an open system (i.e., in the T, V, ~ en-
semble) w h e r e the n u m b e r of particles in the system are not fixed. The
normalization condition for GAB is (4,5)

GAB = i [gAB(R) -- 1]4~rR2dR

0
= V ([(NA NB -- /q/ANB)/NA/CqBI -- (SAB/NA)) [2.5]
w h e r e ~AB is the Kronecker delta function.
If we were in a closed system (i.e., T,V,N ensemble), then /~/A and
NB are fixed quantities, and NANB = /~/A/~B; hence, the corresponding
normalization condition is

PAf[gAB(R) -- 1]4"~R2dR = - ~AB [2.6]

0
Thus, if A = B, the integral in Eq. [2.6] is - 1 as it should be, since the
total deficiency of As around a fixed A is exactly the one particle that we
have placed at the center. On the other hand, for A ~a B the integral is
zero. Placing of, say, one A at the center does not change the total n u m -
ber of particles in the entire system.
These simple conclusions are valid only in closed systems and do not
apply in o p e n systems for which the normalization condition in Eq. [2.5]
holds. In an o p e n system, placing a particle A at the center can produce
any excess or deficiency of As in the entire volume V. In other words, we
cannot use the conservation of the total n u m b e r of As (or Bs) in the entire
system if the system can exchange molecules with its surroundings.
We n o w exploit the fact that gAB(R) decays to unity beyond some dis-
tance R/> RAB, w h e r e RAB may be referred to as the correlation distance
for the species A and B. Let Rc be the largest correlation distance for any
pair of species in the system. We define the correlation volume as
Vcor = 4~rR3/3 [2.7]
Cell Biophysics Vol. 12, 1988
 Theory of Solvation of Nonelectrolytes 261

Since all the pair correlation functions are practically equal to unity
at R I> Re, we may write the average n u m b e r of A particles in the correla-
tion volume around S as
Re

NA,s(Rc) = P A f gAs(R)4"rrR2dR
0

= PA )
0
[gAs(R) -- 1]4"rrR2dR + PA )
0
4"rrR2dR

= pAGAs + pAVcor [2.8]

Equation [2.8] simply means that the average n u m b e r of As in the
correlation volume is the sum of the average n u m b e r of As in the same
region, before placing S at its center, plus the change in the n u m b e r of As
in the same region caused by placi_ng S in the center of this region.
Using a similar definition for NB,s(Rc)
NB,s(Rc) = pBGBs + pBVcor [2.9]
we can define the local composition in the correlation region a r o u n d S as
xA,s(Iocal) = NA,s(Rc)/[NA,s(Rc) + NB,s(Rc)]
= (xAGAs + XAVcor)/(XAGAs + XBGBs + Vcor) [2.10]
w h e r e XA is the bulk composition in the system.
The local composition XA,s(local) can n o w be compared with XA to
determine the preferential solvation of S. If XA,s(local) > XA, we may say
that S is preferentially solvated by A, and if x~,,s(local) < XA, S is prefer-
entially solvated by B.
Thus, we define the preferential solvation of S with respect to A sim-
ply by
8A,s = XA,s(1ocal) -- XA
= XAXB(GAs -- GBS)/(XAGAs + XBGBs + Vcor) [2.11]
Clearly, the sign and extent of preferential solvation might d e p e n d
on the composition of the solvent. Figure 4 depicts a few possible cases
w h e r e there are positive, negative, or mixed signs of preferential solva-
tion according to whether XA,s(1ocal) is above or below the diagonal line.
Note that in all cases
if xA --+ 0
~A,S --+ 0 or XB --+ 0 [2.12]
or GAS - GBS -'* 0
In the next section we shall develop the theoretical relationship be-
t w e e n GAS and Gss and experimentally measurable quantities. The only
quantity that is left ambiguous in Eq. [2.11] is Vcor- Clearly, if we take
very large correlation volume, we obtain
8A,s ~ 0 for Vco r ---+ r
Cell Biophysics I/ol. 12, 1988
262 Ben-Naim

Fig. 4. Schematic dependence of the local composition xA,s(local) as a

function of the bulk composition XA. The diagonal line corresponds to the case
when there is no preferential solvation of S. Curves a and b correspond to posi-
tive and negative preferential solvation. In curve c, the preferential solvation
changes sign as XA changes.

On the other hand, for too small Vcor, the approximate equality of gAs(R)
1 p r e s u m e d in Eq. [2.8] (for R ~ Rc) will not hold. In practice, we can
choose, for each specific system, a reasonable Rc (and hence Vcor) accord-
ing to the behavior of the functions gij at large distances. Theoretically,
however, we can get rid of Vcor by taking the first order term in the ex-
pansion of 8A,s in Eq. [2.11] in p o w e r series about e - V -cot, 1
thus
~A,S = 0 -}- ~ - X A X B ( G A s -- GBS) + .-- [2.13]
We define the limiting linear preferential solvation as

8Os OSA,s[
-- OE. = XAXB(GAs -- GBS) [2.14]
=0
Since XAX B > 0, the sign of 8 ~ is determined by the sign of GAs -
GBS, and this is, of course, i n d e p e n d e n t of the correlation volume. Thus,
we have defined in Eq. [2.13] a quantity that unambiguously measures
the preferential solvation of S with respect to a two-component solvent.
As noted earlier (5), GAS measures the affinity of S toward A. Thus, the
difference GAS -- GBS measures the difference between the affinities of S
toward A and B. We next turn to the question of measurability of the
quantity G A S -- GBS.

RELATIONS BETWEEN PREFERENTIAL SOLVATION AND

MEASURABLE QUANTITIES IN A THREE-COMPONENT
SYSTEM
Let S be a solute highly diluted in a solvent mixture of A and B at
composition XA. Let I denote the liquid mixture of A and B. In this system
the chemical potential of the solute S is given by (5,6)
Cell B i o p h y s i c s VoL 12, 1 9 8 8
 Theory of Solvation of Nonelectrolytes 263

t~(T,P,ps,XA) = la*J(T,P,XA) + kT In psA3q-s l [3.1]

The quantity F%l is referred to as the pseudochemical potential (but note
that this is not the conventional standard chemical potential [6]). This
quantity depends on the composition of the solvent mixture xa but not
on the solute density Ps = Ns/V. This follows from the assumption that
the system is very dilute with respect to S. A3 is the m o m e n t u m partition
function of S and qs is the internal partition function (including rota-
tional, vibrational, and so on) of S. In the following we shall assume that
neither A3 nor qs depends on the composition of the solvent mixture.
The derivative of ~- with respect to, say, NA is
( 3 ~ s / 3 N A ) T , P , NB,N s = (C~I/3NA)T,P,NB,Ns -- (kT/V)V A [3.2]
where VA is the partial molar volume of A in the mixture. The derivative
in Eq. [3.2] may be expressed in termed of Kirkwood-Buff integrals (4).
The following is an exact result (7).

( o~ \
9 ,
kT
PSPAV
IEKS,A)] [3.3]

where IE(S,A)I and IDI are two determinants, the definition of which for
any mixture is available (7). Here we shall be interested only in the case
of very dilute solutions of S in the mixture. For this case we have

i 1 0 1 1
1E(S,A)I = GSA 1 GAB = --(1 + pBGBB -- pBGAB) + GSA -- GSB [3.4]
PB
I GSB 1 GBB+PB-1
= .rlVAPB -I + GSA -- GSB

where we have used the Kirkwood-Buff result for the partial molar
volume in the pure mixture of A and B, i.e.,
V A -- (1 q- pBGBB -- pBGAB)/'rl [3.5]
and where ~l is defined for any mixture of A and B as
"q = PA + PB + PAPB(GAA + GBB -- 2GAB) [3.6]
The determinant in the denominator of [3.3] is defined by
0 1 1 1
ID} = 1 Gss + Ps- I GSA GSB
1 GSA GAA + PA- I GAB [3.7]
1 GSB GAB GBB q- PB- 1

which has the limiting form

IDI [3.81

Substituting [3.4] and [3.8] in [3.3], we obtain

Cell Biophysics Vol. 12, 1988
264 Ben-Naim

-kr -

O-~A,JT.P.NB,N s -- V.FI ['qVA + pB(GsA -- GSB)] [3.9]

From [3.2] and [3.9] we obtain

- k r "nCA + pB(CSA--CS.) + - -k T V A
_ [3.10]
/ r,p,No,Ns V n V
However, for practical purposes, it is more convenient to transform into
a derivative with respect to the mole fraction xA:
Olx~! Op.~1 OXA Op,~1 NB
aNA - OXA ONA OXA (NA + NB)2 [3.11]
and hence we have the final result
lim ( 0_~'~ _ kT(pA +pB) 2
(GBs-- Gas) [3.12]
PS--'>0 k 0XA "//T,P 3]
Let ix;g be the pseudochemical potential of S in an ideal gas phase
(6). Clearly, ix;g is i n d e p e n d e n t of XA. Therefore, we may rewrite Eq.
[3.12] as
lim (OAG;~ kT(pA + pB)2 (GBs-- GAs) [3.13]
Ps-'~0 \--~-A-A/p,T = n "

w h e r e AG~ = ix~I - ix~g is the s o l v a t i o n free e n e r g y of S i n o u r s y s t e m

(6).
Thus, by measuring the slope of the solvation free energy as a func-
tion of xA, we can extract the difference GAs - GBS. As we have seen, this
is a measure of the limiting preferential solvation of S with respect to A.
Note that ~q is a calculable quantity through an inversion of the
Kirkwood-Buff theory (8). However, since "q > 0 (8), the entire quantity
kT(pA + pB)2HIis always positive. Therefore, the sign of the derivative is
the same as the sign of GBS GAs. -

In the next section we present some numerical results of preferential

solvation in three-component systems. In the following we shall derive
another equation relating GBS and GAs with experimental quantities. The
latter, together with relation [3.13], may be used to compute GBS and GAs
individually.
Before deriving the second equation, we note that relation [3.13] is
useful for a solute for which the solvation free energy can be determined.
If we are interested in proteins as solutes, then Eq. [3.13] is impractical.
However, we can still measure the solvation free energy of S in I relative
to, say, pure A (e.g., solvation of protein in a solution relative to the
solvation in pure water). The relevant relation is
AAG; = AG~ l - AG~ A = kT ln(p~/p~)eq [3.14]
Thus, by measuring the density of S in I and in pure A at equilibrium
with respect to the pure, say solid, S, we can determine AAG; from Eq.
Cell Biophysics VoL 12, 1988
Theory of Soivation of blonelectrolytes 26.5

[3.14]. Since AGgA is i n d e p e n d e n t of XA, we can apply Eq. [3.14] in Eq.

[3.13] to obtain

lim [a(aao ) 1 _ 2 (GBs - GAS) [3.15]

ps--,0 L ax A J "n
which is the useful relation for complex solutes, such as proteins. Note
that if A and B form a symmetrical ideal solution (5), i.e., w h e n
GAA + GBB -- 2GAB = 0 [3.16]
in the entire range of composition, Eq. [3.13] as well as [3.15] reduces to
(note 3.6)
lim ( OAG*s~
ps---~0 3xA ] = kT(pA + PB)(GBs -- GAS) [3.17]

i.e., there could still be preferential solvation of a third component S. It is

only in the case w h e n Gss = GAs that there is no preferential solvation,
and the solvation free energy of S becomes i n d e p e n d e n t of composition.
The second equation that we shall now derive is for the partial molar
volume of S at infinite dilution in the solvent I, of composition xA.

I~,~ = \ON~ ,v,r,N' - V~--~- BrV [3.18]

where N' includes all Ni's except N~, the determinant jBl, for our case, is
defined by
Ps + p~Gss PsPAGAs PsPBGBs
[B[ = PsPAGAs PA + p2AGAA PAPBGAB [3.19]
PsPBGBs PAPBGAB PB + p2GBB

which has the following limiting behavior at ps--~O

IB[ ~ PSPAPB~ [3.20]
with ( defined by
= 1 + pAGAA + pBGBB + PAPB(GAAGBB -- GAB) [3.21]
The quantities B~ in [3.18] are the cofactors of the determinant B, which
for the limit ps--*0 are
BSA--> --psPAPB[GsA(1 + pBGBB) -- GSBPBGAB] [3.22]
BSB-'~ --PsPAPB[GsB(1 + pAGAA) -- GSAPAGAB] [3.23]
Bss ~ PAPB~ [3.24]
Using the Gibbs-Duhem equation
PAISA + PB~St~ + PSI~SS = 0 [3.25]
and the Kirkwood-Buff results for the isothermal compressibility and the
partial molar volumes of A and B
Cell Biophysics VoL 12, 1988
266 Ben-Naim

[37 = ~lkT'q [3.26]

WA = [1 4- pB(GBB -- GAB)]/~q [3.27]
WB = [1 4- p A ( G a a -- GAB)]/'rl [3.28]
we obtain the following equality

pa
[kTB VSVA]
j
[kTB sB VsVA]
+ PB LV---~ pTv J
[kTB ss
.] = o [3.29]
+ psL l-g ] PTV /
which may be reduced to
-- pAGsAWA~q -- pBGsBVB"q -- pAWsVA'q -- pBVsVBI"I = 0 [3.30]
From [3.30] we may eliminate Vs (for ps---~0) to obtain the final result:

lim 17 s = kTPT
F~s - ps--*0 - pAVAGsA -- pBVBGsB [3.31]

In Eqs. [3.13] and [3.31] all the quantities PT, "q, PA, PB, VA, VB, ~S,
and 3 A G ~ a X A a r e experimentally determinable. Hence, these two equa-
tions may be used to eliminate the required quantities GSB and GsA. (Note
that (PA + pB)2/'q in Eq. [3.13] may be computed directly from the vapor
pressure as a function of composition in the mixture of A and B (5)).
Thus, denoting
3AG~ kT(pA 4- pB)2,
a -- aXA b = - a n d c = kT[3T [3.32]
' ~1
we may solve for Gas and GBS. The results are

GBS = C -- ~ S + a paVa [3.331

a
Gas = c - ~ s + ~ pBVB [3.34]
which are the required quantities.

RESULTS AND CONCLUSIONS

The two Eqs. [3.33] and [3.34] relate GBs and GAS to experimental
quantities (all measured either in pure liquid mixture of A and B or in the
limit of very dilute solution of S in such mixture). At present, there seems
to be no complete data on any three-component systems. Therefore, we
cannot use these equations to evaluate GAs and GBS separately. However,
as noted previously, the important quantity that determines the prefer-
ential solvation of S is GAS -- GBS. The sign of this quantity may be deter-
m i n e d from the slope of AG~l with respect to XA.

Ceil Biophysics Vol. 12, 1988

Theory of SolvatJon of t'lonelectrolytes 267

Fig. 5. Solvation free energy of methane in mixtures of water and etha-

nol, as a function of the mole, fraction of ethanol, at two temperatures.

In Fig. 5 we s h o w the c o m p o s i t i o n d e p e n d e n c e of the solvation free

e n e r g y of m e t h a n e as a function of the mole fraction of ethanol in
w a t e r - e t h a n o l mixtures. Similar data for m e t h a n e in mixtures of water
a n d p-dioxane are s h o w n in Fig. 6. D e n o t i n g by W and E the water a n d
the s e c o n d c o m p o n e n t (ethanol or p-dioxane), we have the relation
aAG*s kT(pE + pw) 2
= (Gws - GES) [4.1]
OXE ~1
T h u s , a positive slope of AG~ as a function of XE corresponds to a positive
difference Gws - GES > 0. This is equivalent to the s t a t e m e n t that S has
preference for water relative to ethanol. From Figs. 5 a n d 6, we observe
that the slope 3AG~/Ox E is almost always negative; i.e., S has preference

Cell Biophysics VoL 12, 1988

268 Ben-hlaim

Fig. 6. Solvation free energy of methane in mixtures of water and p-diox-

ane as a function of the mole fraction of p-dioxane at two temperatures.

for the organic component. This is true except for a relatively small re-
gion between say 0.1 ~< XE ~< 0.15 for the water--ethanol system at low
temperature (see Fig. 5) where the slope is positive. Note that at higher
temperature in this region, there is almost an inflection point where the
slope is almost zero, indicating no preferential solvation in this region.
A search in the literature reveals that there exists almost no system
for which a complete set of experimental data as required in Eqs. [3.33]
and [3.34] is available. This is unfortunate, since the required data are
quite easily measurable. We hope that this paper will encourage experi-
mentalists to undertake such measurements.
In particular, it could be interesting to examine the preferential
solvation of large biopolymers such as proteins and nucleic acids. AI-

Cell Biophysics VoL 12, 1988

Theory o f Soivation of Nonelectrolytes 269

t h o u g h the information that we can extract from such m e a s u r e m e n t s per-

tains to the average local c o m p o s i t i o n of the solvent a r o u n d the protein,
such studies could help in elucidating the concept of hydrophobicity
w h e n applied to the entire protein molecule.

ACKNOWLEDGMENT

O n the occasion of Terrell Hill's 70th birthday, it is a great pleasure

for me to a c k n o w l e d g e the major influence his books have h a d on m y
d e v e l o p m e n t as a scientist. The clarity of his t h i n k i n g a n d writing has
always b e e n a goal toward which I have striven in m y o w n work, a n d has
contributed greatly to m y enthusiastic interest in the field of statistical
mechanics.

REFERENCES
1. Engberts, J. B. F. N. (1979), in Water, A Comprehensive Treatise F. Franks, ed.,
Plenum, New York, vol. 6, ch. 4, 139.
2. Grunwald, E., Baughman, G., and Kohnstam, G. (1960), J. Am. Chem. Soc.
82, 5801.
3. Covington, A. K., and Newman, K. E. (1976), Adv. Chem. Ser. 155, 153.
4. Kirkwood, J. G., and Buff, F. P. (1951), J. Chem. Phys. 19, 774.
5. Ben-Naim, A. (1974), Water and Aqueous Solutions, Plenum, New York, ch. 4.
6. Ben-Naim, A. (1987), Solvation Thermodynamics, Plenum, New York.
7. Ben-Naim, A. (1975)]. Chem. Phys. 63, 2064.
8. Ben-Naim, A. (1977)J. Chem. Phys. 67, 4884.

CellBiophysics Vol. 12, 1988