JOURNAL OF CHEMICAL PHYSICS VOLUME 117, NUMBER 13 1 OCTOBER 2002

Grand canonical Monte Carlo simulation of hydration forces between

nonorienting and orienting structureless walls
Tomohiro Hayashi, Alexander J. Pertsin, and Michael Grunze
Angewandte Physikalische Chemie, Universität Heidelberg, INF 253, D-69120 Heidelberg, Germany
共Received 9 May 2002; accepted 10 July 2002兲
The hydration forces between structureless model walls were calculated using the grand canonical
Monte Carlo technique. Several wall–water interaction potentials were tried, including both
orientation independent and strongly directional potentials which reflected the preference of water
for tetrahedral hydrogen bonding coordination. Primary attention was given to large wall-to-wall
separations 共4 nm and more兲, where the oscillations of the hydration force due to layering effects
decayed. The hydration force was found to be highly sensitive to the presence of
orientation-dependent terms in the wall–water interaction potential. Strongly directional potentials
led to hydrophobic attraction of the walls even when the wall–water interaction was substantially
stronger than the water–water interaction. The reason had to do with the orientational ordering
induced by the walls in the adjacent water layers. © 2002 American Institute of Physics.
关DOI: 10.1063/1.1504436兴

I. INTRODUCTION surfaces, the number of simulations concerned with hydra-

The origin of forces occurring between various surfaces
in water and aqueous solutions has been discussed for a long
time, mainly in the context of the interactions in colloidal
and biological systems.1– 8 A first explanation of these forces
was given by the Derjaguin, Landau, Verwey, and Overbeek
共DLVO兲 theory3 in terms of direct van der Waals attraction
between the surfaces and screened electrostatic mean-field
repulsion between the ions adsorbed on or concentrated near
the surfaces. In more recent studies,9 the DLVO theory was
substantially improved by including the correlation contribu-
tion to the electrostatic interaction energy. As the experimen-
tal techniques for measuring water-mediated forces have be-
come available, forces of different nature have been found.
These non-DLVO forces, which are usually referred to as
hydration forces, have nothing to do with the presence of
ions and they would occur even in ideally deionized and
nondissociable water. The source of hydration forces is the
surface-induced changes in the structure and density of adja-
cent water. Hydrophobic surfaces usually attract each other
in water, whereas hydrophilic ones show water-mediated
repulsion.2 For this reason, the attractive and repulsive water-
mediated forces are frequently referred to as ‘‘hydrophobic
attraction’’ and ‘‘hydrophilic repulsion,’’ respectively. A de-
tailed discussion of water-mediated forces and their impor-
tance in colloid chemistry, biology, and other areas can be
found in numerous review articles.4 –7
Inasmuch as hydration forces originate from microscopic
structural changes in the interfacial region, the specific mo-
lecular mechanisms behind these forces are difficult to elu-
cidate experimentally. This imparts importance to the meth-
ods of computer simulation, which allow evaluation of
hydration forces based on the principles of statistical me-
chanics and an assumed form of the water–water and water–
surface interaction potentials. Despite a large body of litera-
ture on computer simulation of water in contact with solid
tion forces is very limited. The reason is the necessity of
simulating an open confined water system which is allowed
to exchange molecules and is in chemical equilibrium with a
bulk water reservoir 共Fig. 1兲. The standard molecular-
dynamics 共MD兲 technique is not well suited for such simu-
lations because it requires an explicit simulation of both the
confined water region and the bulk water reservoir. This im-
poses severe restrictions on the size of the confined system,
since it should be much smaller than the bulk water reservoir.
In this respect, the Monte Carlo 共MC兲 techniques based on
the implementation of the grand canonical 共GC兲 and isoten-
sion 共IT兲 ensembles have a great advantage because the bulk
water reservoir is present in such simulations implicitly.10–12
Nevertheless, these techniques are still of limited use be-
cause the maintenance of a constant chemical potential
within the GC and IT ensembles is computationally expen-
sive.
Nearly all of the few simulations of hydration forces
available in the literature are concerned with hydrophobic
attraction.13–16 Thus Wallqvist and Berne13 used the MD
technique to simulate the hydration force between two large
hydrophobic ellipsoids interacting with water through a re-
pulsive inverse power potential. The simulations were re-
stricted to very short separations: The largest width of the slit
between the ellipsoids was about 10 Å. As the ellipsoids
were moved together, an oscillating hydration force was ob-
served, until the constrained water between the ellipsoids
underwent capillary evaporation 共cavitation兲 leading to at-
traction between the ellipsoids due to the pressure imbalance.
The interpretation of hydrophobic attraction in terms of cap-
illary evaporation 共or ‘‘drying’’兲 can also be found in recent
publications by Chandler and co-workers.17–19 By contrast,
the ITMC simulation by Forsman et al.14 of water confined
between two hard walls at separations ranging from 10 to 23
Å revealed a strong hydrophobic attraction due to a density

0021-9606/2002/117(13)/6271/10/$19.00 6271 © 2002 American Institute of Physics

Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
6272 J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
FIG. 1. System of two parallel plates of area A, which are immersed in
water and repel each other with force Ap h ; p h is the hydration pressure and
f is the external force applied in surface-force experiments to keep the sur-
faces at a given separation H apart.
depression between the walls, with no cavitation observed.
Similar results were obtained in an early study by Luzar
et al.20 for a simple one-site model of water confined be-
tween hard walls.
To the best of our knowledge, the only published com-
puter simulation of hydration forces on hydrophilic surfaces
is the one reported by Forsman et al.21 The water–surface
interaction potential comprised a short-range exponential at-
tractive term and an inverse ninth power repulsion term. An
orientation dependent potential was also tried, which in-
cluded an additional term proportional to the cosine of the
angle between the molecular dipole moment and the surface
normal. The authors21 concluded that the hydration force was
mainly determined by the range of the water–surface poten-
tial: A strong repulsion was only observed when the potential
decay length was greater than about half the molecular di-
ameter. The inclusion of the orientation dependent term in
the potential made the hydration interaction more repulsive,
though this effect was asserted to be of minor importance.
The general objective of this work is to gain a better
understanding of the factors responsible for the sign and
magnitude of hydration forces. As in all reported simulation
studies of hydration forces,13–16,20,21 we restrict ourselves to
structureless model walls. In contrast to the previous work,
however, the focus of our simulations will be on large wall-
to-wall separations 共4 nm and more兲, where the oscillations
of the hydration force have decayed and the sign of the hy-
dration force reflects the thermodynamic affinity of the walls
for water 共see the next section兲. Both hydrophobic and hy-
drophilic walls will be considered. Similar to the Forsman’s
et al. study,21 the walls will be described using both nonori-
enting and orienting potentials. In the latter case, however,
more realistic potentials, which reflect the preference of wa-
ter for tetrahedral hydrogen bonding coordination, will be
used, including potentials which model proton-acceptor sur-
faces and also surfaces bearing both proton acceptors and
proton donors in equal amounts.22 For each particular surface
type, several discrete values of the potential well depth will
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Hayashi, Pertsin, and Grunze
be tried to follow the effect of the surface–water interaction
strength on the hydration force. In addition to changes in the
hydration force, the behavior of various distribution func-
tions and order parameters will be monitored to see how the
changes in the water–surface interaction potential affect the
structure of the adjoining water layers. Of particular interest
will be the behavior of the average density of confined water.
Since the position of a water molecule close to a hydrophilic
surface is favorable, it can intuitively be expected that the
average water density between two such surfaces will be
always enhanced. Reciprocally, the average water density be-
tween two hydrophobic surfaces can be expected to be al-
ways depressed. Whether these expectations are justified or
not will be seen from the simulations described below.
II. TERMINOLOGY
As far as the interaction of solid surfaces with water will
be concerned, we will inevitably have to employ the terms
‘‘hydrophilic’’ and ‘‘hydrophobic.’’ Despite the widespread
use of these terms in surface science, colloid chemistry, bi-
ology, and other areas, their meaning involves ambiguity be-
cause of the lack of a well-defined and commonly accepted
criterion that distinguishes surfaces into hydrophilic and hy-
drophobic. For a detailed analysis of the hydrophilic/
hydrophobic terminology, we refer the reader to a recent re-
view by Vogler23 and here we only make few comments
necessary to avoid confusion in the subsequent discussion.
The customary understanding of the terms hydrophilic
and hydrophobic is associated with the strength of the inter-
action of a surface, colloid particle or solute with water. Thus
‘‘hydrophilic 共hydrophobic兲 solute’’ is usually understood as
a solute which attracts water molecules more 共less兲 strongly
than water molecules attract one another. A similar criterion
is frequently applied to surfaces: A surface is referred to as
being hydrophilic 共hydrophobic兲 if it is capable 共incapable兲
of forming strong hydrogen bonds with water. It is clear,
however, that the hydrogen bond strength is not the only
factor responsible for the affinity of the surface for water.
Equally important are the areal density, lateral arrangement,
orientation, and flexibility of the surface groups involved in
the hydrogen bonding with water. Similar arguments apply to
hydrophobic surfaces, whose effect upon the adjacent water
may substantially differ from that of small hydrophobic
solutes.17
A thermodynamically sound criterion for the hydrophi-
licity of a surface is the condition that the water–surface
interfacial tension, ␥ ws , is negative, i.e., that an increase in
the area of the water–surface interface lowers the free energy
of the system. The criterion for hydrophobicity is opposite:
␥ ws⬎0. To relate ␥ ws to the hydration pressure p h consider
two parallel plates of an area A, which are immersed in a
water reservoir of volume V and held in equilibrium at a
separation H apart by an external force f.24 The hydration
pressure experienced by the plates can be expressed in terms
of the tension of the water film between the plates ␥. If the
system is treated using the grand canonical ( ␮ VT⫽const)
ensemble, then ␥ ⫽⌬⍀/A, where ⍀⫽U⫺TS⫺ ␮ N is the
grand potential and ⌬ denotes the difference between the
systems with and without plates. The hydration pressure is
J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
then given by differentiation, p h ⫽⫺( ⳵ ␥ / ⳵ H) ␮ T . 24 When
the surfaces are in contact (H⫽0), there is no water tension,
␥ (0)⫽0, while at H→⬁, ␥ (H)→2 ␥ ws . If ␥ ws⬍0 and ␥ is
a monotonic function of separation, then p h ⬎0, i.e., the
plates repel each other. That is the hydrophilicity criterion
␥ ws⬍0 is equivalent to the condition that p h is repulsive.
Similarly, the hydrophobicity criterion ␥ ws⬎0 reduces to the
condition that p h is attractive.
At short separations, the monotonicity of ␥ (H) can be
violated due to water layering effects, which may lead to
oscillations in p h . 3 In this case, however, the sign of p h can
still be used as a hydrophilicity/hydrophobicity criterion pro-
vided that H is large enough for the oscillations in p h to
decay.25
Based on an analysis of the available surface force lit-
erature, Vogler2 suggested to include the reference to the sign
of p h in the definition of hydrophilicity and hydrophobicity.
Considering, however, that the true hydration contribution to
the surface force may be masked by forces due to the pres-
ence of ions and impurities, p h can hardly serve as a practi-
cal criterion for hydrophilicity/hydrophobicity. By contrast, it
can well be used as such a criterion in computer simulations,
which deal with ideally pure and ion-free water.
The hydration pressure can also be related to the water
contact angle ␪, based on the Young equation, ␥ w ␯ cos(␪)
⫽␥s␯⫺␥ws , where the subscript ‘‘␯’’ refers to water vapor.
When employing ␪ or the so-called ‘‘adhesion tension’’
␥ w ␯ cos(␪) as a hydrophilicity/hydrophobicity criterion, the
dividing line between hydrophilic and hydrophobic is usually
taken to be ␪ ⫽90° or ␥ w ␯ cos(␪)⫽0 共see Vogler’s review23
for discussion of alternative choices兲, which corresponds to
the null change in free energy upon immersing the surface in
water. The offset of this dividing line from that based on ␥ ws
or p h is equal to ␥ s ␯ . Since ␥ s ␯ is always positive, the
hydrophilicity/hydrophobicity criteria based on ␪ and
␥ w ␯ cos(␪) as if overestimate the surface’s hydrophilicity
compared to the ␥ ws- and p h -based criteria: In the interval
0⬍ ␥ w ␯ cos(␪)⬍␥s␯ , ␪ and ␥ w ␯ cos(␪) show that the surface
is already hydrophilic, while it is still hydrophobic from the
viewpoint of ␥ ws and p h ( ␥ ws⬎0, p h ⬍0). In this interval,
the terms ‘‘hydrophilic repulsion’’ and ‘‘hydrophobic attrac-
tion’’ make no sense if the hydrophilicity/hydrophobicity of
the surface is understood in terms of the water contact angle.
The occurrence of an offset between the different
hydrophilicity/hydrophobicity criteria can well be seen in the
plot discussed by Vogler,2 where the characteristic decay
length of surface force for partially silanized silicas was de-
picted as a function of ␪. In this plot, the changeover from
attractive to repulsive forces occurred at ␪ ⬵62°
关 ␥ w ␯ cos(␪)⫽34 dyn cm⫺1 兴 . That is, in the range between
62° and 90° the surface was hydrophobic with regard to ␪,
while being hydrophobic with regard to p h . Similar ex-
amples were found by Besseling8 in his theoretical treatment
of hydration forces.
In the simulations described in the following sections,
the direct interaction between the constraining walls is ne-
glected, so that no work is required to spread the walls in
vacuum from contact to infinity. Hence the term ␥ s ␯ in the
Young equation vanishes. That is, for the walls studied in our
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Simulation of hydration forces between walls 6273
work, the hydrophilicity/hydrophobicity dividing lines based
on p h and ␪ are coincident.
III. MODELS
In all our simulations, the interactions of water mol-
ecules with themselves are described with the four-site
TIP4P model,26 which is perhaps the most reliable of rigid
nonpolarizable water models. 共The effect of the assumed
nonpolarizability on the behavior of water near solid surfaces
was studied by Wallqvist27 and found to be of minor signifi-
cance.兲 The summation of water–water interactions is per-
formed using a spherical cutoff 共SC兲 scheme, with a cutoff
radius of 7.5 Å. Test simulations with the cutoff radius in-
creased to 8.5 Å gave very similar results: The deviations for
the hydration pressure and average density were within the
statistical uncertainty of the calculations 共0.05 kbar and
0.004 g cm⫺3, respectively兲. The preference of the SC sum-
mation scheme over the minimal image and cylindrical cut-
off ones in simulations of confined water was demonstrated
by Shelley and Patey.28 The SC scheme was shown to pro-
vide nearly the same results as the Ewald method, while
being computationally much cheaper.
The water–wall interaction potentials are all constructed
based on a 共9-3兲 inverse power function,
冋冉 冊 冉 冊 册
␧ zm 9
zm 3
u共 z 兲⫽ ⫺3 , 共1兲
2 z z
where z is the separation of the water oxygen atom from the
wall, ␧ is the potential well depth, and z m is the equilibrium
separation, i.e., ␸ ⬘ (z m )⫽0. In all potentials, z m is fixed at
2.97 Å,29 a typical O¯O separation in the O–H¯O hydro-
gen bond. It is to be noted that the individual terms of the
共9-3兲 potential, as derived from the 共12-6兲 Lennard-Jones
potential by integration,3 are usually associated with the ex-
change repulsion and dispersion attraction, respectively. We
will, however, treat the 共9-3兲 potential in a purely formal
way, without assigning the original physical meaning to its
individual terms. In the interaction range important to our
simulations, this potential can well be fitted by the sum of
two exponential functions, with the largest deviation as small
as 0.02␧. The decay length of the fitted exponential that de-
scribes attraction proves to be 1.69 Å, which is larger than
half the molecular diameter of water. That is, based on the
simulation results of Forsman et al.,21 we can expect that the
共9-3兲 potential will allow simulation of hydration repulsion,
provided that the potential well depth ␧ is large enough to
model hydrogen bonding.
In describing nonorienting walls, isotropic water–wall
potentials are used. For these walls, hereafter referred to as
walls of the I type, ␧ is treated as a constant. A total of four
distinct values are tried for ␧. One value, ␧⫽␧ 0
⫽0.46 kcal mole⫺1 , represents a typical interaction energy
of a water molecule with hydrophobic paraffinlike surfaces.29
For walls capable of forming hydrogen bonds with water, ␧
is assigned the meaning of the hydrogen bond energy be-
tween a water molecule and the wall, ␧⫽␧ H . Three discrete
values are tried for ␧ H : 6.24, 10, and 15 kcal mole⫺1. The
first value is the binding energy of a linear TIP4P water
6274 J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
FIG. 2. Orientation dependence of the hydrogen bond potential, as specified
by h( ␪ ) in Eq. 共2兲.
dimer,26 whereas the two higher values are selected some-
what arbitrarily just to follow the effect of the water–wall
bond strength on the hydration force. From the standpoint of
the water–wall interaction energy, all the three nonorienting
walls with ␧⫽␧ H can be regarded as hydrophilic.
For orienting walls, ␧ is considered to be dependent on
the orientation of the water molecule. In constructing an ap-
propriate model for ␧, we assume that the preferred hydrogen
bonding coordination of the water molecule is tetrahedral.
With this choice, ␧ is represented as the sum of orientation-
dependent contributions from the individual water protons
and lone electron pairs:
␧⫽␧ 0 ⫹␧ H 兺i h i共 ␪ i 兲 .
In this equation h i ( ␪ i ) varies from 0 to 1 and describes the
orientation dependence of the energy of the hydrogen bond
formed by the ith water proton or lone electron pair and the
surface; ␪ i is the angle between the O–H bond or O–lone
pair vector and the outward-pointing vector normal to the
wall surface (0⭐ ␪ i ⭐180°, see Fig. 2兲. It is assumed, for
simplicity, that the angle between the two O–lone pair vec-
tors in the water molecule is the same as between the O–H
bonds 共104.52° for the TIP4P model兲.26 The numerical val-
ues of the potential parameters ␧ 0 and ␧ H in Eq. 共2兲 are taken
to be the same as discussed above for the non-orienting
walls. For the orienting walls, ␧ H takes on the meaning of
the energy of a single hydrogen bond formed by a water
molecule with the wall. The parameter ␧ 0 is introduced in the
expression for ␧ to avoid penetration of the water molecule
into the wall when no hydrogen bonds are formed, i.e., when
all h i ( ␪ i )⫽0. In modeling proton acceptor walls 共hereafter,
the walls of the A type兲, the summation in Eq. 共2兲 is per-
formed over the two water protons, while for the walls bear-
ing both proton acceptors and proton donors 共hereafter, the
AD-type walls兲, the lone electron pairs are included as well.
In this way, the ability of the wall to carry either proton
acceptors only or both proton acceptors and proton donors is
modeled.
The function h( ␪ ) in Eq. 共2兲 is in essence a somewhat
smoothed on-or-off function. As seen from Fig. 2, where
h( ␪ ) is depicted, the hydrogen bond forms only if the direc-
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Hayashi, Pertsin, and Grunze
tion of the O–H bond or O–lone pair vector is partly down,
i.e., ␪ ⬎90°. In the transition range, 90°⬍ ␪ ⭐180°⫺⌰,
h( ␪ ) is represented by a cosine function and it rapidly in-
creases from 0 to 1 as the hydrogen bond geometry becomes
closer to linear. At ␪ ⬎180°⫺⌰, h( ␪ )⫽1, so that the hydro-
gen bond energy is independent of ␪ and takes its maximum
value, ␧ H . The width of the latter interval, ⌰, is an important
parameter which specifies the directionality of the hydrogen
bond and also the relationship between the binding energies
of single, double, and triple hydrogen bonds formed by the
water molecule with the wall. The presence of an orientation
independent range in ␧共␪兲 at 180°⫺⌰⬍ ␪ ⬍180° formally
reflects the flexibility of the surface groups, which allows
them to adjust their orientations so as to meet water mol-
ecules at the most favorable angles for the formation of hy-
drogen bonds. The more flexible the surface groups, the
greater ⌰. For this reason, ⌰ will hereafter be referred to as
the flexibility parameter. As ⌰ is decreased, the orientation
independent range in ␧共␪兲 becomes narrower and the orient-
ing effect of the wall is enhanced.
For walls of the A type, the energies of the strongest
double and single hydrogen bonds can be easily shown to be
in the ratio 2:1 when ⌰⬎⬔HOH/2⬵52°. At ⌰⫽30°, this
ratio reduces to 1.4:1, which is close to the ab initio elec-
tronic structure calculation results for the double and single
hydrogen bonds formed by water with the oxygen atoms of
an oligo共ethylene oxide兲 chain.30 For walls of the AD type,
the ratio 3:2:1 for the triple, double, and single hydrogen
bond energies is attained when ⌰ is chosen to be greater than
⬃70°.
共2兲
IV. SIMULATION METHOD
As already mentioned in the Introduction, there are two
basic MC techniques suitable for simulating open confined
fluid systems, as studied in surface-force experiments 共Fig.
1兲. In the GCMC technique, the chemical equilibrium of the
confined system with the 共fictitious兲 bulk reservoir is main-
tained by allowing density fluctuations through particle in-
sertions and deletions. The main problem involved in the use
of this technique is a too low probability of performing suc-
cessful insertions and deletions at densities typical of fluids.
In the past decade, however, this problem has been more or
less successfully solved using Swendsen–Wang filters,31 cav-
ity bias,32 and related methods11 based on simple and com-
putationally inexpensive predictors for the success of inser-
tion and deletion moves. The alternative to the GCMC
technique is provided by ITMC simulations. In these simu-
lations, the density fluctuations are implemented by allowing
area fluctuations parallel to the confining walls at a fixed
lateral pressure P L . As the separation between the walls is
changed, the chemical potential is maintained constant by an
appropriate adjustment of P L using the free-energy differ-
ence method.12 The ITMC technique, however, cannot be
applied to simulation of fluids near structured solid sub-
strates because the area fluctuations in the fluid are inconsis-
tent with the condition that the lateral dimensions of the
simulation cell must be commensurate with the substrate lat-
J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
tice. Considering our parallel simulations of the behavior of
water near structured surfaces,33 we prefer the GCMC tech-
nique over the ITMC one.
The computational procedure used in our GCMC simu-
lations is similar to that described in our previous work.33 To
improve the efficiency of insertions, the excluded volume
mapping method34 and a Swendsen–Wang filter31 based on
evaluation of the van der Waals energy of the system are
employed. The usual length of the MC runs is 3⫻106 to 1
⫻107 passes, each comprising N moves, where N is the cur-
rent number of water molecules in the system. For a typical
system size used, this corresponds to a total of 1.5⫻109 to
5⫻109 attempted configurations.
The simulation box represents a rectangular prism with
dimensions L x , L y , and L z . The lower confining wall is
placed at z⫽0 parallel to the x-y plane. In the x and y di-
mensions, the system is replicated periodically, with periods
L x ⫽30 Å and L y ⫽26 Å. Along the z axis, the ‘‘gliding
plane’’ boundary 共GPB兲 conditions,33 which combine a mir-
ror plane at z⫽L z with a half period translation along x, are
used. That is, the space above the top face of the simulation
cell is filled with a mirror image of the system shifted along
x by L x /2. 共The shift is necessary to avoid straightforward
correlations between the displacements of a particle and its
image across the mirror plane.兲 Because of the mirror plane
at z⫽L z , the upper confining wall occurs at z⫽H⫽2L z .
The transformation of coordinates by the GPB conditions is
the following:
x ⬘ ⫽x⫹kL x /2
共 k⫽1 for x⬍L x /2 and k⫽⫺1 otherwise兲 , 共3兲
y ⬘ ⫽y, 共4兲
z ⬘ ⫽2L z ⫺z. 共5兲
As a molecule leaves the simulation cell through the top
face, its image, whose coordinates are given by Eqs. 共3兲–共5兲,
enter the cell through the other half of the top face.
Although the GPB conditions save nearly half the CPU
time, they may, in principle, affect the simulation results be-
cause of the artificial correlations introduced by Eqs. 共3兲–共5兲
in the configuration of the system near the midplane z
⫽L z . To assess the importance of these correlations, we
made a comparative simulation of two bulk water systems.
One system was simulated using the usual periodic boundary
conditions in all three dimensions. In the other system, the
periodic boundary conditions were only applied along the x
and y axes, while along the z axis the system was replicated
by applying the GPB conditions at z⫽0 and z⫽L z . The
three lengths of the simulation box were set equal to 30 Å.
The calculated thermodynamic and structural quantities
proved to be practically independent on whether the periodic
or GPB conditions were used along the z axis.
V. QUANTITIES CALCULATED
The set of thermodynamic and structural quantities cal-
culated in our simulations is basically the same as described
in our previous work.33 In addition, we calculate the hydra-
tion pressure
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Simulation of hydration forces between walls 6275
p h ⫽ 具 f 典 /A⫺ p b , 共6兲
where p b is the bulk water pressure and 具f典 is the mean force
exerted by water molecules upon the lower surface,
具 f 典⫽ 冓兺 冔 i
⳵u
⳵ z ic
. 共7兲
In this equation, the angular brackets denote GC ensemble
averaging, u is the wall–water potential, as given by Eq. 共1兲;
the differentiation is performed with respect to the z coordi-
nate of the center of mass of the ith water molecule. With our
choice of the z axis along the outward-pointing vector nor-
mal to the lower confining surface, positive values of p h
would correspond to attraction between the confining sur-
faces. To make our results consistent with the conventional
choice of the sign of p h , 21 we will hereafter refer to the
values of p h on the upper confining surface, so that positive
p h will correspond to repulsion, while negative p h to attrac-
tion.
The orientational structure of water at different separa-
tions from the wall is monitored by calculating the distribu-
tion of the angles, ␪ ␮ and ␪ OH , formed, respectively, by the
molecular dipole moments and O–H bonds with the z axis.
Also calculated are the orientational order parameter,
S 10 共 z 兲 ⫽ 具 cos ␪ ␮ 共 z 兲 典 , 共8兲
and the integrated density-weighted order parameter,
⌶⫽ 冉冕 0
H/2
dz ␳ 共 z 兲 冊 冕
⫺1
0
H/2
dzS 10 共 z 兲 ␳ 共 z 兲 . 共9兲
While S 10 (z) is the average of cos ␪␮ over the molecules ly-
ing at distance z from the wall, ⌶ represents the average of
cos ␪␮ over all molecules between the wall and the midplane
at z⫽H/2.
The average water density in the confined region is cal-
culated in a straightforward way as ¯␳ ⫽ 具 N 典 /V c , where V c
⫽L x L y L z is the volume of the simulation cell. In the calcu-
lations of the density profiles, orientational distributions, and
other quantities dependent on the distance from the wall, the
simulation box is divided into 100 slices of thickness ⌬z
⫽L z /100 lying parallel to the wall. During the GCMC run,
the z-dependent quantities are averaged within each indi-
vidual slice and then referred to the z coordinate of its center.
VI. RESULTS AND DISCUSSION
A. Bulk water
The knowledge of the chemical potential of bulk water is
a prerequisite of simulating confined water in the chemical
equilibrium with the bulk water reservoir 共Fig. 1兲. For rigid
water models, such as TIP4P, only the excess 共nonideal兲 part
of the chemical potential, ␮ ⬘ , is required. An early MD
simulation by Hermans et al.35 using the thermodynamic in-
tegration method resulted—for the TIP4P model at room
temperature—in a value of ⫺5.3 kcal mole⫺1, in satisfactory
agreement with an experimental estimate of ⫺5.7
kcal mole⫺1 based on the known ratio between the molar
volumes of water vapor and liquid. Note that Hermans
et al.’s simulations35 were made with a very small periodic
6276 J. Chem. Phys., Vol. 117, No. 13, 1 October 2002 Hayashi, Pertsin, and Grunze

system (N⫽80) and a fairly short cutoff distance 共6 Å兲.

More recently, Shelley and Patey28 reported a value of ⫺6.0
kcal mole⫺1 as best reproducing the room-temperature bulk
water density when the TIP4P model and Ewald summation
method were employed.
To find ␮ ⬘ suited to the particular summation scheme,
cutoff distance, and typical system size used in our simula-
tions of confined water, we performed a series of bulk water
simulations as a function of ␮ ⬘ . In these simulations, the
periodic boundary conditions were used in all three dimen-
sions. The experimental bulk water density at room tempera-
ture, ␳ b ⫽0.997 g cm⫺3 , was reproduced at ␮ ⬘ ⫽
⫺6.10 kcal mole⫺1 , close to the Shelley and Patey result.28
However, the respective bulk pressure, as calculated from the
virial formula, proved to be 0.13⫾0.03 kbar, somewhat
higher than the atmospheric pressure. The desired pressure FIG. 3. Water density distributions for the hydrophobic I-type wall at two
was obtained at ␮ ⬘ ⫽⫺6.15 kcal mole⫺1 . The corresponding wall-to-wall separations H.
equilibrium density was 0.991 g cm⫺3, 0.6% lower than re-
quired. The observed small inconsistency between the calcu-
lated equilibrium pressure and density of TIP4P water is The water density distribution ␳ (z) near the ␧⫽␧ 0 wall
hardly surprising because the original fitting of the TIP4P is shown in Fig. 3 for the largest separation tried 共65 Å兲 and
model26 to the experimental properties of water employed a also for a separation of 58.6 Å just beyond H c . The vanish-
different cutoff distance, number of molecules in the simula- ing density at z⬍2 Å is associated with the sharply ascend-
tion cell, and statistical ensemble 共NPT兲 共Ref. 36兲 than our ing water–wall repulsion and it is observed for all the walls
simulations. In deciding between the two alternative values studied in this work. By contrast, the substantially depressed
found for ␮ ⬘ , our reasoning was that the ‘‘overpressure’’ of density in the range 3⬍z⬍6 Å is a distinguishing feature of
0.13 kbar could be readily compensated for through Eq. 共6兲, the ␧⫽␧ 0 I-type wall, which can be compared with the the-
unlike the shift in density, which might have an unpredict- oretical predictions by Lum et al.17 for water in contact with
able effect on the structural organization of water near con- hydrophobic walls. As H approaches H c , the density depres-
fining surfaces. So, the excess chemical potential of ⫺6.10 sion progresses. The average water density between the walls
kcal mole⫺1, which exactly reproduced ␳ b , and somewhat is also substantially depressed 共Table I兲.
overestimated p b was adopted in all subsequent calculations. Although the water–wall interaction potential for the
I-type surfaces contains no explicit orientation-dependent
terms, the ␧⫽␧ 0 wall does have a perceptible orienting ef-
B. Nonorienting walls fect on the neighboring water molecules. This can be seen
We begin the discussion of the I-type walls with the one from Fig. 4, which shows the distribution of angles formed
characterized by the most shallow potential (␧⫽␧ 0 ). The by the O–H bond vector and the z axis for different separa-
behavior of water between two such walls was studied in the tions from the wall. The orientation of a typical water mol-
separation range 30⭐H⭐65 Å. At the wall-to-wall separa- ecule at the wall is such that one of its O–H bonds is pref-
tions less than a critical separation H c ⫽58.4 Å, the confined erentially directed toward the surface ( ␪ OH⫽180°), in
water experienced capillary evaporation 共cavitation兲, typical agreement with the results of MD simulations by Lee and
of hydrophobic confining surfaces: Starting with a certain Rossky29 who used the same potential model for water and
length of the MC run, the number of water molecules rapidly the wall. With increasing separation, the molecules reorient
decreased until nearly all of the molecules left the confined so that one of their O–H vectors look preferentially outward
region. At separations in a ⬃0.5 Å range below H c , the state from the surface ( ␪ OH⫽0). The orientational ordering near
of the system was dependent on the starting configuration the ␧⫽␧ 0 wall was found to be short range and lost almost
used to initiate the MC run: Some configurations resulted in completely at separations greater than two molecular diam-
cavitation, while the other remained liquid up to 107 MC eters.
passes. It is therefore quite likely that H c would shift to
larger separations if the MC runs would be longer. Note that
the observed H c ⫽58.4 Å agrees with a theoretical mean- TABLE I. Average density, hydration pressure, and integrated order param-
field estimate of the limit of metastability reported by Lum eter for water confined between I-type walls.
et al.17 for water confined between two hard walls 共⬃50 Å兲.
␧, kcal mole⫺1 H, Å ¯␳ , g cm⫺3 p h , kbar ⌶
On the other hand, H c is about an order of magnitude larger
than the critical slit width found by Wallqvist and Berne13 for 0.46 65 0.898 ⫺0.51 0.011
two hydrophobic ellipsoids 共⬃6.4 Å兲. A likely reason is a 0.46 58.6 0.885 ⫺0.53 0.010
6.24 40 0.980 ⫺0.12 0.018
comparatively small size and large curvature of the ellip-
10 40 1.013 0.13 0.017
soids, as well as the absence of an attractive contribution in 15 40 1.040 0.28 0.018
the water–ellipsoid interaction potential.

Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
FIG. 4. Orientational distribution of O–H bonds for some selected separa-
tions from I-type walls with ␧⫽0.46 and 6.24 kcal mole⫺1.
The simulation results for p h at H⬎H c 共Table I兲 show a
strong hydrophobic attraction between the ␧⫽␧ 0 walls,
which can be associated with the reduction in water density.
That is, as the separation between the walls is increased, the
driving force behind the hydrophobic attraction changes
from cavitation to density depression. Importantly, the mag-
nitude of p h due to the density depression substantially ex-
ceeds the upper limiting value of p h due to cavitation, p h
⬍p b 共0.13 kbar in our case兲. This result can be roughly
understood in terms of the fact that the compressibility of
liquid water is very low, so that a small reduction in the
water density may result in a significant stretching of the
confined water. The substantial exceeding of p h over p b also
means that the treatment of the hydrophobic attraction solely
in terms of cavitation or ‘‘drying’’ can hardly provide a com-
prehensive description of this phenomenon.
The increase in the wall–water interaction strength on
going from the ␧⫽␧ 0 to ␧⫽␧ H walls is accompanied by
substantial changes in the structure and properties of con-
fined water. Already at ␧ H ⫽6.24 kcal mole⫺1 , the water den-
sity distribution becomes sharply structured and shows two
distinct hydration layers next to the wall 共Fig. 5兲. The orien-
FIG. 5. Water density distribution for the I-, A-, and AD-type walls with
␧ H ⫽6.24 kcal mole⫺1 .
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Simulation of hydration forces between walls 6277
FIG. 6. Hydration pressure as a function of wall-to-wall separation for the
I-, A- (⌰⫽30°), and AD-type walls with ␧ H ⫽6.24 kcal mole⫺1 .
tational ordering perceptibly increases, which can be judged
from the increase of the density-weighted order parameter ⌶
in Table I. A typical water molecule in the first hydration
layer now has one of its O–H bonds directed outward the
wall 共Fig. 4兲.
The behavior of p h as a function of H is shown in Fig. 6
together with the data for the ␧ H ⫽6.24 kcal mole⫺1 walls of
the other types. Because of the high computational cost of
these simulations, we restricted ourselves only to few se-
lected separations and did not undertake calculation of the
detailed profile p h (H). It can nevertheless be seen that p h
ceases changing its sign at H⬇18 Å, long before the sepa-
ration of 40 Å used in most of our simulations. 共Note that at
␧ H ⭓6.24 kcal/mole, all the walls showed no cavitation at all
separations tried.兲
From the negative sign of p h exhibited by the I-type
␧ H ⫽6.24 kcal mole⫺1 wall at large separations, we con-
clude that it is hydrophobic. The origin of its hydrophobicity
can be understood from a simple analysis of the individual
components of the water film tension, ␥ ⫽(⌬U⫺T⌬S
⫺ ␮ ⌬N)/A. At the given ␧ H and H⫽40 Å, ¯␳ is fairly close
to ␳ b 共see Table I兲 and hence ⌬N vanishes. For the
ensemble-averaged potential energy, the simulations give
⫺11.4 kcal mole⫺1, which is substantially lower than that for
bulk water 共⫺10.1 kcal mole⫺1兲. Considering that these en-
ergies refer to the systems with nearly the same N, we obtain
that ⌬U⬍0. Thus we can conclude that the positive sign of ␥
can only result from a negative ⌬S. The obvious cause of the
entropy loss in the confined region is the surface-induced
ordering. The entropic factor was earlier found to be domi-
nant for water confined between hydrophobic ellipsoids13
and it turns out to be important in our case, too.
As ␧ H is increased to 10 and then 15 kcal mole⫺1, the
average potential energy decreases to ⫺12.3 and ⫺13.5
kcal mole⫺1, respectively. This more than compensates for
the loss in entropy with the result that p h goes repulsive and
the wall becomes hydrophilic 共Table I兲. Simultaneously, ¯␳
becomes greater than ␳ b , as intuitively expected.
6278 J. Chem. Phys., Vol. 117, No. 13, 1 October 2002 Hayashi, Pertsin, and Grunze

TABLE II. Average density, hydration pressure, and integrated order param-
eter for water confined between A-type walls (H⫽40 Å).

␧ H , kcal mole⫺1 ⌰, deg ¯␳ , g cm⫺3 p h , kbar ⌶

6.24 30 0.938 ⫺0.37 ⫺0.052

10 30 0.965 ⫺0.21 ⫺0.090
15 30 0.992 ⫺0.09 ⫺0.123
6.24 75 0.982 ⫺0.15 ⫺0.050
10 75 1.007 ⫺0.07 ⫺0.078
15 75 1.040 ⫺0.02 ⫺0.108

drophobic at all ␧ H tried. From Tables I and II, one can also
see that at a given ␧ H the average water density between the
A-type walls is higher than that between the I-type walls, so
FIG. 7. Orientational order parameters S 10 for the I-, A- (⌰⫽30°), and
that the more attractive hydration forces observed between
AD-type walls with ␧ H ⫽15 kcal mole⫺1 . the A-type walls cannot be explained in terms of density
depression.
A comparison of the I-and A-type walls in terms of the
internal energy shows that the latter are 0.7–1.1 kcal mole⫺1
C. Proton-acceptor walls
inferior. The reason can be understood from Fig. 8, which
The simulations of water near the walls of the A type presents the average interaction energy of a water molecule
were performed for two different flexibility parameters, ⌰ with its surroundings as a function of its separation from the
⫽30° and 75°. We first discuss the results for ⌰⫽30°, I- and A-type walls with ␧ H ⫽15 kcal mole⫺1 . Both the total
which correspond to a stiffer hydrogen bond and hence a interaction energy and its constituents due to the water–
stronger orienting effect of the wall on the neighboring water water and water–wall interactions are shown. One can see
molecules. Compared to the I-type wall with ␧ H that the residence of a water molecule near the A-type wall is
⫽6.24 kcal mole⫺1 , the water density distribution near the energetically less favorable, mainly because of an appre-
A-type wall with the same ␧ H is less structured 共Fig. 5兲. Both ciable loss in the water–water interaction energy. The energy
the first and second density maxima of ␳ (z) are substantially loss is particularly pronounced in the first hydration layer,
lower in intensity and the gap between them is partly filled. where the orientational ordering is strongest and each water
An analysis of the angular distribution of the molecular di-
pole moments and O–H vectors shows that the water mol-
ecules near the wall of the A type have two distinct preferred
orientations associated with the formation of double and
single hydrogen bonds with the wall. The presence of two
kinds of hydrogen bonds, which differ noticeably in strength,
leads to a wide variation in the wall–water bond lengths and
the associated smearing of ␳ (z) between the first and second
maxima.
The property of the A-type wall to bind water molecules
in a highly asymmetric way, with their O–H bonds directed
preferentially towards the wall, brings about strong and fairly
long-ranged orientational ordering. This can be appreciated
from Fig. 7, which compares the order parameters S 10 (z) for
water in contact with the I-, A-, and AD-type walls. In the
case of the I-type walls, which are characterized by oscilla-
tions of S 10 (z) about zero, ⌶ is fairly small, positive, and
practically independent of the strength of the water–wall in-
teractions, as specified by the magnitude of ␧ H 共Table I兲. By
contrast, water in contact with the A-type walls shows large
negative ⌶, strongly dependent on ␧ H 共Table II兲.
As noted in Sec. III, the A-type wall with ⌰⫽30° is
capable of forming double hydrogen bonds with a water mol-
ecule with a total binding energy of 1.4␧ H . That is, based on
the wall–water interaction strength, the A-type walls can be
expected to be more hydrophilic than the I-type walls with
the same ␧ H . In spite of this, however, the former are found FIG. 8. Average interaction energy of a water molecule with its surround-
to result in lower p h 共cf. Tables I and II兲. Moreover, the ings and the individual contributions to this energy for the I-type and A-type
A-type walls all show attractive p h and hence they are hy- (⌰⫽30°) walls with ␧ H ⫽15 kcal mole⫺1 .

Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
J. Chem. Phys., Vol. 117, No. 13, 1 October 2002
molecule has to sacrifice at least one hydrogen bond with its
neighbors in order to bind to the wall. The constraints im-
posed by the orientational ordering upon the freedom of the
water molecules to form hydrogen bonds with each other
remain important in the second hydration layer and at farther
separations. In this separation range, no hydrogen bonds with
the wall can be formed, yet the loss of the water–water in-
teraction energy is quite perceptible.
The presence of sharply directional interactions at the
wall–water interface affects not only the water–water but
also the water–wall contribution to the average interaction
energy. From the respective curve in Fig. 8, one can see that
the interaction energy of the A-type wall with a typical water
molecule in the first hydration layer is only slightly greater in
magnitude than the energy of a single hydrogen bond (␧ H
⫽15 kcal mole⫺1 for the example shown in the figure兲. This
suggests that the too stiff hydrogen bond potential used in
this simulation impedes the formation of double hydrogen
bonds with the wall. All the above-mentioned factors make
the A-type walls energetically less favorable compared to the
I-type walls. The stronger orientational ordering induced by
the A-type walls should also make them unfavorable from
the entropic standpoint, too.
The increase of the flexibility parameter ⌰ from 30° to
75° makes all O–H bond orientations with ␪ i ranging from
105° to 180° energetically equivalent. This reduces the ori-
entational ordering 共see the changes in ⌶ in Table II兲 and
affords more orientational freedom for the water molecules
in forming hydrogen bonds with their neighbors and the
wall. As a consequence, the wall becomes less hydrophobic
共or more hydrophilic兲 which manifests itself in an increase in
ph .
Turning to the calculated average densities in Table II,
one can see that the changes in the density and hydration
pressure are symbate: With increasing ␧ H , both ¯␳ and p h
increase. At the same time, the intuitive expectation that for
hydrophobic confining walls ¯␳ should always be lower than
␳ b proves to be fallacious: At ⌰⫽75° and ␧ H
⭓10 kcal mole⫺1 , the system shows an enhanced ¯␳ along
with an attractive 共hydrophobic兲 p h . A similar example has
been reported by Besseling8 in his theoretical work.
D. Walls bearing both proton acceptors
and proton donors
With the accepted orientation of the lone pairs in the
water molecule, an AD-type wall sees the molecule as pos-
sessing a V d symmetry. This means, in particular, that the
wall–water potential is invariant with respect to a fourfold
inversion axis transformation of the water molecule and the
associated change in the sign of the molecular dipole mo-
ment. The water–water potential, however, does not possess
this symmetry. As a consequence, the whole system may
well show a nonvanishing S 10 even though the wall itself does
not distinguish between the symmetrically related configura-
tions with opposite dipole moments. 共Note that similar argu-
ments apply to the I-type surfaces, which show noticeable S 10
and ⌶ despite the fact that the I-type walls perceive the water
molecule as being isotropically symmetric.兲
Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp
Simulation of hydration forces between walls 6279
TABLE III. Average density, hydration pressure, and integrated order pa-
rameter for water confined between AD-type walls (H⫽40 Å, ⌰⫽75°).
␧ H , kcal mole⫺1 ¯␳ , g cm⫺3 p h , kbar ⌶
6.24 1.023 0.05 0.000
10 1.059 0.12 ⫺0.003
15 1.093 0.15 ⫺0.033
The orientational order generated in water by an AD-
type wall can be appreciated from Fig. 7 for the case ␧ H
⫽15 kcal mole⫺1 where the magnitude of S 10 is greatest. The
ordering is seen to be particularly pronounced in the range
from 6 to 7 Å and from 8.5 to 9.5 Å, where S 10 assumes large
negative values. These ranges, however, correspond to
minima in the water density distribution ␳ (z), while the
positive portions of S 10 correspond to maxima of ␳ (z) 共Fig.
5兲. As a consequence, the integrated density-weighted pa-
rameter ⌶ proves to be fairly small 共Table III兲. As ␧ H is
decreased, ⌶ vanishes.
For all tried ␧ H , p h ⬎0 and hence the AD-type walls are
all hydrophilic. The average density ¯␳ is greater than ␳ b for
all ␧ H 共Table III兲. An analysis of the energetics of the con-
fined water region in terms of the water–water and water–
wall interactions shows that the AD-type walls are interme-
diate in the water–water interaction energy between the I-
and A-type walls with the same ␧ H . However, the ability of
the AD-type walls to form triple hydrogen bonds with water,
with a total binding energy of 3 ␧ H , makes this type of wall
much more favorable for water, compared to the I- and
A-type walls.
VII. CONCLUDING REMARKS
In this work we used the GCMC technique to study the
hydration forces between structureless flat walls. Most atten-
tion was given to large wall-to-wall separations, where the
force oscillations due to the water layering effects decayed
and the sign of the hydration pressure can be used as a cri-
terion for the hydrophilicity/hydrophobicity of the walls. For
the wall with the weakest interaction with water (␧
⫽0.46 kcal mole⫺1 ), we observed capillary evaporation at
H⭐58.4 Å and a strong attraction due to density depression
at larger H. In the latter case, the magnitude of p h substan-
tially exceeded p b 共the upper limiting pressure due to evapo-
ration兲, which cast some doubt on the treatment of capillary
evaporation as the main source of hydrophobic attraction.
Our results for the hydrophilic walls are, in some re-
spects, opposite to the findings reported by Forsman et al.21
In contrast to the cited work,21 we found that the inclusion of
orientation dependent terms in the water–wall interaction po-
tential added an attractive contribution to the hydration force
by making the confined region less favorable for water from
both the entropic and energetic points of view. While the loss
in entropy can be well understood in terms of the surface-
induced orientational ordering and confinement of the libra-
tional motion of water molecules, the reasons of the energy
loss are not so apparent. One is the reduction of the water–
water contribution to the internal energy due to the distur-
bance of the hydrogen bonding network natural for water.
Another reason has to do with constraints imposed by the
6280 J. Chem. Phys., Vol. 117, No. 13, 1 October 2002 Hayashi, Pertsin, and Grunze

directionality of the hydrogen bond upon the ability of water
molecules to form multiple hydrogen bonds with the surface.
In the model potentials used in describing the water–surface
interaction, the orientation dependence of the hydrogen bond
energy reflects not only the intrinsic trend of the hydrogen
bond to linearity but also the flexibility of the surface groups
involved in the hydrogen bonding, i.e., the ability of these
groups to adapt their orientations to the approaching water
molecules. The confinement of this ability should obviously
be energetically unfavorable.
ACKNOWLEDGMENTS
This work was supported by the Deutsche Forschungs-
gemeinschaft, the Office for Naval Research, and the Fond
der Chemischen Industrie. The authors thank Professor E. A.
Vogler, Professor N. Spencer, Dr. P. Schmidt, and Dr. G.
Hähner for critical reading of the manuscript and helpful
comments.
1
J. Israelachvili and H. Wennerström, Nature 共London兲 379, 219 共1996兲.
2
E. A. Vogler, Adv. Colloid Interface Sci. 74, 69 共1998兲.
3
J. Israelachvili, Intermolecular and Surface Forces 共Academic, London,
1992兲.
4
J. Lyklema, Fundamentals of Interface and Colloid Science I: Fundamen-
tals 共Academic, London, 1991兲.
5
J. Lyklema, Fundamentals of Interface and Colloid Science II: Solid-
Liquid Interfaces 共Academic, London, 1991兲.
6
R. P. Rand and V. A. Parsegian, Biochim. Biophys. Acta 778, 224 共1989兲.
7
V. A. Parsegian, Adv. Colloid Interface Sci. 16, 49 共1982兲.
8
N. A. Besseling, Langmuir 13, 2113 共1997兲.
9
R. Podgornik, J. Chem. Phys. 91, 5840 共1989兲.
10
M. P. Allen and D. J. Tildesley, Computer Simulation of Liquids 共Claren-
don, Oxford, 1993兲.
11
J. C. Shelley and G. N. Patey, J. Chem. Phys. 102, 7656 共1995兲.
12
B. Svensson and C. E. Woodward, J. Chem. Phys. 100, 4575 共1994兲.
13
A. Wallqvist and B. J. Berne, J. Phys. Chem. 99, 2893 共1995兲.
14
J. Forsman, B. Jönsson, and C. E. Woodward, J. Phys. Chem. 100, 15005
共1996兲.
15
J. Forsman, B. Jönsson, and T. Åkesson, J. Phys. Chem. B 102, 5082
共1998兲.
16
J. Forsman, C. E. Woodward, and B. Jönsson, J. Colloid Interface Sci.
195, 264 共1997兲.
17
K. Lum, D. Chandler, and J. D. Weeks, J. Phys. Chem. B 103, 4570
共1999兲.
18
D. M. Huang, P. L. Geissler, and D. Chandler, J. Phys. Chem. B 105, 6704
共2001兲.
19
P. G. Bolhuis and D. Chandler, J. Chem. Phys. 113, 8154 共2001兲.
20
A. Luzar, D. Bratko, and L. Blum, J. Chem. Phys. 86, 2955 共1987兲.
21
J. Forsman, C. E. Woodward, and B. Jönsson, Langmuir 13, 5459 共1997兲.
22
Surfaces bearing solely proton donors are of little practical importance as
far as strong hydrogen bonds such as O–H¯OH2 or N–H¯OH2 are
concerned. In both cases, the proton-donor group can also serve as a
proton acceptor for the water molecule, so that the respective surfaces
cannot formally be regarded as purely proton donating.
23
E. A. Vogler, in Water in Biomaterials Surface Science, edited by M.
Morra 共Wiley, New York, 2001兲, pp. 149–182.
24
R. Evans and U. M. B. Marconi, J. Chem. Phys. 86, 7138 共1987兲.
25
As shown by Besseling 共Ref. 8兲 in his mean-field lattice theory of hydra-
tion forces, ␥ (H) may have an extremum regardless of the presence of
layering effects. Such a situation, which does not allow the use of p h as a
hydrophilicity/hydrophobicity criterion at long separations, can, however,
be readily recognized both in real and computer experiments.
26
W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L.
Klein, J. Chem. Phys. 79, 926 共1983兲.
27
A. Wallqvist, Chem. Phys. Lett. 165, 437 共1990兲.
28
J. C. Shelley and G. N. Patey, Mol. Phys. 88, 385 共1996兲.
29
S. H. Lee and P. J. Rossky, J. Chem. Phys. 100, 3334 共1994兲.
30
R. L. C. Wang, H. J. Kreuzer, and M. Grunze, Phys. Chem. Chem. Phys.
2, 3613 共2000兲.
31
R. H. Swendsen and J.-S. Wang, Phys. Rev. Lett. 58, 86 共1987兲.
32
M. Mezei, Mol. Phys. 61, 565 共1987兲.
33
A. J. Pertsin and M. Grunze, Langmuir 16, 8829 共2000兲.
34
M. R. Stapleton and A. Panagiotopoulos, J. Chem. Phys. 92, 1285 共1990兲.
35
J. Hermans, A. Pathiaseril, and A. Anderson, J. Am. Chem. Soc. 110, 5982
共1988兲.
36
The difference in the statistical ensembles could introduce some discrep-
ancy in view of the relatively small system size used in deriving the TIP
4P model (N⫽125) 共Ref. 26兲.

Downloaded 17 Oct 2002 to 129.206.85.25. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/jcpo/jcpcr.jsp