NANO

LETTERS

Molecular Dynamics Simulation of

Contact Angles of Water Droplets in
Carbon Nanotubes

2001
Vol. 1, No. 12
697-702

Thomas Werder,*, Jens H. Walther, Richard L. Jaffe, Timur Halicioglu,

Flavio Noca,| and Petros Koumoutsakos,
Institute of Computational Science, ETH Zurich, Weinbergstr. 43,
CH-8092, Switzerland, NASA Ames Research Center, Moffett Field, California 94035,
Eloret Corporation, 690 W. Fremont AVenue, SunnyVale, California 94086, and
Jet Propulsion Laboratory, 4800 Oak GroVe DriVe, Pasadena, California 91109
Received October 3, 2001; Revised Manuscript Received October 20, 2001

ABSTRACT
We study the behavior of water droplets confined in a carbon nanotube by means of parallel molecular dynamics simulations. We report radial
density profiles, radial hydrogen bond distributions, and contact angles for tube radii ranging from 12.5 to 37.5 and for droplets containing
up to 4632 water molecules. Our results indicate nonwetting behavior of the pristine CNT at room temperatures.

Introduction. Since the discovery of carbon nanotubes

(CNT) by Iijima,1 the field of prospective applications
continues to expand. A particular area of interest is the use
of carbon nanotubes in nanofluidics applications. Nanofluidics is the study of fluid (gas, liquid) flow around and inside
nanoscale systems. Nanofluidics is envisioned as a key
technology for designing engineering devices for biological
applications, such as biomedical devices (e.g., nanoexplorers,
cell manipulators, etc.) in which the dominant biomolecular
transport process is carried out by natural and forced
convection. Other applications include the usage of CNT as
nanopipets,2 as sieves for DNA sequencing applications, and
(in arrays of carbon nanotubes) as acoustic sensors in the
form of artificial stereocilia.3 As a starting point for these
applications, we have been studying the canonical problem
of the interaction of water with carbon nanotubes, and in
particular in this paper we consider the behavior of water
droplets confined in carbon nanotubes. A fundamental issue
is the ability of a fluid to wet the interior of a CNT as this
would facilitate solution chemistry inside CNT.4 Dujardin
et al.5 found experimentally that liquids with surface tension
exceeding the limit of approximately 130-170 mN/m do
not wet single-wall CNT (SWNT) bundles. According to
these arguments water should wet CNT, as the surface
tension of water at room temperature is 72 mN/m. However,
* Corresponding author. E-mail: werder@inf.ethz.ch.
Institute of Computational Science, ETH Zurich.
NASA Ames Research Center.
Eloret Corporation.
| Jet Propulsion Laboratory.
Also at Center for Turbulence Research, NASA Ames/Stanford
University.
10.1021/nl015640u CCC: $20.00
Published on Web 11/10/2001

2001 American Chemical Society

the purity of the nanotube samples and the chemical

composition of the liquid samples were not specified. Gogotsi
et al.6,7 observed wetting of CNT by aqueous inclusions
formed during the process of hydrothermal synthesis of
multiwall CNT. In these experiments the estimated composition of the fluid inclusion is 85.2% water, 7.4% CO2, and
7.4% CH4, which may exhibit different behavior from pure
water inside a pristine SWNT. In a related study of water
droplets on a flat graphite surface under high relative
humidity, Luna et al.8,9 measured the contact angle of water
on graphite to be 30, which is indicative of pronounced
wetting. However, they state that minuscule amounts of
impurities could drastically influence their measurements.
They also report observation of a 5-nm thick mobile water
layer that slowly transforms into a 2-nm thick ice-like layer.
This apparent wetting of CNT and graphite by water is
contrasted by the general notion that the surface of graphite
is hydrophobic, that the contact angle of water on graphite
is 80-90,10 and that CNTs have a low solubility in water.11
However, it should be noted that the measured contact angle
of liquid drops on a smooth surface appears to decrease as
the size of the drop decreases (see Rieutord and Salmeron12).
The intrinsic difficulties in experimental characterization
of thermodynamic and structural properties of nanoscale
structures have motivated the extensive use of computational
techniques such as molecular dynamics simulations (MD).
Recently, Gordillo and Mart 13 conducted numerical simulations of SWNTs completely filled with water. In particular,
they find that for SWNTs with diameters in the range
between 8.2 and 16.3 , the hydrogen bond network of
the water is distorted due to the confinement of the water.

Table 1: Overview of the CNT-Water Systemsa

case

chirality

A
B
B0
B0+
B0C
G1
G2
G02

(32, 0)
(64, 0)
(64, 0)
(64, 0)
(64, 0)
(96, 0)
graphite 1L
graphite 2L
graphite 2L

Db [] NH2Oc
25.0
50.0
50.0
50.0
50.0
75.0

NCd

typee

360 4224 flexible

1872 8448 flexible
1872 8448 rigid
1872 8448 rigid
1872 8448 rigid
4632 12672 flexible
1210
576 flexible
1210 1152 flexible
1210 1152 rigid

COf [kJ mol-1]

0.3135
0.3135
0.3135
0.3762
0.2508
0.3135
0.3135
0.3135
0.3135

a All systems are periodic in the direction of the tube axis. The length
of the tubes is 138.9 and the chirality corresponds to a zig-zag CNT.b D
denotes the diameter of the CNT. c NH2O is the number of water molecules.
d N is the number of carbon atoms in one periodic image of the system.
C
e Flexible and rigid refer to the force field description of the CNT. f The
Lennard-Jones parameter CO determines the interaction strength between
the water and the graphite. The cases B0+ and B0- have a 20% larger and
lower value for CO, respectively, than the reference values. Case Gi refers
to the simulation of the planar graphite-water system, and i indicates the
number of graphite layers.

The impurities encountered in experiments (e.g., presence

of amorphous carbon and catalyst particles on the CNT
surface and ions and other liquid contaminants in the water)
can change the wetting behavior. On the other hand, the
behavior exhibited in MD simulations of pure components
is subject to the uncertainties in the choice of parameters
for the interaction potentials. In this letter we focus on
wetting aspects and on structural properties of water droplets
confined in SWNT using large scale parallel MD simulations.
We examine the equilibrium shape of the water droplet and
deduce its contact angle with the CNT surface. This is done
using isochore line plots of the water following the algorithm
suggested by Nijmeijer.14 Additionally, radial water density
profiles and radial hydrogen bond distributions are reported.
The systems considered include zigzag CNTs with diameters
of 25.0, 50.0, and 75.0 respectively and initially flat 23.7
thick drops (this corresponds to 12 layers of water molecules), see the overview in Table 1. The simulations indicate
that pure water does not wet pristine single wall carbon
nanotubes.
Methodology. We implement parallel molecular dynamics
(MD) simulations. The MD simulations are based on the
following force field: (i) The water-water interaction is
modeled by a flexible SPC water model15 that accounts for
intramolecular degrees of freedom including harmonic O-H
bond potentials and a H-O-H angle potential. The intermolecular interactions consist of an O-O Lennard-Jones
potential and a Coulomb potential between point charges
residing on the atoms. (ii) The Valence force field for the
CNT includes a Morse potential, harmonic angle potential,
and a torsion potential.16 A C-C Lennard-Jones interaction
is added with the standard 1-2 and 1-3 nearest neighbor
exclusion rules applied. (iii) The carbon-water interaction
consists of a Lennard-Jones term between the carbon and
oxygen atoms where the parameters of the potential, CO )
0.3135 kJ mol-1 and CO ) 3.19 , are obtained from the
experimental data of Bojan and Steele.17
The nonbonded interactions (including van der Waals and
Coulomb potentials) are both considered within a cutoff
698

distance rc of 10 using smooth truncation of the potentials.18 For a complete description of the potential models
and parameters used in the present study, we refer to the
work of Walther et al.16 and the references therein. The
electrostatic energy term between the carbon atom quadrupole moments and the water partial charges is not included
in this work.
The equations of motion are integrated using a leapfrog
scheme with a time step of 0.2 fs. Periodic boundary
conditions are imposed in the direction parallel to the tube
axis, and free space conditions prevail in the normal
directions. The length of the CNT is 138.9 . During the
first 4 ps of equilibration (total equilibration time is 30 ps),
the system is coupled to a Berendsen thermostat19 at a
temperature of 300 K. In the second part of the equilibration
and for the sampling, the thermostat is turned off and the
system is allowed to evolve according to the force field
described earlier. In all cases, statistics are collected during
the remaining 170 ps and samples are taken every 0.04 ps.
The present parallel implementation is based on a spatial
domain decomposition method using the message passing
interface (MPI). The bonded interactions are handled by
maintaining a global bond list that is shared by all processors
and which allows all processors to extract the local information needed while the non bonded interactions are treated
with a standard ghost layer approach. A single time step of
the largest system considered in this study (case C with 4632
water molecules and 12672 carbon atoms, see below)
requires 0.56 s on a Linux Beowulf cluster using 8 Pentium
III/500 MHz processors (100 MBit/s UTP Ethernet) and
2.93 s on a single processor, corresponding to a speed-up of
5.2.
Results and Discussion. The present study considers
nanometer-size droplets of water in carbon nanotubes of
diameters 25.0 , 50.0 , and 75.0 respectively (the cases
A, B, and C). Twelve layers of randomly oriented water
molecules form the initially flat drops. The spacing of the
water molecules on a cubic lattice is chosen such that the
initial water density amounts to 1 g cm-3. The cases B0,
B0+, and B0- have identical initial configurations as case B,
but the zero superscript signifies that the carbon nanotube
is rigid, i.e., that the position of the carbon atoms is fixed
during the simulation. Thus, the influence of the CNT on
the droplet shape is limited to the Lennard-Jones interaction,
which is controlled by the parameters CO and CO. In case
B0, the reference value of 0 ) 0.3135 kJ mol-1 is used,
while this value is increased by 20% for case B0+ and
lowered by 20% for case B0- (CO is 3.19 in all cases).
The cases G1, G2, and G02 involve a slab of water
covering a planar graphite sheet where the zero superscript
again denotes rigid graphite atoms and the subscript
encodes the number of graphite layers. The size of the
computational domain is 39 39 100 , and periodic
boundary conditions are imposed in all spatial directions.
The water slab is initially 10 molecules thick. The overall
simulation length is 200 ps for the cases A, B, and C and
100 ps for the cases B0, B0+ , B0-, G1, G2, and G02. A
summary of these systems is given in Table 1.
Nano Lett., Vol. 1, No. 12, 2001

Figure 1. Schematic of the CNT-water system. R denotes the

tube radius and z the tube axis. The contact angle is determined
from the radii d and s by the relation ) - cos-1 (d/s), where
d is an inner radius that excludes the near wall water layering and
s is the radius of the fitted sphere.

The radial density and hydrogen bond distributions are

measured in the central part of the drop using 50 radial bins
of constant volume for case A and 100 and 150 bins,
respectively, for cases B and C. This binning is centered
vertically at the center of mass of the drop, and the height
of the bins is 19.8 in all cases. For the calculation of the
water isochore profiles, a total of 250 bins is used along the
z axis while the resolution in the radial direction is similar
to that used for the radial profiles. The resulting bin spacing
along z is 0.56 and the bin volume is 5.5, 11, and 16 3
for cases A, B, and C, respectively. All quantities that are
plotted for the CNT-water cases are averaged circumferentially and are displayed as a function of the radius r. A
schematic of the CNT-water systems is shown in Figure 1.

The water radial density profiles for the three different

diameter tubes are shown in Figure 2a and 2b. Note that the
r axis corresponds to the distance from the CNT wall toward
the center of the tube. Using a standard sampling technique
(Figure 2a), the density profiles display an apparent reduction
of the density peaks for the large diameter CNTs and a shift
of position of the peaks toward the center of the tubes.
However, this significant reduction of the peaks is partially
an artifact of the binning as demonstrated in Figure 2c, where
a conditional binning has been applied. The CNTs are
modeled by a force field that accurately reflects their
mechanical behavior, e.g., it is able to reproduce the correct
vibrational modes of the CNT as reported in ref 16. Thus,
the CNT does not retain the initial cylindrical shape that is
assumed for the standard circumferential averaging. This
leads to a smearing of the radial density peaks since the
binning, as applied to a CNT with an ellipsoidal cross section,
spreads the near-wall water molecules over more bins than
it would for a CNT with a circular cross section. A
straightforward way to circumvent this problem is to use a
conditional binning, where a sample is accepted when the
standard deviation of the C atom positions from the cylindrical shape is less than 0.4 . This threshold was determined
by inspection of the evolution of the CNT cross sections in
time. Using the conditional binning, 88.8% of the samples
for case B and 26.3% of the samples for case C are accepted.
The cases where a conditional binning is applied are marked
with a prime. The comparatively small tube in case A is not
affected by the conditional binning, since it fulfills the
regularity criterion for every sample. As illustrated in Figure
2b, the lowering and shifting of the peaks is largely reduced.
The remaining differences are attributed to curvature effects;

Figure 2. Water radial density profiles and hydrogen bond profiles of a water drop inside a carbon nanotube are presented. All profiles
are measured at the center of the drop, but while in (a) and (c) a standard sampling technique is applied (cases A, B, and C), we used a
conditional binning in (b) and (c) (cases A, B, and C). The tube diameters are 25.0 (), 50.0 (s), and 75.0 (--), and R denotes the
radius of the individual tubes.
Nano Lett., Vol. 1, No. 12, 2001

699

Figure 3. Water radial density profiles and hydrogen bond profiles are presented for a water drop inside a carbon nanotube (a, c) and for
a water layer on graphite (b, d). The profiles in (a) and (c) are from the cases B0 (), B0+ (s), and B0- (- -), which denote the same
water-CNT geometry but have a different water-CNT Lennard-Jones interaction strengths. The reference value of CO ) 0.3135 kJ mol-1
for case B0 is increased by 20% in case B0+ and lowered by 20% in case B0-. Figures (b) and (d) display analogous plots for the graphite
cases, which include a single flexible layer in case G1 (- -), a double flexible layer in case G2 (), and a double rigid layer in case G02
(s).

in other words, a water molecule in a small diameter CNT

is subject to an increased influence from the wall and to a
reduction of the intermolecular water interaction.
The hydrogen bond distributions shown in Figures 2 and
3 are found using the geometrical definition of Mart,20 which
assumes that two water molecules are hydrogen bonded if
(i) the oxygen-oxygen distance is smaller than 3.6 , (ii)
the distance between the oxygen of the donor and the
hydrogen of the acceptor is less than 2.4 , and (iii) the
bond angle between the O-O direction and the molecular
O-H direction of the donor, where H is the hydrogen that
forms the bond, is less than 30. In Figure 2c the plots are
generated using all samples, while in Figure 2d only the
samples that fulfill the conditional sampling criterion are
used. In both cases the number of hydrogen bonds amounts
to approximately 3.7 in the central part of the drop, which
corresponds to the value for bulk water. It can be observed
that the number of hydrogen bonds decreases more slowly
near the wall for the larger diameter CNTs, reflecting the
fact that water molecules have fewer neighbors close to a
strongly curved wall. A molecule in the outermost water layer
forms on average approximately 2.6 hydrogen bonds.
To study the role of the water-carbon interaction, three
additional cases (B0, B0+, and B0-) were studied for systems
with rigid walls. The Lennard-Jones potential well depth
(between an oxygen and a carbon atom) was increased or
reduced by 20% from the reference value, corresponding to
-1
CO ) 0.3135 kJ mol-1, +
CO ) 0.3762 kJ mol , CO )
700

0.2508 kJ mol-1 for the cases B0, B0+, and B0-, respectively.
In Figure 3, the rigid cases B0, B0+, and B0- are compared.
The differences induced by the variation of the interaction
strength manifest themselves in the peak height and also in
the value of the contact angle, as will be seen later. The
higher the value of the parameter CO, i.e., the stronger the
water-carbon interaction, the higher the peak. The hydrogen
bond number and distribution are only slightly affected by
the difference in the interaction strength, supporting the
conjecture that these properties are governed by confinement
effects. The cases with the water slab on graphite (G1, G2,
and G02) are the limiting case of a CNT-water system
where the CNT has an infinite radius. In that sense, the
double-layer graphite cases might be perceived as models
for multiwall CNTs. The double layer cases G2 and G02
exhibit higher peaks in the radial density functions than the
single layer case G1, see Figure 3b, due to the increased van
der Waals attraction between the water and the graphite.
The notion of a contact angle in its classical meaning is a
macroscopic concept that is not fully applicable on the
nanometer scale.21 Nevertheless, Nijmeijer et al.14 suggested
a technique to visually determine the contact angle of a fluid
between two walls, which is based on the observation that
the liquid-vapor interface would relax to a spherical shape.
Following their work, we measure the contact angle by fitting
a circle to the isochore lines of 0.45, 0.55, 0.65, and 0.75 g
cm-3 using a least-squares method, see Figure 1. In this
procedure, we neglect the first 7.5 in radius of the water
Nano Lett., Vol. 1, No. 12, 2001

Figure 4. Determination of the contact angle. (a) Isochore line plot for case B at the levels 0.2, 0.4, 0.6, 0.8, and 1.0 g cm-3, respectively.
(b) Plot of the isochore lines 0.6 and 0.7 g cm-3 for the cases A, B, and C. The least squares fit of the 0.65 g cm-3 isochore line is
superposed. For the fit, only the part of the isochore lines that lies outside the CNT-water interface is considered (indicated by the vertical
lines).

Table 2: Computed Contact Angles for Tubes of Different

Diameters: 25 (case A), 50 (case B) and 75 (case C)a
isochore [g cm-3]
case

0.45

0.55

0.65

0.75

error

A
B
B
B0
B0+
B0C
C

103.0
104.4
104.3
105.3
102.6
125.3
105.3
104.4

103.0
105.2
104.9
107.2
103.8
129.2
106.1
105.3

100.9
106.4
105.8
109.9
105.0
132.2
107.2
106.2

106.2
107.6
107.0
115.4
107.7
136.4
109.1
107.9

(4.8
(0.9
(0.6
(1.1
(1.5
(2.6
(0.3
(0.4

a Cases B and C denote the results obtained using the conditional

binning. Cases B0, B0+, and B0- involve fixed CNTs with the reference
(cases B0) and perturbed (cases B0+ and B0-) Lennard-Jones parameters.
The angles are obtained by fitting a sphere through isochore surfaces of
the water meniscus. The conditional binning has a negligible influence on
the contact angle for case A, it reduces the number of samples for B and
C to 89% and 26%, respectively. The errors are the maximum errors (for
each row) as computed from the standard deviation of the fit.

meniscus from the wall, as indicated by the dashed vertical

lines in Figure 4b. This effectively excludes the large density
variations that do not contribute to the macroscopic contact
angle. The 0.6 and 0.7 g cm-3 isochore lines for the cases
A, B, and C are shown in Figure 4b with the circular curves
superimposed that fit the at 0.65 g cm-3 isochore. An
inspection of the isochores in Figure 4 reveals that a variation
of the size of the exclusion zone in a range from 5 to 10
would not affect the value of the contact angle significantly.
As an example, the complete water isochore contours for
the levels 0.2, 0.4, 0.6, 0.8, and 1.0 g cm-3 for case B are
shown in Figure 4a. The values obtained for the contact angle
are given in Table 2. The error estimates are the maximum
standard deviation of the four least squares fits in each row.
The results clearly indicate that none of the systems exhibit
wetting and that the smallest contact angle, 103, is found
for the smallest-diameter tube considered (case A). This is
in reasonable agreement with the macroscopic value meaNano Lett., Vol. 1, No. 12, 2001

sured for water on graphite (a value of 80-90 is reported

in ref 10), but it qualitatively contradicts the experimental
observations of Gogotsi et al.7 and Dujardin et al.22 However,
all of the experimental results might be compromised by
contamination of the surface or of the water. A more detailed
study of the cases A-B-C reveals that the value of the
contact angle has a slight tendency to increase with diameter,
but the differences are negligible compared to the estimated
errors. Also, the conditional binning seems not to influence
the value of the contact angle significantly, the largest
difference between case B and its conditionally binned
counterpart B is 0.6. The difference between C and C is
larger, namely 1.2, which is explained by the high rejection
rate of samples in case C. The contact angle for the
conditionally binned cases B and C are consistently smaller
than the ones found for B and C. Together with the
observation that the density peaks are larger for the double
layer graphite case G2 than for the single layer graphite case
G1, one expects that the contact angle is likely to be smaller
for multiwall CNTs than for the corresponding SWNTs. Also
interesting is the comparison between the averaged values
of the contact angle for the cases B0+, B0, and B0-, which are
104.8, 109.5 and 130.8 respectively (averaged over all
considered isochores). Even for the case B0+ with an
increased interaction strength, wetting is not observed. The
fact that a change in the parameter CO of (20% leads to a
consistent change in the contact angle suggests that the
CNT-water system could be used as a model system for
comparison with experiments.
Conclusions. We have conducted parallel molecular
dynamics simulations to investigate the wetting behavior of
water droplets confined in carbon nanotubes. Contrary to the
wetting behavior observed experimentally by Gogotsi et al.7
and by Dujardin et al.,22 the present study of pure water in
pristine CNT indicates a nonwetting behavior. For the
interaction potentials considered, a perturbation of the
carbon-water interaction energy of (20% did not alter the
701

nonwetting behavior of the interface. We do not expect

significant differences in the contact angle for armchair and
achiral tubes (the tubes in the present study are all of zigzag
type). Small changes could be caused due to a chirality
specific mechanical deformation behavior of the CNT and
by the electronic structure, which potentially leads to a
different carbon water interaction. Direct comparisons of our
MD simulations with experimental data would be desirable
but are not possible at the present time due to differences in
length and time scales as well as discrepancies in the
composition of the working materials (impurities). However,
the present parallel implementation will soon allow us to
approach regimes such as the one probed in the experiments
by Gogotsi et al.7 Further investigations are ongoing to assess
the effect of the various terms in the potential functions on
the overall wetting behavior of the CNTs. Also, the possible
influence of the presence of a pressurized inert gas on the
contact angle will be addressed to allow a closer comparison
to experiments. We believe that the present results may offer
a benchmark study for future experiments of carbon nanotube
and water interactions.
References
(1) Iijima, S. Nature 1991, 354, 56-58.
(2) Hong, M.-H.; Kim, K. H.; Bae, J.; Jhe, W. Appl. Phys. Lett. 2000,
77(16), 2604-2606.
(3) Noca, F.; Hoenk, M.; Hunt, B.; Koumoutsakos, P.; Walther, J.;
Werder, T. J. Acoust. Soc. Am. 2000, 108(5), 2494.
(4) Ugarte, D.; Stockli, T.; Bonard, J. M.; Chatelain, A.; de Heer, W. A.
Appl. Phys. A 1998, 67(1), 101-105.

702

(5) Dujardin, E.; Ebbesen, T. W.; Krishnan, A.; Treacy, M. M. J. AdV.

Mater. 1998, 10(17), 1472-1475.
(6) Gogotsi, Y.; Libera, J. A.; Yoshimura, M. J. Mater. Res. 2000, 15(12),
2591-2594.
(7) Gogotsi, Y.; Libera, J. A.; Guvenc -Yazicioglu, A.; Megaridis, C. M.
Appl. Phys. Lett. 2001, 79(7), 1021-1023.
(8) Luna, M.; Colchero, J.; Baro, A. M. J. Phys. Chem. B 1999, 103,
9576-9581.
(9) Gil, A.; Colchero, J.; Luna, M.; Gomez-Herrero, J.; Baro, A. M.
Langmuir 2000, 16, 5086-5092.
(10) Adamson, A. W.; Gast, A. P. Physical Chemisry of Surfaces, 6th
ed.; John Wiley & Sons: New York, 1997.
(11) Liu, J.; Rinzler, A. G.; Dai, H.; Hafner, J. H.; Bradley, R. K.; Boul,
P. J.; Lu, A.; Iverson, T.; Shelimov, K.; Huffman, C. B.; RodriguezMacias, F.; Shon, Y.-S.; Lee, T. R.; Colbert, D. T.; Smalley, R. E.
Science 1998, 280, 1253-1256.
(12) Rieutord, F.; Salmeron, M. J. Phys. Chem. B 1998, 102, 3941-3944.
(13) Gordillo, M. C.; Mart, J. Chem. Phys. Lett. 2000, 329, 341-345.
(14) Nijmeijer, M. J. P.; Bruin, C.; Bakker, A. F.; van Leeuwen, J. M. J.
Phys. ReV. A 1990, 42(10), 6052-6059.
(15) Teleman, O.; Jonsson, B.; Engstrom, S. Mol. Phys. 1987, 60(1), 193203.
(16) Walther, J. H.; Halicioglu, T.; Jaffe, R.; Koumoutsakos, P. J.
Phys.Chem. B 2001, 105, 9980-9987.
(17) Bojan, M. J.; Vernov, A. V.; Steele, W. A. Langmuir 1992, 8, 901908.
(18) Levitt, M.; Hirshberg, M.; Laidig, K. E.; Daggett, V. J. Phys. Chem.
B 1997, 101, 5051-5061.
(19) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; DiNola,
A.; Haak, J. R. J. Chem. Phys. 1984, 81(8), 3684-3684.
(20) Mart, J. Phys. ReV. E 2000, 61(1), 449-456.
(21) Israelachvili, J. N. Intermolecular and Surface Forces. With Applications to Colloidal and Biological Systems, 2nd ed.; Academic
Press: New York, 1992.
(22) Dujardin, E.; Ebbesen, T. W.; Hiura, H.; Tanigaki, K. Science 1994,
265, 1850-1852.

NL015640U

Nano Lett., Vol. 1, No. 12, 2001