Journal of Colloid and Interface Science 324 (2008) 9–14

Contents lists available at ScienceDirect

Journal of Colloid and Interface Science

www.elsevier.com/locate/jcis

Monte Carlo simulations of phase transitions and adsorption isotherm

discontinuities on surface compression
C.L. Charniak, T.E. Wetzel, G.L. Aranovich, M.D. Donohue ∗
Johns Hopkins University, Baltimore, MD 21218, USA

a r t i c l e i n f o a b s t r a c t

Article history: Low temperature, Grand Canonical Monte Carlo simulations were used to study the adsorption of fluid
Received 19 February 2008 layers on cubic, hexagonal, and atomically smooth substrates to determine the effects of registry and
Accepted 28 April 2008 surface compression on the system. The size of the fluid molecules was fixed to be 20% larger than the
Available online 1 May 2008
substrate molecules in order to observe the transition from an expanded to commensurate and finally
Keywords:
to an incommensurate monolayer. For relatively weak fluid–substrate interactions, the cubic system un-
Adsorption derwent a first-order phase transition. As the strength of the fluid–substrate interactions increased, the
Monte Carlo molecules became fixed at commensurate locations and the transition from low density to commensu-
Phase transition rate packing became continuous. The strong fluid–substrate interactions lead to the development of a
kink in the adsorption isotherm that showed the increased stability of the commensurate phase. This
kink became more pronounced as the system temperature was decreased. The hexagonal system showed
less dramatic results due to a decrease in the substrate well depth of the relative to the cubic system.
The system did experience a first-order phase transition for a weak fluid–substrate interactions and the
transition became much more gradual as the fluid–substrate interaction increased. The molecules be-
came fixed to commensurate substrate locations, but the surface was not corrugated sufficiently to have
a stable commensurate phase. The atomically smooth substrate showed the first-order phase transition
expected of a low temperature system with no effects of registry.
© 2008 Elsevier Inc. All rights reserved.

1. Introduction substep is observed at high pressures. These authors attributed this

Isotherms have been used to study the adsorption of gases onto
solid surfaces for the past 120 years. Researchers in the field of
adsorption are interested in what determines the shape of the ad-
sorption isotherm, since the shape of an adsorption isotherm often
is used to characterize the macroscopic consequences of molecular
behavior [1]. This paper examines how the shape of the isotherm
changes for system undergoing surface compression.
In 1981, Thomy and Duval studied the isotherms of krypton and
xenon adsorbing onto exfoliated graphite [2]. Their results were
of interest because they showed a second step in the adsorption
isotherm, suggesting that the physisorbed 2D monolayer under-
went successive phase transitions. Similar phase transitions have
been observed for monolayer adsorption of methane [3] and other
rare gases [4,5] onto graphite.
The isotherms for krypton and xenon adsorbing on exfoliated
graphite from Thomy and Duval’s paper show that at low pres-
sures a first-order phase transition occurs, which is exhibited by a
large vertical jump in the isotherm. For low temperatures, a second
* Corresponding author.
E-mail address: mdd@jhu.edu (M.D. Donohue).
step to the reconfiguration of the molecules in the adsorbed phase,
since at this pressure, the adsorbed layer transitioned from com-
mensurate to incommensurate packing. For commensurate pack-
ing, the repeat unit for the adsorbed layer is a simple multiple
of the underlying substrate’s basic unit cell. For incommensurate
packing, the ratio of repeat units between the adsorbed layer
and substrate molecules is irrational and may show continuous
variation. This commensurate–incommensurate transition occurs
in many adsorbed systems in addition to krypton and xenon on
graphite [6].
Since Thomy and Duval published these findings, numerous ex-
perimental studies have observed similar discontinuities in adsorp-
tion isotherms [7]. Sub steps in adsorption isotherms have been
found for nitrogen [8], methane, and argon [9] on graphite, as well
as simple gases on alkali halides, such as CoCl2 [10]. Discontinu-
ities in adsorption isotherms have been observed in several recent
adsorption experiments on carbon nanotubes [11].
Computer simulations are a powerful tool in the study of
adsorption due to the ease of changing interaction parameters
and substrate corrugation. Computer simulations have greatly
enhanced the understanding of two-dimensional melting, com-
mensurate–incommensurate transitions [12], wetting phenom-
ena [13], and critical phenomena [14]. Monte Carlo simulations

0021-9797/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2008.04.063
10 C.L. Charniak et al. / Journal of Colloid and Interface Science 324 (2008) 9–14
have been used to study a variety of surfaces including phase tran-
sitions on graphite [15], hexagonal [16], and cubic surfaces [17]. Pa-
trykiejew and Sokołowski conducted Monte Carlo simulations [18]
that produced similar results to the findings of Thomy and Du-
val’s physical experiments. Like Thomy and Duval, Patrykiejew and
Sokołowski found a discontinuity in the adsorption isotherm at
the first-order phase transition. Using these computer simulations,
they also were able to detect a discontinuity at the liquid–solid
transition, and the commensurate–incommensurate phase transi-
tion.
The purpose of this paper is to determine the effect of phase
transitions in a two-dimensional adsorbed layer on the compres-
sion of molecules in the adsorbed layer. Surface compression
occurs when the nearest neighbor distances between adsorbed
molecules are smaller than those of the corresponding bulk fluid.
Compression of the molecules of the adsorbed phase in the mono-
layer can occur by multiple mechanisms. For highly corrugated
surfaces, molecules adsorb to well-defined sites. If the distance be-
tween these sites is less than distance between the fluid molecules
in a bulk liquid, the adsorbed phase can compress in order to
form a commensurate structure. The unfavorable fluid–fluid inter-
actions are compensated by the energy difference between adsorp-
tion onto commensurate and incommensurate positions.
In weakly corrugated systems, the energy difference between
adsorption on commensurate and incommensurate sites is rel-
atively small. For sufficiently strong fluid–substrate attractions,
the adsorbed phase in a weakly corrugated system can compress
isotropically in order to increase the number of fluid–substrate
interactions. Adsorption onto weakly corrugated systems will con-
tinue and will compress until the decrease in free energy from the
adsorption of an additional molecule to the monolayer is balanced
by the increase in free energy from fluid–fluid repulsions [19].
Compression within the adsorbed layer has been seen exper-
imentally in chemisorbed [20–22] and underpotential deposition
(UPD) [23–25] experiments. In the case of physical adsorption, it
has been conventional wisdom that the separation between ad-
sorbed molecules is greater than their van der Waals diameter
(i.e. the separation corresponding to a potential energy of zero, σ ,
for a Lennard-Jones fluid) [26,27]. However, there is experimen-
tal data showing that adsorbed layers in physisorbed systems can
be compressed and have nearest neighbor separations less than
their van der Waals diameter. The MgCl films that adsorb onto
(111) Pd have been compressed by 10% relative to the minimum
in its potential [28]. The lattice spacing of the saturated low tem-
perature argon monolayer adsorbed on (111) Pt was 1.6% less than
that of the bulk argon at similar temperatures [29]. Neon adsorbed
on graphite was compressed by 2.5%, and xenon on graphite was
compressed by 3% [30]. Aranovich et al. analyzed data from a
large number of physisorbed systems and found that adsorbed
molecules often had repulsive nearest neighbor interactions, which
indicated compression [31]. Wetzel et al. used Monte Carlo sim-
ulations to show that compression of the adsorbed phase occurs
on cubic, hexagonal, and graphitic substrates depending on the
interaction energies and size discrepancy between the fluid and
substrate molecules [32].
This paper extends previous work to examine the effect of
phase changes within an adsorbed monolayer on surface com-
pression [32]. Grand Canonical Monte Carlo simulations were used
to generate equilibrium configurations of adsorbed phases. Ad-
sorption isotherms were created over a range of chemical po-
tential values for differing interaction energies for fixed molecule
sizes. The surface compression was examined specifically at ad-
sorption isotherm discontinuities, which are indicative of phase
changes.
2. Monte Carlo simulation method
The pairwise interaction between fluid molecules was modeled
using a truncated Lennard-Jones potential characterized with pa-
rameters dff corresponding to the distance to the minimum of the
fluid–fluid potential and εff . The fluid–substrate simulations were
pair-wise additive for the cubic and hexagonal substrates with pa-
rameters dfs and e fs and for the atomically flat substrate [33] had
an interaction potential of
  10  4  
2 1 dfs dfs d4fs
u fs ( z) = πρ ε
s fs dfs Δ − − ,
5 z z 3Δ( z + 0.61Δ)3
(1)
where ρs is the density of the substrate and Δ is the distance be-
tween crystalline layers. Equation (1) averages out the periodicity
of the substrate and creates a uniform approximation for the fluid–
substrate potential.
The substrate was rigid throughout the simulation, making
the specification of surface–surface interactions unnecessary. All
lengths in the simulation were scaled by the diameter of an ab-
sorbent molecule (dss ). The distance to the minimum in the fluid–
substrate potential was calculated from dfs = (dff + 1)/2. The cutoff
radius in the simulations was set to a value of 4dfs .
The Monte Carlo simulations were run in the grand canonical
ensemble over a range of chemical potential values with periodic
boundary conditions in the x and y directions but not in the z di-
rection. A structured, semi-infinite surface was specified at z = 0
and a noninteracting, hard wall was used for the upper bound of
the simulation box, at z = 4dff . In all cases studied, the height of
the simulation box was sufficiently large that the hard wall had
no effect on the properties of the adsorbed phase. The size of the
simulation boxes in the x and y directions were scaled to 30dss .
The initial arrangements of fluid molecules were chosen to have
a random configuration with uniform density. Insertion, deletion,
and translation moves were attempted with equal probability and
the acceptance criteria followed the Metropolis algorithm.
The simulation equilibrated after 20 × 106 steps, after which
equilibrium snapshots of the molecular configuration were ob-
tained to construct the quantities reported in this work. After
obtaining an equilibrium snapshot, the simulation volume was ran-
domly increased and the molecular locations were rescaled. The
system was allowed to equilibrate for 5 × 106 steps and then the
system was returned to the original size. Following this expansion
and contraction, the system was brought to equilibrium and an-
other snapshot was obtained. This procedure was used to prevent
multiple sampling of structures in the same frozen configuration.
Three substrate geometries were examined; a (100) face which
is cubic, a (111) hexagonal substrate, and an atomically smooth
substrate that has a potential that varies only in the direction
perpendicular to the substrate. For the cubic substrate, previous
results showed that molecules with a size less than dff = 1.1 ad-
sorbed in a c (1 × 1) fashion, whereas molecules that had diameter
larger than 1.3 adsorbed in a c (2 × 2) packing [32]. At intermedi-
ate sizes of around dff = 1.2 the molecules experience a frustration
between the two commensurate packings and a higher density
orientation that is incommensurate with the underlying lattice.
Therefore, dff was set to 1.2, to study the transition from commen-
surate c (2 × 2) packing to incommensurate packing as a function
of the fluid chemical potential.
The previous study showed that the adsorption of molecules
onto a hexagonal surface did not show such dramatic changes in
surface structure as was seen for adsorption on a cubic surface,
since both commensurate and incommensurate structures have
hexagonal packing. The transition from adsorption onto commen-
surate sites on the substrate to an incommensurate packing occurs
 C.L. Charniak et al. / Journal of Colloid and Interface Science 324 (2008) 9–14 11

at approximately dff = 1.2, depending on the heat of adsorption. 3. Simulation results

The simulations for the adsorption onto cubic and hexagonal sub-
strates were compared to an atomically flat surface that shows no 3.1. Cubic substrate
effects of registry.
The adsorbed density in the simulations is defined relative to Fig. 1 shows the adsorption isotherm for εff = εfs = 2. The den-
commensurate packing for cubic and hexagonal, and relative to sity on the left axis is relative to a completely filled commensurate
closed packing for the atomically smooth surface. The densities for layer. The force shown on the right axis measures the compression
the three surfaces are defined as of the adsorbed layer. The system undergoes a simple first-order
phase transition from an expanded phase to a disordered con-
densed phase. The interlayer forces increase from the expected
2N
ρ∗ = , for cubic substrates, value of zero for the expanded phase and then increases to the
X size Y size value for a full monolayer. Fig. 2 gives snapshots of configura-
√ tions before and after the phase transition. As expected from the
N 3
ρ∗ = , for hexagonal substrates, isotherm, the molecules in the full monolayer are not bound by
4 X size Y size locations that are commensurate with the substrate lattice but in-
√ stead form an irregular packing.
2N 3
ρ∗ = , for atomically smooth surfaces, (2) As the fluid–substrate interaction is increased to εfs = 6 (for
π X size Y size constant εff ) the adsorption isotherm changes as shown in Fig. 2.
Instead of the first order phase transition, the system changes con-
where N is the number of molecules in the system and ρ ∗ is a re- tinuously from an expanded to condensed phase. The increases in
duced density. The value ρ ∗ is scaled in both cubic and hexagonal interlayer force are also more gradual than the previous case.
simulations so that a value of unity corresponds to all commensu- As the interaction strength is further increased to εfs = 10, the
rate sites being occupied. For the atomically smooth substrate, the transition from an expanded to condensed phase becomes even
density was scaled so that a density of unity corresponds to close more gradual as shown in Fig. 3. The interlayer force is approxi-
packing of molecules with diameters equal to the minimum of the mately zero until the density reaches a value of approximately 1.0
fluid–fluid interaction potential. and then experiences a significant jump in compression. This jump
A major focus of this work is to discuss compression of ad- corresponds to molecules packing in incommensurate positions on
sorbed layers of Lennard-Jones molecules on a structured surface. the substrate. After this jump in interlayer force, the isotherm and
The bulk phase of a Lennard-Jones fluid has intermolecular spacing forces change continuously corresponding to continuous changes
approximately equal to the minimum in the potential well [34,35]. in surface packing.
Compression was examined using the negative derivative of the
potential function F (r ) which is defined to be a scalar quantity
that measures the magnitude of the force between two molecules.
F (r ) is positive for molecule separations of r < dff and negative
when r > dff . Compression was studied only in the first layer of the
adsorbed phase; therefore, the fluid–substrate interactions were ig-
nored in the compression calculation. F i is defined to be the total
force on molecule i and is calculated by summing F (r i j ) for ev-
ery possible molecule j in the first layer. Since F (r ) is a scalar
and not a vector force, the value of F i can be significantly greater
than 0 if nearest neighbor distances to molecule i are less than dff
at equilibrium. From the equilibrium configurations generated by
the simulation, an adsorbed phase was defined to be compressed
if
 
k i F i ,k Fig. 1. The isotherm, ρ , shown as the dashed line with boxes, and an eighth of
F  = > 0, (3) the force (or compression),  F , shown as the dashed line with triangles for the
εff ik interaction energies εff = 2 and εfs = 2 calculated over a range of μ values.

where the index k was summed over all configurations analyzed,

and the index i was summed over all molecules within a con-
figuration. For this analysis, the contributions of the substrate
molecules to the compression of the adsorbed layer are ignored.
There will be some effect in the plane of the adsorbed mono-
layer, but previous studies showed that for εfs  εff the adsorbed
molecules occupied positions with the substrate molecules that
maximized the potential, thus corresponding to a force value that
was approximately zero.
The greater the value of  F  for the simulation, the greater
the compression of the adsorbed phase.  F  was normalized by
εff and has units of (length)−1 and (molecule)−1 . The length scale
was nondimensionalized by the diameter of the substrate molecule
(dss ) and therefore  F  has units of (molecule)−1 . For this study,
compression is only observed in the monolayer. The substrate was Fig. 2. The isotherm, ρ ∗ , shown as the dashed line with boxes and  F , shown as
sufficiently attractive that the formation of a multilayer were not the dashed line with triangles for the interaction energies εff = −2 and εfs = −6
observed in the chemical potential ranges that were studied. calculated over a range of chemical potential values.
12 C.L. Charniak et al. / Journal of Colloid and Interface Science 324 (2008) 9–14
Fig. 3. The isotherm, ρ ∗ , shown as the dashed line with boxes and  F , shown as
the dashed line with triangles, for the interaction energies εff = 2 and εfs = 10 cal-
culated over a range of μ values.
Fig. 4. Finite size effects for the εff = 2 and εfs = 10 system with the heavy line cor-
responding to a 30 × 30 substrate, and the thin line corresponding to a 20 × 20. The
curves for the isotherm, ρ ∗ , and force values are shown without the data points.
The curves are sufficiently close that finite size effects can be neglected.
The isotherm does not exhibit a vertical substep but does show
a kink in the vicinity of ρ = 1.0. This corresponds to the formation
of a stable c (2 × 2) layer that transitions slowly to have molecules
fill incommensurate locations.
The influence of system size for the εff = 2, εfs = 10 system is
observed by reducing the box length from 30 substrate molecules
to 20 molecules. The curves for the two isotherms and compres-
sion values are shown in Fig. 4 with the heavy line correspond-
ing to the 30 × 30 system. The two sets of curves show that the
simulation results are within numerical error, implying that the
correlation length between molecules is shorter the length of the
substrate. The 30 × 30 is sufficiently large for these systems that
the finite system size can be neglected.
The influence of system temperature was observed by fixing the
ratio of εfs /εff to 5, and then changing εff to 1 and 3. The results
for these two systems are shown in Fig. 5. The higher temperature
system (εff = 1) showed a smooth continuous increase from an ex-
panded phase to a condensed phase. The interlayer force showed a
similar steady increase indicating that the molecules are not con-
fined to site adsorption. The low temperature simulation (εff = 5)
experienced a smooth transition from an expanded phase to a con-
densed phase until reaching a density of ρ = 1.0. The interlayer
force for the system is negative until all commensurate sites are
occupied and thereafter shows a steady increase to the value for a
completed monolayer. Similar to Fig. 3, Fig. 5b shows a kink in the
isotherm as the beginning of the incommensurate region. This kink
is the beginning of a horizontal step observed in repulsive systems
with site interactions
The adsorption of Lennard-Jones molecules onto a cubic sub-
strate showed a range of behaviors. When the interaction between
(a)
(b)
Fig. 5. The isotherm, ρ , shown as the dashed line with boxes and  F , shown as the
dashed line with triangles for the interaction energies (a) εff = 1 and εfs = 5, and
(b) εff = 3 and εfs = 15 calculated over a range of μ values.
the fluid and a substrate molecule is equal to the interaction be-
tween a pair of fluid interactions the isotherm shows a phase
transition from an expanded to a condensed phase. The well depth
of the substrate is not strong enough to restrict the system to site
bonding. The resulting monolayer is incommensurate with the un-
derlying substrate. As the relative strength of the fluid–substrate
interaction increases, the isotherm changes and there is no discon-
tinuity with the transition from an expanded to condensed phase.
The increase in the substrate attraction causes the fluid molecules
to occupy commensurate positions. The adsorbed molecules oc-
cupying the commensurate sites are sufficiently separated that
isotherm becomes Langmuir-like in nature without a phase tran-
sition. As the relative strength of the substrate–fluid further in-
creased the isotherm became increasing Langmuirian and exhib-
ited a kink in the region around ρ = 1. The kink formed because
the commensurate phase became stable over a region of chemical
potential and resisted the formation of a complete monolayer. De-
creasing the system temperature resulted in a continuous variation
from an expanded phase to a complete monolayer. Conversely, in-
creasing the system temperature made the kink more pronounced.
This kink is the beginning of the formation of a horizontal substep
that can be observed in systems that exhibit site adsorption with
repulsive nearest neighbor interactions [36].
The interlayer force gives a quantitative measure of the distri-
bution of adsorbed molecules. For the systems with weak fluid–
substrate interactions, the interlayer force increases with increas-
ing adsorbed density. The molecules are not restricted to commen-
surate sites and the repulsions in the system are proportion to the
density of the adsorbed layer. As the relative fluid–substrate inter-
action increases, the molecules become pinned to specific locations
on the substrate. Since these commensurate sites are separated
by distances greater than the minimum in the fluid–fluid poten-
 C.L. Charniak et al. / Journal of Colloid and Interface Science 324 (2008) 9–14 13

tial, significant repulsions do not occur until the adsorbed den-

sity reaches values greater than unity. After achieving a density of
unity, the adsorption of additional molecules causes repulsions in
the monolayer resulting in large increases with small changes in
density.

3.2. Hexagonal substrate

Simulations were performed on a hexagonal substrate to com-

pare the effects of the underlying structure. In a previous pub-
lication [32], a hexagonal system experienced a transition from
commensurate to incommensurate packing at monolayer comple-
tion when dff = 1.2. The fluid molecules were too large to have
a completely commensurate packing and experienced frustration
(a)
between balancing the minimization of the energy of adsorption
for a single molecule by occupying a commensurate site, and the
global minimization by an additional molecule adsorbing to the
surface. The registry effects on the hexagonal surface are much less
than the cubic substrate for two major reasons. The first is that
the hexagonal system has a higher density of substrate molecules
so that the minimums in potential energy for adsorption are less
steep and much less shallow for a given energy. The other is that
the natural packing for fluid molecules is hexagonal. With the
commensurate and incommensurate structures both being hexag-
onal, the frustration between the two packings is much less than
similar adsorption on cubic substrates.
Fig. 6 shows the adsorption isotherm and interlayer force for
the hexagonal substrate and a range of interaction parameters.
(b)
When εff = εfs (Fig. 6a) the system experienced a rapid change
from an expanded to a condensed phase. As the relative strength
of the substrate interaction increases the rate of change from low
to high density decreases. Similarly, as the relative strength of the
fluid–substrate interaction increases, the interlayer forces increase
less gradually before ρ = 1.0. This is similar to the cubic substrates
of having a more gradual transition due to molecules adsorbing to
commensurate locations. However, as expected the effect is less
pronounced than the cubic simulations and no substep or kink
were observed in the simulations.

3.3. Atomically smooth substrate

The third substrate examined was the atomically flat surface

that uniformly averages the interactions of the Lennard-Jones sub-
(c)
strate molecules. The fluid–substrate interaction is only a function
the distance perpendicular to the surface. The interaction strength
was set to be εfs = 4 and εfs = 8 (with εff = 2) to give compa-
rable results to the cubic and hexagonal surfaces. The isotherms
and interlayer forces are shown in Fig. 7. Both systems showed a
first-order phase transition from an expanded to condensed phase.
There are no registry effects binding molecules to specific sites,
allowing the molecules to form clusters which grow from a two-
dimensional gas to a liquid. The system density initially reaches
unity, but with the strong fluid–substrate interaction, it becomes
favorable to compress the monolayer to bring additional molecules
to the surface. The interlayer force jumps at the phase transition,
and then steadily increases with small increases in monolayer den-
sity.
(d)
4. Conclusions
Fig. 6. The isotherm, ρ ∗ , shown as the dashed line with boxes and  F , shown
as the dashed line with triangles for the interaction energies: (a) εfs = 4, εff = 2;
Grand Canonical simulations were performed for cubic, hexag-
(b) εfs = 6, εff = 2; (c) εfs = 8, εff = 2; (d) εfs = 10, εff = 2.
onal, and atomically smooth substrates. Both cubic and hexag-
onal substrates showed that as the relative strength of the
fluid–substrate interaction increased, the isotherm became more these sites were sufficiently separated that they adjacent molecules
Langmuir-like in nature. This was because molecules adsorbed were only weakly interacting. For cubic surfaces, a kink was ob-
strongly to the highly coordinated commensurate sites, and that served for systems with relatively strong fluid–substrate interac-
14 C.L. Charniak et al. / Journal of Colloid and Interface Science 324 (2008) 9–14

The interlayer force was useful in quantifying the structure

of the adsorbed monolayer. Both the cubic and hexagonal sys-
tems experienced significant compression after the formation of
the commensurate layer. Adsorption on the atomically smooth sur-
face showed significant compression upon undergoing a first-order
phase transition and then showed increasing compression upon
densification of the adsorbed monolayer.

References

[1] T.L. Hill, P.H. Emmett, L.G. Joyner, J. Am. Chem. Soc. 73 (1951) 5102.
[2] A. Thomy, X. Duval, J. Regnier, Surf. Sci. Rep. 1 (1981) 1.
[3] J.W. Riehl, K. Koch, J. Chem. Phys. 57 (1972) 2199.
[4] C.G. Shaw, S.C. Fain, Surf. Sci. 83 (1981) 1.
[5] S. Calisti, J. Suzanne, J.A. Venables, Surf. Sci. 115 (1982) 445.
[6] K. Horikoshi, X. Tong, T. Nagao, S. Hasegawa, Phys. Rev. B 60 (1999) 13287.
[7] A. Thomy, X. Duval, Surf. Sci. 299 (1994) 415.
[8] S. Villar-Rodil, R. Denoyel, J. Rouquerol, A. Martínez-Alonso, J.M.D. Tascón,
J. Colloid Interface Sci. 252 (2002) 169.
[9] K. Miura, H. Yanazawa, Carbon 41 (2003) 151.
[10] J.P. Coulomb, in: H. Taub, G. Torzo, H.J. Lauter, S.C. Fain Jr. (Eds.), Phase Transi-
tions in Surface Films, Plenum, New York, 1991, p. 113.
[11] L. Belyakova, A. Kalpakian, A. Kiselev, Chromatographia 7 (1974) 14.
[12] C. Taut, A.J. Pertsin, M. Grunze, Langmuir 12 (1996) 3481.
[13] K. Binder, D.P. Landau, Phys. Rev. B 37 (1998) 1743.
[14] E. Stoll, K. Binder, T. Schneide, Phys. Rev. B 8 (1973) 3266.
[15] S.Y. Jiang, K.E. Gubbins, J.A. Zollweg, Mol. Phys. 80 (1993) 103.
[16] M. Christov, K. Sundmacher, Surf. Sci. 547 (2003) 1.
[17] A. Patrykiejew, S. Sokołowski, K. Binder, Surf. Sci. Rep. 37 (2000) 207.
[18] A. Patrykiejew, S. Sokołowski, Langmuir 17 (2001) 938.
[19] G.L. Aranovich, M.D. Donohue, Colloids Surf. A 187 (2001) 95.
[20] N. Al-Sarraf, D.A. King, Surf. Sci. 307–309 (1994) 1.
Fig. 7. The isotherm, ρ ∗ , shown as the dashed line with boxes and  F , shown as
[21] R. Kose, W.A. Brown, D.A. King, J. Phys. Chem. B 103 (1999) 8722.
the dashed line with triangles for atomically flat substrates. A cubic spline is drawn
[22] C. Ebner, C. Rottman, M. Wortis, Phys. Rev. B 28 (1983) 4186.
to guide the viewer.
[23] E. Herrero, L.J. Buller, H.D. Abruna, Chem. Rev. 101 (2001) 1897.
[24] M.F. Toney, J.G. Gordon, M.G. Samant, G.L. Borges, O.R. Melroy, Phys. Rev. B 45
tion and that this kink formed due to the formation of a stable (1992) 9362.
c (2 × 2) monolayer. The kink became more stable as the temper- [25] Y.S. Chu, I.K. Robinson, A.A. Gewirth, Phys. Rev. B 55 (1997) 7945.
ature was increased and is suspected to be the beginning of a [26] G.A. Somorjai, Introduction to Surface Chemistry and Catalysis, Wiley, New
York, 1994, p. 73.
formation of a horizontal substep that has been observed in lattice
[27] G.A. Kimmel, M. Persson, Z. Dohnálek, B.D. Kay, J. Chem. Phys. 119 (2003) 6776.
systems with repulsive interactions. [28] D.H. Fairbrother, J.G. Roberts, S. Rizzi, G.A. Somorjai, Langmuir 13 (1997) 2090.
The hexagonal substrate was continuous for all parameters [29] P. Zeppenfeld, U. Becher, K. Kern, G. Comsa, Phys. Rev. B 45 (1992) 5179.
measured. The lack of a kink in the isotherm is due to differences [30] D.A. Huse, Phys. Rev. B 29 (1984) 6985.
in the hexagonal and cubic systems. The fluid–substrate potential [31] G.L. Aranovich, C. Sangwichien, M.D. Donohue, J. Colloid Interface Sci. 227
energy well is less steep in the hexagonal system which leads to a (2000) 553.
[32] T.E. Wetzel, J.S. Erickson, P.S. Donohue, C.L. Charniak, G.L. Aranovich, M.D.
less corrugated surface. Both the incommensurate and commensu-
Donohue, J. Chem. Phys. 120 (2004) 11765.
rate structures are hexagonal leading to a much smaller difference [33] W.A. Steele, Surf. Sci. 36 (1973) 317.
in densities. However, as compared to the atomically smooth sur- [34] F. Tsien, J.P. Valleau, Mol. Phys. 27 (1974) 177.
face, the effect of registry was still very important for the adsorp- [35] F. Cuadros, A. Mulero, Chem. Phys. 159 (1992) 89.
tion onto the hexagonal substrate. [36] T.E. Wetzel, G.L. Aranovich, M.D. Donohue, J. Phys. Chem. B 109 (2005) 10189.