Adsorption of polymers on a brush: Tuning the order of the wetting phase transition

Luis G. MacDowell and Marcus Müller

Citation: The Journal of Chemical Physics 124, 084907 (2006); doi: 10.1063/1.2172597
View online: http://dx.doi.org/10.1063/1.2172597
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/124/8?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Wetting of nanopatterned surfaces: The striped surface
J. Chem. Phys. 118, 816 (2003); 10.1063/1.1524159

Polymer melt droplets adsorbed on a solid wall: A Monte Carlo simulation

J. Chem. Phys. 114, 8610 (2001); 10.1063/1.1362164

First-order wetting transitions of neon on solid CO 2 from density functional calculations

J. Chem. Phys. 112, 4768 (2000); 10.1063/1.481033

Response to “Comment on ‘Surface diffusion near the points corresponding to continuous phase transitions’ ” [J.
Chem. Phys. 111, 11232 (1999)]
J. Chem. Phys. 111, 11234 (1999); 10.1063/1.480480

Comment on “Surface diffusion near the points corresponding to continuous phase transitions” [J. Chem. Phys.
109, 3197 (1998)]
J. Chem. Phys. 111, 11232 (1999); 10.1063/1.480479

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 THE JOURNAL OF CHEMICAL PHYSICS 124, 084907 共2006兲

Adsorption of polymers on a brush: Tuning the order

of the wetting phase transition
Luis G. MacDowella兲
Departamento de Química Física, Facultad de Ciencias Químicas, Universidad Complutense,
28040 Madrid, Spain
Marcus Müller
Institut für Theoretische Physik, Friedrich-Hund-Platz 1, Georg-August Universität,
37077 Göttingen, Germany
共Received 29 November 2005; accepted 10 January 2006; published online 28 February 2006兲

We develop a computational methodology for the direct measurement of a wetting transition and its
order via the effective interface potential. The method also allows to estimate contact angles in the
nonwet state and to study adsorption isotherms. The proposed methodology is employed in order to
study the wetting behavior of polymers on top of a brush consisting of identical polymers. In the
absence of long-range forces, the system shows a sequence of nonwet, wet, and nonwet states as the
brush density is increased. Including attractive long-range interactions we can make the polymer
liquid wet the bush at all grafting densities, and both first- and second-order wetting transitions are
observed. The latter case is limited to a small interval of grafting densities where the melt wets the
brush in the absence of long-range interactions. Second-order wetting transitions are preceded by a
first-order surface transition from a thin to a thick adsorbed layer. The interval of second-order
wetting transitions is limited at low grafting densities by a surface critical end point and at high
grafting densities by a tricritical wetting point. Our study highlights the rich wetting behavior that
results when competing adsorbent-substrate interactions of different scales are tuned over a broad
range. © 2006 American Institute of Physics. 关DOI: 10.1063/1.2172597兴

I. INTRODUCTION cases, a sequence of nonwet, wet, and nonwet states may

As a result of the increased ability to control and modify be observed as the brush density is increased.11,15 The role
surface structure,1 theoretical and computational works on of entropic factors in this wetting sequence has been
wetting and adsorption have received abiding interest.2–8 emphasized,13 and several self consistent field theoretical
Among the different materials considered, polymers are one studies have been devoted to the study of the autophobic
of the most studied, because of their great versatility and regime.15–19
practical interest 共e.g., protective coatings兲. Moreover, the Obviously, the wetting properties of a film will also de-
vapor pressure is vanishingly small for long chain molecules pend on the nature of the substrate where the brush is
and evaporation effects can be neglected. Many properties grafted. From an applied or experimental point of view, this
such as structure, composition, or chain length may be varied provides further control over the system 共e.g., by using a
during the synthesis providing control over the required wet- surface coating or a thin adsorbed layer one can modify the
ting properties. Furthermore, polymers can also be adsorbed wetting behavior兲. From the theoretical point of view, one
or irreversibly grafted onto substrates,9–11 and the properties expects an intriguing interplay between forces that act on
of the resulting brush may be further modified via the quality different length scales. 共i兲 Short-range interactions that stem
of the surrounding solvent.12 from the distortion of the narrow liquid-vapor or substrate-
Although grafting polymers on a substrate is expected to liquid interface. The width of these interfaces is dictated by
increase the compatibility of polymer films because it adds the correlation length of density fluctuations ␰, which is on
attractive interactions to the substrate, it has been found that the order of the statistical segment length for dense polymer
the brush’s density plays an important role.13 For too high melts. 共ii兲 There are weaker interactions in polymer melts
grafting densities of the brush 共or large crosslink density of that extend up to the size of the polymer coil, Re.20–22 For
polymer networks兲, polymers of the same material as the long polymer chains these two short-ranged contributions
brush have been found to be autophobic, i.e., the chemically have well separated length scales, but for the rather short
identical molecules in the brush and the melt repel each other chains considered in our simulations we do not attempt to
and the melt dewets from the brush.9–11,14 For smaller brush distinguish between the two contributions. 共iii兲 There are
densities, however, suitably prepared brushes are known truly long-range interactions that arise from the tails of the
to enhance the wetting properties and reduce the contact dispersion interactions between the polymer segments and
angle of polymers that do not wet the bare substrate. In such the constituents of the substrate. The strength of these long-
range power-law interactions is parametrized by the
a兲
Electronic mail:[email protected] Hamaker constant A. In the case of a polymer layer on top of

0021-9606/2006/124共8兲/084907/13/$23.00 124, 084907-1 © 2006 American Institute of Physics

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-2 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

a polymer brush the distinction between long-range and tion we introduce our coarse-grained polymer model and
short-range forces becomes particularly clear because the simulation methodology. Then we present our simulation re-
liquid-vapor interface does not come close to the substrate sults for the different regimes of the grafting density and
and difficulties in distinguishing between long- and short- analyze them within a simple phenomenological expression
range forces do not arise. for the interface potential. Section IV contains our conclu-
Interestingly, not only the intensity but also the attractive sions.
or repulsive character of long-range forces can be modified
in some situations. In dewetting studies of polymers, these
interactions could be greatly modified by coating the surface II. MODEL AND SIMULATION METHODOLOGY
with an oxide layer of varying thickness. This control of the
surface properties can lead to a new wetting behavior. Stud- A. Model system
ies of polystyrene films adsorbed on silicon wafers show a We consider a simple coarse-grained polymer model
two-stage dewetting process. At a high temperature, mi- which we have utilized previously.32–34 Polymers consist of a
crometer scale droplets are found on top of a uniform ultra- linear string of N beads, which interact with every other bead
thin film, a few nanometers thick. Upon cooling, rupture of in the system by means of a truncated and shifted Lennard-
the thin film is observed.23–25 This reveals a transition be- Jones potential,
tween two different equilibrium film thickness, none of
which correspond to a fully wet state. A similar two-stage
wetting process due to the interplay of short- and long-range VLJ共r兲 =
冦 再
4⑀冉 冊 冉 冊冎
␴ 12
r
−
␴
r
6
+
127⑀
4096
, r 艋 rc
冧 共1兲
interactions has been observed in studies of alkanes adsorbed 0, r ⬎ rc ,
at the air/water interface. Under suitable conditions, it is
found that these systems exhibit a first-order surface transi- where the truncation distance rc is set equal to rc = 2 ⫻ 21/6␴.
tion between a thin and a thick film.26,27 This unexpected In the following, we will employ the Lennard-Jones diameter
transition is then followed by a second-order or critical wet- ␴ as unit of length and the Lennard-Jones potential depth ⑀
ting transition, where the thickness of the film gradually as unit of energy, unless otherwise stated.
diverges.28 Contiguous monomers along the same chain additionally
In this work we employ computer simulations in order to interact via a bonding potential of finite extensible nonlinear
explore the wetting behavior of polymers on a brush consist- elastic 共FENE兲 type,
ing of grafted polymer of the same chemical nature and
length. Interactions between polymer segments are consid- 冉 冊
VFENE共r兲 = − kR20 ln 1 −
r2
R20
. 共2兲
ered to be strictly short range. The strength of the short-range
interactions between polymer and substrate is modified by In the above equation the maximum allowed displacement
changing the grafting density of the brush. Laterally inte- between bonded monomers is R0 = 1.5␴, while k, which plays
grated Lennard-Jones interactions between polymer and sub- the role of an elastic constant, takes the value k = 15⑀ / ␴2.
strate and the tails of the dispersive interactions between Free polymers of the system are placed within a cuboidal
polymer segments that have been truncated in the model of simulation box, whose sides Lx, Ly, and Lz are such that
the direct interactions result in long-range interactions. Their Lx = Ly and Lz Ⰷ Lx. We impose periodic boundary conditions
strength is controlled by an effective Hamaker constant A. along the lateral x and y directions, and place perfectly flat
Previously, this system was studied using self consistent field walls at z = 0 and z = Lz. Additionally, a number of ng chains
theory and a very rich wetting behavior was observed, ex- are grafted onto each wall. Their first monomers are located
hibiting inter alia a two-stage dewetting process similar to at z0 = 1.2 and the grafting points are arranged on a square
that observed experimentally for polystyrene on silicon wa- lattice with 1 / ⌺ = ng / 共LxLy兲 being the number of grafted
fers and alkanes at the air/water interface.15 chains per unit area. The grafting substrate additionally in-
Despite this a priori information, the calculation of a teracts with the chain segments via a long-range potential of
the form,

再 冉 冊 冉 冊冎
wetting phase diagram by means of computer simulations is
not a simple matter.29 Even for simple Ising spin models the ␴ 9
␴ 3
growth of the film thickness and the concomitant slow, long- Vwall共z兲 = A sign共A兲 − , 共3兲
z z
wavelength fluctuations at weakly first- or second-order wet-
ting transitions pose a challenge for simulations.29,30 In view where z denotes the perpendicular distance of the polymer
of the much longer molecular relaxation times of polymeric site to the closest substrate. A is the effective Hamaker con-
materials we have only been able to investigate the wetting stant characterizing all long-range interactions between sub-
behavior of rather short chains utilizing advanced simulation strate and polymer and the residual tails of the polymer-
and analysis techniques and a substantial amount of com- polymer interactions that are omitted by the truncation of the
puter time. Here we propose a new technique which allows Lennard-Jones potential.
us to directly measure the effective interface potential, con- In the following grafted and free polymers are identical,
taining a wealth of information about the surface thermody- and they are comprised of N = 10 segments. The simulations
namics. A brief account of this technique was reported are all carried out at the reduced temperature kBT / ⑀ = 1.68. In
recently.31 terms of the monomer number density ␾, the coexistence
Our manuscript is arranged as follows. In the next sec- liquid and vapor densities are ␾l = 0.61 and ␾␷ = 0.0083, re-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-3 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

spectively. The interface tension is ␥ / kBT = 0.0953 and the malize g such that g共l → ⬁兲 → 0, the knowledge of the film
mean end-to-end distance of the polymers is Re = 3.66.32 thickness where the interface potential exhibits its global
The optimal size of the simulation box has to meet two minimum lmin allows us to measure the contact angles, be-
requirements. A large substrate area is needed for simulating cause then
low grafting density due to the small number of grafted
chains. Additionally, Lz must be large enough such that the ␥LV共1 − cos ␪兲 = g共⌫ = ⬁兲 − g共⌫min兲 = − g共⌫min兲, 共4兲
interfaces at z = 0 and z = Lz do not interact. Obviously, a too where ⌫min is the adsorption at the minimum of g. Further-
large system makes the simulations computationally infea- more, g contains information on the order of the transition
sible and density fluctuations in the vapor might significantly and the stability of thin films.
contribute to the total density fluctuations. Typically we con- In order to measure free energy differences we perform
sidered system sizes with Lx between 14␴ for the low graft- Monte Carlo simulations in the grand canonical ensemble. In
ing densities and 8␴ for the highest grafting densities. In this way, we are able to monitor the probability of finding a
order to avoid interaction of the two interfaces, we typically given number of polymers within the system. Since the
employed Lz = 90␴, but occasionally Lz = 180␴ was required grand potential of a state with n particles is directly propor-
for simulating very thick films. The number of grafted chains tional to −kBT ln P共n兲, the probability distribution allows us
on a wall varied between ng = 16 for the lowest grafting den- to directly calculate the free energy differences. To be more
sities and ng = 36 for the highest grafting densities. specific, we would like to estimate the interface potential by
A symmetric simulation box with two identical sub- monitoring the probability P1/2共n兲 of finding n polymers in-
strates allows to gather statistics from both halves of the side the half of the simulation box that contains the brush-
simulation. In practice, however, having two brush-covered covered substrate. We can then define a surface adsorption ⌫
substrates results in very large relaxation times, in particular, as the total number of monomers in that half box n1/2 less the
when thick liquid films form. Therefore, we often use a amount found in an equal bulk volume, i.e.,

冉 冊
purely repulsive substrate at z = Lz to sample the fluctuation
of a single liquid-vapor interface, to reduce possible correla- 1
⌫ = Nn1/2 − ␾vV /LxLy . 共5兲
tions, and to obtain better statistics. 2
In this way, we can determine the interface potential as
B. Simulation methodology
k BT
Locating a wetting transition by computer simulation g共⌫兲 = − ln P1/2共⌫兲. 共6兲
L xL y
poses a challenge because of the divergence of the thickness
of the liquid film at the transition. Both, protracted long To a crude approximation the interface excess is related to
equilibration times as well as significant finite size effects, the thickness l of a uniform liquid film via ⌫ = ⌬␾l with
have to be accounted for. ⌬␾ = ␾l − ␾v. This approximation assumes a sharp, rectangu-
An accurate method to locate first-order wetting transi- lar density profile at the substrate and the liquid-gas interface
tions consists in calculating the surface free energy differ- and it neglects that density fluctuations in the bulklike re-
ence of the liquid and vapor in contact with the substrate and gions of the profile. Thus, plotting the interface profile versus
the liquid-vapor interface tension in separate simulations. ⌫ / ⌬␾Re yields an approximation for the interface potentials
Then, the Young equation is utilized to calculate the macro- as a function of the film thickness in units of the end-to-end
scopic contact angle and to locate the wetting transition. This distance, Re.
method is particularly advantageous at strong first-order tran- This procedure poses both conceptual and technical
sitions because the thickness of the fluid layer at the nonwet challenges. If we fix the lateral system dimensions, Lx and
substrate virtually vanishes which strongly reduces finite size Ly, of our cuboidal simulation cell, and increase the perpen-
effects.32 dicular distance Lz, the interface of a thick film at coexist-
This method, however, gradually becomes more difficult ence will not be bound to the substrate and will wander
as the strength of the wetting transition decreases. Although around as to explore the whole volume. In the limit Lz → ⬁
the liquid layer in the nonwet state remains microscopic it there is no stationary probability distribution even if g共l兲 has
comprises several molecular diameters and the relaxation a finite minimum binding the interface to the substrate. In
times of long-wavelength fluctuations of the film thickness order to obtain a meaningful normalizable probability distri-
are protracted long. Moreover it is difficult to estimate the bution one first has to consider the limit Lx , Ly → ⬁ and then
strength of the first-order transition from the analysis of the can consider the limit of large film thickness, Lz → ⬁. The
surface and interface tensions that enter Young’s equations. fact that the interface explores the whole volume of the sys-
An alternative is to directly measure the interface poten- tem is also a technical problem because it imparts extremely
tial in the simulations. The interface potential g共l兲 measures long correlation times of interface fluctuations and the simu-
the free energy of a liquid film per unit area as a function of lation spends a lot of time exploring regions where the inter-
its thickness l or another suitably defined measure of the face and the substrate do not interact and the interface po-
substrate adsorption. This interface potential or the concomi- tential is flat.
tant disjoining pressure can also be measured experimentally, A simple remedy is to simulate at slight under-
but the experiments are only able to probe a limited region saturation,30 ⌬␮ = ␮ − ␮coex. In this case the effective inter-
corresponding to rather large film thickness.35,36 If we nor- face potential g⌬␮ is exactly obtained as described above but
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-4 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

large film thicknesses are suppressed and the interface al-

ways returns to the substrate. The interface potential at co-
existence as defined in Eq. 共6兲 is simply recovered by means
of a Legendre transformation,
g0共⌫兲 = g⌬␮共⌫兲 − ⌫ f ⌬␮ , 共7兲
where ⌫ f is the adsorption of free chains 共as opposed to ⌫,
the adsorption of both grafted and free chains兲.
This procedure works well if g0 exhibits only a shallow
minimum 共wet, small contact angle or weakly first-oder tran-
sition兲. In the opposite case, g0 showing a deep well 共large
contact angle or strongly first-order transition兲, the simula-
tion explores only the narrow region where the interface is
bound to the substrate and it is difficult to extract the limit
l → ⬁ from the simulation data. In order to avoid this prob-
lem we utilize successive sampling,37 a variant of the well
FIG. 1. Calculation of interface potentials. Top figures show the probability
known umbrella sampling technique.38 of finding n molecules inside the whole simulation box at kBT / ⑀ = 1.68 and
In our simple implementation of successive sampling, the coexistence value of the chemical potential. Top left panel presents the
we perform standard grand canonical simulations but rather piecewise distributions obtained within each window. Top right panel is the
resulting distribution obtained after matching the piecewise functions. Bot-
than sampling over the actual grand canonical probability tom figures show the probability distribution of finding n molecules a dis-
distribution P共n兲 for a given temperature T and chemical tance less than Lz / 2 away from the substrate. Left panel illustrates the cu-
potential ␮, we subdivide the interesting interval of the num- mulative distributions that are obtained as an increasing number of windows
ber of chains into overlapping subintervals, 关ni−1 , ni兴. Monte is considered. Right panel shows how the cumulative distributions in the left
collapse into a single master curve, the interface potential.
Carlo moves that try to leave the interval are rejected as to
fulfill detailed balance at the boundaries. Performing differ-
ent simulations in the subintervals we obtain an estimate of each, but the number of subintervals required can be in-
the distribution Pi共n兲 in the ith subinterval. Then, the results creased as needed.
of the subintervals are matched at their boundaries to provide Figure 1 illustrates the methodology proposed for the
an estimate for the grand canonical distribution function over calculation of interface potentials. The top left panel shows
the entire range of the number of chains. Let n be in the the partial distributions Pi obtained within each window.
interval 关n−i , n+i 兴 and n⬘ be in the the neighboring interval Shifting the logarithms of the piecewise probability distribu-
关ni−1
− +
, ni−1 兴 which overlap at n̄, then tions according to Eq. 共8兲 as to match them in the overlap-
P共n兲 Pi共n兲 Pi−1共n̄兲 ping windows we obtain a continuous function P共n兲, inde-
⬇ . 共8兲 pendent of the specific choice of subintervals employed 共top
P共n⬘兲 Pi共n̄兲 Pi−1共n⬘兲 right兲. For each of the windows, we also monitor the condi-
Successively we construct an estimate for the unnormalized tional probabilities which, using Eq. 共9兲, allow us to estimate
distribution, P共n兲, for all the considered n relative to some P1/2共n兲. A distribution obtained in this way from i successive
arbitrarily chosen probability, P共n0兲. windows shows a spurious increase close to the maximum
Actually, P共n兲 contains information on two interfaces, number of particles allowed in the system, i.e., ni, due to
that at z = 0 and that at z = Lz. The interface potential of one truncation effects in Eq. 共9兲. However, increasing the number
such substrate is determined from P1/2共n兲, the probability of of windows employed allows to shift the spurious feature to
obtaining n particles in the corresponding half box. This in- larger particle numbers as desired 共bottom left兲, and the re-
formation may be recovered by monitoring P1/2共n 兩 k兲, the sulting distributions differ by a simple additive normalization
probability of finding n particles in a half box, given that the constant within the unperturbed interval 共bottom right兲.
total number of polymers within the box is k. This condi- From the knowledge of the interface potentials, we can
tional probability is independent of the chemical potential determine several interface properties, including the value of
and may be measured during the simulations. At the end of the effective Hamaker constant Awet at which the liquid wets
the simulation, we calculate the information as the brush. We estimate the location of first-order wetting
transitions using two different methods. First, we plot P1/2共n兲
⬁
for a system slightly above a first-order wetting transition.
P1/2共n兲 = 兺 P共k兲P1/2共n兩k兲. 共9兲 The probability distribution is bimodal and by decreasing the
k=n
bulk chemical potential, ⌬␮pw = ␮pw − ␮coex ⬍ 0, into under-
In principle, all states of the system with more than n mol- saturation the area below the two peaks can be made equal.
ecules will contribute to the summation. In practice, obtain- The equal weight rule of phase coexistence39 asserts that at
ing meaningful probabilities for very large n is not feasible. this undersaturation and substrate interaction, A, a thin and a
Fortunately, it was found that the sum converges fairly fast thick film thickness coexist with each other. In contrast to the
共cf. Fig. 1兲 and could be truncated, provided the convergence wetting transition itself, only finite layer thickness is in-
is carefully monitored. Typically, we studied a range of 250 volved and finite size effects can be controlled. This prewet-
molecules divided in about five subintervals with 50 particles ting coexistence is the hallmark of a first-order wetting tran-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-5 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

sition. Once the line of prewetting points 共A , ␮pw兲 has been that exp共−␤共U + ⌽兲兲 may be considered as a probability den-
determined we estimate the location of the wetting transition sity, we obtain32
by extrapolating the prewetting line back towards the coex-
istence chemical potential according to32,40 P␮A⬘VT共n兲 ⬀ P␮A VT共n兲具exp共− 共A⬘/A − 1兲⌽/kBT兲典A , 共14兲
where the average is performed over a canonical distribution
⌬␮ ⬀ 共A − Awet兲3/2 . 共10兲
with Hamaker constant A. This expression means that aver-
Thus, a fit to the available data for the prewetting line allows ages corresponding to the thermodynamic states ␮, V, T, and
us to accurately estimate Awet. A⬘ can be constructed from a sample of microstates gener-
The calculation of the prewetting line as described above ated 共according to the Boltzmann weight兲 at the states ␮,
becomes difficult for systems with small lateral size because V, T, and A by assigning each microstate the weight
the tail of the peak that corresponds to a thick film decays as exp共−共A⬘ / A − 1兲⌽ / kBT兲.
exp共−LxLy⌬␮⌫ / kBT兲. For small system sizes this tail decays An important point regards our order parameter, the ad-
very slowly and acquires a large weight. In large systems, the sorption ⌫. This corresponds to the integral criterion for lo-
states between fully developed thin and large systems consist cating the position of the liquid-vapor interface, but other
of coexisting thin and thick films, with an extra energy pen- schemes based on the detailed knowledge of the profiles
alty arising from the line tension ␭. The line between the 共crossing criteria兲 can also be utilized to determine the film
films will run across the smallest lateral dimension, either Lx thickness l.42,43 Moreover, monitoring the adsorption we can-
or Lz, and will have a total cost ␭ min共Lx , Ly兲. For small not distinguish between the formation of a uniform layer and
systems, however, this contribution is relatively small, and a droplet. At a fixed adsorption the system will adopt the
the minimum of the probability distribution between the two configuration that minimizes the free energy—be it a uni-
stable peaks is not very well developed. In such cases, we form film, two coexisting films of different thickness, or a
determine the wetting transition by direct inspection of the droplet. This can be verified explicitly via configurational
interface potential, i.e., by estimating the substrate interac- snapshots. The configurations that a system adopts at fixed
tion at which g共⌫min兲 = 0, where ⌫min is the adsorption at the excess number of particles and their dependence on the sys-
minimum of the interface potential. This procedure is remi- tem size have been discussed both for thin films on sub-
niscent of locating the prewetting via the equal height rule strates 44 as well as for the periodic simulation cells 共i.e.,
共instead of the equal weight rule兲, and we expect stronger bulk simulations兲.45,46 In the latter cases 共coexistence be-
finite size effects. tween two different film thicknesses or a droplet兲, the mea-
Since the calculation of full interface potentials is rather sured probabilities will not be simply related to the interface
time consuming, it is desirable to exploit the data obtained potential but contain additional information about the line
during a simulation as much as possible. In the course of this tension.44,47 Importantly, states that correspond to the minima
study, extrapolation of one calculated interface potential in of the interface potential 共i.e., the configurations that are per-
order to predict interface potentials for other values of A and tinent to the equilibrium wetting behavior兲 will correspond to
T is very useful. The expressions required for this histogram laterally uniform configurations and free energy differences
extrapolation 41 are obtained by writing the grand canonical between these states have the usual meaning.
probability of observing a given number of particles as fol-
lows: III. RESULTS
P␮A VT共n兲 ⬀ exp共␮n/kBT兲QnVT
A
, 共11兲 Figure 2 shows the interface potentials for brushes of
where A
QnVTis the canonical partition function of a system increasing grafting density in the absence of long-range
with Hamaker constant A. The probability distribution for forces, i.e., A = 0. As expected, for low grafting densities the
another value of the Hamaker constant A⬘ can be expressed interface potential shows a rather deep minimum at low ad-
by the previous distribution, sorption. Due to steric reasons, the almost bare surface
greatly reduces the configurational entropy of chains close to
A⬘ the substrate, and a thick film is not stable. Actually, for a
QnVT
P␮A⬘VT共n兲 ⬀ P␮A VT共n兲 A . 共12兲 bare hard surface, previous results for the same model sug-
QnVT
gest a drying transition at A = 0.32 Gradually increasing the
Noting that the wall energy of a given microcanonical con- brush density we decrease the depth of the minimum. As a
figuration differs in both systems by a multiplicative factor result of attractive interactions, grafting chains on the sub-
A / A⬘ only, the ratio of canonical partition functions may be strate somewhat compensate for the loss of entropy. Eventu-
written as ally, at intermediate grafting densities the interface potential
becomes a monotonically decreasing function of the adsorp-
A⬘
QnVT 兰 e−␤共A⬘/A−1兲⌽e−␤共U+⌽兲drn tion, and no trace of a minimum remains. This shows that the
= , 共13兲
A
QnVT 兰 e−␤共U+⌽兲drn attraction provided by the grafted chains can lead to wetting
even in the absence of long-range forces. However, upon
where U is the sum of all pair interactions, while ⌽ is the further increasing the grafting density, the brush becomes so
sum of all wall-monomer interactions. The equation above dense that penetration of the chains is at a great cost and the
has been arranged in a manner such that its resemblance with system dewets. The dewetting of a fluid on top of a substrate
perturbative free energy calculations is made explicit. Noting made of the same kind of molecules is known as autopho-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-6 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

FIG. 3. Density profiles for a brush of low grafting density, 1 / ⌺ = 0.0816,

with three different adsorptions, corresponding to a bare brush, a brush at
the minimum of the interface potential, and a brush at a 共large兲 plateau value
FIG. 2. Interface potentials g共⌫兲 as a function of the total adsorption ⌫ for of the interface potential. The lines are the density profiles of grafted chains
several brush densities and A = 0. From left to right, g共⌫兲 of increasing 共full, bare brush; long dashed, brush at minimum; dotted, brush at plateau兲.
grafting densities, and Re2 / ⌺ = 1.09 共full line兲, 1.23 共long dashed兲, 1.31 共short The empty symbols correspond to the free chains 共circles, brush at mini-
dashed兲, 2.68 共full兲, 5.63 共long dashed兲, and 6.70 共 short dashed兲. mum; squares, brush at plateau兲. The filled symbols correspond to the total
共graft+ free兲 density profiles 共circles, brush at minimum; squares, brush at
plateau兲. The arrow on the z axis marks the position of the first monomers of
bicity, a phenomenon which has been studied both ex- the grafted chains which have been excluded from the density profiles. The
inset shows the detail of density profiles for the grafted chains.
perimentally10,11,14 and theoretically.16–19
It would be interesting to relate the equilibrium adsorp-
tion 共i.e., minimum of the interface potentials兲 to the well liquidlike adsorbed film, corresponding to the asymptotic
known properties of the system, such as end-to-end distance, part of the interface potential 共⌫ = ⌫⬁兲. For each situation, we
grafting density, and wall strength. For extremely simplified consider the density profile of grafted polymers ␳g共z兲 共full
models—assuming incompressible brushes, no enthalpic in- lines兲, that of free chains ␳ f 共z兲 共dashed lines兲, and the result-
teractions, and Gaussian distribution of polymers—analytic ing total density profile ␳t共z兲 共symbols兲. The contribution of
expressions may be obtained.13 However, for the short poly- the first monomer of grafted chains is a sharp peak at
mers considered here, typical polymer scaling laws are only z0 = 1.2 which is not included but marked by an arrow on the
qualitatively obeyed. Both short-range and long-range inter- horizontal axis.
actions must be considered, and packing effects on the Figure 3 shows the density profiles for a brush with
monomer length scale are relevant. A simple relation be- R2e / ⌺ = 1.09 in the low density regime. The density profile for
tween equilibrium film thickness and fluid polymers is there- the bare brush vanishes at the substrate and exhibits a maxi-
fore not available, but self consistent field theories may be mum around z ⬇ 0.6Re. For the short chain length considered,
employed to obtain numerical estimates of the interface however, the different length scales—segment size ␴, that
potential.15–19
The interface potential in the absence of long-range
forces allows us to classify the behavior into three regimes:
sparsely grafted brushes, ranging from R2e / ⌺ = 0 to 1.34,
show a behavior similar to that of a bare substrate. The at-
traction provided by the grafted chains is not enough to sta-
bilize a thick polymer film and the system remains nonwet.
Brushes with intermediate grafting density, ranging from
R2e / ⌺ = 1.34 to about R2e / ⌺ = 5.36, show a monotonically de-
creasing interface potential, i.e., the polymer melt wets the
brush. Finally, for high grafting densities, beyond R2e / ⌺
= 5.36, the brush is so dense that liquid films become again
unstable and autophobic dewetting occurs.
The existence of different regimes and their behavior
may be intuitively explained and rationalized by inspecting
the polymer density profiles of the system. Figures 3–5 show
the density profiles for brushes in each of the three regimes.
Each of the plots includes the results for three different film
thicknesses: 共1兲 a bare brush 共⌫ = ⌫g兲, 共2兲 a thin film that
FIG. 4. Density profiles for an intermediate grafting density brush,
corresponds to the equilibrium adsorbed amount 共i.e., the 1 / ⌺ = 2.68. Symbols as in Fig. 3. The density of the grafted chains for ⌫min
minimum of the interface potential, ⌫ = ⌫min兲, and 共3兲 a thick and ⌫⬁ looks equal on the scale of the graph.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-7 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

density profiles as a function of perpendicular distance to the

substrate. Therefore, the in-plane structure is averaged out.
In bad solvents both simulations 49,50 and experiments 51 ob-
serve that sparsely grafted brushes do not form a laterally
homogeneous dense layer but rather form dimples whose
characteristic lateral size is dictated by Re. This effect has not
been observed in our self consistent field theory 共SCFT兲
calculations,15 which assumed a laterally homogeneous pro-
file from the outset.
In our simulations we find that for very low grafting
density at each instant in time large patches of the substrate
FIG. 5. Density profiles for a dense brush, Re2 / ⌺ = 6.70 at A = 0 共left兲 and are not covered by the brush, i.e., there are strong density
A = 1.2 共right兲. Symbols as in Fig. 3, except for total density profiles, which fluctuations. The assumption of a spatially homogeneous
are shown in the inset: long dashed line, brush at the minimum; short dashed
line, brush at the plateau. The grafted chain density profile 共full line兲 is also
density in the lateral direction is only a very crude descrip-
included in the inset for comparison. The arrow on the z axis marks the tion of the system. As we increase the grafting density, how-
position of the first monomers of the grafted chains which have been ex- ever, the spatial inhomogeneity decreases. Note that our
cluded from the density profiles. chain length is rather short and that the temperature is not
very far below the ⌰ point, T / ⌰ ⬇ 0.51. For brushes with
determines the packing, excluded volume screening length ␰, R2e / ⌺ 艌 1.09, the lateral structure of density profiles is mainly
that characterizes density correlations, and the molecular size due to the first two monomers of the grafted chains, the
Re—are not well separated. The maximal density is much remainder of the chain producing a rather uniform back-
lower than the liquid density ␾l. The profile results from a ground.
balance between attractive interactions, promoting the col- In order to relate the structure of the brushes and the
lapse of chains, and conformational entropy, penalizing the adsorbed films with the wetting behavior, it is also important
compression of chains into an ultrathin film. to consider the effect of changing the Hamaker constant.
Adding a small amount of free chains to this brush de- Clearly, the long-range wall potential may influence the wet-
creases the free energy, for the free chains help building a ting behavior as a result of interactions with the adsorbed
uniform layer and thus reduce the stretching of the grafted film, but also indirectly via the influence on the brush struc-
chains. As can be observed from the profile, the laterally ture. For dense brushes, however, the effect is fairly small,
averaged surface to the vapor sharpens and gradually adopts because the brush is so compact anyway. Figure 5 共right兲
the shape of a narrow liquid-vapor interface. shows the density profiles for a brush with R2e / ⌺ = 6.70 and
In Fig. 4 we show the density profiles observed for a A = 1.2. Comparison with results for A = 0 shows that the
brush R2e / ⌺ = 2.68 characteristic of the intermediate density structure remains rather similar. Actually, the only significant
regime. The brush is now considerably more dense and some difference results from the z−9 repulsive part of the long-
density oscillations 共packing兲 in ␳g共z兲 already occur in the range potential 关cf. Eq. 共3兲兴, which expels the first few brush
bare brush. However, the density is not so large as to prevent monomers away from the neighborhood of z = 0. The much
free polymers to penetrate the brush. Therefore, increasing longer range contribution, z−3 has almost no effect and does
the adsorption decreases the overall free energy, and the not produce qualitative changes of the in plane structure.
maximum possible penetration occurs for ⌫ → ⬁. There is an The limit of a dry, dilute brush is more rich and has been
extremely shallow minimum in the interface potential which investigated for a similar model both below50,52 and above
is beyond the statistical accuracy of our simulations and its is the theta temperature.52 In such cases, the resulting structure
therefore disregarded in the following discussion.48 is laterally inhomogeneous and density profiles as a function
The autophobic dewetting at high grafting densities may of the perpendicular distance to the wall do not provide a
be understood from the resulting density profiles shown in complete description of the brush.
Fig. 5 共left兲 for a brush with R2e / ⌺ = 6.70. In this case, the After this qualitative discussion of the density profiles,
very dense brush results in a highly packed and structured we now study the nature of the wetting transitions that occur
brush, as revealed by ␳g共z兲 for the bare brush. The brush is in each of the three regimes, and the effect of long-range
stretched and the density of brush segments is typically forces.
higher than that of the liquid at coexistence. Thus, only a
very limited amount of the free polymers can penetrate the
A. Low grafting densities
brush and the adsorption ⌫min is very small. It is just suffi-
cient to make the surface of the bare brush adopt the shape of Previously, we have studied the wetting behavior of our
a narrow liquid-vapor interface. When more chains are model polymer on a flat surface without brush.32 We found
added, they are confined to the top of the brush. Since the that the polymer fluid will wet the substrate if the Hamaker
penetration is so small, they do not sufficiently benefit from constant is larger than the wetting value Awet = 3.22. We ex-
attractive interactions with the brush, but suffer a rather large pect that grafting chains on the substrate will increase the
entropy loss due to the steep density gradient at the brush- extent of attractive interactions, i.e., a wetting transition be-
melt interface. low Awet = 3.22 can be anticipated.
The previous discussion is based on the analysis of the In Fig. 6 we show the interface potentials obtained for a
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-8 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

FIG. 6. Interface potentials as a function of total adsorption for a brush of

FIG. 7. Interface potentials as a function of the total adsorption for A = 0 and
density Re2 / ⌺ = 1.09 and several values of the Hamaker constant. From bot-
several grafting densities approaching the wetting transition. From bottom
tom up, A = 0.3, 0.5, 0.7, and 0.85.
up, Re2 / ⌺ = 1.23, 1.28, and 1.31. The symbols show the simulation results.
The full lines are fits to the data 共see text兲. The symbols “X” mark the
location of the minima. The dashed line connects all such points and shows
grafting density R2e / ⌺ = 1.09 and increasing values of the Ha- no evidence of divergence.
maker constant. For small Hamaker constants, g共⌫兲 exhibits
a rather deep minimum at low adsorption. The depth of the
minimum gradually decreases as we increase the Hamaker diverges upon approaching a second-order wetting
constant. Eventually, the minimum becomes metastable, i.e., transition40 but this is not at all what we observe in our
simulations.
a film of infinitely large adsorption is stabilized and the melt
Second, we study the shape of the interface potential as
of free chains wets the brush. From the interface potential we
the wetting transition at A = 0 is approached. From theoretical
infer the location of the wetting transition at about Awet
considerations, we expect that an interface potential for a
= 0.7, well below Awet = 3.22. Furthermore, the structure of
system with short-range forces be described approximately
the interface potentials, with a well developed maximum by a sum of exponentials,40
separating thin and thick films, shows that the wetting tran-
sition is of first order.
⬁
As we increase the grafting density, the value of
Awet共R2e / ⌺兲 gradually decreases. Indeed, from Fig. 2 we ex- ␤g共⌫兲 = 兺 Cke−k⌫/␬ , 共15兲
k=1
pect that for large enough R2e / ⌺ the brush becomes wet for
A = 0. Therefore, the line of first-order wetting transitions
eventually intersects the A = 0 axis. Figure 7 shows a se- where Ck are coefficients depending on the properties of the
quence of interface potentials calculated in the absence of substrate 共i.e., the grafting density兲, while ␬ is a constant that
long-range forces as the wetting transition is brought about only depends on the nature of the fluid. For such a model, the
by increasing R2e / ⌺. In order to assess how close these three first few coefficients of the above expansion vary continu-
systems are from the wetting transition, we calculate the cor- ously with the system parameters. A minimal model of a
responding contact angles according to Eq. 共4兲. second-order wetting transition requires two coefficients: C1
positive and C2 negative, with the latter exactly changing
The values obtained for the three brushes of Fig. 7 are
sign at the wetting transition. The simplest possible descrip-
16°, 12°, and 6°, for grafting densities R2e / ⌺ = 1.23, 1.28, and
tion of a first-order phase transition with the above model
1.31, respectively, i.e., these brushes are very close to the
consists of a positive C1, a negative C2, and a third, positive
wetting transition. Unfortunately, the statistical accuracy of
coefficient C3.
our simulation data is not sufficient to decide whether or not We performed nonlinear regression analysis for the in-
a maximum that separates the nonwet state at ⌫min and a terface potentials using this simple model. A sum of three
macroscopic thick film persists as we increase the grafting exponentials provides an excellent fit to the simulation re-
density. Such a maximum would indicate that the wetting sults 共see Fig. 7兲, while a sum of two exponentials yields a
transitions remain first order. clearly unsatisfactory fit with root-mean-squared deviations
Despite this difficulty, we point out two strong features that are about twice as large 共fit not shown兲. Furthermore,
in favor of a first-order phase transition. First, if a second- taking into account that the parameter ␬ is related to the bulk
order wetting transition occurred, we would expect the correlation length via ␬ = ␰共␾l − ␾v兲, the fit suggests
minima to gradually shift towards macroscopic adsorption. ␰ = 2.24␴ = 0.61Re which corresponds to the length scale of
For this reason we mark the location of the minima ⌫min with density fluctuations in the liquid or the intrinsic width of the
an X in Fig. 7. For short-range forces ⌫min logarithmically liquid-vapor interface 共without capillary waves兲.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-9 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

TABLE I. Results for the fit of interface potentials to Eq. 共15兲. Top: values
of the coefficients ak and bk obtained from the fit to the linear model
Ck = ak + bkRe2 / ⌺ 共a value of ␬ = 1.350 263 is obtained for the decay rate of
the exponentials兲. Bottom: values of the preexponential terms Ck obtained
for the three interface potentials employed in the fit.

1 2 3

ak −0.021 54 −0.9662 0.7576

bk 0.027 82 0.5977 −0.2255

Re2 / ⌺ C1 C2 C3

1.230 0.012 68 −0.2312 0.4803

1.278 0.014 02 −0.2024 0.4694
1.307 0.014 84 −0.1848 0.4628

Furthermore, we can exploit the fit in order to estimate

the actual location of the transition which is dictated by the FIG. 8. Interface potentials as a function of the total adsorption for a brush
condition, of grafting density Re2 / ⌺ = 2.68 and several 共negative兲 values of A as indi-
cated in the figure. Lines with circles: simulation results. Full black lines:
C22 = 4C1C3 . 共16兲 long-range contribution of the interface potential as obtained from the sharp-
kink approximation. Dashed lines: superposition approximation, obtained as
Assuming a linear dependence of Ck on the grafting density a sum of g共⌫兲 at A = 0 and the calculated long-range contribution.
共see Table I兲, we estimate that the liquid will wet the brush
for grafting densities larger than R2e / ⌺ = 1.334 in the absence
of long-range forces 共A = 0兲. tribution in ⌫−8 is extremely small and can be neglected for
⌫ / Re⌬␾ ⬍ 0.5. In all cases, the agreement between the model
and the simulations is good. In the same figure we show
B. Intermediate grafting densities interface potentials obtained by adding the computed g共⌫兲
Next we consider the behavior for grafting densities for A = 0 and Eq. 共17兲 共dashed lines兲. In all cases the agree-
larger than R2e / ⌺ = 1.334, where the brush is wet already in ment is again very good, supporting the superposition ap-
the absence of long-range forces. Including long-range forces proximation employed in previous work.24
with a positive Hamaker constant will make the adsorption At low grafting densities, a sufficiently large A produces
of a thick film even more favorable and the melt will wet the a first-order wetting transition. For intermediate grafting den-
brush. However, including repulsive long-range forces, sities, we have shown evidence for a critical, second-order
A ⬍ 0, we destabilize thick films. In that case, the interface wetting transition as A changes sign. Both regimes are sepa-
potential will exhibit an extra monotonously increasing con- rated by a critical end point located at R2e / ⌺ = 1.334. The
tribution whose leading term takes the form g ⬀ −⌫−2. Thus, question then arises in which form the first-order transitions
no macroscopically thick film is stable for A ⬍ 0. The com- continue beyond R2e / ⌺ = 1.334. Unfortunately, the wetting
petition between a monotonously decreasing short-ranged transitions observed as the grafting density approaches
contribution of the interface potential and the monotonously R2e / ⌺ = 1.334 are so weakly first order, that a direct monitor-
increasing negative contribution from the long-range contri- ing of the interface potential with sufficient accuracy be-
butions produces a second-order wetting transition as A comes very difficult: the possible maximum will be hardly
changes sign.40 detectable within the statistical uncertainty of our simula-
Figure 8 presents the interface potentials for a brush with tions. Therefore, we exploit the smooth dependence of the
grafting density R2e / ⌺ = 2.68 and various Hamaker constants. coefficients in the simple model of the interface potential,
For A = 0, the interface potential is a monotonically decreas- Eq. 共15兲, in order to explore the wetting behavior in the
ing function of the adsorption. As the Hamaker constant de- vicinity of the critical end point. A simple analysis shows
creases, however, a shallow minimum appears and the cor- that the interface potential will exhibit a minimum followed
responding adsorption ⌫min decreases with A. In order to by a maximum as long as C22 − 3C1C3 艌 0. Thus, for an ap-
illustrate that the appearance of the minimum stems from the propriate value of A ⬍ 0 there is a coexistence between a thin
effect of long-range forces, we consider the contribution of and a thick liquid layer on op of the brush. Note that this
the segment-substrate interaction, Eq. 共3兲, to the interface continuation of the line of first-order wetting transition dif-
potential within the sharp-kink approximation,

再冉冊 冉 冊冎
fers from prewetting because it occurs at bulk coexistence
1 ␴ 8
1 ␴ 2 and not at undersaturation. The unfavorable long-range inter-
glr共l兲 = 兩A兩共␾l − ␾v兲 + sign共A兲 , 共17兲 actions play a similar role as the shift of the chemical poten-
8 l 2 l
tial from the bulk coexistence value in prewetting, i.e., they
where we have assumed a film thickness, l = ⌫ / 共␾l − ␾v兲. The limit the growth of a macroscopically thick wetting layer.
leading term of order ⌫−2 is plotted in Fig. 8 and compared Our extrapolation shows that the above condition is fulfilled
with the tails of the computed interface potentials; the con- for R2e / ⌺ 艋 1.3670, just slightly above the predicted wetting
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-10 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

TABLE II. Results for the fit of interface potentials to Eq. 共18兲. Top: values
of the coefficients ai and bi obtained from the fit to the linear model
Ci = ai + biRe2 / ⌺ 共a value of ␬ = 1.2217 is obtained for the decay rate of the
exponentials兲. Bottom: values of the preexponential terms Ci obtained for
the four interface potentials employed in the fit.

1 2 3

ai −0.005 430 0.062 47 0.026 94

bi 0.001 557 −0.016 20 0.003 134

Re2 / ⌺ C1 C2 C3

5.358 0.002 914 −0.024 31 0.043 73

5.626 0.003 332 −0.028 65 0.044 57
6.028 0.003 958 −0.035 16 0.045 83
6.698 0.005 001 −0.046 01 0.047 93

FIG. 9. Interface potentials slightly beyond the autophobic dewetting tran-

sition. The symbols are the results from simulation 共circles, squares, up forces, the stable minimum is raised, and a maximum ap-
triangles, and down triangles correspond to Re2 / ⌺ = 5.36, 5.63, 6.03, and pears leading to a first-order wetting transition at about
6.70, respectively兲. The lines are a fit to the three exponential model, Awet = 1.1. Since a further increase of the brush density at
Eq. 共18兲. The inset shows a series of interface potentials for a brush A = 0 results in deeper nonwet minima and larger contact
Re2 / ⌺ = 6.03 as A is increased, from bottom to top, A = 0.6, 0.8, 1.1, 1.4, and
1.6. angles, we expect that the value of A required to wet a denser
brush will steeply increase.
In order to locate more quantitatively the transition from
transition at R2e / ⌺ = 1.334. Beyond this grafting density the
critical wetting to autophobic behavior and to determine the
two minima structure disappears, i.e., the line of thin-thick
order of the wetting transition, we again perform an analysis
transitions at the coexistence ends in a critical point. Fluc-
based on the simple model for the interface potential, Eq.
tuations which are neglected in the mean field type of ex-
共15兲. The difference in this case is that we do not expect the
trapolation are expected to considerably shorten the line of
adsorbed polymers to significantly penetrate into the brush.
thin-thick transitions and the critical point at the end of the
Therefore, the adsorption of free polymers, rather than the
coexistence line will exhibit two-dimensional 共2D兲 Ising
total adsorption, is the relevant variable and we consider a
critical behavior with strong finite size effects. In view of the
model of the form
very limited regime of grafting density where this mean
field-type extrapolation predicts thin-thick transitions at ⬁
A ⬍ 0 the continuation of the first-order wetting transition to ␤g共⌫ f 兲 = 兺 Cke−k⌫ f /␬ . 共18兲
A ⬍ 0 is virtually unobservable in our simulations. k=1

We find that a fit with two exponentials produces root-mean-

C. High grafting densities
squared deviations which are a factor of 1.4 larger than those
As shown previously in Fig. 2, for A = 0 and sufficiently of a fit to three exponentials, the latter providing a rather
high grafting densities the interface potential developed good description of the simulated data 共cf. Fig. 9, full lines兲.
again an absolutely stable minimum indicative of autophobic Studying the coefficients of the fit 共see cf. Table II兲, we
dewetting of the polymer film on top of the brush consisting find that the simple model suggests that first-order wetting
of identical polymers. Figure 9 shows a more detailed view transitions occur for R2e / ⌺ 艌 5.171 while the wetting transi-
of the interface potentials close to the onset of autophobicity. tion is second-order for smaller values of the grafting den-
The depth of the minimum increases gradually as the brush sity. This is in agreement with the previous SCF
becomes denser. There seems to be no evidence of related calculations.15 The fit of the simulation data does not suggest
maxima in the interface potentials from visual inspection. In that the location of ⌫min on the nonwet side of the first-order
this case, the wetting transition that would occur as R2e / ⌺ transition increases as one approaches R2e / 兺tcp ⬇ 5.171 from
varied while keeping A = 0 is tricritical. At larger grafting above. We note, however, that our estimate of the transition
densities the melt dewets from the brush and an attractive from second- to first-order wetting has substantial uncertain-
long-range interaction is required to stabilize a macroscopi- ties. 共i兲 The line of first-order wetting transitions Awet共⌺兲
cally thick liquid film.31 The dependence of the Hamaker approaches A = 0 with a very small, possibly vanishing slope.
constant Awet on the grafting density in the autophobic re- 共ii兲 The prediction of the fit for the location of the wetting
gime is much stronger than in the limit of low grafting den- transition does not match very well the location of direct
sity, because the melt is displaced from the substrate and estimate obtained by the simulations at larger grafting densi-
only the tails of the attractive segment-substrate interactions ties. In view of the statistical accuracies of the simulation it
stabilize the melt. The inset of Fig. 9 shows a series of in- is again very difficult to accurately locate the point which
terface potentials obtained for a brush of R2e / ⌺ = 6.03 as the marks the end of the second-order wetting transitions and the
Hamaker constant is increased. As a result of the long-range onset of autophobicity directly from the simulation.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-11 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

2␻A⌫0
⌫min共A兲 = ⌫0 + , 共20兲
g0⬙⌫40 + 6␻A

where g0⬙ is the curvature of gsr at ⌫0. Given the new location
of the minimum, the value of the interface potential at that
point is given by
␻A
g共⌫min兲 = g0 + . 共21兲
关⌫0 + 共2␻A/g0⬙⌫30兲兴2

Considering that a wetting transition occurs for g共⌫min兲 = 0,

we find, to the first approximation,

g0⌫20
Awet ⬇ − . 共22兲
␻关1 + 共4g0/g0⬙⌫20兲兴

This simple result qualitatively illustrates the effect of long-

range forces in the neighborhood of a first-order short-range
FIG. 10. Wetting phase diagram as a function of Hamaker constant A and wetting transition. Below the short-range wetting transition
grafting density Re2 / ⌺. The symbols are the wetting transitions as calculated
from the results of this work. The full lines are a guide to the eyes. Labels g0 is negative, and long-range forces with positive A are
W, C, and D indicate the first-order wetting transitions, second-order wetting required to wet the brush. Above the short-range wetting
transitions, and autophobic dewetting transitions, respectively. The dashed transition, and as long as gsr exhibits a metastable minimum,
lines are predictions from the superposition model, Eq. 共19兲. The inset is the long-range forces with a negative A can pull the minimum g0
phase diagram magnified in the region inside the circle.
down and stabilize it.
In order to get a quantitative description of the transi-
D. Discussion of the wetting phase diagram tions discussed above, we solve Eq. 共19兲 numerically, using
The full wetting phase diagram in the A − R2e / ⌺ plane the fit to the interface potentials to interpolate or extrapolate
that results from the above discussion is shown in Fig. 10. the results as required. Included as dashed lines, Fig. 10
The proposed methodology allows us to locate the wetting shows the results obtained for the wetting transitions close to
transitions directly from our simulations 共empty squares兲. We the intersections of the first-order and second-order lines.
find a direct evidence of first-order wetting transitions at low The inset of Fig. 10 magnifies the scale of the phase
共W兲 and high 共D兲 grafting densities and positive values of diagram in the region where line W meets line C. The point
the effective Hamaker constant, as well as wetting of the of intersection is a critical end point, because the second-
melt on top of the brush at an intermediate range of grafting order wetting transitions along C terminate. Below that
densities 共C兲. In the latter case a second-order wetting tran- point, at negative A, we find the continuation of line W. The
sition occurs at Awet = 0. transitions remain first-order, but they occur between a thin
It is difficult to accurately locate the grafting densities at and a mesoscopic film, the latter gradually diverging as
which the wetting behavior changes and to determine the A → 0−. As we decrease A, eventually, the two states become
type of the wetting transition at those points. To this end we identical and the line of transitions ends in a critical point
have fitted the simulation data with a simple model for the 共CP兲 in accord with the predictions of the SCF theory.15
interface potential, Eqs. 共15兲 and 共18兲, at both ends of the From an experimental point of view, the interface poten-
line of critical wetting transitions. For finite values of A the tials are usually not available and one measures either the
effect of long-range interactions is accounted for by a super- adsorption or the contact angle. What would be the experi-
position approximation. Assuming a continuous dependence mental signature of this complex wetting behavior? This
of the fit parameters on the grafting density, we obtain ex- question can be answered by employing our interface poten-
trapolations and we have interpreted them in a mean field tials along with Eqs. 共4兲 and 共7兲. In principle, the adsorption
style. may be obtained as an average over a distribution of the form
The superposition approximation allows us to discuss P共⌫兲 ⬀ exp共−␤LxLyg兲. In practice, for the systems studied the
the qualitative effect of long-range interactions, expected isotherms are rounded off, because of the small
lateral size and the concomitant small free energy differ-
␻A ences. It is therefore more convenient to employ our super-
g共⌫兲 = gsr共⌫兲 + , 共19兲 position model with coefficients as obtained from the fit, Eq.
⌫2
共15兲, and determine adsorptions in the mean field sense, i.e.,
where gsr is the short-range part of the interface potential, as as the extrema of g共⌫兲. From Fig. 10, we have seen that the
obtained from Eq. 共15兲, ␻ = 21 共␾l − ␾v兲3 is a constant, and we superposition model works rather well so we expect our re-
have neglected the ⌫−8 contribution. Let ⌫0 be the adsorption sults to be at least semiquantitative.
at the minimum of gsr共⌫兲. Then, taking the derivatives of Figure 11 共top兲 shows the adsorption isotherms for three
g共⌫兲 and expanding in a Taylor series about ⌫0, we can ob- different brushes and several values of the Hamaker con-
tain a first approximation for the shift of ⌫min due to long- stant. The left panel displays the adsorption isotherms for a
range interactions, brush below the critical end point 共CEP兲 for several values of
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-12 L. G. MacDowell and M. Müller J. Chem. Phys. 124, 084907 共2006兲

lated adsorption at coexistence as a function of Hamaker

constant for the three brushes discussed previously. Again,
the left and right panels show the well known behavior char-
acteristic of first- and second-order wetting transitions. For
the brush in between the CEP and the CP, however, the con-
tinuous divergence of the adsorption 共second order兲 is pre-
ceded by a discontinuous but a finite increase 共first order兲, a
behavior known from experiments of the air/water/oil
interface.26 Since adsorption is related to the derivative of
contact angles, the discontinuity is revealed as a change of
slope in contact angle measurements. The expected values of
the contact angles are displayed as dashed lines in the right
ordinate axis of Fig. 11—bottom 共right ordinate axis兲.54,55
Indeed, a change of slope in the contact angle is observed in
FIG. 11. Typical adsorption isotherms and contact angles in the neighbor-
hood of the critical end point 共CEP兲. Top: plot of adsorption vs chemical the middle panel. However, the change occurs for very small
potential for three grafting densities: below the CEP 共Re2 / ⌺ = 1.206, left兲, contact angles of about 2°. Experimentally, the transition be-
between the CEP and the CP 共Re2 / ⌺ = 1.393, middle兲, and above the CP tween the scenarios depicted in the left and middle panels,
共Re2 / ⌺ = 1.741, right兲. The results are shown for several values of the and the discontinuity in the contact angle was only measured
Hamaker constant. From top to bottom A = 0, and 0.1, 0.244, and 0.32 共left兲;
A = −0.5, −0.2, −8.106⫻ 102, and −6.012⫻ 103 共middle兲; A = −0.5, −0.35, recently.53 The transition between isotherms in the middle
−0.2, and −2.0⫻ 103 共left兲. Bottom: Contact angles and adsorption at coex- and right panels has not yet been found, however.
istence for the previously studied brushes. Contact angles in the middle The nature of the intersection between line D and line C
figure are magnified by 10, while those of the left figure are magnified by cannot be fully elucidated by our simulations. From the ex-
100.
trapolation of the fit, cf. Eq. 共18兲, a continuation of the first-
order autophobic dewetting line also occurs at negative A. In
the Hamaker constant, from A = 0, where the brush is not wet, that case, the intersection point would also be a critical end
to A above the wetting transition. These isotherms clearly point. However, as noted previously, it is unlikely that the
reveal the first-order nature of the transition. A similar set of extrapolation is reliable. We rather expect that the line D
results is displayed for a brush to the right of the CP. In this ends at A = 0 in a tricritical point.
case, the Hamaker constant is increased from negative values
towards A = 0, and the behavior observed is typical of a
IV. CONCLUSION
second-order wetting transition. The peculiarity of the behav-
ior described above is seen when we study brushes in be- In this work we have proposed a powerful simulation
tween the CEP and the CP. This is displayed in the top technique to study wetting properties. The method deter-
middle panel. The isotherms obtained in this case show fea- mines the effective interface potential containing all thermo-
tures common to the left and right panels. For large negative dynamic information. From the interface potential one can
values of A, the system is not wet, and the adsorption re- determine whether the system is wet or nonwet, estimate
mains small up to coexistence. However, when the continu- contact angles, and infer the order of the wetting transitions.
ation of the W line at negative A is crossed, a discontinuous Furthermore, combining the technique with histogram re-
jump is observed in the adsorption. This is very similar to a weighting one can compute adsorption isotherms and locate
first-order wetting transition, as in the right panel. However, wetting transitions.31
the adsorption does not actually diverge at coexistence, but The proposed methodology has been employed in order
rather remains finite, though large compared to that typical of to study the wetting behavior of brushes varying the grafting
adsorbed thin films. As A → 0, the adsorption increases con- density and strength of long-range forces. We have found
tinuously and diverges at A = 0, in very much the same way that the interplay of these two control variables results in an
as the right panel. The wetting isotherm at A = 0 may show a interesting and rich wetting phase diagram, in qualitative
discontinuity, if the brush is close to the CEP, or be continu- agreement with the previous predictions of self consistent
ous all the way up to coexistence, for brushes beyond the field theory.15 In the absence of long-range forces, we find
CEP but yet below the CP. In the figure, we show the latter transitions from nonwet to wet and back to nonwet with
case, where the wetting isotherm at A = 0 still displays an increasing brush density, in agreement with recent experi-
inflexion reminiscent of the discontinuity than can be ob- mental observations.11 The first transition is short-range first-
served for brushes of smaller grafting density. order wetting, while the second is long-range critical wet-
It is important to note that the salient features described ting. Within a very small interval of brush densities beyond
in Fig. 11 occur within an extremely small range of chemical the short-range first-order wetting transition, the adsorption
potentials very close to coexistence. For this reason, it would isotherm at bulk coexistence exhibits a transition from a thin
be very difficult to observe them directly, either in an experi- to a thick layer. Increasing the grafting density further the
ment or in our simulations. In practice, it has been found difference between the thin and the thick state vanishes and
more convenient to focus either on the measurement of the adsorption increases continuously as we approach the
contact angles or on the study of adsorption at co- second-order wetting transition at A = 0.
existence.23,26,28,53 In Fig. 11 共bottom兲, we show the calcu- Long-range forces lead to a line of first-order wetting
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15
 084907-13 Adsorption of polymers on a brush J. Chem. Phys. 124, 084907 共2006兲

22
transitions at low brush density, and to a line of autophobic S. P. Obukhov and A. N. Semenov, Phys. Rev. Lett. 95, 038305 共2005兲.
23
P. Müller-Buschbaum, P. Vanhoorne, V. Scheumann, and M. Stamm,
dewetting transitions at large brush density. At negative
Europhys. Lett. 40, 655 共1997兲.
Hamaker constants, the continuation of the former produces 24
M. Müller, L. G. MacDowell, P. Müller-Buschbaum, O. Wunnike, and M.
the line of first-order transitions between a microscopic and Stamm, J. Chem. Phys. 115, 9960 共2001兲.
25
mesoscopic thin film discussed above. The coexistence be- P. Müller-Buschbaum, O. Wunnicke, M. Stamm, Y.-C. Lin, and M.
Müller, Macromolecules 38, 3406 共2005兲.
tween nonwet thin and mesoscopic films has also been found 26
N. Shahidzadeh, D. Bonn, K. Ragil, D. Broseta, and J. Meunier, Phys.
for short alkanes at the air/water interface,26,53 and for poly- Rev. Lett. 80, 3992 共1998兲.
styrene on silicon wafers.23 The very different nature of all 27
E. Bertrand, H. Dobbs, D. Broseta, J. Indekeu, D. Bonn, and J. Meunier,
such systems suggests that this behavior could be of a rather Phys. Rev. Lett. 85, 1282 共2000兲.
28
K. Ragil, J. Meunier, D. Broseta, J. O. Indekeu, and D. Bonn, Phys. Rev.
general nature.
Lett. 77, 1532 共1996兲.
29
K. Binder, D. P. Landau, and M. Müller, J. Stat. Phys. 110, 1411 共2003兲.
ACKNOWLEDGMENTS 30
K. Binder and D. P. Landau, Phys. Rev. B 37, 1745 共1988兲.
31
L. G. MacDowell and M. Müller, J. Phys.: Condens. Matter 17, S3523
It is a great pleasure to thank Kurt Binder and Claudio 共2005兲.
32
Pastorino for stimulating discussions. One of the authors M. Müller and L. G. MacDowell, Macromolecules 33, 3902 共2000兲.
33
共L.G.M.兲 would like to thank the Universidad Complutense L. G. MacDowell, P. Virnau, M. Müller, and K. Binder, J. Chem. Phys.
117, 6360 共2002兲.
de Madrid and the Spanish Ministerio de Ciencia y Tecnolo- 34
P. Virnau, M. Müller, L. MacDowell, and K. Binder, J. Chem. Phys. 121,
gia for the award of a Ramon y Cajal fellowship and the 2169 共2004兲.
35
MCYT for financial support under Grant No. FIS2004- R. Seemann, S. Herminghaus, and K. Jacobs, J. Phys.: Condens. Matter
06227-C02-02. Another author 共M.M.兲 received support 13, 4925 共2001兲.
36
R. Vazquez, R. Nogueira, S. Busquets, J. L. Mata, and B. Saramago, J.
from the priority program “Mikro- und Nanofluidik” 共Mu Colloid Interface Sci. 284, 652 共2005兲.
1674/3兲 and the Volkswagen foundation. 37
P. Virnau, Dissertation, Johannes Gutenberg-Universität, 2003.
38
G. Torrie and J. Valleau, J. Comput. Phys. 23, 187 共1977兲.
39
1
A. Cabañas, D. P. Long, and J. J. Watkins, Chem. Mater. 16, 2028 C. Borgs and R. Kotecky, Phys. Rev. Lett. 68, 1734 共1992兲.
40
共2004兲. M. Schick Les Houches lectures on Liquids at Interfaces 共Elsevier Sci-
2
M. Müller and L. G. MacDowell, J. Phys.: Condens. Matter 15, R609 ence, Amsterdam, 1990兲, pp. 1–89.
41
共2003兲. A. M. Ferrenberg and R. H. Swendsen, Phys. Rev. Lett. 61, 2635 共1988兲.
42
3
A. I. Milchev and A. A. Milchev, Europhys. Lett. 56, 695 共2001兲. M. Müller and M. Schick, J. Chem. Phys. 105, 8885 共1996兲.
43
4
J. R. Errington, Langmuir 20, 3798 共2004兲. E. Chacon and P. Tarazona, J. Phys.: Condens. Matter 17, S3493 共2005兲.
5 44
C. Rascon and A. O. Parry, J. Chem. Phys. 115, 5258 共2001兲. L. G. MacDowell, M. Müller, and K. Binder, Colloids Surf., A 206, 277
6
M. Müller, L. G. MacDowell, and A. Yethiraj, J. Chem. Phys. 118, 2929 共2002兲.
45
共2003兲. K. Binder, Z. Phys. B: Condens. Matter 43, 119 共1981兲.
7 46
P. Bryk and S. Sokolowsky, J. Chem. Phys. 121, 11314 共2004兲. L. G. MacDowell, P. Virnau, M. Müller, and K. Binder, J. Chem. Phys.
8
S. Tripathi and W. G. Chapman, J. Chem. Phys. 112, 094506 共2005兲. 120, 5293 共2004兲.
47
9
G. Reiter and R. Khanna, Phys. Rev. Lett. 85, 5599 共2000兲. Y. Djikaev, J. Chem. Phys. 123, 184704 共2005兲.
10 48
T. Kerle, R. Yerushalmi-Rosen, and J. Klein, Europhys. Lett. 38, 207 Such minima have been predicted by self consistent field calculations
共1997兲. 共Ref. 17兲 and are indicative of weak first-order wetting transition. In our
11
A. Voronov and O. Shafranska, Langmuir 18, 4471 共2002兲. simulations, however, the minimum is too shallow to be resolved within
12
S. M. Sirard, R. R. Gupta, T. P. Russell, J. J. Watkins, P. F. Green, and K. the statistical accuracy and it occurs at rather large adsorption.
49
P. Johnston, Macromolecules 36, 3365 共2003兲. P. Lai and K. Binder, J. Chem. Phys. 97, 586 共1992兲.
50
13
L. Leibler, A. Ajdari, A. Mourran, G. Coulon, and D. Chatenay, OUMS L. Wenning, M. Müller, and K. Binder, Europhys. Lett. 71, 639 共2005兲.
51
Conference on Ordering in Macromolecular Systems, Osaka 共Springer- W. Zhao, G. Krausch, M. Rafailovich, and J. Sokolov, Macromolecules
Verlag, Berlin, 1994兲. 27, 2933 共1994兲.
14 52
G. Reiter, P. Auroy, and L. Auvray, Macromolecules 29, 2150 共1996兲. S. Metzger, M. Müller, K. Binder, and J. Baschnagel, J. Chem. Phys.
15
M. Müller and L. G. MacDowell, Europhys. Lett. 55, 221 共2001兲. 118, 8489 共2003兲.
16 53
P. G. Ferreira, A. Ajdari, and L. Leibler, Macromolecules 31, 3994 E. B. S. Rafaï, D. Bonn, and J. Meunier, Phys. Rev. Lett. 92, 245701
共1998兲. 共2004兲.
17 54
M. W. Matsen and J. M. Gardiner, J. Chem. Phys. 115, 2794 共2001兲. The only long-range contribution to the interface potentials stems from
18
J. H. Maas, G. J. Fleer, F. A. M. Leermakers, and M. Cohen-Stuart, the explicit long-range monomer-wall interaction with strength A because
Langmuir 18, 8871 共2002兲. all interactions are truncated. It gives rise to a term proportional to A / l2.
19
F. A. M. Leermakers, J. H. Maas, and M. Cohen-Stuart, Phys. Rev. E 66, Upon approaching the second-order wetting transition, A → 0−, this long-
051801 共2002兲. range contribution has to be balanced by an exponentially decaying short-
20
A. N. Semenov, J. Phys. II 共France兲 6, 1759 共1996兲. range contribution 共and not, as usual, a term which is proportional to
21
J. P. Wittmer, H. Meyer, J. Baschnagel, A. Johner, S. Obukhov, L. 1 / l3, cf. Ref. 55兲. Therefore, 1 − cos ⍜ linearly vanishes even at the
Mattioni, M. Müller, and A. N. Semenov, Phys. Rev. Lett. 93, 147801 second-order wetting transition in our model.
55
共2004兲. V. C. Weiss and J. O. Indekeu, Physica A 301, 37 共2001兲.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Tue, 11 Aug 2015 00:36:15