A model for contact angle hysteresis

Cite as: J. Chem. Phys. 81, 552 (1984); https://doi.org/10.1063/1.447337

Submitted: 10 February 1983 . Accepted: 14 March 1984 . Published Online: 31 August 1998

J. F. Joanny, and P. G. de Gennes

ARTICLES YOU MAY BE INTERESTED IN

A thermodynamic model of contact angle hysteresis

The Journal of Chemical Physics 147, 064703 (2017); https://doi.org/10.1063/1.4996912

Dynamic contact angle of spreading droplets: Experiments and simulations

Physics of Fluids 17, 062103 (2005); https://doi.org/10.1063/1.1928828

Generalization of the classical theory of capillarity

The Journal of Chemical Physics 66, 5464 (1977); https://doi.org/10.1063/1.433866

J. Chem. Phys. 81, 552 (1984); https://doi.org/10.1063/1.447337 81, 552

© 1984 American Institute of Physics.

A model for contact angle hysteresis
J. F. Joanny and P. G. de Gennes
College de France 75231 Paris Cedex 05, France

(Received 10 February 1984; accepted 14 March 1984)

We discuss the behavior of a liquid partially wetting a solid surface, when the contact angle at
equilibrium 00 is small, but finite. The solid is assumed to be either flat, but chemically
heterogeneous (this in turn modulating the interfacial tensions), or rough. For weak
heterogeneities, we expect no hysteresis, but the contact line becomes wiggly. For stronger
heterogeneities, we first discuss the behavior of the contact line in the presence of a single,
localized defect, and show that there may exist two stable positions for the line, obtained by a
simple graphic construction. Hysteresis shows up when the strength of the defect is above a
certain threshold. Extending this to a dilute system of defects, we obtain formulas for the
°
"advancing" and "receding" contact angles a' 0" in terms of the distribution of defect strength
and defect sharpness. These formulas might be tested by controlled contamination of a solid
surface.

I. INTRODUCTION the triple line,3.4 are not incorporated here. Since the long
Wetting phenomena have been studied quantitatively range component is independent of surface contaminations,
during two centuries (at least). I The specific observable one may, in a first approximation, ignore it; this is in fact a
°
which we want to discuss here is the contact angle for the
fluid-gas-solid system represented in Fig. 1.
safe assumption if, and only if, the size of the perturbed re-
gions is larger than the range of the van der Waals forces
If the solid surface is flat, smooth, and chemically ho- (e.g., 300 A).
mogeneous we are dealing with an ideal situation. Then, in In Sec. II we discuss the elasticity of the triple line, i.e.,
°
equilibrium, the angle = 00 is related to the interfacial ten-
sions defined in Fig. l(a) through the Young-Dupre equa-
what sort of deformations does the line display when it is
subjected to arbitrary external forces, and in particular to a
tion very localized force. Then in Sec. III we consider the effects
of weak inhomogeneities (small h values in Eq. 1.2), they
YSG - YLS = Y cos °0, (1.1) impose a certain wiggling of the line, but they do not create
This implies, for instance, that a vertical capillary con- any hysteresis when h is small. To obtain hysteresis we need
taining a finite column of fluid, and displaying the angle 00 at "strong" inhomogeneities. This regime is much more com-
both menisci (see Fig. 2) should always drain out by gravity plex. To reach definite conclusions we focus our attention
forces (the Bertrand theorem). Ie In practice we all know first on a single defect (see Sec. IV). Provided that the defect
that, for thin capillaries, the theorem is not valid, the column is small in size, we can construct the various allowed states,
does not move. An equilibrium is set between hydrostatic describing a line which is anchored to the defect, or free from
and capillary pressures, and this imposes that the angles at it. Then we extend this to a distribution of defects on the
the rear and at the front be different. Equation (1.1) is then surface (see Sec. V) and predict the macroscopic contact an-
violated, and the natural interpretation of this violation is gles for an advancing line (Oa) and for a receding line (0,) [see
based on irregularities of the solid surface. There are two Fig. lIb)].
main types of irregularities: The special case of "mesa defects" (see Fig. AI) is dis-
(i) surface roughness 2- 8 and (ii) inhomogeneous conta- cussed at length in the Appendix. The word "mesa" means
mination of the surface. that the perturbation h (x ,y) has step-like singularities on the
Our first aim, in the present paper, is to focus on irregu- defect boundary. This, with a simple choice of defect shape
larity (ii), which is expected to be present often, and which is (rectangle) leads to a relatively simple model, and is instruc-
also amenable to a relatively simple theoretical treatment.
The local inhomogeneity is described by a single function
-h(x,y) = YSL(X ,y) - YSG(x ,y) - [ - YSG + Ysdo,
where the brackets [ ]0 denotes the unperturbed value, and
where the xy plane coincides with the solid surface.
In turns out, however, that most of the physics can be
(1.2)
G
+ + +
S
A11. + + (1 s Ll
(a)

extended to incorporate a discussion of the effects of rough-

A;,~
ness. This is explained later in Sec. VI. Thus, for many prac-
tical situations our analysis covers both problems (i) and (ii).
All our work is based on macroscopic (19th century)
concepts. The special effects of long range van der Waals FIG. 1. Definition of the interfacial tensions (a) and of the receding and
forces, which are known to command the fine structure of advancing contact angles (b).

552 J. Chern. Phys. 81 (1). 1 July 1984 0021·9606/84/130552-11 $02.1 0 @ 1984 American Institute of Physics
 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis 553

imposing that z vanishes on the line

+ + + + z[x,7](x)] = O. (11.2)
Introducing Fourier components through
+ + + +
1 f+oo
7j(q)e1qx dq
.
+ + + + 7](x) = -
21T - 00

+ + + + we have to first order in 7]:

aq = - (Jo7j(q)· (11.3)
p (0) P (b) Inserting this into Eq. (11.1) we arrive at an explicit specifica-
tion for the profile. Inverting the Fourier transforms it can
FIG. 2. A column ofliquid in a capillary. In a, the rear and front contact be written explicitly as
angles are equal the column should drain out. In reality the rear and front
contact angles are different and the column is pinned in the capillary b.
z(x ,y) = (JoY - (Jo + f+ 00 7](X') 2 y ,2 dx'.
1T - 00 Y + (x - x )
(11.4)
tive, But, on the other hand, mesa defects are misleading,
Equation (11.4) may also be derived directly from a two-
because even for very small h, they always lead to some hys-
dimensional electrostatic analog, treating z as a potential
teresis. Finally in Sec. VI, the practical implications and
and (Jo as a charge density imposed on the contact line. Here
limitations of our general model are reviewed.
we are mostly interested in the energy associated with a giv-
en line shape.
II. A SLIGHTLY DEFORMED CONTACT LINE The correction to the capillary energy is
We consider for the moment a perfectly smooth solid
surface, but we assume that the contact line is deformed by
some weak external forces. We want to compute the re-
Ucap = ~ J dxdyy[(VZ)2_(J6],

sponse of the line to such forces. Later (in Sec. III) we shall integrated on all the regiony > 7](x):
replace these forces by the direct effect of inhomogeneities.
The line is described by a displacement function 7](x) defined Ucap = -1y (J 02
q
f -
+ 00

00
2
aqq - - ,
2 1 dq
Iql 21T
(11.5)

6f
in Fig. 3. Note that we assume only weak distortions (in
particular, we do not allow for separate loops). The displace- Ucap = ~ Y (J + 00
dq
Iqll7j(qW 2 . (11.6)
ment 7](x) (when measured for the average line position q -00 1T
y = 0) is small, and we compute all relevant properties to The reader may check that there is no linear term in 7j(q)
lowest order in 7]. in this capillary energy, when one takes into account the
Let us construct first the shape of the interface z(x ,y) Young equilibrium equation for the contact angle (Jo'
associated with a given 7](x). This satisfies the Laplace condi- The unusual Iq I dependence comes from the integration
tion of a q2 energy over a thickness Iq I-I. Again we may Fourier
transform this result and write formally

U
cap
= y(J
21T
6 J dx dx' 7](X)7](X') .
(x _ X')2
(11.7)

Of course cutoffs must be introduced in Eqs. (11.6) and

(11.1) (11.7) to prevent the singularities at large q (or x~x'):the
linearization leading to Eq. (11.3) assumes q7j< 1. Keeping
The first term is the unperturbed profile. The second term these restrictions in mind, we may write down the external
describes corrections of wavelength 217'1lql, extending only force/which is required to create the distortion 7]:
up to a distance Iql-l from the average line position. We can
then relate the amplitudes a q to the imposed line shape 7](x) 8U
F(x)=~= y(J2o
__ J dx' 7]X(') . (11.8)
87](x) 1T (x - X')2
Physically the most interesting object is the response
function R (x - x') giving the deformation field 7](x') for a
localized force/Ix') = /18(x - x'). From Eq. (11.5) we find
.. J(q) = y(J 6Iql7j(q)· (11.9)

y
This gives 10
1 L
R(x)=--ln-, (11.10)
1T'Y(J 6 Ixl
FIG. 3. Wiggly three phase contact line: 17(X) is the line displacement func-
tion in the y direction. where L is a large scale cutoff, provided by some macroscop-

J. Chem. Phys., Vol. 81, No.1, 1 July 1984

554 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis

x ruled by diffusion, with a diffusion coefficient D:

([1](x) - 1](O)P) = 2D lxi, (111.4)
where D is related to the correlation function of velocities by
a classical formula
_d_I_2+--_ _ _ _-+-_ 11 (x)
-dl2
D = i+ «0 (v(O)v(x)dx

= A 2 i + «0 (((O)f(x)dx

(+«0
= A 2)0 (h (0,0), h (x,O)dx. (111.5)
FIG. 4. Distortion of the contact line induced by a force]; (per unit length)
acting over a distance d of the contact line. Equation (111.4) describes a very strong wiggly motion.
For a macroscopic size of droplet L, the amplitudes of fluc-
tuation behave like 1]2 -DL, i.e., they increase likeL 1/2. This
ic sample size. The general aspect of the response is shown in result has also been obtained independently by Vannimenus
Fig. 4, when the forcefl is not strictly restricted to a point, and Pomeau (private communication).
but rather spread out over a small region of linear dimen- As a specific example, let us choose a one-dimensional
sions d, with a central value correlation function
(h (O,O)h (x,D) = h 2e- 1xIIS-, (111.6)
(11.11)
where 5 is a correlation length for the surface inhomogene-
The associated energy is ities. Then we have

Ucap = J fl d1](d) = + k1]2(d), (11.12) (111.7)

k may be called the spring constant of the contact line for In many practical cases we expect h IrO ~ :S 1 and thus
localized peturbations. It will be an essential parameter for 1]2-L5' For a droplet of millimetric size, with defects of
our later discussions on hysteresis. correlation length 5- 10 A we would then expect fluctu-
ations 1] of order one micron. Note that the opposite limit
III. WEAK HETEROGENEITIES h IrO 6> 1 is not compatible with our assumptions. There is
a natural limit corresponding to a local contact angle on the
Let us now consider a heterogeneous solid surface, de- defect 0 = 0, which is
scribed by a random fluctuation h (x ,y) of the work of adhe-
sion, as explained in Sec. I. For the moment we take h (x ,y) to
be very small, so that no hysteresis occurs. But, because of
the heterogeneity, the contact line is distorted. We want to We end up this section with a brief discussion of the line
discuss the statistics of this wiggly line. energy (per unit length) or, more precisely, of the contribu-
As explained in the introduction, h (x ,y) acts as a force tions to this energy originating from the heterogeneity.
pulling the contact line. To a first approximation, for weak From the capillary term (111.6) and the coupling term
distortions 1](x) we may simplify the force as follows: ( - fJ(qJ;q)) we find a correction per unit length of the form

fIx) = h [x,1](x)]-+h (x,D). (111.1) JJ.F = - ~ J!!!L If(q) 12

We can then insert this force [or its one dimensional 2 2rr rO~ Iql
Fourier transformj(q) into Eq. (11.9)], solve for the deforma- 1 5h2 L
- ----In- (111.8)
tion fJ(q), and obtain 2 rO~ 5
[for the one-dimensional correlations see Eq. (111.6)].
(111.2) It is not easy to see, however, how this energy could be
detected in practice. Weak heterogeneities give small line
energies and large heterogeneities are dominated by hystere-
Equation (111.2) has a certain similarity to the viscous tical effects, which we now begin to discuss.
motion of a particle [with position 1](x) at "time"x] under a
random forcef(x). The instantaneous "velocity" v = d1]ldx IV. STRONG HETEROGENEITIES-THE SINGLE
is assumed proportional to the force DEFECT PROBLEM
iqfJ(q) = Aj(q) (111.3) A. The defect force
and Eq. (111.2) is satisfied if we take A = lIrO &(because the We now consider specifically a distribution h (x ,y) with
factor i drops out when we take an amplitude squared). We a peak (at x =xd , Y =Yd) and a certain small width
expect that the long range behavior of this particle will be JJ.x = JJ.y = d. A contact line anchored on such a "defect" is

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis 555

y f

elastic force

y+ y;;, Y
ltd It
FIG. 6. Geometrical construction of the eqUilibrium positions of the con-
FIG. S. A localized Gaussian defect of width d.
tact line in the presence of a localized defect. The nominal line position is Y L •
The maximum advancing position has two possible values y;" and y;;'. In a
receding experiment (retreating line) the anchoring by the defect ends for
shown in Fig. 5. Far from the defect, the line returns to YL = Y +. An advancing line is captured by the defect when YL = Y _.

Y=YL·
For our purpose the important parameter is the total
force exerted by the defect on the line derived in Eq. (II. II ). Thus the balance of forces is simply

11= f ds cos fP h (xs'Y.)

k (Ym - YL) =/1(Ym - Yd).
This leads to a simple graphical construction (see Fig.
(IV.5)

+ co 6). For a given distance between line and defect (Yd - YL)

f
= _ co dx h [x,YL + 7J(x)] ,
where ds is the element of contact line (of position Xs' Ys).
(IV.I) we can find the equilibrium pointsYm. For weak defects (/1/
k small) there is in general only one equilibrium point (no
hysteresis). For stronger defects, we can have three equilibri-
(The study is limited here to attractive defects h > O. The case um points. The smallest (Y;") and largest (Y;;') of these are
of repulsive defects is perfectly symmetric.) stable, while the intermediate one is unstable.
fP is the angle between this element and the average line The onset of hysteresis corresponds to the particular
orientation (x) and 7J = Ys - YL is the distortion. The inte- case where the inflection point in/l(Y) has a slope just equal
gral (IV. I ) is dominated by the central region, where 7J (x) is to the spring constant k. For instance in the Gaussian model,
close to its maximum 7Jm = Ym - YL. In what follows we this corresponds to
approximateII by the simpler form
+ co
_ _ (...!.-.)
ho -h -
112
k -_ (...!.-.) 112 1TY() ~
. (IV.6)
f
J;'.:::::f. _ co dx h (x ,Ym)·
The great merit ofEq. (IV.2) is to generate a force which
(IV.2) e 21T 21T InL/d
Defects with ho < he do not contribute to the hysteresis.
Let us now concentrate on the interesting case ho> he' and
does not involve the whole profile 7J(x), but only its peak assume that we gradually decrease the nominal position of
value (related to Y m). For a given functional form of the de- the line Y L· Then we find anchoring with Y m = Y;;' up to the
fect structure h (x - x d , Y - Yd) we can then compute expli- point where two rootsYm merge and disappear. At this mo-
citly a force/l(Ym - Yd) by Eq. (IV.2). A simple example, to ment the line position is YL = Y +, and suddenly the line
which we shall sometimes refer, is a Gaussian defect snaps to a very weakly distorted profile Ym = y;". Similarly,
h(x-Xd'Y-Yd) in the reverse process (y L is increasing) we get anchoring on
= ho exp - [(x - Xd)2 + (y - Yd)2]/2d 2. (IV.3) Y = Y;" up to a certain position YL = Y _ at which the line
A nice feature of Eq. (IV.3) is that the simplified force jumps off to the other configuration Ym = Y;;' .
(IV.2) is also Gaussian, and thus simple
c. Energy function
II = fiiihod exp - (Ym - Yd)2/2d 2. (IV.4)
The energy associated to the defect is the sum of two
Another merit ofEq. (II.3) is that, if the chemical contamin- contributions:
ants which create the defect were spreading from an initial
point source, simple diffusion would indeed generate a Ucap = !(Ym - yd 2k (IV.7)
an elastic energy discussed in Sec. II;
Gaussian form. However, in most of what follows, we can
pursue our discussion without choosing a specific form for
J;, and this gives a much broader generality to our model. Ud = - J~~ dy;"/I(Y;" - Yd) (IV.8)
a defect energy.
B. Balance of forces
Minimization of this energy with respect to Y m (at fixed no-
Far from the defect, the line has a fixed ordinate Y = Y L. minal position of the contact line YL) gives back the force
The line tip (aty = Ym) is then in equilibrium under the ac- balance (IV.5). The absolute minimum of the energy deter-
tion of two forces. One is the force II described above, and mines the stable line conformation. We may define a line
the other is a restoring force, tending to bring the line back at position YL = Y e , where the two line conformations have
Y = Y L· The spring constant for this second force has been exactly the same energy. This is defined by the Maxwell con-

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

556 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis

YL > Y +. and Y m = Y;" wherever YL < Y +. Integration over

.k.
Ye Y
FIG. 7. Maxwell construction
for the commuting position
YL = Y. at thermal equilibri-
um: the shaded areas are equal.
the defect position Y d at a fixed line position YL is equivalent
to an integration over the line position YL at a fixed defect
position Yd' The result is expressed in terms of the energy
function
y(cos () - cos eol
= m[ - U( + 00) + U(Y+) + U( - 00) - U(Y_)],
struction of Fig. 7. Ifwe were able to shake the line gently (by (V.2)
thermal agitation for small defects or by mechanical vibra- where Y + and Y _ are the positions of the two roots for Ym at
tions for larger defects) the line would then commute reversi- the commutation of the contact line. In Eq. (V.2) the angle ()o
bly from Y;" to Y::' at YL = Y•. corresponds to the absence of all defects.
In real experiments such an agitation does not exist. In At thermodynamic equilibrium the commutation
receding experiments, the line does not commute atyL = Ye , between anchored line and free line occurs at Y L = Ye and
but remains in a metastable position up to YL = Y +. The the two positions have the same energy
jump of the line to its stable position is associated with an
energy dissipation E (Y+) equal to the shaded area in Fig. 8.
y(cos ()E - cos ()o) = n[ U( - 00) - U( + 00 I]. (V.3)
The corresponding dissipated energy for the advancing line The difference U ( - 00) - U ( + 00) depends only on
E (Y _) is the shaded area of Fig. 8'. Ud [Eq. (IV. 8)] because the line is undistorted when it is far
Finally a simple expression can be obtained for the ener- from the defect (Ucap = 0). Then Eqs. (IV.8) and (IV.2) show
gy function by taking its derivative with respect to the no- that
minalline position YL :
"'L U ( - 00) - U ( + 00) = f dx dy h (x - Xd, Y - Y d)'
U = Ucap + Ud = cst -
iY.
fl(Ym - Yd)dYL, (IV.9)
Thus, Eq. (V.3) describes simply the renormalization of
interfacial tensions due to the average density of defects.
Ym is a function of YL with two determinations Y;" and y;;'·
According to the experiment studied one adequate deter- In the receding experiment the jump occurs at Y L = Y +
mination is to be chosen. and U (Y +) - U (Y_) represents the dissipated energy E (Y +)
for a single defect
V. MACROSCOPIC CONTACT ANGLES y(cos () - cos ()E) = n E(Y +). (V.4)
A. A dilute system of defects
In the advancing experiment the jump occurs at
We now want to describe the effect of a distribution of YL = Y _ and U(Y+) - U(Y _)isthenegativeofthedissipat-
surface defects on the macroscopic contact angles. Let us ed energy E (Y _) for a single defect
first assume that all defects are identical, and that they are
y(cos()a -COS()E) = -nE(Y_). (V.5)
spread at random on the solid surface, with a number of
defects per cm 2 which we call n. We assume that the defects In this limit of small defect density. we obtain a very
are well separated (nd 2 -( 1). Then it is plausible to assume general relationship between the advancing and receding an-
that the forces are simply additive. The macroscopic force gle and the total energy dissipated by one defect around a
per unit length is by the Young equation hysteresis cycle Wd =E(Y+)+E(Y_):
y(cos () - cos ()o), (V.6)

and the sum of defect forces is B. Renormalizatlon of the spring constant

n f dYdft(Ym - Yd) = y(cos e - cos eo), (V.l) Our formula (11.11) for the spring constant described a
line attracted by one defect and pinned at the sample edges
where Ym and YL are related by the balance of forces [Eq. (x <L). For the many defect problem we may guess that L
(IV.S)] or equivalently by the construction of Fig. 6. The only must be replaced by the average distance between defects (as
delicate point in Eq. (V. 1) is the choice of roots Ym = Y;" or measured along the contact line) which we shall call b. This
Y;;'. For instance, when the line is pulled towards low YL distance is given by
(receding experiment), we must use Ym = Y;;' whenever b = (ndl- I . (V.7)
We shall now give two slightly more detailedjustifica-
tions of this point; one based on a periodic system, the other
based on the discussion of a pair of defects.

~.
y- y
1. A periodic array
Let us consider a line of defects, located at the points
x = mb (m = integer), Y = YD' Each defect gives a forcefto
FIG. 8. Dissipated energy ofa receding line in the presence ofa single defect and we want to analyze the shape of the pinned line.
8 and of an advancing line (shaded areas). The line shape can be studied by an expansion in a Four-

J. Chern. Phys., Vol. 81, NO.1, 1 July 1984

 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis 557

ier series of period b: strength is above a certain threshold

1J =
+
Lco 1Jp exp(i21T"X)
-- . (V.8) e )112 ryO ~
(VI.I)
p= - 00 b ho> he = ( 21T" In Lid
The constant term (p = 0) in this expansion represents [where, to be specific, we have chosen a Gaussian structure
the overall displacement of the line. If we focus on one de- for the defect, defined in Eq. (IV. 3)]. Equation (VI. I) gives a
fect, the other defects play the role of an effective medium threshold he which depends only very weakly on the defect
which create this overall displacement. Weare interested size d. This, however, results from a cancellation between
here in the perturbation to this average line. two opposite effects, which we now discuss in slightly more
The Fourier series of the periodic force is detail. For this discussion, it is illuminating to consider an
anisotropic delect, with two distinct sizes d x along the direc-
I = "r
£.JIp exp(i21T"PX)
-- , tion of the (average) contact line, and dy normal to it. The
p b
maximum pinning force II is then of order hodx. But the
I" = {b -11' p <bd - I (V.9) onset of hysteresis (as explained in Fig. 6) occurs when the
Jp 0, p>bd -I' spring constant k of the line becomes equal to the maximum
since the defect size (d ) provides a cutoff. derivative ailiay. Qualitatively
-:::::=.II-:::::=.ho~,
The relation between force and displacement is still giv-
en by Eq. (11.9), for a wave vector q = 21T"pb -I:
alII
ay max dy dy
1;, = rO~ 1P121T"b -11Jp. (V.8 ' ) thus
The relative distortion of the contact line can be determined dy
he =-k. (VI.2)
by using Eqs. (V.8) and (V.9): dx
Equation (VI.2) agrees with Eq. (VI.1) when d y = d x ,
1J(X=O) = _J;
_1 -
bid -1= - IJ;- l n ( -
L b.)
2ryO~ -bid IPI ryO~ d but gives much longer domain of hysteresis if d y < d x • More
generally, if the defect has sharp edges it will pin strongly. We
Thus this periodic problem leads to a renormalized elastic
insist on this point in the Appendix, which is aimed at the so
constant
called mesa defect, where the perturbation h (x, y) has step
- ryO~ function singularities on the edges. In this mesa limit hyster-
k=---- (V. 10) esis is present even for very weak amplitudes ho.
In(n-Id -2)
in agreement with our qualitative prediction.
B. Defect statistics
Up to now we have assumed that all defects were identi-
2. The two defect problern-a self-consistent argument cal. In practice they will be distributed in strength and in
Let us consider two neighboring defects at a distance size. Let us, for instance, assume that the size d is fixed, but
x = b from each other, and at the samey = YD. We want to that we have a distribution p (ho)dh o for the strengths, nor-
look at the perturbation which they introduce on the average malized by
contact line. This perturbation can be determined from Eq. (+ 00

(11.2) for the response function )0 p(ho)dh o = n, (VI.3)

1J(X) = JL-(lnl D I +
ryO~ k
Inl~I)·
b-x
(V. 11) where n, is always the number of defects per unit area, and is
assumed small (nd 2 -( I). For each defect, the crucial param-
D is not the macroscopic length L but a new integration eter, controlling the deviation from equilibrium contact an-
constant chosen through a self-consistent argument. gles, and defined in Eq. (V.6), is
Between the two defects the contact line goes back to its Wd(ho)=E(Y+)+E(Y_) (ho;.>h e ). (VI.4)
average (perturbed) position. We find D-b 12.
The effective elastic constant is then We have discussed the analytic structure of Wd(h o) in
Sec. IV (see Fig. 9 also). The deviation from Young formula
k=~= 1T"rO~ is then a weighted average on the Wd • s
1J(d) In(b Id)
y(cos 0, - cos ( 0 ) = (+ ""p(hO)Wd(ho)dh o. (VI.5)
and this leads us back to Eq. (V. 10). )h e

To illustrate this, consider the following example: p(h o)

VI. CONCLUDING REMARKS is a Gaussian
A. The role of defect size and shape
p(h o) = ~1T" h -I exp (- h ~/2h )2 (VI.6)
We have analyzed the effects of certain localized "de-
fects" on the solid surface--chemical inhomogeneities, and assume that the defects are rather weak (h < he ).
characterized by a "strength" ho (giving the local modifica- Then, because WAho)-(ho - he )2 near ho-zhe we find
tion of YSL - rSG' and a geometrical size d). We find that this the following structure (this result is simply obtained by cal-
type of defect will effectively pin the contact line, if the culating the shaded area of Fig. 9 using a Taylor expansion

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

 558 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis

FIG. 9. A proposed experiment for the study

of differential susceptibilities [Eq. (VI. I I)] a FIG. 10. A plane surface with
slight change in the reference line (dzo) in- y smooth roughness.
duces a slight change in the contact angle (dO)
and in the position of the contact line (dYL).
The displacements involved are of order of I
contact angle 00 is the angle between the liquid gas interface
p.: the angles 0 and 0 + dO are both inside the
hysteresis interval (8" Oa)' and the actual surface. It is related to the contact angle with
the average planar surface 0 by

00 = 0 +!!!!.. (~< I). (VI.1I)

force inflection point): dy ay
k 2 2d
y( cos 0 - cos 0 ) _ n - - exp _ _
h (h)5
2
c_ - . (V1.7)
The Young equilibrium equation is then
r hc 2h2 hc
+ ~;) = rSG -
a
rcos(o rSL
Thus, all hysteresis effects become exponentially small
when h<hc' or
r cos 0 = rSG - rSL + h. (VI.12)
C. Limitations The effective increase of work of adhesion is then

= cos 00 -au
The use of Equations such as Eq. (VI. 7) must be con-
- h (x, y) • (VI.13)
ducted with care. On the one hand, it is true that for h < hc ay
the number of pinning defects is small, and thus our assump-
Equation (VI. 13 ) shows an equivalence for the two pos-
tion on dilution is easily satisfied. On the other hand, various
sible kinds of heterogeneities leading to hysteresis for
complications may come in. The (few) remaining pinning
smooth surfaces. The main practical difference is that a con-
defects are deforming the contact line very significantly. But
taminated spot leads to an h function with one given sign and
our analysis assumed that the local inclination ofthe contact one peak (as in Fig. 5) while a rugosity bump leads [from Eq.
line (dz/dx in the notation of Sec. II) was small. Returning to (VI. 13 )] to an h function with a maximum and a minimum.
Eq. (11.10) we see that this imposesf) < 1TrO ~ Ix I, where Ix I is But all our constructions remain valid for both cases.
the distance from the observation point to the defect. The
minimum meaningful value of Ixl is d, and thus the pinning E. Systematic experiments
force must be bounded by
The first experimental aim should be to produce con-
f)<1TrO~d. (VI.8) trolled defects:
But we also know from Eq. (VI.I) that hysteresis occurs (i) On the 10 p, scale, they could be generated by the
only when deposition of metallic (or other films) in the form of spots
with well-defined sizes. This being a rather standard pro-
1TrO~d
f»dk=--. (VI.9) gram in microelectronics. In most practical cases, these de-
InL/d fects will have rather sharp edges, but with a residual
Writing f) = hod we see that the range of interest is smoothness often controlled by diffusion processes. The
graphical construction of Fig. 6 should be essential to discuss
1TY0~ 2 such cases.
- - - < ho < 1TrO o' (VI. 10)
InL/d (ii) One could also generate the defects by chemical
If ho becomes comparable to 1TrO ~, we expect more grafting (silanation, ... ) on surfaces which are accessible only
exotic phenomena (a) strong distortions of the contact line, at certain spots (using photoresist or photolabile film cover-
possibly leading to islands of unwetted regions in a wetted ages and optical images to define these spots).
matrix (or reversely, to disconnected droplets ofliquid on a Having well specified the defects, one would also like to
dry solid), (b) for small defects (d$1 p,) thermal agitation vary continously their relative strength (measured by
may allow hopping from one minimum (y m = y;;' ) to the oth- hol1TYO ~). This is extremely delicate, but could possibly be
achieved with suitable mixtures of two liquids.
er minimum (y m = y;"). We hope to return to these more
Let us now list a few possible experiments: (i) the macro-
delicate questions in future work.
scopic contact angles Oa' Or can be determined by relatively
standard procedures. It is much more different to define an
D. Extension to surface roughness
experiment leading to the thermodynamic value Oe' Con-
All our discussion has been limited to heterogeneities trolled mechanical agitation (e.g., sonication) could, lead to
coming from contaminants on the surface. However our Oe' but this point would require further study, (ii) the contact
model here is more general and can be extended to surface line, anchored on defects, may in principle display areversi-
roughness. 8 Planar surfaces which are smooth but not per- ble susceptibility
fect can also be described by an effective reduction of the
work of adhesion h (x, y). X= -
ciYL (VI.14)
Let us for example assume that the roughness of the d (r cos 0)
surface is characterized at each point by a height u(x, y) (Fig. for small modulations of 0 within the hysteresis limits 0 a' Or'
10). The contact line is parallel to the x axis; the Young andYL being the average line position. For instance with the

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis 559

set up of Fig. 9 one could vary () very slightly ()-+() + d() by a. Defect energy U( y, 1)
shifting the reference Zo; then measure the changes in level of
The defect energy is the sum of two terms. (i) An elastic
the contact line (djiL) by some optical or electrical technique.
energy derived in Sec. II:
However, it must be realized that the displacements d 2 in-
volved are very small (smaller than Y + - Y _ in Fig. 6): there ry() ~ 2
Ud=--L- y ; (AI)
experiments appear feasible only with rather large defects
21n-
(> 1O,u). d
Last but not least, one would like to observe directly the (ii) An adhesion energy corresponding to the defect surface
structure induced by the defects: (i) statics: the geometrical covered by the liquid. This adhesion energy is proportional
distortions of the contact near one defect, and the remarka- to the surface covered. To calculate this energy Uadh , we can
ble "Brownian path" 1](x) predicted at larger scales in Eq. in a good approximation ignore the small difference between
(111.4), (ii) dynamics: when a contact line springs from one 1](0) and 1](d /2) and consider that the defect surface covered
equilibrium position to the next, it must generate some form by the drop is a rectangle.
of weak noise, similar in origin to the Barkhausen noise of At at given distance I between the defect and the con-
ferromagnets, 9 or to Haynes jumps for biphasic flows in po- tact line, the value of the total defect energy U( y,l) depends
rous media. It may be that these jumps can be detected opti- on the value ofy.
cally or acoustically; again we hope to return to these ques- If y < 1, the liquid does not cover the defect Uadh = 0 and
tions later.
U = U =..!.. ry() ~ y2.
d
(A2)
ACKNOWLEDGMENTS 2 In!:...
We have greatly benefited from discussions with M. d
Dupeyrat, B. Legait, L. T. Minassian Saraga, Y. Pomeau, J. If 1 <y < 1 + b, the liquid covers the left part of the defect of
Vannimenus, and B. Widom, on various features of contact aread(y - x) and

ry()~ y2-hd(y-x).
angles.
U=..!.. (A3)
2 In!:...
APPENDIX: MESA DEFECTS d
In this Appendix, we study in great detail the contact If I + b > y, the liquid covers the whole defect
angle hysteresis induced by mesa defects. These defects are
small regions of space where the difference rSL - rsa is re- U =..!.. ry() ~ y2 - hdb. (A4)
duced by a given quantity h. We focus here on attractive 2 In!:...
defects h > 0, the discussion would be identical for repulsive d
defects (h < 0).

1. Hysteresis of a single rectangular defect

The defect has the rectangular shape shown in Fig. AI.
It sits at a distance 1 of the unperturbed contact line. We do
not want to focus on the precise profile of the contact line but
UL{.
I bt-I Y
LL Iv b.1 Y
""In
U~
I b+1 Y

f~ltabl.
on its maximum advancing position y = 1](0). For a given
value ofy, we first derive the excess energy due to the defect
U (y, 1). The equilibrium positions for yare the minima of 1= I= stable

I
this excess energy. Knowing these equilibrium positions and
r
~ metaltable =metaltable
their stability, we build a hysteresis cycle for the advancing
and receding experiments.
t= unstable

contact line
/ surface defect
liquid -~==-.....,:;--- ... I d
b
b»d
i stable = I stable
FIG. A2. Excess energy due to a rectangular Mesa defect for different posi-
FIG. Al. Rectangular Mesa defect: length b, width d, distance to the unper- tions of the defect as a function of the maximum distortion due to the defect
turbed contact line l. y.

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

560 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis

According to the different values of the distance between the

contact line and the defect edge 1 we distinguish five differ-
ent situations:
_hl·JD.D' . E'
J 4
C

rc,~
B
:;V
A
~
R
E
(i) I> Ie = ~ In !::.- ,
free line.
1TYO~
d
The energy is sketched in Fig. A2a. There is one equilibrium FIG. A4. Hysteresis cycle for a rectangular Mesa defect: force--<listance
plot.
valuey = O.
I
(ii) Ie =..!.-. < I < Ie' metastable capture.
2
For a thermodynamic equilibrium experiment, the capture
The defect energy shows two minima. The undeformed ends at point E where this energy reaches zero
profiles is stable but a metastable minimum (U > 0) appears RA - RE = b -Ie = b -IJ2. In the receding experiment
(Fig. A2b) corresponding to Ymin = Ie. the equilibrium profile is the metastable one up to point D.
(iii)!e - b <I <Ie' stable capture. RB - RD = b where it becomes unstable. The defect does
When I = Ie the energy of the second minimum be- not perturb the contact line any longer.
comes negative and there is inversion of stability the stable 2. Advancing experiment. At the beginning of the exper-
profile corresponds to a capture of the contact line by the iment the defect is completely uncovered, the liquid does not
defect. The undeformed profile is metastable, it becomes un- know of the existence of the defect. This unperturbed profile
stable for 1 = O. is stable up to point E, then metastable up to point C where
(iv) - b < I < Ie - b pinning of the contact line on the the contact lines reaches the defect and jumps to its stable C'
right edge of the defect. equilibrium conformation RB - Rc = b - Ie. Increasing
When the right edge of the defect is too close to the the radius we follow then the receding experiment.
unperturbed contact line, the minimum of the defect energy This energy plot shows a clear hysteresis cycle
CC' EDD 'EC. Energy is dissipated at the two points where
remains pinned on that defect right edge. The undeformed
the contact line jumps C and D. The total dissipated energy
profile is unstable.
when going around this hysteresis cycle is then
(v) 1< - b the liquid covers the whole defect.
2
b. Hysteresis of a rectangular defect Wd = Ue - U; + UD - U D = (hd ) In!::.-. (A5)
We can now describe a receding and an advancing ex- 1Ty06 d
periment for the perfect plane with one rectangular defect. The hysteresis effects described in this section are very
We study first the defect energy in the corresponding equi- similar to the hysteresis of a first order magnetic transition.
librium state as a function of the contact line position R [Fig. The role of the order parameter is played here by the variable
(A3)]. This energy is the minimum value of U(y,l) for the y (maximum distortion of the contact line). Instead of the
given relative position of the defect and the contact line. plot energy distance, a more familiar pair of variables to look
1. Receding experiment. We start with a contact line at at the hysteresis is the forceJ = d U / dR distance plot. This is
infinite distance R which completely covers the defect, the somehow analogous to the field magnetization plots of mag-
defect energy is U (R ) = - hdb [situation (v)]. When the po- netic transitions (Fig. A4).
sition becomes equal to RA the defect pins the contact line On this diagram, we can use the usual dissipation some
[situation (iv)]. rule: the total dissipated energfy is the area of the hysteresis
The defect energy is U (R ) = - hdb + !1TYO 6 cycle DCC'D'.
X [(R - RA f/ln L /d]. This pinning lasts up to point B 3. Square deject. The rectangular defects are easy to
where RA - RB = Ie. Decreasing the radius, we reach the study but they are not very realistic because of their asymme-
stable capture region (iii) for which the defect energy is try. The same kind of study can be made with square defects.
U = - hdb + !1Ty[(hd )2/1TYO ~] In L /d - hd (R - RA)' The major difference between the two is that the pinning
regime is more important for square defects. The energy-
radius plot is given in Fig. A5.
U D
B A

C'
Dp= CF~--~I--"-A- -
I

UfD~E C A

D'I = C'~

FIG. A3. Hysteresis cycle for a rectangular Mesa defect: energy--<listance

plot. FIG. AS. Hysteresis cycle for a square Mesa defect: energy--<listance plot.

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis 561

R > R A U = - hd 2 defect covered, Inserting Eq. (AS), we get

2 1 2 (R - RA)2
RB <R<RA U= - hd +~(Jo pinning,
2 InL/d
RD<R<RBU= _hd2+ (hd)2 InL/d (All)
21ry(J~ with
+ hd (R B R) metastable capture. + 00

f
-
f](q,l)lql = _ 00 e - jqxn(x,l)dx.
The position of the different points is given by
For a single defect the profile is related to the decrease in
work of adhesion h through Eq. (II. 9) which can be written in
Fourier space
Rc -RE =1; = [ ~lnL/d -l]d,
1ry(J6 _ 1r 1 -
1](q,l) = - - - - h (q,l). (A12)
Rc -RD =Ic' Iql 1ry(J 6
The total dissipated energy is
The slope (J can be then be written
Wd = Uc - U; + UD - U~~ (hd)2 In~. (A6)
21rY(J~ d
(J (y) = (Jo [ 1 - 2~(J 6 J dlJ dqh (q)e - Iql Yejqxf](q,l) l
2. Dilute system of defects-advancing and receding (A13)
angles
a. Dilute system of defects The defect density is separated into two parts. The
A single defect shows hysteresis but is not sufficient to Fourier transform of the average density has a Dirac func-
explain the difference between receding and advancing an- tion singularity at the origin and thus leads to a finite contri-
gles. We need to introduce a certain distribution of defects on bution to (J (y) even at infinite y. The Fourier transform of the
the surface. We will work here in the dilute limit where the fluctuation 8n(r) has no singularity at the origin it contrib-
surface fraction occupied by the defects is small. The centers utes to (J ( y) at finite y but its contribution vanishes at infinity.
of the defects are situated at random (but fixed) point rj and The contact angle is then given by
the defect density is

n(r) = I8(r - r j). (A7)

This defect density can be separated into two parts an or equivalently

average density n and a fluctuation 8n(r).
For a dilute system of defects, there are no interactions r(cos (J - cos (Jo) = n J dlJ dx h [x,1](x),/]. (A14b)
between the defects. The perturbation of the contact line due
We have here directly demonstrated for mesa defects
to the defects; (x) is the sum of the perturbations of all the
the force balance ofEq. (V. 1).
defects.

; (x) = J J du dl nIx - u,l )1](u,I). (AS)

b. Thermodynamic, advancing, and receding angle
The coordinate 1 (Fig. A6) is the distance between the Before calculating the contact angle in the three differ-
defect and the contact line, 1](x,l) is the profile induced by a ent situations, we need take a further precaution. The equi-
single defect at a distance I of the contact line. This profile librium relationship between the profile and the surface ten-
depends on the type of experiment we are studying receding, sion (A12) is valid for stable and metastable capture but not
advancing or thermodynamic equilibrium, through the do- in the so-called pinning regime. The results of Sec. IV sug-
main of interation of the variable 1. gest that in this regime the relevant reduction of the work of
The experimentally measured contact angle is the slope adhesion is not h but a smaller quantity helf = 1ry(J 61ln L /
of the liquid gas interface at a large distance from the per- dxl '. I' being the distance between the contact line and the
turbed contact line. In the small contact angle approxima- right edge of the defect. It is easily seen that using the bare h
tion would produce too strong a force on the defect and the maxi-
mum advancing value 1](0) would be outside the defect area.
(J= lim az. (A9)
Thus we need to use this effective value of h, which gives no
y_,+aay
force in the limit where the defect is completely covered
The slope is given by Eq. (11.1) for a perturbation; (x) of the
I' = 0 and gives back h at the end of the pinning regime.
contact line
With this precaution it is simple to determine the ad-
vancing receding and thermodynamic angle for a surface
with rectangular defects or square defects (see Fig. A4).

J. Chern. Phys., Vol. 81, No.1, 1 July 1984

562 J. F. Joanny and P. G. de Gennes: Model for contact angle hysteresis

For rectangular defects

r(cos (JE - cos (Jo) = nhdb (the variable I is integrated between points A and E)
n (hdf L
r(cos (Ja - cos (Jo) = nhdb - - --In - (the variable I is integrated between C and A )
2 1Tr(J~ d (AlS)
n (hd)2 L
r(cos (Jr - cos (Jo) = nhdb + - - - I n - (the variable I is integrated between D and A ).
2 1TY(J~ d
For square defects (see Fig. AS)
r(cos (JE - cos (Jo) = nhd 2
1Tr(J~
r(cos (Ja - cos (Jo) = n---d 2
(A16)
21n !:....
d
n (hd)2
r(COS (Jr - cos (JO) = - - - I n L /d + nhd 2.
2 1Tr(J~

'(a) An excellent introduction to capillarity problem is the book by Rowlin- SA. W. Neuman, Adv. Colloid Interface Sci. 4,105 (1974).
son and Widom, Molecular Theory o/Capillarity (Oxford University, Ox- 6R. N. Wenzel, Ind. Eng. Chern. 28, 988 (1936).
ford); (b) Contact angles are extensively discussed by A. W. Neuman in 7R. Johnson and R. Dettre, Surface Colloid Science, edited by E. Matijevic
Wetting Spreading and Adhesion, edited by J. F. Padday (Academic, New (Interscience, New York, 1969), Vol. 2, p. 85.
York, 1978); (c) A description of the early knowledge (up to 1920) is the Be. Cox, J. Fluid Mech. 131, 1(1983).
classic book: H. Bouasse, Capillarite (De1agrave, 1924). 9Barkhausen, Phys. Z. 20, 401 (1919).
2(a) R. Johnson, R. Dettre and D. Brandeth, J. Colloid Interface Sci. 62, 205 '''The Fourier transform oflnlxl/Lis
(1977); (b) R. GoodandM. Koo, ibid. 71, 283 (1979). (c)E. Wolfram and R.
+ ~e~jqx1nMdx =
Faust in Ref. l(b), p. 183. (d) Ottewill in Ref. I(b), p. 183.
's. G. Mason in Ref. I(b), p. 321. f -~ L
-11'.
Iql
·C. Huh and S. G. Mason, J. Colloid Interface Sci. 60, II (1977).

J. Chem. Phys., Vol. 81, No.1, 1 July 1984