PHYSICS OF FLUIDS 19, 123601 2007

Achieving large slip with superhydrophobic surfaces:

Scaling laws for generic geometries
Christophe Ybert,a Catherine Barentin, Ccile Cottin-Bizonne,
Pierre Joseph, and Lydric Bocquet
Laboratoire de Physique de la Matire Condense et Nanostructures, UMR 5586,
Universit Claude Bernard Lyon 1 et CNRS, 69622 Villeurbanne Cedex, France
Received 16 February 2007; accepted 16 October 2007; published online 3 December 2007
We investigate the hydrodynamic friction properties of superhydrophobic surfaces and quantify their
superlubricating potential. On such surfaces, the contact of the liquid with the solid roughness is
minimal, while most of the interface is a liquid-gas one, resulting in strongly reduced friction. We
obtain scaling laws for the effective slip length at the surface in terms of the generic surface
characteristics roughness length scale, depth, solid fraction of the interface, etc.. These predictions
are successfully compared to numerical results in various geometries grooves, posts or holes. This
approach provides a versatile framework for the description of slip on these composite surfaces. Slip
lengths up to 100 m are predicted for an optimized patterned surface. 2007 American Institute
of Physics. DOI: 10.1063/1.2815730

I. INTRODUCTION the Navier boundary condition BC11 for the velocity field at
the surface, beffzv = vw, where beff is the effective slip length,
The design and fabrication of micro- and nanopatterned vw is the slip velocity at the wall, and z is the normal coor-
nonwetting surfaces have received much attention in recent dinate to the wall.
years.1,2 This was initially motivated by the peculiar static- Since the first experiments involving rolling drops,12 a
wetting properties of such surfaces, associated with the so- few experimental works have reported the characterization of
called superhydrophobic effect. The natural nonwettability of friction properties of superhydrophobic surfaces.1319
the flat substrate, as quantified by the liquid contact angle, is All experimental studies confirm large slippage on
enhanced by the underlying roughness, reaching values close microtextured13,14 and nanotextured1619 nonwetting sur-
to 180.3 Depending on the characteristics of the liquid-solid faces, which are orders of magnitude higher than what is
interface, two different classes of superhydrophobic states achievable with liquid on smooth nonwetting walls. Typi-
are exhibited, namely the so-called Wenzel and Cassie states. cally, slip lengths in the micrometer range have been dem-
For the Wenzel case,4 the liquid impregnates the roughness, onstrated on superhydrophobic surfaces, while it remains
while for the Cassie fakir state,5 the liquid interface is re- within the tens of nanometers range on a smooth hydropho-
stricted to the top of the roughness, the roughness being oc- bic surface.2023
cupied by a gas phase. The relative stability of the two states On the theoretical side, flows on composite interfaces
depends on the surface structure characteristics height and combining solid and gas areas were first tackled theoreti-
lateral scale, etc.6 and the experimental conditions liquid cally by Philip8,9 and more recently by Lauga and Stone24
pressure, etc..2,7 While a strong dissipation is expected in and Cottin-Bizonne et al.25 Numerical approaches have also
the Wenzel state as the liquid flow follows the contour of the been followed, either at the molecular scale using molecular
roughness, it was predicted by Philip8,9 that a composite in- dynamics6,25 or at larger mesoscopic scales using, e.g.,
terface in the Cassie state should display a low friction- Lattice-Boltzmann methods.26 At the hydrodynamic level,
superlubricating behavior. the composite surface is modeled as a spatially dependent
Such a superlubricating behavior is particularly attrac- boundary condition, with a no-slip BC on the solid surface
tive in the context of micro- and nanofluidic devices. As and a shear-free BC on the liquid-vapor interface.
downsizing leads to an increased surface-to-volume ratio, su- However, the hydrodynamic flow on a mixed BC is dif-
perlubricating properties of textured nonwetting surfaces ficult to solve in practice and analytical results are only
provide a way to bypass the huge increase in hydrodynamic available in simple geometries. These latter essentially re-
resistance that comes with system miniaturization.10 Accord- duce to the case of a flow on stripes either parallel or per-
ingly, surface effects become key factors in the understand- pendicular to the external flow, as solved by Philip and
ing of the motion of liquids at ever smaller scales. A reduced Lauga-Stone, so that results for more complex geometries,
wall friction is associated with a breakdown of the no-slip even the rather simple case of a regular array of posts, are
boundary condition of the hydrodynamic velocity field at the still lacking. In particular, for a given nanotexture character-
surface, leading to wall slippage. Slippage is described by ized by its morphology, height, length scale, etc. investi-
gated experimentally,1719 there exists no prediction for the
a
Electronic mail: [email protected]. URL: http:// amount of slippage that is to be expected. Even simple ques-
lpmcn.univ-lyon1.fr/interfacesfluides/. tions such as whether slippage on an array of posts is ex-

1070-6631/2007/1912/123601/10/$23.00 19, 123601-1 2007 American Institute of Physics

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-2 Ybert et al. Phys. Fluids 19, 123601 2007

mentum transfer between liquid and wall.27 Second, by as-

suming a shear-free BC over the gas/liquid regions, the vis-
cous dissipation taking place in the underlying gas phase has
been neglected. Both effects are expected to increase the
surface friction, i.e., decrease the effective slip length. The
results obtained in this section, therefore, provide an upper
limit for the slip length. How curvature of the liquid-gas
interface and dissipation in the gas phase affect the slip
FIG. 1. Color online Sketch of the liquid interface at a superhydrophobic length will be discussed in Sec. III.
wall in the Cassie state. L, e: roughness periodicity and height; a: typical In the following, our aim is to define an effective bound-
length scale for solid/liquid contact areas. ary condition for the composite interface in the form of an
effective averaged Navier BC:
pected to be larger than for a stripe geometry remain without
any answer up to now. w = l w = effuw, 1
Thus, a quantitative understanding of liquid friction past
superhydrophobic surfaces is still challenging. The design of where l is the liquid dynamic viscosity, eff is the effective
optimized interfaces faces a lack of predictive tools linking surface friction coefficient, and w, w, and uw are, re-
the wall characteristicstexture geometry pattern type and spectively, the averaged shear stress, shear rate, and slip
dimensions and chemistry setting the intrinsic slip length velocity at the interface. This BC here expressed in the form
over the smooth solidto the final slippage properties. of a stress balance at the interface can also be rewritten to
In the present work, we propose analytical expressions introduce the effective slip length beff characterizing the
for the slip length on superhydrophobic surfaces in the form interface,
of scaling laws in terms of the texture properties. We quan-
tify furthermore how dissipation into the gas layer and the l
curvature of the liquid-gas interface may affect the final re- beff w = uw with beff = . 2
eff
sult. These predictions are successfully compared to numeri-
cal calculations of the slip lengths both by direct resolution
Let us emphasize that the effective slip length beff is the
of the hydrodynamic equations and by finite-element meth-
pertinent BC for the hydrodynamic problem at scales larger
ods. Altogether, these results are used to discuss the merits of
than those characterizing the underlying roughness L. The
different generic surface geometriesstripes, pillars,
roughness scales are therefore integrated out in the definition
holeswith respect to the resulting frictional properties. Fi-
of beff.
nally, we show how these simple analytical laws can be used
to link basic surface parameters to the surface slip length in
A. Limit of vanishing solid areas: s \ 0
order to anticipate and optimize the surface frictional
properties. We first consider the case in which the solid fraction s
is very small. This is the interesting limit to obtain superlu-
II. A SCALING LAW APPROACH FOR SLIPPAGE bricating surfaces, for which slippage effects are expected to
be the largest.
We examine in this part the problem of an idealized In such a limit, the wall is almost frictionless, and close
superhydrophobic surface in the Fakir state sketched in to the interface the flow is plug-like and described by the
Fig. 1 where a liquid slab lies on top of the surface imposed plug-flow velocity U. Accordingly, the averaged
roughness. slip velocity simply reads in this situation uw U.
The liquid-gas interface is assumed to be flat no menis- Let us now estimate the averaged viscous stress at the
cus curvature, so that the modeled superhydrophobic sur- wall w in this limit. The residual friction stress is only
face appears as a perfectly smooth surface with a pattern of
imposed on the solid parts, i.e., over a fraction s of the
BCs. The latter BCs are taken as no-slip over solid/liquid
surface. This leads to an averaged shear stress
areas and shear-free over gas/liquid regions Fig. 1. We de-
note as L the roughness periodicity and a the typical length
scale of solid/liquid areas. The fraction of such solid/liquid w = s l ws , 3
areas will be denoted s. We assume that the fluid is de-
scribed by the Stokes equation, i.e., that the Reynolds num- where ws is the local shear rate on the solid surface. To
ber is very small. This is a pertinent limit for microfluidic estimate this quantity, we recall that the Stokes equation has
devices. Overall, the situation we investigate is similar to a Laplacian form, which strongly couples the spatial depen-
those studied in the simple geometries previously considered dence of the velocity profile along the different axes x , y , z.
in Refs. 8, 9, 24, and 25. This implies that ws = vx / zs U / a, with a the typical
In this idealized surface description, it should be recog- size of the solid area.
nized that two of the assumptions have a possible influence Gathering the above results, one obtains that the effec-
on the surface frictional properties. First, by assuming flat tive slip length scales as beff a / s. We thus write in the
menisci, we have neglected an additional mechanism for mo- limit of small solid fraction s

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-3 Achieving large slip with superhydrophobic surfaces Phys. Fluids 19, 123601 2007

a
beff , 4
s0 s
with a numerical prefactor, which is expected to depend on
the underlying geometry of the surface stripes, posts, etc..
This relationship is the main result of this paper.
First, it is interesting to compare this scaling law 4 to
existing analytical predictions in simple geometries. For pe-
riodic grooves oriented parallel8,9 respectively, perpen-
dicular24 to the flow, the exact expression for beff actually
reads

beff =
L



log cos 1 s
2
5

FIG. 2. Normalized effective slip length beff / L for a composite surface made
respectively, 1 / 2 of Eq. 5. In the limit of vanishing solid of solid patches organized on a square lattice; limit of low solid fraction:
fraction s, it therefore predicts beff to depend only logarith- s = 2 30%. beff / L is plotted against 1 / s according to Eq. 4 with s
mically on s through beff L log s. For this groove ge- = a / L2. 2D numerical approach Ref. 25: solid patches of square
ometry, the solid fraction simply reads s = a / L so that the cross section; 3D finite-element calculation: solid patches of square
cross section, circular cross section; linear regression: beff / L
scaling law approach 4 now becomes = 0.325/ s 0.44.
beff L. 6
s0

In agreement with the exact calculation, our approach, there- access the coefficients of the scaling behavior, Eq. 7,
fore, predicts no dependency of the effective slip length on beff / L 0.325/ s 0.44. This formula provides a useful and
the solid fraction s to leading order in s i.e., up to the very simple expression for the slip length on patterned sur-
logarithmic term. faces of posts.
We now consider a more complex geometry of major To finish, we show that it is simple to relax the assump-
practical interest: a bidimensional 2D pattern of posts. In tion of a no-slip BC on the solid areas, assuming a finite
this situation, the solid fraction now reads s = a / L2 so that intrinsic slip length bs. Slippage over a bare, smooth surface
Eq. 4 predicts that here denoted as intrinsic slippage has been intensively
L investigated over the past decade.28 An intrinsic slip length
beff 7
s0 s . bs of a few tens of nanometers is demonstrated over smooth,
hydrophobic surfaces.2023
Since no exact calculation is available for post patterns, we Going back to the above derivation of the scaling law in
have checked the validity of the above scaling law using a the limit s 0, one expects that a finite slip length on the
numerical calculation for the slip length. To this end, we solid will reduce the shear rate ws over the solid regions:
used a previously reported numerical approach.25 We only ws U / a + bs. The averaged shear stress over the total
briefly recall here the basic steps of this approach and refer surface now reads w = slU / a + bs. One gets accord-
the reader to Ref. 25 for a more detailed description. A shear ingly a modified scaling law for the effective slip length,
flow is considered over a composite surface characterized by
a heterogeneous slip length pattern. The boundary is mod- a + bs
eled by a pattern of local slip lengths on a planar surface. beff . 8
s0 s
The characteristics of the flow far away from the surface and
an effective slip length are determined by solving the hydro- This modified scaling is tested on the square lattice ge-
dynamic equations with the hydrodynamic BC given by the ometry of solid patches in Fig. 3. The difference between the
local slip length, using an integral method. effective slip length beffs , a , bs with intrinsic slippage bs
Using this numerical approach, the effective slip length on the solid, and the no-slip prediction ideal case shown in
was computed for a square lattice of solid patches of square Fig. 2 bideal beffs , a , bs = 0, is plotted against bs / s. This
cross section. The computed slip lengths beff / L are plotted in figure confirms the soundness of the scaling law 8, which,
Fig. 2 against 1 / s for different solid fractions in the range therefore, allows us to quantify the impact of an intrinsic
s 30%. The agreement with the prediction 7 is shown to solid slip on the overall effective slip over superhydrophobic
be excellent, therefore validating the proposed scaling law. surfaces.
As an additional check, these results were complemented by Additionally, it is important to note that bs / s does not
numerical results obtained from 3D finite-element methods depend on the geometry of the surface so that the increase of
see Sec. III B for details in the same geometry square lat- the effective slip length due to the intrinsic slip is indepen-
tice of solid patches of either square or disk shape. As dent of the details of the underlying surface texture stripes
shown in Fig. 2, the agreement is again excellent, with all or posts. Moreover, Eq. 8 shows that such a term will play
data collapsing on a single straight line. A linear regression a role only if the size of the solid region, a, is of the order of
performed on the numerical calculation data allows us to the intrinsic slip length: bs a. With bs limited to a few tens

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-4 Ybert et al. Phys. Fluids 19, 123601 2007

FIG. 3. Influence of a finite slippage on solid patches on the normalized

FIG. 4. Normalized effective slip length beff / L for a composite surface of
effective slip length beff / L for a composite surface made of solid patches
various geometries in the limit of high solid fraction: s = 60 100%. beff / L
organized on a square lattice; limit of low solid fraction: s = 2 16%. In-
is plotted against 1 s / L according to Eq. 9 with the geometry-
crease in beff / L compared to the reference for no slip on solid is plotted
dependent typical size of shear-free areas. Symbols are numerical data ob-
against bs / Ls according to Eq. 8. 2D numerical approach for patches of
tained by solving the hydrodynamics equations Ref. 25. Main: = L1
square cross section Ref. 25: s = 2%, s = 5%, s = 16%;
s; square solid patches organized on a square lattice surface struc-
linear regression: beff / L = 0.165bs / Ls.
ture made of pillars, solid stripes perpendicular to the flow, solid
stripes parallel to the flow, linear regressions: slopes are, respectively,
0.21 0.02 and 0.41 0.04. Inset: = L1 s; square shear-free patches
organized on a square lattice surface structure made of holes, linear
of nanometers,28 intrinsic slippage will only play a role for regression: slope is 0.17.
very sharp nanostructures, with a in the tens of nanometer
range.
surface made of pillars as already studied in Sec. II A, and of
B. Limit of vanishing gas areas: s \ 1
square gas shear-free patches, corresponding to a holes sur-
We now consider the opposite limit of vanishing gas face texture.
area: s 1. This limit has a much lower practical impor- It is interesting to note that the two 2D geometries in-
tance than the previous one, s 0. Indeed, in this case the vestigated do not behave the same way in the considered
shear-free area will have a tiny influence on the wall friction limit a point discussed in more detail in Sec. II C. Namely,
and the effective slippage is expected to be much smaller. We for the hole pattern, the characteristic size simply relates to
mention this result for completeness. the area of the hole so that one has = L1 s. On the
Since in this limit the BC is almost everywhere a no-slip contrary, for a post pattern, the characteristic size over
BC, the flow is now appropriately described by the imposed which the flow is modified is associated with the small width
shear rate . Accordingly, the averaged viscous stress at the of the gaseous stripe separating two adjacent posts. Conse-
wall is expected to basically reduce to w l in this quently, scales as L1 s for the post pattern, which
limit. means that this 2D structure behaves as a stripe-like geom-
Now, the averaged slip velocity at the wall only results etry when s 1.
from the 1 s surface fraction of gas areas. In these re- Numerical results for the effective slip length obtained
gions, the typical fluid velocity is of order , with a with the different geometries simple grooves, posts, and
characteristic length scale for the gas areas. Combining those holes are gathered in Fig. 4. The main figure shows data
two estimates leads to the scaling behavior obtained for grooveseither parallel or perpendicular to the
flowand posts. The expected scaling beff L1 s2 is
beff 1 s. 9 very well verified with = 0.41 0.04 respectively,
s1 = 0.21 0.02 agreeing with the prefactor = 0.393 respec-
tively, 0.186 obtained from analytical prediction8,9,24 for
Again, this result 9 can be first compared with analytical grooves parallel respectively, perpendicular to the flow see
predictions for grooves.8,9,24 For this simple one-dimensional Eq. 5. As predicted, in this s 1 limit, the post geometry
geometry 1D, the typical length scale is simply = L1 is equivalent to perpendicular grooves in term of surface fric-
s and Eq. 9 reduces to beff L1 s2. This prediction tional properties.
is in perfect agreement with the analytic expression in this Finally, the inset of Fig. 4 displays data for the hole
limit.8,9,24 pattern. The modified behavior now predicted by the scaling
Turning now to geometries made of 2D patterns, the law, beff L1 s3/2, is perfectly recovered.
scaling law predictions are tested against the numerical re- Overall, the agreement with Eq. 9 is shown to be ex-
sults for the slip length obtained using the approach de- cellent for all geometries and for solid fractions s 60%.
scribed in Ref. 25. Two geometries have been considered: a Eventually, we quote that the effect of intrinsic slippage
square lattice of square solid patches, corresponding to a on the solid surface can be considered along the same lines

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-5 Achieving large slip with superhydrophobic surfaces Phys. Fluids 19, 123601 2007

as above, yielding the prediction beff bs to leading order in

1 s: the main contribution to slippage originates in the
intrinsic slippage on the solid surface, therefore correspond-
ing to a small effective slip length.

C. Influence of the geometry

on the effective slip length
In the experimentally interesting limit s 0, the differ-
ent results can be summarized, restricting to the case of a
no-slip BC on the solid regions, as
stripe geometry beff L logs,
10
L
post geometry beff
s . FIG. 5. Normalized effective slip length beff / L as a function of the gas area
fraction 1 s for a composite surface of geometries. 2D numerical ap-
These predictions show that the geometry has a strong effect proach Ref. 25 square solid patches organized on a square lattice
on the effective slip length. Indeed, for stripe geometries the surface structure made of pillars, square shear-free patches organized
on a square lattice surface structure made of holes, -- solid stripes per-
log-dependency in s practically limits the achievable effec-
pendicular to the flow, solid stripes parallel to the flow. Inset: focus on
tive slip lengths to no better than the surface periodicity L. the low gas fraction region.
On the contrary, the power-law divergence exhibited by the
post pattern as s 0 makes it possible to overcome the L
bound to achieve much higher effective slip lengths.
It is interesting to note that this change in behavior does limit, while in the opposite limit shown in the inset of Fig.
not hold for any 2D pattern as opposed to a 1D pattern. As 5, it is the pillar surface that behaves that way. In the limit
already noticed in the preceding section, 2D holes and post s 1, the surface made of holes is the one having the
patterns may have completely different frictional properties, higher effective slip length for a given solid fraction.
and in the present limit s 0, the pattern of holes surface Altogether, surfaces made of pillars posts are the best
indeed reduces to a stripe-like one. As a matter of fact, for candidates to obtain very large slip lengths in the s 0
holes, s 2a / L with a the thickness of the solid walls limit.
surrounding the holes, so that the general predictions, Eq.
4, lead to beff L, and holes thus behave like stripes.
Of course, this is to compare with the opposite limit III. TOWARD A GENERAL DESCRIPTION
s 1, where we already noticed that the post pattern this FOR THE SLIP LENGTH
time behaved as a stripe-like structure, in contrast to the
In the preceding section, we derived general scaling laws
behavior of the hole pattern.
relating the surface geometry to the effective surface friction
The two limiting cases can be rationalized in the follow-
at the superhydrophobic wall. However, these general scaling
ing way. Slippage in the s 0 limit depends on the typical
laws are obtained within an idealized description assuming a
scale a of the solid phase see Eq. 4, while in the s 1
flat liquid interface and ideal shear-free BC over the gas-
limit it depends on the gas phase extension see Eq. 9. A
eous regions.
structure shifts from what we called 2D behavior to a 1D
The validity of these assumptions is now examined and
stripe-like behavior whenever this typical size a respec-
these effects are quantified. Expressions for the effective slip
tively, no longer reduces to the square root of the solid
lengths taking these effects into account are provided as a
respectively, gas phase area. Not surprisingly, this type of
predictive tool for experimental situations.
criterion amounts to considering a shape factor for the phase
under scrutiny: an elongated phase structure leading to a
A. An interpolation formula for composite slippage
stripe-like behavior.
These points are summarized in Fig. 5, where the effec- We start by considering the situation in which the slip
tive slip lengths are plotted as a function of the gas area length on the gas phase is not infinite, so that the local BC is
fraction 1 s for different patterns: parallel grooves, per- not a shear-free BC as assumed in Sec. II. We denote as bg
pendicular grooves, pillars, and holes. Note that for discreti- the slip length on the liquid-gas interface. We assume that a
zation reasons in the numerical calculations, we were limited no-slip BC applies on the solid regions.
to solid fractions s higher than 1% for pillars, and to the Let us first consider the limiting case bg 0. For a shear
more stringent condition s 10% for holes. Even with this flow with shear rate , the velocity profile may be approxi-
restriction, the effective slip length for surfaces made of pil- mated as ux , y , z = z + bx , y, with bx , y the locally
lars squares is seen to be much higher than for periodic varying slip length. Such an approximation is expected to be
grooves solid and dashed lines and holes circles in the valid when b is smaller than its variation length scale b
limit s 0. Additionally, one verifies that holes behave as 1, expressed here as bg L. The averaged velocity profile
grooves perpendicular to the flow in this low solid fraction thus reads uz = z + b. Using the definition of the ef-

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-6 Ybert et al. Phys. Fluids 19, 123601 2007

In the following, we will make use of this interpolation

to discuss the effect of a finite dissipation in the gas phase.

B. Finite dissipation within the gas subphase

We now discuss how the dissipation in the gas phase
modifies the friction properties and the effective slip length.
In Sec. II, the assumption of a shear-free BC at the
liquid-gas interface implicitly amounts to neglecting such a
dissipation. As a consequence, the predicted effective slip
length at the surface does not depend on the liquid viscosity.
This is to contrast with the opposite situation, which we will
refer to as the gas cushion model, which assumes that the
dissipation at the surface is dominated by the shearing of a
FIG. 6. Effective slip length beff for various surface geometries, compared
continuous gas layer, lying in between the solid and liquid
with the interpolation formula 12. Numerical results for the effective slip phases gas cushion. Such a model predicts29,30
length using the approach in Ref. 25: periodic pattern of pillars with
s = 3 80%; periodic pattern of holes with s = 60 80%. This also in-
corporates the effect of the viscosity of the gas phase g on the effective slip l
length see Sec. III B. In this case, the effective slip length is calculated beff = e, 13
g
using finite-element calculations with various surface heights e / L 0.052
and viscosity ratio l / g: 0.01100: , stripes perpendicular or par-
allel to the flow with s = 20%; same for pillars; same for holes. with g the gas dynamic viscosity and e the thickness of the
-- linear prediction: y = x.
gas layer. Of course in the limit of vanishing s, this should
represent an upper bound value toward which the slip length
will saturate.
fective slip length, this velocity profile is expected to identify In the following, we develop an approach to estimate the
with uz = z + beff. This yields the following expression dissipation resulting from the gas phase in intermediate cases
for the effective slip length in this limit: finite s and g / l. Focusing on a surface structure corre-
sponding to a square lattice of cylindrical pillars, which re-
beff = 1 sbg . 11 semble most of the artificial superhydrophobic surfaces, we
One may show this result more rigorously using a will first define and estimate an averaged slip length bg over
perturbative approach of the flow field in bg. The reference the gas area. Then we will make use of the interpolation
flow field corresponds to bg = 0, which reads, for a shear rate formula, Eq. 12, to obtain the effective slip length.
at infinity, u0 = z. The first-order correction for the flow Before proceeding with this strategy, it should be men-
field over a composite surface with no-slip region S0 and tioned that the problem of Stokes shear flow over a liquid-
partial slip region Sg thus verifies lu1 = P1, together with impregnated roughness g now equals l has been the sub-
the following boundary condition on the liquid-gas ject of different works, especially in the simpler groove
interface: bgzu0 + u1w = u1w. In the limit bg 0, one geometry, where it has been solved analytically by Wang31
has u1w bg so that in the end uw = 1 sbg , and using eigenfunction expansions and matching. In this simpler
geometry, a strategy might then be to follow Wangs ap-
beff = 1 sbg as shown above.
proach by modifying the matching condition to incorporate
For large bg, no such simple prediction could be ob-
the viscosity contrast. To our knowledge, however, no such
tained. However, it is possible to propose an interpolation
route has been developed for the surfaces made of pillars that
formula that will prove to be useful in the following. This
we consider now.
interpolation is based on the two known limits: for bg 0,
We start with the estimate of the averaged slip length bg
beff = 1 sbg; for bg , beff bideal with bideal the effec-
over the gas area. To evaluate the extra dissipation occurring
tive slip length obtained in the idealized case where a shear-
within the gas subphase, we use an effective medium ap-
free BC is assumed at the liquid/gas interface. bideal is given
proach analogous in spirit to Darcy-Brinkmanns law in a 2D
by the expressions discussed in Sec. II typically bideal
porous medium.32 The gas is assumed to flow parallel to the
a / s.
bottom reference surface see Fig. 1 and satisfies a Stokes
A heuristic formula interpolating between these two lim-
equation,
iting cases is
1 1 1
= + . 12 gU = P, 14
beff 1 sbg bideal
This amounts to adding the friction coefficient in the above with P the gas pressure. An effective medium approach
two limits. This interpolation formula is tested in Fig. 6. amounts to replacing the gas+ pillars problem by a homo-
While there is no deep physical justification for such a for- geneous problem with an effective velocity U and an effec-
mula, it is shown to provide a very good description of the tive pressure gradient whose strength is set by the drag on
effect of a finite slip length on the liquid-gas interface. the pillars,

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-7 Achieving large slip with superhydrophobic surfaces Phys. Fluids 19, 123601 2007

P = Fd , 15
with = 1 / L the number density of pillars and Fd the drag
2

force per unit of length experienced by a single pillar. The

, where is the friction
latter is of the form Fd = sU s
coefficient, the s dependency of which accounts for hydro-
dynamic interactions between pillars and can be found in the
literature.33 For example, for a square array of pillars, a se-
ries expansion in s leads to
4g 1
= ln s 0.738 + s 0.887s2 + 2.039s3
s 2
+ Os4, 16
accurate to within 5% in the range of practical interest s
30%. Combining Eqs. 14 and 15 with the definition of FIG. 7. Effective slip length bg over gas areas as obtained from finite-
element calculations, compared with theoretical predictions according to Eq.
s, one obtains the relation 19. The surface is made of cylindrical pillars of height e arranged on a
square lattice and the viscosity ratio l / g between the two phases is varied
= s U
U , 17 from 0.1 to 100; solid and empty symbols correspond, respectively, to e / L
g = 0.25 and e / L = 1. , s = 30%, , s = 20%, , s = 10%,
, s = 3%, -- perfect match: y = x.
complemented by the BCs U z = 0 = u and U
z = e
w g
= 0. It resolves straightforwardly and yields an averaged vis-
cous surface stress of
slip length bg over the gas areas. The latter is defined from
q the averaged shear stress wg and the averaged slip velocity
wg = guwg , 18
tanh qe uwg over the gas phase according to Navier-like relation-
ships of the form 1 and 2.
with q2 = s / g. According to Eqs. 1 and 2, this pre- Let us first discuss the result for the averaged slip length
dicts for the averaged slip length bg over the gas areas bg. The comparison between the theoretical prediction 19
el tanh qe and the numerical results using finite-element computations
bg = , 19 is shown in Fig. 7. For the explored viscosity ratios l / g,
g qe
the gas layer thickness e / L, and the solid fraction s, our
with qs given above, an expression reminiscent of the analytical prediction proves very satisfactory and remains al-
Beavers and Joseph BC on porous media.34 ways within 20% of the numerical value.
To estimate the effective slip length, we now use this We can now test the global result for the effective slip
prediction together with the interpolation formula, Eq. 12, length, in Eq. 20. This phenomenological relationship is
to obtain tested by comparing with the results of the 3D finite-element
calculationwhere both beff and bg are measuredbut also
1 1 1
= + , 20 by comparing with the 2D numerical approach25 with a flat
beff el tanh qe bideal composite interface with mixed BC: no-slip and partial slip
1 s
g qe with slip length bg. This is done in Fig. 6, where beff 1

bideal is plotted against 1 sbg. As already noted, the

1 1
where we recall that bideal stands for the effective slip length
interpolating formula 20 appears to be astonishingly effi-
for the composite surface with a negligible dissipation in the
cient. It describes with very good precision the effect of the
gas shear-free BC, as discussed in Sec. II.
finite dissipation within the underneath subphase. Combining
To assess the validity of this approach, we have con-
Eqs. 16 and 20 allows us to quantify the influence of the
ducted numerical resolution of the Stokes equations using
subphase dissipation on the effective slip length.
finite-element methods implemented through Comsol. The
computed geometry is like the one sketched in Fig. 1 with an
C. Pressure dependency: Meniscus curvature effect
undeformable planar liquid interface separating a liquid
phase and a gas phase, with a given viscosity ratio g / l. A In this section, we discuss qualitatively the effect of a
planar upper wall not sketched encloses the liquid slab pressure drop P = Pg Pl between the liquid and the under-
separated from the liquid interface by a distance H H was lying gas phase. A pressure drop is associated with a curva-
chosen large enough so that the results matched the H ture of the liquid-vapor meniscus, leading to a reduction of
limit. Symmetries were used to reduce the size of the com- slippage. Our aim in this section is to provide an upper limit
putation cell. A Couette flow is simulated by imposing a for the pressure drop P below which curvature effects can
velocity U of the top planar wall. be neglected. To this end, we shall focus on the case in which
The resulting flow field is computed, from which we the gas meniscus is oriented outward protruding by a dis-
have extracted the effective slip length beff over the entire tance into the liquid phase, which can be handled in a
surface, as done previously. We also measured the averaged simple way within the present description. This situation was

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-8 Ybert et al. Phys. Fluids 19, 123601 2007

recently shown to be relevant experimentally whenever no TABLE I. Summary of the effective slip length for flows over a surface
made of pillars in various situations. bg origin and expressions: a sublayer
direct control can be exerted on the gas subphase,35 and we viscous dissipation bg = el / gtanhqe / qe; b curvature of menisci bg
expect the present analysis to capture the general features = 2lv / P.
occurring at curved interfaces.27,36
To obtain how pressure-induced curvature effects Model assumptions beff
modify the slip length, we follow a similar approach to that
Ideal: flat, no-slip, shear-free bideal = L0.325/ s 0.44
in the preceding section. The additional dissipation resulting
Finite solid slip bs bideal + 0.165bs / s
from the boundary curvature will be incorporated as a finite
Finite gas slip ba,b bideal + 1 sbg11
1
g
slip length bg applied on the liquid-gas interface. In the case
of a curved gas meniscus, this effective slip length bg should
correspond to the local radius of curvature,37 which reads
bg 2 / in a parabolic approximation. value beff 2lg / P. To fix ideas, for P 1 bar, beff satu-
This relationship, based on the local expression of the rates at beff 1 m.
stress tensor in the tangential/normal vector frame,37 can ac- These results allow us to estimate the domain of pressure
tually be recovered with a simple scaling argument. Starting difference P, for which the curvature effects do not affect
from the expression for the drag force F = 4lRu experi- the slippage on the superhydrophobic surface. According to
enced by a spherical bubble moving at velocity u,38,39 we Eq. 24, this amounts to L2 / bideal, with bideal the slip
expect the additional drag exerted on the cap to be of the length when curvature effects are negligible. This can be
form rewritten
Fg luwg . 21 lg
P . 25
The averaged shear stress resulting from this drag force is bideal
simply wg = Fg / 2 with the typical length scale for the This condition shows that situations with larger slip
gas areas so that in the end one recovers for the effective slip lengths are more sensitive to curvature effects. However, on
length over the gas areas can note that a reasonable pressure difference in the order of
tens of mbars will not affect a slip length in the hundreds of
b g 2/ . 22
micrometers range. It should therefore be moderately limit-
We are now in a situation similar to the one considered in ing in the design of reduced friction interfaces. This is sup-
Sec. III B except for the origin of bg. The effective slip ported by the absence of pressure effects reported in Ref. 19
length can be obtained using the interpolation formula 12. or the low curvature reported in Ref. 13.
It is first interesting to compare this prediction with a We eventually remark that such curvature effects are ex-
recent analytical work considering the 1D situation of a shear pected to be more important for surfaces with a roughness
flow parallel to grooves.36 In this geometry, a first-order per- made of holes, for which the gas is not connected to a res-
turbative approach was developed that allowed to quantify ervoir. Depending on the condition for surface immersion,
the effect of meniscus curvature in the limit = L gas may be trapped in the holes with a non-negligible excess
a. Considering the limit of large solid fraction s, our pressure. Such effects have indeed been observed in Ref. 35,
scaling approach provides in this 1D situation the following with a geometry of holes of size 650 nm, thus resulting
curvature correction to the effective slip length: in a strongly reduced slip length.

beff bideal 1 s , 23
IV. DISCUSSION
which is obtained by combining Eqs. 9, 12, and 22. This
The results obtained in the previous sections are summa-
result agrees fully with the first-order correction term found
rized in Table I for the pillars surface geometry. Based on
in Ref. 36.
this, we now discuss how to optimize slippage, and what
In the more interesting limit of a 2D surface pattern with
maximum slip length may actually be expected, taking into
s 0, the ideal flat surface slip length is now given by Eq. account the actual limitation in surface engineering nano-
4. We thus obtain
lithography, etc.. In the following, we will restrict our dis-

beff =
L2
+
1
bideal
1
24
cussion to the case of flat interfaces, i.e., the pressure differ-
ence P obeys the nonrestrictive condition in Eq. 25. As
we noted above, a geometry of holes is to be avoided to
with bideal / L 0.325/ s 0.44 see Sec. II. This result reduce these effects. This is, however, not actually limiting,
shows that in the limit bg = L2 / bideal, the effective slip as this geometry is not the one that favors large effective slip
length beff does saturate at the gas spherical cap radius of lengths see Sec. II C.
curvature. Coming back to macroscopic variables, this radius Let us consider a practical surface. Of course from the
of curvature can be traced back to the pressure difference different previous sections, minimizing the friction and
between the liquid and the gas phase through the Laplace maximizing the slip length would require the use of large
equation, P = Pg Pl = 2lg / L2, with lg the liquid/gas sur- periodicity L. However the latter is limited by stability con-
face tension. A pressure difference between the liquid and the siderations: the transition to the impregnated Wenzel state,
gas thus induces a saturation of the effective slip length at a where the liquid fills the roughness and therefore sweeps out

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-9 Achieving large slip with superhydrophobic surfaces Phys. Fluids 19, 123601 2007

the gas lubricating pockets, is indeed associated with the

disappearance of superlubricating properties.6,19 A reason-
able compromise may lie in the range L few m for which
pressure differences of tens of mbarstypical for gravity-
driven microfluidic flowscan be withstood by an air/water
interface. Note that for a precise value of the stability limit,
one should actually take into account all surface geometric
parameters, including, for instance, the structure height.25
Now for microfabricated or nanofabricated surface pat-
terns, the exact geometry of the surface is known a priori
and Eqs. 4, 8, 12, and 19 allow us to estimate the
expected frictional properties of this surfaces. However, for
most superhydrophobic surfaces reported, the structure ge-
ometry is not as controlled, and a key parameter such as the
solid fraction s has to be estimated or measured. It is there-
FIG. 8. Effective slip length beff as a function of the contact angle eff on the
fore very convenient to relate this parameter to a macro- superhydrophobic surface. The considered surface is made of pillars of
scopic observable such as the effective contact angle eff on height e = 5 m arranged on a square lattice of periodicity L = 5 m. The
the surface. With 0 the contact angle on the bare surface in pillars radius is the only remaining geometrical parameter that controls the
the range of 100120 for hydro- or fluorocarbon coatings, overall solid fraction s, the latter being related to eff according to the
Cassie relation 26. The viscosity contrast between the liquid and gas cor-
eff relates to s through the Cassie-Baxter relation,5 responds to the water/air interface: l / g = 55.5. Theoretical expressions are
taken according to Table I: theoretical prediction assuming ideal BCs
cos eff = scos 0 + 1 1. 26 on the composite interface no-slip and shear-free, -- theoretical predic-
tion assuming no slip on the solid and a finite dissipation within the gas
Fixing the typical height of the structures to e = L = 5 m and sublayer -- upper bond of a uniform gas layer s = 0. Inset: close-up of
assuming a pillars-like type of geometry, it is therefore pos- the near 180 contact angle region.
sible to predict the frictional properties beff of the surface as
a function of its macroscopic effective contact angle eff. For
pillars, Eq. 4 gives Finally, we note that the effect of a finite slip length bs
on the solid in the tens of nanometer range does decrease the

beff L cos 0 + 1
cos eff + 1
, 27
overall surface friction but still not in large proportion for
contact angles below 179. In the considered example, and
for eff = 179, it would change the overall slip length from
with = 0.325 and = 0.44 the numerical factors. This ex- 60 to 65 m not plotted in Fig. 8. Such an effect is more
pression shows that close to eff = 180, beff strongly diverges, effective as the effective contact angle eff is very close to
like 180.
In conclusion, we presented here a scaling law analysis
L
beff . 28 that allows us to calculate the frictional properties of super-
180 eff hydrophobic surfaces as a function of the generic geometri-
This is apparent in Fig. 8, where huge slippage in the hun- cal parameters characterizing the surface structure. These
dreds of micrometers range can be achieved for large yet laws were successfully tested against numerical calculations
achievable10 contact angles. for generic geometries and showed that very large slip
It should be noted that for such low solid fractions, the lengths can be obtained for an ultrahydrophobic surface,
dissipation within the gas sublayer, though small, may have a characterized by a contact angle very close to 180 eff
non negligible effect as it is predicted to decrease the slip 178 .
length from 110 solid line to 60 m dashed line for eff Such surfaces may be obtained by present techniques of
= 179. As a consequence, the effective slip length at the nano-engineering. For example, a surface made of pillars of
surface will depend on the liquid viscosity. Let us stress, 50 100 nm in diameter, 5 m apart, is now achievable
however, that this dependency is much weaker than what and corresponds to an effective contact angle eff larger than
would be predicted by the simple gas cushion model, for 179. For such surfaces, our analysis predicts slip lengths
which beff scales as l. For the previous example, eff that can reach hundred of micrometers. Beyond, the slip
= 179, doubling the liquid viscosity only modifies beff from length is ultimately limited by the dissipation in the gas
60 to 80 m. An effect of fluid viscosity on the slip length phase. We hope that our predictions, and the simple analyti-
has been reported in Ref. 17, however with a much larger cal laws we provided will motivate further experimental
magnitude than with the above prediction. work to develop superlubricating surfaces, with ultralow liq-
We note in addition that the sublayer dissipation has uid friction.
been calculated assuming no slip of the gas on the solid
ACKNOWLEDGMENTS
surfaces, while with a mean free path of order 0.1 m we
would anticipate a slip length for the gas onto the solid of the We acknowledge support from ANR PNANO. We thank
same order. We therefore somewhat overestimate the sub- Elisabeth Charlaix and Jean-Louis Barrat for many discus-
layer dissipation together with its influence on beff. sions on the subject.

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
123601-10 Ybert et al. Phys. Fluids 19, 123601 2007

1
J. Bico, U. Thiele, and D. Quere, Wetting of textured surfaces, Colloids surements. 2. Lubrication and the atomic force microscope, Langmuir
Surf., A 206, 41 2002. 19, 1227 2003.
2 21
A. Lafuma and D. Quere, Superhydrophobic states, Nature Mater. 2, C. Cottin-Bizonne, B. Cross, A. Steinberger, and E. Charlaix, Boundary
457 2003. slip on smooth hydrophobic surfaces: Intrinsic effects and possible arti-
3
L. C. Gao and T. J. McCarthy, A commercially available perfectly hy- facts, Phys. Rev. Lett. 94, 056102 2005.
drophobic material thetaa / thetar = 180 degrees/ 180 degrees, 22
L. Joly, C. Ybert, and L. Bocquet, Probing the nanohydrodynamics at
Langmuir 23, 9125 2007. liquid-solid interfaces using thermal motion, Phys. Rev. Lett. 96, 046101
4
R. N. Wenzel, Resistance of solid surfaces to wetting by water, Ind. 2006.
Eng. Chem. 28, 988 1936. 23
P. Huang and K. S. Breuer, Direct measurement of slip length in electro-
5
A. Cassie and S. Baxter, Wettability of porous surfaces, Trans. Faraday lyte solutions, Phys. Fluids 19, 028104 2007.
Soc. 40, 546 1944. 24
E. Lauga and H. A. Stone, Effective slip in pressure-driven Stokes flow,
6
C. Cottin-Bizonne, J. L. Barrat, L. Bocquet, and E. Charlaix, Low- J. Fluid Mech. 489, 55 2003.
friction flows of liquid at nanopatterned interfaces, Nature Mater. 2, 237 25
C. Cottin-Bizonne, C. Barentin, E. Charlaix, L. Bocquet, and J. L. Barrat,
2003. Dynamics of simple liquids at heterogeneous surfaces: Molecular-
7
D. Quere, A. Lafuma, and J. Bico, Slippy and sticky microtextured sol- dynamics simulations and hydrodynamic description, Eur. Phys. J. E 15,
ids, Nanotechnology 14, 1109 2003. 427 2004.
8
J. Philip, Flows satisfying mixed no-slip and no-shear conditions, Z. 26
M. Sbragaglia, R. Benzi, L. Biferale, S. Succi, and F. Toschi, Surface
Angew. Math. Phys. 23, 353 1972. roughness-hydrophobicity coupling in microchannel and nanochannel
9
J. Philip, Integral properties of flows satisfying mixed no-slip and no- flows, Phys. Rev. Lett. 97, 204503 2006.
shear conditions, Z. Angew. Math. Phys. 23, 960 1972. 27
S. Richardson, On the no-slip boundary-condition, J. Fluid Mech. 59,
10
J. W. Kim and C. J. Kim, Nanostructured surfaces for dramatic reduction
707 1973.
of flow resistance in droplet-based microfluidics, Fifteenth IEEE Interna- 28
E. Lauga, M. P. Brenner, and H. A. Stone, Handbook of Experimental
tional Conference on Micro Electro Mechanical Systems, Technical Digest
Fluid Dynamics Springer, New York, 2005.
IEEE, Piscataway, NJ, 2002, pp. 479482. 29
11 O. I. Vinogradova, Drainage of a thin liquid-film confined between hy-
C. Navier, Mmoire sur les Lois du Louvement des Fluides lAcadmie
drophobic surfaces, Langmuir 11, 2213 1995.
Royale des Sciences de lInstitut de France, Paris, 1823, Vol. VI. 30
12 P. de Gennes, On fluid wall slippage, Langmuir 18, 3413 2002.
D. Richard and D. Quere, Viscous drops rolling on a tilted non-wettable 31
C. Y. Wang, Flow over a surface with parallel grooves, Phys. Fluids 15,
solid, Europhys. Lett. 48, 286 1999.
13
J. Ou, B. Perot, and J. P. Rothstein, Laminar drag reduction in micro- 1114 2003.
32
channels using ultrahydrophobic surfaces, Phys. Fluids 16, 4635 2004. H. Brinkman, On the permeability of media consisting of closely packed
14
J. Ou and J. P. Rothstein, Direct velocity measurements of the flow past porous particles, Appl. Sci. Res., Sect. A 1, 81 1947.
33
drag-reducing ultrahydrophobic surfaces, Phys. Fluids 17, 103606 A. S. Sangani and A. Acrivos, Slow flow through a periodic array of
2005. spheres, Int. J. Multiphase Flow 8, 343 1982.
34
15
S. Gogte, P. Vorobieff, R. Truesdell, A. Mammoli, F. van Swol, P. Shah, G. S. Beavers and D. D. Joseph, Boundary conditions at naturally per-
and C. J. Brinker, Effective slip on textured superhydrophobic surfaces, meable wall, J. Fluid Mech. 30, 197 1967.
35
Phys. Fluids 17, 051701 2005. A. Steinberger, C. Cottin-Bizonne, P. Kleimann, and E. Charlaix, High
16
C. H. Choi, U. Ulmanella, J. Kim, C. M. Ho, and C. J. Kim, Effective slip friction on a bubble mattress, Nature Mater. 6, 665 2007.
36
and friction reduction in nanograted superhydrophobic microchannels, M. Sbragaglia and A. Prosperetti, A note on the effective slip properties
Phys. Fluids 18, 087105 2006. for microchannel flows with ultra-hydrophobic surfaces, Phys. Fluids 19,
17
C. H. Choi and C. J. Kim, Large slip of aqueous liquid flow over a 043603 2007.
37
nanoengineered superhydrophobic surface, Phys. Rev. Lett. 96, 066001 P. Panzer, M. Liu, and D. Einzel, The effects of boundary curvature on
2006. hydrodynamic fluid-flow: Calculation of slip lengths, Int. J. Mod. Phys. B
18
L. Bocquet, P. Tabeling, and S. Manneville, Comment on large slip of 6, 3251 1992.
38
aqueous liquid flow over a nanoengineered superhydrophobic surface, W. Rybczyski, On the translatory motion of a fluid sphere in a viscous
Phys. Rev. Lett. 97, 109601 2006. medium in german, Bull. Int. Acad. Pol. Sci. Lett., Cl. Sci. Math. Nat.,
19
P. Joseph, C. Cottin-Bizonne, J. M. Benoit, C. Ybert, C. Journet, P. Tabel- Ser. A 2, 40 1911.
39
ing, and L. Bocquet, Slippage of water past superhydrophobic carbon J. Hadamard, Mouvement permanent lent dune sphre liquide et vis-
nanotube forests in microchannels, Phys. Rev. Lett. 97, 156104 2006. queuse dans un liquide visqueux, C. R. Acad. Sci. Paris 152, 1735
20
O. I. Vinogradova and G. E. Yakubov, Dynamic effects on force mea- 1911.

Downloaded 10 Dec 2007 to 129.187.254.46. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp