Langmuir 2005, 21, 11053-11060 11053

Analysis of Droplet Evaporation on a Superhydrophobic

Surface
G. McHale,*,† S. Aqil,† N. J. Shirtcliffe,† M. I. Newton,† and H. Y. Erbil‡
School of Biomedical and Natural Sciences, Nottingham Trent University, Clifton Lane,
Nottingham NG11 8NS, U.K., and Gebze Institute of Technology, Faculty of Engineering,
Department of Chemical Engineering, Cayırova, Gebze 41400, Kocaeli, Turkey

Received July 11, 2005. In Final Form: September 10, 2005

The evaporation process for small, 1-2-mm-diameter droplets of water from patterned polymer surfaces
is followed and characterized. The surfaces consist of circular pillars (5-15 µm diameter) of SU-8 photoresist
arranged in square lattice patterns such that the center-to-center separation between pillars is 20-30 µm.
These types of surface provide superhydrophobic systems with theoretical initial Cassie-Baxter contact
Downloaded via INDIAN INST OF TECHNOLOGY KHARAGPUR on December 5, 2018 at 13:54:12 (UTC).

angles for water droplets of up to 140-167°, which are significantly larger than can be achieved by smooth
hydrophobic surfaces. Experiments show that on these SU-8 textured surfaces water droplets initially
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

evaporate in a pinned contact line mode, before the contact line recedes in a stepwise fashion jumping from
pillar to pillar. Provided the droplets of water are deposited without too much pressure from the needle,
the initial state appears to correspond to a Cassie-Baxter one with the droplet sitting upon the tops of
the pillars. In some cases, but not all, a collapse of the droplet into the pillar structure occurs abruptly.
For these collapsed droplets, further evaporation occurs with a completely pinned contact area consistent
with a Wenzel-type state. It is shown that a simple quantitative analysis based on the diffusion of water
vapor into the surrounding atmosphere can be performed, and estimates of the product of the diffusion
coefficient and the concentration difference (saturation minus ambient) are obtained.

I. Introduction Picknett and Bexon1 investigated the mass and profile

Free evaporation of completely spherical droplets of
water is an apparently simple problem with widespread
relevance to drying problems such as ink-jet printing,
spraying of pesticides, and the spotting of DNA microarray
data. These types of applications involve small droplets,
dominated by surface tension rather than gravity, de-
posited on solid substrates. The involvement of the
substrate complicates the evaporation process in obvious
ways, such as altering the droplet shape to a spherical
cap with a contact angle determined by various interfacial
surface tensions. Less obviously, when the droplet evapo-
ration is governed by the diffusion of the liquid into the
surrounding atmosphere, the presence of a substrate
restricts the space into which vapor may diffuse and so
reduces the evaporation rate. Thus, it has been theoreti-
cally predicted that a small droplet resting on a flat solid
surface but possessing a 180° contact angle so that it is
completely spherical will have a rate of mass loss reduced
by loge 2 compared to an identical, completely spherical
droplet far removed from any solid surface.1 Investigating
the free evaporation of almost spherical droplets from solid
substrates has not previously been attempted because of
the obvious difficulty of obtaining a model physical system
with an initial contact angle higher than that of the water-
Teflon system of 110-120°. As has become clear over
recent years from work on superhydrophobic surfaces,
nature has not refrained from creating leaves on which
contact angles approaching 180° are observed.2-4
* Corresponding author. E-mail: glen.mchale@ntu.ac.uk. Tel:
+44 (0)115 8483383.
† Nottingham Trent University.
‡ Gebze Institute of Technology.
(1) Picknett, R. G.; Bexon, R. J. Colloid Interface Sci. 1977, 61, 336-
350.
(2) Neinhuis, C.; Barthlott, W. Ann. Bot. 1997, 79, 667-677.
(3) Blossey, R. Nat. Mater. 2003, 2, 301-306.
(4) Quéré, D.; Lafuma, A.; Bico, J. Nanotechnology 2003, 14, 1109-
1112.
evolution of a slowly evaporating liquid (methyl acetoac-
etate) droplet on a poly(tetrafluoroethylene) (Teflon)
surface in still air. They reported three modes of evapora-
tion: mode 1, corresponding to a constant solid-liquid
contact area; mode 2, corresponding to a constant contact
angle, and mode 3, corresponding to changes in both the
contact angle and contact area. They also developed a
diffusion-based theory to predict the evaporation rate.
Since then, various authors have considered the evapora-
tion of small droplets of liquid in constant contact area
mode,5-9 constant contact angle mode,8-12 and mixed
mode.8,9 In each case, models were developed to enable
data on the product D∆c, where D is the diffusion constant
and ∆c ) cS - c∞ is the difference in concentration of vapor
at the droplet surface (assumed to equal the saturated
vapor concentration) and that far removed from the droplet
surface, to be estimated. We emphasize that these
estimates were not intended to be accurate measurements
but were used to demonstrate the consistency of theory
with experiment. In a sequence of papers, Erbil and co-
workers investigated how deviations from a spherical cap
droplet shape might influence the evaporation rate13,14
and how the rate of evaporation might be used to determine
(5) Birdi, K. S.; Vu, D. T.; Winter, A. J. Phys. Chem. 1989, 93, 3702-
3703.
(6) Rowan, S. M.; Newton, M. I.; McHale, G. J. Phys. Chem. 1995,
99, 13268-13271.
(7) Rowan, S. M.; McHale, G.; Newton, M. I.; Toorneman, M. J. Phys.
Chem. 1997, 101, 1265-1267.
(8) Shanahan, M. E. R.; Bourgès, C. Int. J. Adhes. Adhes. 1994, 14,
201-205.
(9) Bourgès-Monnier, C.; Shanahan, M. E. R. Langmuir 1995, 11,
2820-2829.
(10) Birdi, K. S.; Vu, D. T. J. Adhes. Sci. Technol. 1993, 7, 485-493.
(11) McHale, G.; Rowan, S. M.; Newton, M. I.; Banerjee, M. K. J.
Phys. Chem. B 1998, 102, 1964-1967.
(12) Erbil, H. Y.; McHale, G.; Newton, M. I. Langmuir 2002, 18,
2636-2641.
(13) Erbil, H. Y.; Meric, R. A. J. Phys. Chem. B 1997, 101, 6867-
6873.
(14) Meric, R. A.; Erbil, H. Y. Langmuir 1998, 14, 1915-1920.

10.1021/la0518795 CCC: $30.25 © 2005 American Chemical Society

Published on Web 10/14/2005
11054 Langmuir, Vol. 21, No. 24, 2005
the initial contact angle.15,16 However, in all of these studies
the contact angles involved have been lower than 110°
and so have not tested the higher contact angle regime.
In their work on the influence of evaporation on contact
angle measurements, Bourgès and Shanahan9 noted that
various physical factors, particularly surface roughness,
had still not been incorporated into the understanding of
evaporation. At that time, relatively little had been
published on superhydrophobic surfaces. From the work
of numerous authors, we now know that rough or textured
hydrophobic surfaces providing contact angles of 150-
180° are common in nature and can be produced artificially
in a wide variety of ways. (See the review by Blossey.3)
Moreover, these surfaces can be divided into two distinctive
types, Wenzel17 and Cassie-Baxter,18,19 depending on
whether the liquid fully penetrates the surface features
or whether the liquid bridges across the surface features.
Thus, the development of techniques for producing high
contact angle surfaces and our understanding of the
equilibrium and metastable states of droplets on these
surfaces provide new regimes for the study of droplet
evaporation. It is therefore the purpose of the present
study to report qualitative features of the evaporation of
sessile droplets of water from solid surfaces with initial
contact angles for water greater than 120° and to
quantitatively relate the rate of mass loss to diffusion
models for the evaporation of sessile droplets.
II. Theoretical Development
II.a. Models of Evaporation from Solid Surfaces.
The rate of change of mass of an evaporating droplet can
be modeled on the basis of the diffusion of water molecules
from the droplet surface into the surrounding atmosphere.
For a completely spherical droplet of liquid of density FL,
spherical radius RS, and volume VC, the rate of mass loss
depends on the liquid-vapor surface area, ALV ) 4πRS2,
the diffusion coefficient, D, and the difference, ∆c ) (cS -
c∞), between the vapor concentration at the droplet surface
(assumed to be equal to the saturation concentration cS)
and the ambient value far removed from the droplet
surface (c∞). Assuming a spherical droplet far removed
from any boundaries and solving the diffusion equation,
the concentration of vapor is found to be inversely
proportional to the radial distance from the droplet. The
rate of mass loss is then given by
dVC
FL ) -4πRSD∆c (1)
dt
When the completely spherical droplet is just in contact
(i.e., contact angle of 180°) with a plane boundary, it is
predicted that the evaporation rate will decrease because
of the reduction in space into which vapor can diffuse.1
The reduction can be calculated theoretically and is given
by eq 1, but with an extra factor of loge 2 on the right-hand
side of the equation. By using the analogy between the
diffusive flux and the electrostatic potential, Picknett and
Bexon1 obtained an exact equation for the rate of mass
loss that is valid for all droplets of a spherical cap shape
(15) Erbil, H. Y. J. Phys. Chem. B 1998, 102, 9234-9238.
(16) Erbil, H. Y. J. Adhes. Sci. Technol. 1999, 13, 1405-1413.
(17) Wenzel, R. N. Ind. Eng. Chem. 1936, 28, 988-994.
(18) Cassie, A. B. D.; Baxter, S. Trans. Faraday Soc. 1944, 40, 546-
551.
(19) Johnson, R. E.; Dettre, R. H. Contact Angle, Wettability and
Adhesion; Advances in Chemistry Series 43; American Chemical
Society: Washington, DC, 1964; p 112.
McHale et al.
resting on a solid plane boundary
dVC
FL ) -4πRSD∆cf(θ) (2)
dt
where f(θ) is a function of the contact angle θ. Picknett
and Bexon’s solution involves a complex function (Snow’s
solution), but to aid calculations, they gave two useful
polynomial fits covering the angular ranges of 0-10° and
10-180°
{
2fPB(θ) )
0.6366θ + 0.09591θ2 - 0.06144θ3
0° < θ < 10°
0.00008957 + 0.6333θ + 0.116θ2 - 10° < θ < 180°
0.08878θ3 + 0.01033θ4
(3)
where the contact angle θ is in radians. For a completely
spherical droplet with a vanishingly small contact area
with respect to the solid, this approximation gives 2fPB(π)
) 1.387 compared to an exact solution of 2 loge 2 ) 1.386.
For a hemispherical droplet having a contact angle of 90°
with respect to the solid substrate, the Picknett and Bexon
value of 2fPB(π/2) ) 0.999 compares well to the exact
solution of eq 1.
For comparison to previous studies of the evaporation
of sessile droplets on solid surfaces, it is useful to review
some approximate models and relate them to eq 2. (See
also ref 12.) Equation 2 recovers the evaporation of a
completely spherical droplet neglecting any boundary, eq
1, by taking fBV(θ) ) 1, where fBV(θ) is the function from
the work of Birdi and Vu5,10 used in their study of the
evaporation of sessile water droplets from glass surfaces.
The limitation of Birdi and Vu’s model for contact angles
close to 90° was addressed in the work of Rowan et al.,6
who also studied the evaporation of water droplets from
solid (PMMA and Teflon) substrates by taking into account
the reduced liquid-vapor interfacial area of the spherical
cap droplet for diffusion through a factor fR(θ) ) (1 - cos
θ)/2. Although this formula provides the correct numerical
result at 90°, it overestimates the evaporation rate at 180°
through its neglect of the effect of the solid substrate on
reducing the space for the diffusion of water vapor.
Bourgès-Monnier and Shanahan addressed this neglect
in their self-consistent model of the concentration gradient
occurring during evaporation.9 Assuming a radial diffusion
of vapor into the surrounding atmosphere and a spherical
cap droplet supported by a plane solid substrate, they
derived
-cos θ
fBMS(θ) ) (4)
2 loge(1 - cos θ)
The Bourgès-Monnier and Shanahan result is 0.5 at 90°,
in agreement with the exact solution, and gives 2fBMS(π)
) 1/loge 2 ) 1.4427 compared to 2 loge 2 ) 1.386 at 180°.
II.b. Constant Contact Radius Mode. In the present
study, our patterned substrates enable contact angles
greater than the range of 30-100° studied by Bourgès-
Monnier and Shanahan to be obtained while possessing
the simplifications of a spherical cap droplet and a pinned
contact area; this corresponds to the stage II evaporation
that they identified. Our analysis broadly follows that of
Bourgès-Monnier and Shanahan9 but uses the more
accurate Picknett and Bexon1 formula, eq 3, rather than
eq 4. This allows us to give an analytical form for a function
of the contact angle that is linear with time rather than
Droplet Evaporation on a Superhydrophobic Surface Langmuir, Vol. 21, No. 24, 2005 11055

providing a form requiring a numerical integration. To do expression involving polylogarithm functions. When this
so, consider a spherical cap droplet with a volume result is expanded about u ) -1, an expression similar
to eq 9 is obtained, but with only the first two terms in
πrb3(1 - cos θ)2(2 + cos θ) HPB(θ) and with the coefficients eo and e1 replaced by the
VC(rb, θ) ) (5) numerical values loge 2 ) 0.693 and (loge 2 - 1/2) )
3 sin3 θ 0.193147; this approximate expression is an accurate
approximation to HPB(θ) to around 10.5% over the range
where rb is the contact (or base) radius of the droplet. of 90-180°. If the solid boundary is completely ignored
Assuming the contact radius is constant, the rate of change for the diffusion and the Birdi and Vu approximation fBV-
of the volume can be written in terms of the rate of change (θ) ) 1 is used in eq 7, then the equivalent result to eq 9
of the cosine of the contact angle is
dVC -πrb3 du -1 -2D∆ct
dt
)
(1 - u2)1/2(1 + u)2 dt
( ) (6) HBV(θ) ≡
2(1 + u)
)
FLrb2
+ HBV(θo) (11)

where u ) cos θ. Because rb ) RS sin θ and sin θ ) (1 -
u2)1/2, substituting eq 6 into eq 2 gives
1 du 2D∆c
( )
2f(θ)(1 + u)2 dt
)
FLrb2
(7)
which cannot be integrated exactly in its present form.
However, terms of the form un/(1 + u)2 are integrable, so
we can develop a polynomial fit to 1/2fPB(θ), accurate to
0.04% at all angles, for our range of interest of 90-180°,
1 3
2fPB(θ)
≈ ∑
n)0
dnun ) 0.999766 + 0.481517u +
0.292040u2 + 0.089118u3 (8)
Substituting eq 8 into eq 7 and integrating gives
-e0
HPB(θ) ≡ + e1 loge(1 + u) + e2u + e3u2 )
1+u
-2D∆ct
+ HPB(θo) (9)
FLrb2
where u(t) ) cos θ(t) and the constants en are given by
eo ) d0 - d1 + d2 - d3 ) 0.721171
e1 ) d1 - 2d2 + 3d3 ) 0.164791
e2 ) d2 - 2d3 ) 0.113804
d3
e3 ) ) 0.044559 (10)
2
and are valid over the contact angle range of 90-180°.
The constant HPB(θo) is a constant of integration and
represents the function evaluated at the contact angle for
the initial time t ) 0. Thus, the evaporation of spherical
cap droplets with constant contact radii and contact angles
greater than 90° can be analyzed by plotting the function
HPB(θ) against time, t. On such a graph, the data for any
given droplet should lie on a straight line, and the slope,
multiplied by -FLrb2/2, will provide an estimate of the
diffusion constant-vapor concentration difference product
(i.e., D∆c). Although the specific results in eqs 8-10 are
based on a polynomial expansion valid in the range of
90-180°, a similar approach can be adopted for other
contact angle ranges.
For completeness in comparing different approaches to
analyzing data on evaporation with pinned contact lines,
we have evaluated the effect of different approximations
to f(θ) in eq 7. Using the Bourgès-Monnier and Shanahan
function fBMS(θ) in eq 7 and integrating gives a complex
Similarly, using the Rowan et al. approximation fR(θ) )
(1 - cos θ)/2 gives
HR(θ) ≡
-1
2(1 + u)
1
- loge
4
u-1
u +(1
)
-2D∆ct
)
FLrb2
+ HR(θo)
(12)
The exact result for a completely spherical droplet just in
contact with a solid surface is similar to eq 11 modified
by an extra factor of 1/loge 2 in the H(θ) function. The
numerical factor multiplying the term -1/(1 + u) is then
1
/2 loge 2 ) 0.7213 compared to eo ) 0.721171.
II.c. Surface Texture and Initial Contact Angle.
On a flat, smooth surface, the highest contact angles for
droplets of water that are achievable are determined by
the surface chemistry and are around θes ) 110-120° for
poly(tetrafluoroethylene) (PTFE). To achieve higher con-
tact angles, superhydrophobic surfaces can be created by
roughening or texturing the surfaces.3 For a textured
surface, the liquid may either bridge across surface
protrusions (Cassie-Baxter case18,19) or fully penetrate
into the surface features (Wenzel case17). In the first case,
the droplet sits on a composite solid-air interface, and
the contact angle is given by the Cassie-Baxter equation
cos θeCB ) φs cos θes - (1 - φs) (13)
where φs is defined as the solid fraction upon which the
droplet rests. In the second case, the droplet is in contact
with the whole solid surface, and the contact angle is given
by the Wenzel equation
cos θeW ) r cos θes (14)
where r is the surface roughness defined as the ratio of
the true surface area to the horizontal projection of the
surface area. In these approaches to modeling the contact
angles for the macroscopic droplet, the effect of the
substrate heterogeneities are taken into account in an
overall manner. Clearly, at a more local level, each element
of the triple line must lie on either the vapor or the solid
component beneath the droplet.
In principle, the lower of the two contact angles predicted
by eqs 13 and 14 is the equilibrium value, but in practice,
edges in surface features can lead to a metastable Cassie-
Baxter state even when the Wenzel state is energetically
preferred; pressure is then needed to cause a conversion
from the Cassie-Baxter to the Wenzel state.4 In addition,
the Cassie-Baxter state reduces the true contact area
between the liquid droplet and the solid and so tends to
reduce any intrinsic contact angle hysteresis arising from
the substrate material and leads to a “slippy” state. In
contrast, the Wenzel state increases the true contact area
11056 Langmuir, Vol. 21, No. 24, 2005
Figure 1. (a) Schematic of the surface pattern, (b) scanning
electron micrograph of a textured SU-8 surface (dp ) 11.5 µm,
Lp ) 18.5 µm, hp ) 25 µm), (c) side profile of a droplet of water
on a flat SU-8 surface, and (d) side profile of a droplet of water
on a textured SU-8 surface.
between the liquid droplet and the solid and so tends to
increase contact angle hysteresis leading to a “sticky”
state.4,21
The patterns for the present study consist of circular
cross-sectional pillars of diameter dp and height hp
arranged in a square lattice of period Lp in each of the two
in-plane directions (Figure 1a). The solid fraction φs
required for the Cassie-Baxter equation is then
πdp2
φs ) (15)
4Lp2
and the roughness factor for the Wenzel equation is
πdphp
r)1+ (16)
Lp2
When θes < 90°, the Wenzel formula always predicts a
decrease in contact angle whereas the Cassie-Baxter
equation can still predict an increase in contact angle.
In the specific case when the separation between pillars
is equal to their diameter (i.e., Lp ) 2dp), eq 15 gives φs
) π/16 ) 0.1963 and r ) 1 + πhp/4dp. Thus, using eq 13 and
θes ) 80°, typical of an SU-8 polymer surface, would give
a Cassie-Baxter angle of 140°. The equivalent Wenzel
contact angle is dependent on the aspect ratio of pillar
height to diameter and for a ratio of unity (i.e., hp/dp ) 1)
is 72°. Increasing the aspect ratio to 2 gives a Wenzel
angle of 63°.
III. Experimental Development
SU-8 is a negative photoresist that can be deposited in a thick
layer and patterned using photolithography; the vertical sides
of the features in those patterns can be smooth. SU-8 is epoxy-
based and becomes strong, stiff, and chemically resistant after
processing with the typical contact angle to water of θes ≈ 80°
on a smooth, flat surface and large contact angle hysteresis. The
photolithographic technique for producing these substrates has
(20) Shirtcliffe, N. J.; Aqil, S.; Evans, C.; McHale, G.; Newton, M. I.;
Perry, C. C.; Roach, P. J. Micromech. Microeng. 2004, 14, 1384-1389.
(21) McHale, G.; Shirtcliffe, N. J.; Newton, M. I. Langmuir 2004, 20,
10146-10149.
McHale et al.
been described in detail by Shirtcliffe et al.20 In this paper, it was
shown experimentally that droplets of water on SU-8 textured
surfaces with a pillar separation twice that of the pillar diameter
and pillar diameters in the range 5-40 µm arranged in a square
lattice obey the Cassie-Baxter equation, provided the pillar
height is sufficiently high. Importantly, the observed contact
angle hysteresis was around 50°, thus providing a system with
both a superhydrophobic Cassie-Baxter-type contact angle and
conditions for a pinned contact line during free evaporation; the
contact angle hysteresis is reduced by the Cassie-Baxter effect
but not sufficiently to give a truly slippy state. The pinning of
the contact line corresponds to our analysis in section II.b, which
is equivalent to the Picknett and Bexon mode 1 evaporation mode
and so simplifies the quantitative analysis of the evaporation.
We constructed surfaces with pillar diameters in the range of
5-15 µm arranged in a square lattice and with spacings between
the centers of pillars in the range of 20-30 µm in each in-plane
direction; the heights of the pillars ranged from 12 to 35 µm
(Figure 1a and b). The surfaces used had a Cassie-Baxter solid
fraction area parameter φs of between 0.064 and 0.135 predicting
Cassie-Baxter angles of between 157 and 147°, respectively.
Ideally, the pillars have vertical sides, but in practice, on some
of the samples with the highest pillars a slight outward taper
occurred, leading to a marginally larger diameter at the top of
the pillar as evident in Figure 1b. This effect resulted in a larger
solid fraction area parameter φs than in the design and hence
a lower Cassie-Baxter angle, but it did not prevent a super-
hydrophobic surface from being created. All experiments were
carried out at room temperature (23-26 °C) and relative humidity
(45-58%) in an enclosure to shield from drafts.
Contact angles, contact base diameters, heights, liquid-vapor
surface areas, and volumes were obtained from side profile images
of a droplet and its reflection in the substrate using a Krüss
DSA10 contact angle meter (video profilometry with drop shape
analysis software). The drop shape analysis provides volume
and surface area data using the measured profile and by assuming
axial symmetry; it does not assume any particular droplet shape.
In our case, top view observation suggested that the droplets we
studied did possess axial symmetry during the evaporation
process. It is conceivable that such symmetry could be lost for
droplets of sizes more comparable to the period of the substrate
pattern or in the later stages of evaporation when droplets became
smaller and entered the substrate pattern; we did not pursue
these aspects of small droplet evaporation or late stage evapora-
tion.
There are several subtleties in the measurement of the contact
diameter in the superhydrophobic droplet on a patterned
substrate system that need to be understood when analyzing
and interpreting the data and forming conclusions. The first is
that the accuracy of contact diameter measurements for a droplet
with a high contact angle of say 160° is more limited than one
for which the contact angle is low, say 60°, if the optical
magnification is chosen to keep the full droplet in the image. In
the latter case, the initial contact diameter can be the full field
of view and so provide optimal accuracy in the measurement of
the contact diameter from the digital image. In the former case,
the initial contact diameter is necessarily only a fraction of the
field of view (rapidly becoming less so as the droplet becomes
more superhydrophobic), and the accuracy of contact diameter
measurements from the digital images is therefore reduced. The
second subtlety is that accurate measurements of contact
diameter depend on accurate identification of the vertical position
of the solid-liquid interface (i.e., substrate surface) in the image,
and as the droplet becomes superhydrophobic, this becomes a
more significant problem. For contact angles approaching 180°,
small changes in the vertical position of the identified baseline
can provide large changes in the apparent contact diameter.
Similarly, if the baseline is misidentified by, say, one vertical
pixel, then an evaporating superhydrophobic droplet with a
decreasing contact angle will give an apparent (rather than real)
decrease in contact diameter. The baseline issue has a particularly
strong effect on the ability to make accurate measurements of
contact diameters once the droplet enters the substrate pattern.
In our case, the substrates have periods of 20-30 µm, and the
heights of the pillars range from 12 to 35 µm. This means the
placements of baselines are more difficult, and we strongly
Droplet Evaporation on a Superhydrophobic Surface
Figure 2. Evolution of the contact angle with time ()) and
height (solid curve) and contact diameter (+) during the
evaporation of a droplet of water where no obvious collapse
occurs during the initial evaporation period.
Figure 3. Evolution with time of the contact angle ()) and
height (solid curve) and contact diameter (+) during the
evaporation of a droplet of water where an obvious collapse
occurs during the initial evaporation period.
emphasize that small changes in measurements of contact
diameters, particularly of less than a substrate lattice period,
and late stage evaporation measurements once the droplet has
penetrated the substrate pattern should be interpreted cautiously
and the data should not be overinterpreted.
Figure 1c shows a typical side profile image of a droplet of
water on a flat SU-8 substrate (θs ≈ 80°), and Figure 1d shows
an equivalent image on a textured substrate (dp ) 11.5 µm, Lp
) 18.5 µm, hp ) 25 µm, θs ≈ 146°). Provided the droplet was
deposited without too much pressing to detach it, the droplet
tended to deposit in a Cassie-Baxter-type state bridging the
gaps between pillars; the Cassie-Baxter state is visually
apparent from the pattern of light penetrating between the pillars
below the droplet.
IV. Results and Discussion
Figures 2 and 3 show the contact angle, base diameter,
and droplet height during a typical evaporation sequence
for droplets on a surface with a pillar center-to-center
spacing of 30 µm and pillar diameter and height of 8 and
15 µm, respectively. In Figure 2, the initial contact angle
is greater than 150°, and after the brief initial period (100
s in this specific case) it steadily decreases (up to t ) 600
s in this specific case), during which time the contact
diameter remains approximately constant. Between 100
and 500 s, the change with respect to contact diameter is
within one period of the substrate pattern. The contact
radius then clearly depins as evidenced by the beginning
of a stepwise retreat shown in the contact diameter data.
During this period, the contact angle shows a net decrease,
but with increases and decreases accompanying each step
change in contact diameter. For this particular example,
the droplet becomes too small for accurate measurements
after 1100 s. Figure 3 shows a second type of sequence
that is qualitatively different from Figure 2 in that a very
obvious discontinuity occurs in the data during the period
corresponding to the pinned contact diameter evaporation
(in this particular example just after 800 s). Such a
Langmuir, Vol. 21, No. 24, 2005 11057
discontinuity has been observed during the detachment
of droplets from the syringe shortly after deposition
(several seconds) and a long time (minutes) after deposition
as shown in Figure 3. Prior to the discontinuity, the
evaporation proceeds in a manner similar to that shown
in Figure 2, but afterward the contact angle continues to
decrease smoothly and no obvious stepwise changes occur
in the contact diameter. The discontinuity appears rapidly.
In these data, the time between recorded frames was 2 s
for the first 2 min and subsequently 10 s, but visually the
collapse occurred in less than 1 s. As in Figure 2, the
droplet becomes too small for accurate measurements of
contact diameter after around 1000 s. All superhydro-
phobic droplet evaporation sequences observed on these
SU-8 pillar surfaces corresponded to one of the behaviors
in Figure 2 or 3.
Figure 4 shows a sequence of four images (panels a-d)
illustrating the behavior in Figure 2 and four images
(panels e-h) illustrating the behavior in Figure 3. For
both cases, the pattern of light visible beneath the droplets
indicate the initial droplets sit upon the pillars and bridge
the gaps between the pillars, so that the droplets are in
the Cassie-Baxter state. As the evaporation proceeds,
the volumes decrease with either no change or a small
change in contact area (Figure 4b and f). In the first case
(Figure 4c and e), the contact diameter eventually
decreases in a stepwise manner; this is visually obvious
when viewing the video of the evaporation. We anticipate
that the period of the steps should correlate with the period
of the lattice of pillars consistent with work reported on
the spreading of silicone oils on this type of surface.22 In
the second case (Figure 4g and h), a discontinuity occurs
in the evaporation sequence; this is a rapid process as
illustrated by Figure 4g and h, which are successive images
in the recording. A close examination of the image before
the discontinuity (Figure 4g) appears to show a shadow
beneath the droplet at the center of the contact area. The
pattern of light below the droplet in the image immediately
after the discontinuity (Figure 4h) appears to show that
the droplet is no longer suspended upon the pillars but
has penetrated into the substrate pattern. The contact
angle immediately after the discontinuity remains higher
than observed on the flat surface. We therefore conclude
that the discontinuity represents a collapse of the droplet
from the Cassie-Baxter state to a Wenzel-type state.
Visual examination of the video sequences shows that
subsequent evaporation appears to have a pinned nature
(for around 2.5 min in the case of Figure 3) until the droplet
becomes small and accurate measurement difficult. Care
is needed in interpreting the contact diameter data in
Figure 3 in this later stage of evaporation because a 30
µm change in contact diameter corresponds to the period
of the pattern and with the lower contact angle and volume
there is more sensitivity in the measurement accuracy of
the diameter, which depends on the placement of the
baseline in the video profile. Qualitatively, we believe that
a droplet collapse also probably always occurs in droplets
showing the evaporation sequence in Figure 2 but that it
occurs for a such a small volume that accurate measure-
ments to confirm it are difficult.
Figure 5 shows the stepwise retreat in the later stages
of evaporation for a larger initial volume droplet. The
surface had a pillar center-to-center spacing of 30 µm and
a pillar diameter and height of 8 and 14 µm, respectively,
which resulted in an initial contact angle of 145.6°. The
time period from 2500 to 3100 s clearly shows eight
(22) McHale, G.; Shirtcliffe, N. J.; Aqil, S.; Perry, C. C.; Newton, M.
I. Phys. Rev. Lett. 2004, 93, article 036102.
11058 Langmuir, Vol. 21, No. 24, 2005
Figure 4. Images of droplet evaporation. Panels a-d show a droplet where the collapse from the Cassie-Baxter state occurs late
in the evaporation. Panels e-h show a droplet where the collapse from the Cassie-Baxter to the Wenzel state occurs at a middle
stage during the evaporation.
Figure 5. Evolution with time of the contact angle ()) and
contact diameter (+) during the stepwise retreat of a droplet
evaporation sequence.
stepwise decreases reducing the base diameter from 857
to 600 µm, and this gives an average step of 32 µm
compared to the lattice center-to-center spacing of 30 µm.
In fact, the period from 1950 to 3060 s is consistent with
16 steps each of 30 µm. Each stepwise retreat event is also
accompanied by a temporary increase in the contact angle.
Because the stepwise retreat event occurs at a given
volume, we would expect a temporary increase in the
contact angle as observed. Quantitatively, this idea can
be formalized by requiring the volume given by the
spherical cap approximation in eq 5 to remain constant
when rb changes by an amount ∆rb and the contact angle
changes by an amount ∆θ (in radians),
( )
∆rb
∆θ ) -sin θ(2 + cos θ) (17)
rb
For a 3.5% decrease in contact radius occurring at a contact
angle of 131°, eq 17 predicts an increase in contact angle
of around 2°, which is of the same order of magnitude as
the observed changes in the data. Such an increase would
reduce the horizontal (inward) component of the liquid-
vapor surface tension force (the reduction in force can be
estimated from eq 17), thus allowing the droplet to again
become temporarily pinned. Further volume reduction by
evaporation would reduce the contact angle and so increase
the horizontal (inward) component of the liquid-vapor
surface tension force until sufficient force occurs to cause
another stepwise retreat of the contact radius in a stick-
slip manner.
It is also possible to analyze the change in surface free
energy for a stepwise retreat event at constant volume,
McHale et al.
Figure 6. Contact angle data ()) reproduced with the measured
liquid-vapor surface area (divided by 3; ‚‚‚, in mm2), volume
(×, in mm3), and height (+, in mm) for droplet data shown in
Figure 3. The solid curves are the estimates calculated for a
spherical cap using the measured contact angle and base
diameter.
and Shanahan has previously considered such an analysis
for stick-slip on a flat surface.23 He derived a minimum
step size necessary to overcome an intrinsic energy barrier
causing contact angle hysteresis on a flat surface. Applying
this interpretation to our substrates, an intrinsic energy
barrier would exist for the substrate material, but the
pattern coupled with the Cassie-Baxter effect of a reduced
solid-liquid interfacial area would then reduce this energy
barrier and impose a stepwise retreat distance.21
Figure 6 reproduces the contact angle data from Figure
3 but now includes the measured data for the liquid-
vapor surface area (‚‚‚ and in mm2), the volume (× and
in mm3), and the droplet height (+ and in mm; the liquid-
vapor surface area has been divided by a factor of 3 so
that it fits onto the same axis as the other two parameters).
The solid lines are the spherical cap values for these
parameters calculated from the measured contact angle
and base diameter and are clearly slight overestimates.
Initially, the spherical cap calculation overestimates the
surface area by 2-3%, but this decreases as time
progresses toward the droplet collapse at 850 s. Similar
overestimates occur in data for drops not showing the
collapse (e.g., Figure 2), and in these cases the overestimate
decreases as the stepwise retreat regime is approached.
From the full data set of observations, we conclude that
the measured liquid-vapor surface area, ALV, is well
approximated by a linear change with time up to the point
(23) Shanahan, M. E. R. Langmuir 1995, 11, 10411043.
Droplet Evaporation on a Superhydrophobic Surface
Figure 7. Tests of eqs 20 and 21 for the data in Figure 2
(lowest pair of curves), Figure 3 (middle pair of curves), and
Figure 5 (upper pair of curves); in each case, eq 21 is the lower
one of the pair at t ) 0. The solid lines are fits over the range
to when the droplet collapses or a stepwise retreat begins.
of either droplet collapse or stepwise retreat of the base
diameter. Using the spherical cap approximation, we can
understand why this may be the case. In this approxima-
tion, the surface area is
2πrb2
ALV ) (18)
1 + cos θ
and assuming a constant contact radius, this can be used
in eq 7 to obtain
( )
dALV -8πD∆cf(θ)
) (19)
dt FL
During the initial evaporation phase when the contact
angle is slowly varying, f(θ) can be approximated to a
constant. For example, in Figure 6 in the period before
droplet collapse the contact angle changes from 149.7 to
138.0°, and the corresponding values of fPB(θ) are 0.6723
and 0.6528 so that approximating f(θ) by 0.6626 gives a
value accurate to (2% over this experimental angular
range and eq 19 is expected to be a constant as observed
in the experiments.
To examine the theory in section II.b describing the
constant contact radius mode with diffusion-controlled
evaporation, we analyzed the data for the validity of eq
9 and consistency with eq 19. To compare these two
equations with data in the same Figure, we use the
rearrangement of eq 9 into
FLrb2HPB(θ)
) -D∆ct + k2 (20)
2 the droplet and hence reduces evaporation. Another
and eq 19 into
FLALV
) -D∆ct + k1 (21)
8πfav
where k1 and k2 are constants and it has been assumed
that rb and f(θ) ≈ fav are approximately constant over the
time period considered during a droplet evaporation; eq
20 is valid only for contact angles of 90° or greater. Figure
7 shows eqs 20 and 21 calculated from the data for rb, θ,
and ALV presented in Figures 2, 3, and 5; the average of
the initial and final values has been used for fav. The solid
curves are straight line fits over the range before either
the droplet collapsed or a stepwise retreat of the contact
line began. The fits show excellent agreement with the
expected linearity having R2 parameters of better than
0.999, although eq 20 gives a slightly higher estimate
Langmuir, Vol. 21, No. 24, 2005 11059
than eq 21 for the diffusion constant-vapor concentration
difference product (i.e., D∆c).
To verify the quantitative agreement of the diffusion
constant-vapor concentration difference product (i.e.,
D∆c), a series of experiments on two surfaces with pillar
diameters of 10 µm and center-to-center separations of 20
and 30 µm were analyzed. The data range used in the
analysis was such that rb is constant to within 4% of its
initial value and the contact angle is greater than 90°; we
have also taken fav as the average of the initial and final
values. The fits of eqs 20 and 21 to the data were straight
lines with regression coefficient parameters of R2 ) 0.9999
or better. Data for four droplets, including temperature
and relative humidity, on each of these two surfaces are
provided in Table 1. This Table also provides a comparison
to values of the diffusion coefficient estimated by correcting
the CRC Handbook24 value of 2.39 × 10-5 m2 s-1 at 8 °C
to the measured temperatures using a T3/2 temperature
dependence to provide reference values. The values of
diffusion constants using both eqs 20 and 21 are to within
8% of these reference values and within the uncertainty
arising from our estimates of concentration difference ∆c
from the temperature and relative humidity. These
quantitative estimates of diffusion coefficients are im-
pressive and provide confidence in the theoretical inter-
pretation of the evaporation process from superhydro-
phobic surfaces.
The agreement in estimates of the diffusion coefficient
using droplets that start with almost completely spherical
shapes (i.e., contact angle approaching 180°) but with a
plane boundary in close proximity supports the prediction
that the evaporation rate will decrease compared to that
of a spherical droplet far removed from any surfaces. The
physical mechanism is the reduction in space into which
vapor can diffuse.1 There are other possible mechanisms
related to droplet evaporation from superhydrophobic
surfaces that we have not examined in the current work.
Examples include the precise triggers for the Cassie-
Baxter to Wenzel transition, evaporation from slippy-type
Cassie-Baxter surfaces, and whether heat flow from the
substrate into a droplet influences evaporative cooling
and if so whether this differs from the Cassie-Baxter to
Wenzel states as would be expected because of different
microscopic contact areas. It is also conceivable that in
some superhydrophobic situations a droplet would ef-
fectively sit upon a layer of vapor. This latter effect could
be analogous to the Leidenfrost effect in which a droplet
placed on a very hot surface creates a layer of vapor in an
initial rapid evaporation of the contact area. The layer of
vapor then prevents further heat being transmitted into
possibility is that condensing a layer of vapor that exists
below a droplet into a liquid layer on the substrate features
below a droplet could trigger a Cassie-Baxter to Wenzel
transition. However, it is not clear whether any of these
mechanisms occur. Because superhydrophobic surfaces
are common in nature and the conservation of water
droplets or vapor near the surface of the leaves of plants
is of importance, further study of evaporation from
superhydrophobic surfaces could be relevant to biological
systems as much as physical systems involving heat and
mass transfer.
V. Conclusions
Free evaporation of small sessile droplets of water from
textured surfaces composed of a lattice of tall pillars of
(24) CRC Handbook of Chemistry and Physics, 62nd ed.; Weast, R.
C., Astle, M. J., Eds.; CRC Press: Boca Raton, FL, 1981-1982.
11060 Langmuir, Vol. 21, No. 24, 2005 McHale et al.

Table 1. Estimates of Diffusion Coefficients for Water into Air for Droplets on Two SU-8 Patterned Surfaces: Surface A
with 10-µm-Diameter Pillars with 20-µm Center-to-Center Spacing and Surface B with 10-µm-Diameter Pillars with
30-µm Center-to-Center Spacing
D∆c (× 10-7 kg m-1 s-1) D (× 10-5 m-2 s-1)
surface T (°C) RH (%) ∆c (g m-3) θi° θf° eq 20 eq 21 eq 20 eq 21 ref
A 23 58.0 8.62 133.3 106.3 2.396 2.358 2.780 2.735 2.584
A 24 57.0 9.34 136.6 103.1 2.559 2.481 2.740 2.657 2.597
A 26 57.0 10.45 140.0 94.8 2.961 2.588 2.833 2.476 2.623
A 26 47.5 12.76 136.2 92.3 3.258 3.189 2.553 2.499 2.623
B 23 51.0 10.06 125.9 90.6 2.664 2.588 2.648 2.572 2.584
B 23 51.0 10.06 141.3 129.5 2.714 2.399 2.698 2.385 2.584
B 25 48.3 11.88 125.0 90.3 3.194 3.153 2.688 2.654 2.610
B 26 46.0 13.12 138.9 118.4 3.318 3.068 2.529 2.339 2.623

polymer photoresist SU-8 has been studied. It has been
shown that such surfaces allow a starting contact angle
for water droplet evaporation exceeding 150° to be
achieved. On our regularly textured surfaces, evaporation
initially proceeds in a pinned contact area mode followed
by a steplike contact line retreat, which mirrors the
underlying lattice structure of the substrate. In some cases,
a sharp steplike decrease in the contact angle with an
accompanying steplike increase in the contact radius
occurs, and the contact area is then completely pinned for
the remainder of the evaporation period. This steplike
change may also be occurring late in the evaporation
process for all droplets, but this has not been possible to
determine because of the resolution of the experimental
system. It has been suggested that the initial phase of
evaporation, consisting of pinned contact line followed by
steplike contact line retreat, corresponds to the droplet
evaporating while suspended across the tops of the pillars
(i.e., a Cassie-Baxter state). The subsequent collapse and
apparent complete pinning of the contact line then
corresponds to a Wenzel state. The initial pinned contact
line phase of evaporation has been analyzed quantitatively
using a model depending of the diffusion of vapor into the
surrounding atmosphere. This model provides a linear
form analysis equation (eqs 9 and 20) that takes into
account a loge 2 correction factor due to the solid boundary
restricting the space into which vapor may diffuse. The
experiments are well described by the model, and this
allows a simple method for the estimation of the diffusion
constant-vapor concentration difference product (i.e.,
D∆c) and hence the coefficient of diffusion D.
Acknowledgment. We acknowledge the financial
support of the UK EPSRC and MOD/Dstl (grant GR/
R02184/01) and of the EU-COST Chemistry Action D19,
WG-007, and TUBITAK-MISAG-COST.D19 projects.
LA0518795