Papers in Physics, vol. 13, art.

130006 (2021)

www.papersinphysics.org
Received: 9 April 2021, Accepted: 27 October 2021
Edited by: L. Kondic
Reviewed by: J. Dijksman
Licence: Creative Commons Attribution 4.0
DOI: https://doi.org/10.4279/PIP.130006

ISSN 1852-4249

The wetting properties of frosted glass

S. Dorbolo1∗
Frosted glass is a common, low cost material. Its roughness can be used to control how
it is wet by water. In this paper, the wetting properties of silicone oil and water are
investigated. For the oil, wetting is total since the oleophilic character of the glass is
enhanced by its roughness. Due to the remarkable optical properties of frosted glass, the
spreading of oil droplets on its surface was recorded over three months. Frosted glass is
a parahydrophilic surface because of its large contact angle hysteresis (up to 80◦ ). The
behaviour of oil and water droplets was compared on a long piece of inclined frosted glass.
The trajectories (and the spreading) of the droplets were studied and phenomenological
laws were deduced to describe the dependence of the droplet speed on the initial volume
of the droplet and the angle of inclination. Such dependences of speed at long travel
distances (100 times the capillary length) were deduced and rationalised with a simple
model that takes into account the thickness of the wake. Moreover, we analysed the flow
inside the wake of water droplets sliding on inclined frosted glass. Suggestions are given on
how to exploit drainage of the water droplet wake and the high hysteresis of water within
the framework of open microfluidics.

I Introduction tems were then qualified as “semi-open” systems

since the liquid presents a free interface. More re-
Microfluidics is probably the most successful scien- cently, microfluidic devices were developed based
tific story of recent decades [1,2]. The development on a high hysteresis surface covered by a “neutral”
of the lab-on-a-chip spirit has led to a breakthrough liquid. The liquid of interest is directly deposited
in fluid science. These developments require clean- on a solid surface in the manner of a fountain pen
rooms and engraving systems which are costly from that literally writes down the circuit [9].
both financial and environmental points of view, In the present paper a rough glass surface is con-
which hampers the growth of the technique world- sidered. This low cost material has interesting wet-
wide. Recent work has focused on low cost and re- ting properties, particularly the high hysteresis of
liable devices like paper microfluidics [3, 4], textiles the contact line of a water droplet. Consequently,
[5] and fiber networks [6–8]. Our aim was to per- when a droplet moves along such a surface, it al-
form basic microfluidic operations (mixing, routing, ways releases some liquid. The wetting properties
analysing) using the imbibition of a droplet by the of a commercial frosted glass were studied when the
fiber network of the medium. The microfluidic sys- plate was horizontal or inclined. In the latter case,
since the droplet loses some liquid on the surface,
∗ [email protected]
the water droplet motion was studied in detail. In
1
order to evaluate the influence of this loss of liquid
FNRS-CESAM-GRASP, Département de Physique B5,
Allée du six Août, 17, University of Liége, B-4000 Liége,
on the speed, the motion of the droplet was mea-
Belgium. sured over a long distance, namely about 100 times
the capillary length. Two liquids were considered:

130006-1
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

silicone oil and water. The aim of the paper was to sively discussed, [20] including the release of small
obtain a detailed description of the physical prop- droplets in the wake. For a large volume the droplet
erties of frosted glass with regard to the wetting takes on the shape of a whale [21]. On the whole,
of these two liquids. This simple system is a very the motion of the droplet was shown to be more
interesting playground for further applications. a rolling than a sliding movement [22, 23] and was
Frosted glass Rough hydrophilic surfaces are modelled by several groups [24, 25]. The case of
known as parahydrophilic surfaces; as summarised a rivulet is a natural extension of droplet sliding
by Marmur [10], we can decouple the contact angle [26]; this rivulet may present meandering due to
due to chemical effects (the nature of the solids, the pinning of the contact line.
liquids and gases) from those caused by the sur- On the other hand, if the surface is highly hy-
face texture. This kind of surface, like frosted drophobic, the droplet slides. Finding inspiration
glass, is wettable and characterised by high con- in natural surfaces like those of lotus leaves, sci-
tact angle hysteresis (as shown in this paper). The entists and engineers have focused on tailoring su-
crevices and peaks play the role of pinning cen- perhydrophobic surfaces: these surfaces are micro-
tres that cause the droplet to behave in a particu- textured with pillars and coated with hydropho-
lar way. Indeed, a water droplet sitting on frosted bic polymers [27]. Anti-adhesion is also the aim of
glass does not spread through the crevices. We can numerous studies (a good summary can be found
consider the droplet as being in a situation between in Ref. [28]). The aim is to decrease as much as
the wicking of the surface crevices and the pinning possible the contact angle hysteresis; i.e., the dif-
of the peaks. On the other hand, for oil the rough- ference between the advancing and receding angles
ness enhances wetting, as demonstrated in the sem- as predicted by the Furmidge law. The influence of
inal paper by Cazabat [11]. Roughness is therefore roughness was also investigated by Miwa [16] in the
agood way to enhance wicking between two plates case of superhydrophobic surfaces. In addition, the
[12] or hemiwicking in patterned channels of micro- fractal dimension of the roughness was also consid-
pillars, as described in Belharz’s thesis [13]. The ered [29]. They found that the higher the contact
chemical nature of the surface can even increase angle, the smaller the tilt angle required for the
the phenomenon of hemiwicking [14]. droplet to slide. Finally, the sliding conditions on
Sliding along an inclined plane If a droplet a superhydrophobic surface depend on the state of
is placed on a flat surface which is then tilted at the droplet: if the droplet remains on the top of the
an angle α, the droplet starts moving when a given asperities (Cassie state), it may roll easily down the
angle α? is reached. The attachment force F of a inclined plate; if the droplet is impaled on the as-
droplet on a surface was investigated by Furmidge perities (Wenzel state), it will stick to the surface
in the 60s [15–17]. This force is related to the ad- [27].
vancing contact angle θa , the receding contact angle Furthermore, making adjustments to the con-
θr and the surface tension of the liquid γ. We have tact angle enables us to obtain counter-intuitive
behaviours. For example, when the inclined plate
F = wγ(cos θr − cos θa ) (1) is made to vibrate, the droplet may move uphill
[30,31]. Some authors inserted hydrophilic patterns
where w is the lateral width of the droplet. Conse- on a hydrophobic plane [32, 33]. They managed to
quently, on an inclined plane with a tilt angle α that control the retention of liquid at precise spots. This
increases, if the driving force given by ρΩg sin α (ρ kind of patched surface opens up the possibility of
is the density of the liquid and g is the gravity ac- depositing particles contained in the liquid at these
celeration) exceeds F , the droplet with volume Ω hydrophilic spots after evaporation. This process
starts moving. The absence of hysteresis implies may lead to a coffee stain effect; that is, prefer-
that the droplet can slide on any inclination. ential deposition along the contact line [34] which
On the one hand, sliding conditions on a hy- can be controlled using non spherical particles [35].
drophilic surface were detailed by Kim [18] and re- This opens up novel perspectives regarding particle
visited by Le Grand [19]. The droplet was found to deposition.
have a teardrop shape with a singularity (the cusp) Objectives Most of the literature on sliding
at the rear of the droplet. This cusp has been inten- droplets reports experiments on very controlled sur-

130006-2
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

(a)

(b)

Figure 1: Water droplets – (a) Pictures taken using Keyence KV-X: on the left, the frosted glass is covered by a
film of water, and on the right the frosted glass is dry. (b) Pictures taken using Filmetrics during evaporation.
The surface started dewetting in the middle picture. On the right, we can see a complex network of water
threads. The side dimension of the picture is about 1 mm.

faces and a small length scale (10-20 times the cap- pose to study the sliding of water and oil droplets
illary length). In consequence, the droplets do not over long distances on a frosted glass plate. As the
release any liquid on the surface (or at least, small droplet moves, its weight changes due to the liquid
droplets do not [20]). Droplet mass is considered film released or invasion of the tortuous surface of
to be conserved and the threshold angle for sliding the frosted glass.
and the initial speed have been discussed. We pro- To achieve these objectives we used a commercial
frosted glass and considered two liquids: silicone oil
(20 cSt) and water. The experimental details can
(a) (b) be found in section II.
First, the wetting properties on a horizontal plate
were studied. The case of the water is reported
in section III.i and the oil in section III.ii. Sec-
ond, the motion of droplets on a tilted frosted glass
plate was investigated as a function of the volume
Ω of the droplet, and as a function of the tilt an-
gle α. The results for oil are reported in section
IV.i and for water in section IV.ii. A first approach
model is detailed for the case of water in section
Figure 2: Water droplets – Contact angle of a 5 µL IV.iii. In the case of water on the inclined plate,
water droplet on (a) the smooth side of the glass (top) the droplet releases a thin film that drains back
and on the rough side of the glass (bottom). (b) Water into the droplet: this film is known as the wake.
droplet sliding along the rough side of the frosted glass. The flow in the wake is discussed in section V. Fi-
This illustrates the advancing contact angle (about 80◦ ) nally, section VI presents some perspectives that
and the receding contact angle (0◦ ). exploit the high contact angle hysteresis of a wa-

130006-3
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

ter droplet to create microchannels and generate a III Horizontal frosted glass plate
microflow. Furthermore, the microflow inside the
wake of the sliding water droplet was also investi- i The case of water
gated: we show that color pigment can be extracted
from a water solution. This conclusion is reported The contact angles of water on the smooth side
in section VII. and on the frosted sides were measured using a
CAM200 goniometer. In this case, on the smooth
side the contact angle of θS = 32◦ ± 5◦ indicates
a tendency to be wet by water; the advancing and
receding angles were found to be 50◦ and < 10◦
respectively. On the frosted glass side, the angle
II Experimental details of contact was found to be θF = 50◦ ± 8◦ . The
advancing and receding angles were evaluated on
an inclined plate and were found to be equal to
The frosted glass was a commercial material found 80◦ ± 2◦ and 0◦ , respectively. In Fig. 2a, the pic-
under the name “Satinovo Maté” (Saint-Gobain) ture of a water droplet is shown on a piece of frosted
[36]. This material was chosen because it is a stan- glass (bottom) and on the same glass plate but on
dard glass developed by Saint-Gobain and because the smooth side (top). We can see that the rough-
the surface state is homogeneous on a large scale of ness increases the contact angle. This is counter
the order of one meter, since the frosted surface is intuitive because the roughness is known to en-
obtained by an acid attack. hance the natural wetting tendency of the material.
According to Ref. [14], the roughness of the con-
Roughness was measured using a Bruker laser
tact angle on the frosted glass should be corrected.
profilometer and a VK-X Keyence microscope. The
We can define the roughness factor r > 1 as the
arithmetic mean deviation was found to be 1.99 µm
ratio between the actual surface area and the ap-
for a root mean square of 2.34 µm. On the sam-
parent surface area [37]. The relation between the
ple under consideration, the maximum vertical dis-
smooth side and frosted side contact angles is given
tance between a valley and a peak was hm = 9.33
by cos θR = r cos θpS . In the present case, factor r
µm; the peaks were separated by a distance dm of
is about 1.14 (= 2 h2m + (dm /2)2 /dm ). However,
34 ± 10 µm. Figure 1a shows two micrographs
the roughness in this case plays the role of pinning
taken using the VK-X Keyence of 400 by 400 µm
centers that confine the droplet to a limited area,
of the frosted glass. The colors correspond to the
and consequently, the contact angle is increased.
height of the roughness: red corresponds to 17 µm
The pinning centers are so strong that the receding
and blue to altitude zero. On the left, the sample
angle is equal to zero; indeed, the droplet releases a
was wet by water. The liquid fully covered the sur-
thin film of liquid after its passage. The high con-
face and the small asperities were wet completely.
tact angle hysteresis was clearly visible when the
Fig. 1b shows a picture of the frosted glass taken
frosted glass plate was inclined to observe the mo-
while water was evaporating, using a Filmetrics in-
tion of the droplet (see Fig. 2b). The front angle
strument (F40-NIR with 15X objective). These
was close to 80◦ ± 2◦ , and the wake consisted of a
pictures were taken when water started dewetting
film of water that drained back towards the droplet.
the surface in some spots, leading to this entangle-
ment of water rivulets. When the surface was horizontal, in the case of
the rough surface the contact line remained pinned
Concerning the liquids used, 20 cSt silicone oil during evaporation of the droplet. Some pictures
was chosen because of its very low saturation pres- were taken during the evaporation process (Fig. 1b)
sure (less than 5 mmHg). This prevents any evapo- using a Filmetrics device. It can be seen that
ration during experiments of long duration. On the dewetting occurs in different places of the film
other hand, working with high viscosity increases (black dots in the picture in the middle and on the
the relaxation time. For this reason the 20 cSt oil right). The dry spots increase in size, resulting in
was a good compromise. The water used was bidis- the entanglement of small liquid threads (picture
tilled. in the middle).

130006-4
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

105 s

104 s

103 s (b)
102 s
101 s

20 mm

(a)

(c)

Figure 3: Silicone oil spreading – (a) Superimposition of 5 pictures taken at 10, 102 , 103 , 104 and 105 s intervals
fromabove a horizontal frosted glass plate on which a droplet (10 µL) of silicone oil was initially released. (b)
Evolution of the oil stain diameter D as a function of time (log-log plot) for a droplet of 5, 7.5 and 10 µL (green
diamonds, red bullets and blue bullets, respectively). The straight lines are power laws for which the exponents
are indicated in the figure. The dashed line corresponds to the saturation regime for which the stain did not
expand any longer. (c) Evolution of the average thickness h of the oil stain as a function of time for three droplet
volumes (the same volumes and symbols are used as in (b)). Note that the data in (b) were not obtained from
the pictures (a).

ii The case of silicone oil ment was performed on a large plate of 580 × 1300
mm, also over a long period (up to three months).
When a droplet of silicone oil was released onto A simple ruler was used to measure the oil stain
the frosted glass, the droplet spread rapidly. The diameter D.
position of the contact line was easy to detect be- In Fig. 3b the oil stain diameter D is presented
cause frosted glass becomes transparent when liq- as a function of time in a log-log graph. Three
uid covers the surface (it enters the crevices). The different initial volumes Ω of the droplet were con-
dynamics of the spreading was recorded by mea- sidered: 5, 7.5 and 10 µL. The growth of the stain
suring the diameter of the circular oil stain on the behaved similarly whatever the initial volume of
frosted glass. In Fig. 3a successive pictures are the droplet. Several regimes can be identified in
shown between the moment of release and 10000 the short, medium and long term. During the first
s (see caption). The droplet spread uniformlyin a 105 s (one day) the droplet spread rapidly, then
disk of growing diameter. It is notable that the the growth rate decreased as indicated by the log-
circular stain at 105 s was still easily observable log plot of Fig. 3b. Finally, a plateau was reached
thanks to the presence of a very tiny amount of oil after about 120 days. The contour of the stain be-
which was sufficient to make the plate transparent. came more difficult to see after 3 months. In the
As oil does not evaporate and as the oil properties medium term, we found that D followed two power
are not sensitive to dust, the experiments ccould laws of time: D ∝ t0.3 between 102 and 105 s and
be performed over a long period. A second experi- D ∝ t0.15 between 105 and 107 s.

130006-5
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

The spreading of oil on a rough surface was stud- i The case of silicone oil
ied by Cazabat and Cohen Stuart [11]. In this work,
three regimes were identified. (i) The first to ap- As shown in the previous section, oil literally in-
pear, just after the positioning of the droplet, wass vades the frosted glass very rapidly during the first
the capillary regime. During this regime, the ra- 10 seconds. On the other hand, on an inclined
dius of the stain increased following the power law plane, gravity is responsible for movement of the
D ∝ t1/10 (this is Tanner’s law [38]). (ii) After- center of mass of the droplet. This force increases
wards, the oil invadeed the crevices, which are con- with the angle of inclination. Consequently, the
sidered randomly distributed. This regime can be sliding of an oil droplet results in a competition
modelled as a diffusion process, D ∝ t0.5 . However, between spreading and sliding. Fig. 4 shows the
because of the complexity of the crevice network, shape of the trace released by the droplet for five
the experimental exponent belongs to the interval different angles of inclination, for a droplet of 75
[0.25, 0.5]. (iii) Finally, the thickness of the droplet µL. The direction of sliding is towards the right of
(or the oil stain) was of the same order of magni- the images. The pictures shown are the result of
tude as the roughness; the capillary force and the the superimposition of pictures taken at different
spreading speed both decreased. times (see caption). According to the angle (and
size of the droplet), a change in morphology was
Applied to the present case, we observed the in-
observed: (i) when the center of mass of the droplet
vasion regime between 102 and 105 s D ∝ t0.3 , then
slides faster than the spreading, the stain takes on
a decrease in the spreading speed D ∝ t0.15 , as also
a comet-like shape, and (ii) when spreading is the
observed in Ref. [11]. The plateau that we observed
dominant effect, the stain is oblong in shape. The
is new. We can assume that the oil stain stopped
comet shape can be observed for the angles 25, 45,
spreading when the plateau was reached. Consid-
65 and 85◦ until the time t=500 s in Fig. 4. For a
ering the initial volume Ω of oil and the surface of
longer time period the tip of the trajectory becomes
the oil stain, we can determine the average thick-
rounded. Indeed, the comet shape cannot hold for
ness h of the oil stain: h = 4Ω/(πD2 ). In Fig. 3c
a long time because the droplet becomes lighter
we show thickness as a function of time. The av-
and lighter as some oil is released in the wake of
erage thickness h started at 10 µm and gradually
the droplet. The loss of weight is due to spreading
decreased until it reached about 100 nm. This value
of the oil in a lateral direction. The oblong shape
is obviously not realistic. The silicone oil invades
is predominant for small angles (and for small vol-
the valleys of the frosted glass and probably does
umes), as observed in Fig. 4; α = 5◦ .
not cover the tops of the hills. Additional observa-
tions are required to investigate the presence of the The sliding position of the droplet can only be
silicone oil at microscopic level. evaluated by plotting the position x of the tip of the
oil stain (x = 0 is the starting point), i.e. the fur-
thest point reached by the oil stain from the start-
ing position. In Fig. 5a, the position of the tip of
IV Inclined frosted glass plate the oil stain is presented as a function of time for
four angles: α =5, 25, 65 and 85◦ (see legend). The
A large frosted glass plate (580 × 1300 mm) was volume Ω of the droplet was 75 µL. In Fig. 5b, the
placed on an aluminium frame. The angle was fixed angle was kept constant (25◦ ) while the droplet vol-
and measured using an inclinometer. Then, the ume was varied between 25 and 100 µL. For each
droplets were released at the top of the inclined angle considered, the behaviour of the droplet was
plane. Their movements were recorded using a Go- similar. First the droplet slid with a velocity that
Pro camera which enabled us to obtain a large field increased with the angle and the droplet mass. Af-
of view with sufficiently high resolution (4000 × terwards, the speed decreased. Note that there was
3000 pixels). By thresholding the images, it was no threshold angle for the oil motion since the oil
then possible to follow the motion of the droplet spread on the frosted glass whatever the angle.
(in the case of water) and the position of the fur- On long time scale, friction acts in addition to
thest point (tip) reached by the droplet (in the case the lateral spreading. Moreover, the volume of the
of oil). droplet at the tip decreases. These processes have

130006-6
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

↵=5
100 s 500 s 1000 s
0s 10 s

↵ = 25
10 s 100 s 500 s 1000 s 2000 s 3000 s
0s

↵ = 45
10 s 100 s 500 s 1000 s 2000 s 4000 s
0s

10 mm

↵ = 65
500 s 1000 s 2000 s
10 s 100 s
0s

↵ = 85
0s 10 s 100 s 500 s 1000 s 1500 s

1000

Figure 4: Silicone oil – Superimposition of several pictures of the oil stain, taken from above, for five angles α of
inclination (5, 25, 45, 65, and 85◦ , from top to bottom). In each case the volume of the droplet was 75 µL. The
height of each image is 40 mm. The times at which the pictures were taken are indicated in the different figures
and range from 0 to 4000 s. The complete set of images (taken every second) was analysed to obtain the area
and the speed of the spreading presented in Fig. 5.

to be taken into account as they decrease the speed quently, the droplet started sliding during produc-
of the tip. In Fig. 5a, the position x(t) was fitted tion. The method based on extrapolation of the ini-
by a phenomenological law tial speed using the logarithmic fit deals with this
problem. Actually, a sine dependence on the angle
x(t) = a1 ln(a2 (t + a3 )) (2) was expected, as the force acting on the droplet is
where a1 , a2 and a3 are the fit parameters. Note ρo Ω sin α, where ρo is the density of the oil. This
that over a long period of time (Fig. 5b) the speed trend is not clearly observed and the actual data
tends to decrease to zero (x(t) is constant). does not allow this conclusion. On the other hand,
Starting from Eq. (2), we can obtain the expres- a linear trend was observed for the initial speed
sion of the speed ẋ as a function of x, dependence on droplet volume. This behaviour is
compatible with the fact that the driving force is
proportional to the mass. As for x? , this value in-
?
ẋ = a1 a2 exp(−x/a1 ) = v0 exp(−x/x ). (3) creases with the angle and the volume. This trend
is obviously expected. However, the value of x?
The initial speed is then given by v0 = a1 a2 and does not seem to go to zero for small angles. This
we can define a characteristic length x? = a1 . reflects the spontaneous spreading observed in the
This characteristic length represents the distance flat configuration.
at which the speed decreases significantly. The
initial speed v0 and the parameter x? were eval- Instead of considering the position of the tip, the
uated and reported as a function of the angle α area A of the oil stain is consequently a good pa-
and the droplet volume Ω in Figs. 5c and 5d, rameter to measure as a function of the parameters
respectively. The initial speed seems to increase α and Ω. In Figs. 5e and 5f, the evolution of area
linearly with the angle and the volume (blue bul- A is shown as a function of time for a constant vol-
lets). The problem with measuring initial speed de- ume (Ω =75 µL) with different angles (5, 25, 45, 65
pendence on the angle and the volume is the time and 85 degrees), and for a constant angle (α=25◦ )
taken to create the droplet. Indeed, it took typi- with four volumes (25, 50, 75 and 100 µL). The re-
cally one second to generate the droplet and conse- sults are reported in a log-log plot. On a long time

130006-7
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

(a) (b)

(c) (d)

(e) (f)

Figure 5: Silicone oil – (a)-(c)-(e) concern data taken for a fixed volume Ω of the droplet equal to 75 µL, for
five angles of inclination α =5, 25, 45, 65 and 85◦ (see legend). For (b)-(d)-(f), the angle was fixed at α = 25◦
for four droplets with volumes Ω =25, 50, 75 and 100 µL (see legend). (a) and (b): Position of the tip of the
oil stain. (c) and (d): Initial speed v0 and characteristic length x? . (e) and (f): Evolution of the area of the oil
stain. The solid line is a power law fit, A ∝ t0.6 , while the dashed line is for A ∝ t0.37 . The insets (e) and (f)
correspond to close-ups of x(t) in a log-log plot between t=1 and 100 s. The dashed lines correspond there to
x ∝ t0.26 .

scale, the behaviour of A(t) suggests that spread- primarily due to spreading , given that gravity no
ing behaves along a power law A(t) ∝ t0.6 after longer has any effect. Therefore, the oil stain on the
the time t=100 s, as shown by the solid black lines inclined plane behaves like the oil stain in the flat
in Figs. 5e and 5f. The droplet no longer moves configuration. In the latter case, the diameter was
downwards. The increase in the oil stain area is found to increase following a power law, D ∝ t0.3 ,

130006-8
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

between 102 and 105 s (section III.ii). The area The trajectories are mostly rectilinear and the wake
should follow the scaling A ∝ D2 ∝ t0.6 , which is is clearly visible. The plate was cleaned after each
not far from the exponent found experimentally. experiment with isopropanol and then rinsed with
On a short time scale (for t <100 s) the spread- bidistilled water. The same experiments were con-
ing of the oil stain on the inclined plate, imaged by ducted on the smooth side of the plate. In contrast
measurement of the area A, behaves according to a to the experiments on the frosted side, the trajecto-
power law. We fitted the different data from Figs. ries were very sensitive to the state of the surface,
5e and 5f between 1 s and 100 s with the law A ∝ tγ . particularly to any imperfections left by the clean-
We found γ = 0.374±0.038 and γ = 0.355±0.03 for ing process. This is a well-known problem when
the data corresponding to a constant volume and to a large glass plate must be cleaned in a domestic
a constant angle, respectively. These scalings are context. On the frosted glass side the results were
represented by dashed black lines in Figs. 5e and much more reproducible.
5f. Within this time range we expect that area A
should scale as the product x(t)w(t), where w(t) is
the evolution of the lateral width of the oil stain.
The dependence of the length (x(t)) of the stain on
time is not linear; moreover, the logarithmic law
does not capture very well the first instants of slid-
ing. Between t =1 and 100 s, a power law x ∝ tβ
is more appropriate, as shown in the inset of Figs.
5e and 5f. Surprisingly, the exponent β found is
not very sensitive to the volumes and angles con-
sidered; on average, we found β = 0.252 ± 0.022
for the data for which the volume is constant, and
β = 0.262 ± 0.025 for the data corresponding to a
constant angle. This exponent β = 0.26 is related
to the friction that counteracts the projected weight
on the plate, and to the weight loss of the droplet
with time. The precise mechanism has yet to be dis-
covered, but is certainly related to the geometry of
the oil flow inside the droplet. The scaling x ∝ t0.26
is shown by the dashed lines in the inset of Figs. 5e
and 5f. Finally, the same exercise was performed
to characterise w(t) ∝ tδ between 1 and 100 s. The
width w was obtained by considering the width of
the smallest rectangle that captured the oil stain.
We found δ = 0.12 ± 0.036 and 0.096 ± 0.005 for
the data at constant volume and constant angle, re-
spectively. This exponent δ = 0.11, on average, is
similar to the characteristic of the capillary regime
as discussed by Cazabat and Cohen Stuart. Finally,
we found that the scaling A ∝ x(t)w(t) is consistent
with the experimental data, since γ ≈ β + δ.

ii The case of water

Figure 6: Water droplet sliding along frosted glass in-
In Fig. 6 a typical picture taken during the sliding clined at 65◦ . The droplet volumes are 30, 40, 50, 60
of water droplets is presented. The position x of and 70 µL from bottom to top. The downhill direction
the droplets was found by image analysis, taking is towards the bottom of the figure. The numbers in
advantage of the reflection of light on the droplets. the figure come from the image analysis.

130006-9
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

In Fig. 7a the trajectories x(t) are shown for the volume being released. Evidence of this process
8 droplet volumes between 30 and 100 µL for is found in the width of the traces released by the
α = 45◦ . The trajectories are very similar to those droplets. The lateral extension w of the droplet,
of the oil droplet: the speed is highest during the and consequently of the wake, was measured and
first 10 s and then decreases. The trajectories are reported in Fig. 7c as a function of the angle for two
represented in a semi-log plot to enhance the slow droplet volumes (50 µL and 100 µL). The width of
decrease in speed. The same phenomenological fit, the wake was larger for the largest volume consid-
Eq. (2), was applied to the data. An example ered, especially for small angles (this was observed
is shown in Fig. 5a on the data corresponding to particularly when the plate was horizontal, α = 0).
Ω = 50 µL (black curve). Exactly like the silicone On the other hand, the difference became increas-
oil case, the speed decreases exponentially with the ingly smaller as the angle increased. Indeed, the
position irrespective of the droplet volume Ω or in- droplet started sliding before all the volume had
clination angle α. This exponential behaviour can been released. This explains why the initial speeds
be interpreted as follows: The droplet releases a v0 are not very different when the angle of inclina-
thin film of water as it moves. The film drains this tion is above 45◦ (see Fig. 7b).
liquid back to the droplet, but some of it remains on The characteristic length x? is less sensitive to
the plate and evaporates. The driving force there- the initial conditions and is reported as a func-
fore decreases, since the weight of the droplet is tion of droplet volume for the same set of angles
decreasing. The same applies to the oil case, but as in Fig. 7d. No clear trend seems to be present.
instead of evaporation, the loss process is caused All the values of x? are located between 200 mm
by spreading in a lateral direction relative to the and 400 mm for all the angles between 30 and 85◦ .
movement. Even more surprising, the value for smooth glass
The initial speeds v0 are reported as a function seems to be compatible with the value found for
of droplet volume Ω in Fig. 7b: seven different an- frosted glass. The exponential decrease in speed
gles between 25 and 85◦ are considered in the case with position is well interpreted as being due to
of the frosted side, and three different angles be- the weight loss caused by the film of water released
tween 30 and 75◦ in the case of the smooth side. in the wake.
The data concerning the smooth side are linked by
lines and are noted with the letter ’S’ in the legend. iii Model of the sliding dynamics
The droplets are found to move when the droplet
volume is sufficient. This threshold value depends On an inclined plane, the speed of the droplet is
on the angle, since the driving force (ρw Ωg sin α, proportional to the driving force; that is, the mass
where ρw is the density of water) should exceed a of the droplet times the apparent gravity (corrected
value that depends on the hysteresis angle and the for the angle of inclination). When the droplet be-
length of the droplet [15] (see Eq. (1)). As soon as gins to move it loses some liquid in its wake. The
the droplet can move, the initial speed increases lin- model to explain the logarithmic time dependence
early with droplet volume Ω, as expected regarding of the droplet position must account for this loss of
the driving force. mass. If we consider only the loss of mass in the
Two points should be noted here. (i) Concerning speed deceleration, i.e., if we disregard friction, the
measurement of the droplet volume threshold, the speed ẋ is proportional only to the driving force:
experimental set-up was not designed to measure
the threshold volume precisely. To find the precise ẋ ∼ Ω(t)g sin α (4)
value, we should start at the horizontal position, The loss of volume depends on time, speed and the
place the droplet on the glass plate and then tilt it thickness of the wake, as follows:
slowly. The system did not allow direct measure- Z t
ment of the angle, and more different droplet vol- Ω(t) = Ω0 − ẋ(t0 )e(t0 )wdt0 (5)
umes should have been tested. (ii) Concerning the 0
initial speed measurements, the system suffers from where Ω0 is the initial volume of the droplet and
the same problem as in the case of silicone oil. Po- e is the thickness of the wake. In other words, we
sitioning the droplet took some time, depending on can write the derivative of the volume

130006-10
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

(a) (b)

(c) (d)

Figure 7: Water droplet – (a) Position of the droplet as a function of time (semi-log) for eight droplet volumes
between 30 and 100 µL (see legend) for an inclination of 45◦ . The solid line is the fit of Eq. (2) on the data
Ω =50 µL. (b) The initial speed v0 of the droplet is reported as a function of the volume Ω for different angles
α between 25 and 85◦ . The data points linked by solid lines correspond to the smooth side of the glass for
three inclination angles: 30, 445 and 75◦ (see legend). (c) Width w of the water droplet wake as a function of
the inclination angle for two droplet volumes (50 and 100 µL). (d) Dependence of the parameter x? = a1 as a
function of droplet volume. The same parameters and symbols are used as in Fig. 7b.

the thickness e is proportional to the speed, µ = 1,

Ω̇ ∼ −ẋew (6) resolving the system of equations (4), (6) and (7)
Following the Landau-Levich-Derjaguin problem
[39], we can assume that the thickness of the wake ẍ ∼ −(ẋ)2
depends on speed ẋ, namely

e ∼ (ẋ)µ (7) 1
ẋ ∼ (8)
The power coefficient µ is equal to 2/3 in the clas- t
sical Landau-Levich-Derjaguin problem. However, this relationship leads to
the present situation is particular since the lateral
size of the droplet is limited and the film released by
the droplet drains back to the droplet continuously. x(t) ∼ ln(t) (9)
In a first step, if we consider that the exponent µ
is equal to zero, which corresponds to a constant This simple model is able to justify the logarithmic
thickness of the wake, we find an exponential de- time dependence of the position of the droplet if
crease in speed with time. This behaviour is not the wake thickness is predicted to be linear with
observed. On the other hand, if we consider that the droplet speed.

130006-11
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

Figure 9: Water droplet – The three images are sep-

arated by 1 s. The real size of the images is 20 mm
width. On the left, a line of blue methyl was drawn.
The droplet came from the top and crossed the line. On
the side of the wake we can see that the blue coloring
went upwards. The angle of inclination was 30◦ and
the volume of the droplet 75 µL.

the line. We can see that the flow speed profile is

Figure 8: Close-up of a water droplet sliding along a not completely oriented downhill. The flow in the
frosted glass plate. middle of the wake goes downhill, but at the side
of the wake the flow goes upwards. In summary,
the flow inside the droplet, close to the surface of
V Flow in the wake the frosted glass, is represented by the arrows on
the right of Fig. 8. The flow goes downwards in
As shown in Figs. 2b and 6, while moving on the the middle of the droplet but upwards at the sides.
frosted glass the water droplet releases some water Note that the flow is three dimensional; however,
in its wake. The film is thick at first, depending more information is needed for a complete picture
on the speed of the droplet [39]. Then some liquid of water flow.
drains back towards the droplet but a thin film re- In Fig. 8 (left) we can see a thin white line in the
mains. Fig. 8 shows a close-up of a sliding droplet middle of the droplet. Actually, the blue color was
(the droplet volume was 100 µL and the angle 35◦ ). disappearing: after 400 mm of travel, the droplet
The wake is about 1 cm wide. The droplet was was nearly transparent. The methylene blue was
colored using methylene blue. The wake can be adsorbed on the Saint-Gobain glass material. The
seen more clearly in this picture. On the right, company does not provide information on the com-
the picture was sketched over to enhance the con- position of the glass, but whatever the reaction,
trast. The droplet, in dark blue, is rounded along this phenomenon clearly shows that a chemical or
the advancing contact line, while the rear looks like adsorption reaction may occur during movement
a flame. The liquid film is presented in light blue. of the droplet. The droplet also decolored on the
The shape of the droplet shows that a thick film is smooth side, but to a lesser extent. This obser-
released just after the passage of the droplet. This vation raises an interesting question regarding the
film retracts towards the axis of symmetry that re- purification or the chemical reaction that can be
leases the thin film. induced in a single droplet. The frosted glass al-
To obtain information about the movement of lows close interaction between the surface and the
the liquid in the wake, a line of methylene blue was droplet because (i) roughness increases the contact
drawn on the surface of the frosted glass (see image surface (r > 1) and (ii) when the droplet slides
on the left in Fig. 9). When the line had dried the along the surface, the thin film interacts for a long
frosted glass was inclined at 30◦ ; a droplet made time before draining back to the droplet. This tech-
of pure water with a volume of 75 µL was then nique could possibly be extended if we consider the
placed on the glass. In the pictures in the middle frosted glass as a 2D trickle-bed reactor [40]; that
and on the right the droplet had already crossed is, a vertical reactor that uses the motion of the

130006-12
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

(a) (b)

Figure 10: Water droplet – (a) Typical dirt released

after drawing on frosted glass. The liquid evaporates
leaving dirt along the contact line. (b) Superimposition
of 5 pictures obtained after the passage and complete
evaporation of 5 successive droplets.

liquid induced by gravity to flow through a granu- Figure 11: Water droplet – (a) After the passage of a
lar material made of particles (3D) or through the droplet, a liquid film is released on the frosted glass
channel network (frosted glass 2D) on the surface (in the picture on the left). After evaporation we can
of which the liquid reacts. The frosted glass surface observe a thin trace due to the accumulation of dust
could be made functional by silanization or grafting (picture on the right). (b) Due to the pinning of the
[41]. contact line, a micro-channel can be drawn between two
droplets (1). The liquid can then flow from the smallest
to the biggest droplet (2). The flow rate changes during
VI Perspectives the evaporation process. Picture 3 shows the result
after complete evaporation of the droplet.
Besides the chemical or adsorption reactions that
can be exploited, the high hysteresis characteristics
of the surface are important. The passage of one was less than two tenths of millimeter. As soon as
droplet on the frosted glass may release some traces the droplets were connected a flow was established
of dirt, as can be observed in our homes (Fig. 10a). from the smaller to the bigger droplet, due to the
Actually, the dust present in the room gets trapped difference in Laplace pressure (some blue coloring
on the surface of the frosted glass. The droplet can be seen in the initial pure water droplet on the
collects it, and after evaporation two parallel lines right). This very flexible way of drawing micro-
can be seen that correspond to the boundaries of channels resembles a system exploited by Walsh et
the liquid film released by the droplet (Fig 11a). al [9]. However, in the case of frosted glass, the
On complete evaporation these lines become invis- system is open and subject to evaporation. The
ible. If we breathe close to the surface, the two bottom picture of Fig. 11b shows the result after
lines become visible. These lines may change the complete evaporation, which limits the duration of
trajectory of another droplet. Successive droplets the experiment and changes the balance of pressure
were released when the previous droplet had totally between the droplets.
evaporated. The starting position was identical for
all the droplets. In Fig. 10b, five pictures of the VII Conclusion
wake of five successive droplets are superimposed.
The trajectories are no longer straight lines during Frosted glass has been shown to be a very simple
the first centimeters of motion. The droplet tends system with peculiar wetting properties when the
to move laterally to “avoid” a previous trajectory. plate is horizontal. For silicone oil, macroscopic
Because the contact angle of the frosted glass is spreading can be studied over a very long period
pinned, it is also possible to draw a path with a using very simple instruments. The microscopic be-
simple feather pen. In Fig. 11b two droplets were haviour has yet to be discovered. The way the oil
joined using a pen. The width of the channel was invades the complex network of crevices and hills
about 1 mm and the length, 20 mm. The thickness could aid our understanding of the isotropic be-

130006-13
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

haviour observed. Concerning the water droplet, [2] N Convery, N Gadegaard, 30 years of microflu-
the high contact angle hysteresis and the pinning of idics, Micro Nano Eng. 2, 76 (2019).
the contact line could be exploited to draw micro-
channels and trigger micro-flow in this low-cost [3] A W Martinez, S T Phillips, B J Wiley,
open microfluidic system. M Gupta, G M Whitesides, Flash: A rapid
On the inclined plate, the trajectory over a long method for prototyping paper-based microflu-
distance of a droplet of oil and and one of water idic devices, Lab. Chip 8, 2146 (2008).
has been shown to follow a logarithmic law for the
intermediate range of time (between 10 s after re- [4] F Ghaderinezhad, R Amin, M Temirel, B Ye-
lease and 2000 s). This behaviour was rationalised nilmez, A Wentworth, S Tasoglu, High-
using a simple model that takes into account the throughput rapid-prototyping of low-cost paper-
dependence of the thickness of the wake on droplet based microfluidics, Sci. Rep. 7, 3553 (2017).
speed.Droplet speed has been shown to decrease ex-
[5] A Nilghaz, D R Ballerini, W Shen, Exploration
ponentially with distance from its origin. In both
of microfluidic devices based on multi-filament
cases (water and oil), this decrease is due to the
threads and textiles: A review, Biomicroflu-
loss of mass of the liquid located at the advancing
idics 7, 51501 (2013).
front. Lateral spreading is the main loss mecha-
nism for the oil, while for water, loss on the surface
[6] M Lismont, N Vandewalle, J Joris, L Dreesen,
and evaporation of the film in the wake are the
Fiber based optofluidic biosensors, Appl. Phys.
dominant effects.
Lett. 105, 133701 (2014).
Finally, we studied flow in the wake of the water
droplet. The slow drainage and the large area of [7] F Weyer, A Duchesne, N Vandewalle, Switch-
contact due to roughness lead to good interaction ing behavior of droplets crossing nodes on a
between the liquid and the solid. This property has fiber network, Sci. Rep. 7, 13309 (2017).
been shown to cause discoloration of the methylene
blue. Some perspectives of this work belong to do- [8] F Weyer, M Ben Said, J Hotzer, M Berghoff,
mains for which a chemical reaction with a surface L Dreesen, B Nestler, N Vandewalle, Com-
is needed. The flow in the network of channels of pound droplets on fibers, Langmuir 31, 7799
the frosted surface enables efficient contact between (2015).
the solid and the liquid.
[9] E J Walsh, A Feuerborn, J H R Wheeler,
A Na Tan, W M Durham, K R Foster, P
R Cook, Microfluidics with fluid walls, Nat.
Acknowledgements - SD thanks the FNRS for Commun. 8, 816 (2017).
financial support as a senior research associate.
This work was also support by PDR-31226867 [10] A Marmur, Hydro- hygro- oleo- omni-phobic?
WOLFLOW (F.R.S.-FNRS). The Author would Terminology of wettability classification, Soft
like to warmly thank Pr. Denis Terwagne for fruit- Matter 8, 6867 (2012).
ful discussions about this fascinating system and
for his vision on potential biomimetic applications. [11] A M Cazabat, M A Cohen Stuart, Dynamics of
The Author would like also to thank Pr. Nicolas wetting: Effects of surface roughness, J. Phys.
Vandewalle for the use of GRASP laboratory equip- Chem. 90, 5845 (1986).
ment. Finally, SD thanks Mathieu, Chloé and Alice
Dorbolo for revealing the remarkable properties of [12] B Darbois Texier, P Laurent, S Stoukatch,
frosted glass. S Dorbolo, Wicking through a confined mi-
cropillar array, Microfluid. Nanofluid. 20, 53
(2016).

[1] G M Whitesides, The origins and the future of [13] D Beilharz, PhD Thesis: Liquids guided by
microfluidics, Nature 442, 368 (2006). texture, ESPCI, Paris (2018).

130006-14
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

[14] D Quéré, P-G de Gennes, F Brochard-Wyart, [25] G Karapetsas, N T Chamakos, A G Pap-

Capillarity and Wetting Phenomena: Drops, athanasiou, Efficient modelling of droplet dy-
Bubbles, Pearls, Waves, Springer, New York namics on complex surfaces, J. Phys. Condens.
(2013). Mat. 28, 085101 (2016).

[15] C G L Furmidge, Studies at phase interfaces. [26] S Couvreur, A Daerr, The role of wetting het-
I. The sliding of liquid drops on solid surfaces erogeneities in the meandering instability of
and a theory for spray retention, J. Colloid In- a partial wetting rivulet, Europhys. Lett. 99,
terf. Sci. 17, 309 (1962). 24004 (2012).

[16] M Miwa, A Nakajima, A Fujishima, [27] A Lafuma, D Quéré, Superhydrophobic states,

K Hashimoto, T Watanabe, Effects of Nat. Mater. 2, 457 (2003).
the surface roughness on sliding angles of
water droplets on superhydrophobic surfaces, [28] P Bourrianne, PhD Thesis: Non-mouillant
Langmuir 16, 5754 (2000). et température: Application aux revêtements
culinaires, ESPCI, Paris VI (2016).
[17] M F Mei, B M Yu, L Luo, J C Cai, A
[29] T Onda, S Shibuichi, N Satoh, K Tsui, Super-
model for the contact angle of liquid droplets
water-repellent fractal surfaces, Langmuir 12,
on rough surfaces, Chinese Phys. Lett. 27,
2125 (1996).
076802 (2010).
[30] P Brunet, J Eggers, R D Deegan, Vibration-
[18] H Y Kim, H J Lee, B H Kang, Sliding of liquid
induced climbing of drops, Phys. Rev. Lett. 99,
drops down an inclined solid surface, J. Coll.
144501 (2007).
Interf. Sci. 247, 372 (2002).
[31] P Sartori, D Quagliati, S Varagnolo, M Pierno,
[19] N Le Grand, A Daerr, L Limat, Shape and mo- G Mistura, F Magaletti, C M Casciola, Drop
tion of drops sliding down an inclined plane, J. motion induced by vertical vibrations, New J.
Fluid Mech. 541, 293 (2005). Phys. 17, 113017 (2007).
[20] J Snoeijer, E Rio, N Le Grand, L Limat, Self- [32] S Varagnolo, D Ferraro, P Fantinel, M Pierno,
similar flow and contact line geometry at the G Mistura, G Amati, L Biferale, M Sbragaglia,
rear of cornered drops, Phys. Fluids 17, 072101 Stick-slip sliding of water drops on chemically
(2005). heterogeneous surfaces, Phys. Rev. Lett. 111,
066101 (2013).
[21] U Thiele, K Neuffer, M Bestehorn, Y Pomeau,
M G Velarde, Sliding drops on an inclined [33] B Chang, Q Zhou, R H A Ras, A Shah,
plane, Colloids Surf. A 206, 87 (2002). Z Wu, K Hjort, Sliding droplets on hy-
drophilic/superhydrophobic patterned surfaces
[22] S P Thampi, R Adhikari, R Govindarajan, Do for liquid deposition, Appl. Phys. Lett. 108,
liquid drops roll or slide on inclined surfaces?, 154102 (2016).
Langmuir 29, 3339 (2013).
[34] R D Deegan, O Bakajin, T F Dupont, G Hu-
[23] M Kim, E Lee, D H Kim, R Kwak, Decoupled ber, S R Nagel, T A Witten, Capillary flow as
rolling, sliding and sticking of a viscoplastic the cause of ring stains from dried liquid drops,
drop on a superhydrophobic surface, J. Fluid Nature 389, 827 (1997).
Mech. 908, A41 (2020).
[35] J Vermant, When shape matters, Nature 476,
[24] J P Dupont, D Legendre, Numerical simu- 286 (2011).
lation of static and sliding drop with contact
angle hysteresis, J. Comput. Phys. 229, 2453 [36] Saint-Gobain, Float standard glass for building
(2000). purpose.

130006-15
 Papers in Physics, vol. 13, art. 130006 (2021) / S. Dorbolo

[37] D Quéré, Wetting and roughness, Annu. Rev. [40] Y Jiang, M R Khadilkar, M H Al-Dahhan, M
Mater. Res. 38, 71 (2008). P Dudukovic, Two-phase flow distribution in
2D trickle-bed reactors, Chem. Eng. Sci. 54,
[38] L H Tanner, The spreading of silicone oil drops 2409 (1999).
on horizontal surfaces, J. Phys. D: Appl. Phys.
12, 1473 (1979). [41] B G Abdallah, A Ros, Surface coatings for
microfluidic-based biomedical devices, In: Mi-
[39] M Maleki, M Reyssat, F Restagno, D Quéré, crofluidic devices for biomedical applications,
C Clanet, Landau-Levich menisci, J. Colloid Eds. X Li, Y Zhou, Pag. 63, Woodhead Pub-
Interf. Sci. 354, 359 (2011). lishing, Cambridge (2013).

130006-16