Professional Documents
Culture Documents
1. Introduction
Many animals have evolved adhesive organs on their feet enabling them to
Author for correspondence: strongly attach to and quickly detach from nearly any kind of surface. These
Marleen Kamperman organs come in two basic designs: (i) pads densely covered with micro- or nano-
e-mail: marleen.kamperman@wur.nl sized fibrils (hairs) with a wide variety of tip shapes, observed in, for example,
spiders, geckos, beetles and flies, and (ii) pads with macroscopically smooth
surface profiles, observed in, for example, tree frogs, ants and grasshoppers [1].
Regardless of the design, both types use non-covalent surface forces to achieve
adhesion, and research suggests that they rely primarily on van der Waals
forces [2]. The surface structure, not the chemistry, is therefore dominating the
function of natural adhesive systems.
Over the last decade, synthetic patterned surfaces have been developed to
mimic the geometry of natural attachment systems [3,4]. Initial research focused
on the development of vertical pillars with diameters in the micrometre range
[5 7]. To fabricate more complex structures, closer in design and performance
to natural systems, relatively expensive semiconductor technologies and multi-
step patterning techniques have been used [8 13]. Alternatively, wrinkling
techniques have been explored as a scalable, potentially cheaper production
method. Wrinkled surfaces have been shown to increase or decrease adhesion
depending on the testing geometry and wrinkle dimensions [1418]. Ideally,
however, scalable methods should not be limited to two-dimensional patterns,
as more complex, three-dimensional sub-micrometre structures would enable
Electronic supplementary material is available the fabrication of surfaces with different property profiles [19].
In this study, we describe a novel approach based on colloidal lithography
at http://dx.doi.org/10.1098/rsif.2014.1061 or
to make bioinspired, patterned, dry adhesives. Colloidal lithography is a techni-
via http://rsif.royalsocietypublishing.org. que that uses self-assembled colloidal monolayers for fabricating patterned
& 2014 The Author(s) Published by the Royal Society. All rights reserved.
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
surfaces with nanometre precision. This technique has proved The product was filtered over glass wool to discard 2
to be a fast and cost-effective alternative to other nanofabrica- some coagulum that was formed during the reaction. The
rsif.royalsocietypublishing.org
tion processes such as electron beam lithography or focused filtered particles were treated using two different protocols.
ion beam lithography [20]. Recently, several groups have Colloidal dispersion CP (carboxylated particles) was directly
reported methods to produce large-area, highly ordered, col- exchanged to ethanol by three cycles of centrifugation at
loidal crystals. For example, Pan et al. [21] introduced a 4500 r.p.m. for 1 h followed by resuspension in ethanol.
dynamic interface, by stirring the aqueous subphase, to facili- Colloidal dispersion CP-W (carboxylated particles
tate colloidal crystal formation at the air/water interface. washed) was washed in water by three cycles of centrifugation
Retsch et al. [22] reported a two-step procedure in which at 4500 r.p.m. for 1 h followed by resuspension in Milli-Q
colloids are applied to a glass slide by spin-coating. water. After this washing step, CP-W was exchanged to
Subsequently, a colloidal monolayer was formed upon immer- ethanol, using the same procedure. The smaller particles
sion of the glass slide in an aqueous surfactant solution [22]. were treated similarly to CP-W.
(a) (b) 3
(i)
rsif.royalsocietypublishing.org
water polystyrene
(ii)
itaconic
acid
(c)
(iii)
PDMS
water
(iv)
PDMS
Figure 1. Fabrication of nanopatterned PDMS using colloidal lithography directly at the air/water interface. (a) Schematic outline of the experimental procedure.
(i) Spreading of the particles using a glass slide. (ii) Application of the PDMS against the side of the Petri dish. (iii) Cross-linking of the PDMS. (iv) Nanopatterned
PDMS prepared by extraction of the particles. (b) Sketch of the charge distribution on a particle when submerged in water at sufficient pH. (c) Photograph of a
PDMS sample with a colloidal monolayer. Note the clear PDMS spot in the upper left corner, where the PDMS first contacted the water, thereby pushing the particles
away. (Online version in colour.)
allowed to cross-link at room temperature for 2 days, resulting exactly six nearest neighbours in these hexagonal lattices.
in PDMS layers of 0.20.6 mm thickness (figure 1a(iii)). This method of assessing crystallinity is very strict: for
To remove the template particles from the PDMS example, all six particles surrounding a vacancy are not
mould, we first applied Scotch Magic Tape, which removed regarded as crystalline, while in fact they are part of the crystal
the largest portion of templating particles. The remain- lattice. However, assessing the crystallinity by using the more
ing particles were removed by soaking the sample in elaborate Steinhardt orientational bond order parameter, C6
N-methyl-2-pyrrolidone (NMP) for 1 h, followed by dipp- [30], showed the same trends as simply counting the number
ing the samples 30 times in NMP in an ultrasonic bath of nearest neighbours [31]. Therefore, the nearest neighbour
(figure 1a(iv)). NMP was chosen as the solvent because it dis- method was used here.
solves polystyrene, yet does not swell PDMS. The samples
were stored at room temperature.
2.6. Immersion depth determination
We put Petri dishes with monolayers in a desiccator, added
2.5. Crystallinity determination some droplets of ethyl cyanoacrylate in a separate dish and let
Monolayers were scooped from the air/water interface with the system react under reduced pressure for 2 h. Ethyl cyano-
wet, piranha-cleaned coverslips. The area of the monolayers acrylate starts to polymerize when it comes into contact
on these coverslips ranged between 0.5 and 2 cm2. Twenty stat- with water. The apparent radii of the particles embedded in
istically independent images were acquired with bright field poly(ethyl cyanoacrylate) was determined from scanning elec-
microscopy at random locations across each sample. The cen- tron microscopy (SEM) images (6878 particles per image)
troid positions of all particles in the images were determined using the Analyze Particles function in IMAGEJ. See the electronic
using standard particle-tracking algorithms [29] with a lateral supplementary material, figure S3a, for a definition of apparent
resolution of approximately 20 nm. Approximately 10 000 par- radius. Six two-tailed t-tests were performed to test whether the
ticles were found per image; per experiment we thus analysed differences in apparent radii were significant (with p , 0.001)
20 10 000 200 000 particles to assess the crystallinity of the and the Bonferroni method was used to correct for multiple test-
monolayers. We then computed the number of nearest neigh- ing. This apparent radius was used to calculate the immersion
bours, using a cut-off distance of 1.25 the particle diameter depth, which is the height percentage of the particle that is
(for Matlab code, see the electronic supplementary material, immersed in the subphase (see the electronic supplementary
S7). We considered a particle crystalline only when it had material, figure S3b).
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
4
glass surface
parallel
rsif.royalsocietypublishing.org
perpendicular XYZ
or normal spherical probe strain gauges translational
stage
sample
pH 3 pH 7 pH 10 5
(a) (b) (c)
rsif.royalsocietypublishing.org
10 m 10 m 10 m
10 m 10 m 10 m
Figure 3. Light microscopy images of CP-W (for a definition, see text) colloidal monolayers self-assembled onto water at (a) pH 3, (b) pH 7 and (c) pH 10.
Crystallinity diagrams obtained from light microscopy images at (d ) pH 3, (e) pH 7 and (f ) pH 10. Particles in light grey have six neighbouring particles and
are considered crystalline, while particles in red (dark grey in printed version) do not have six neighbouring particles. SEM images of colloidal monolayers embedded
in PDMS at (g) pH 3, (h) pH 7 and (i) pH 10. (Online version in colour.)
and is a well-characterized material. PDMS completely wet analysis [41], Fourier transformation [24,42] and microscopy
the air/water interface [36], and thus formed a homogeneous to see grain sizes [43]. In these methods, subtle differences in
film at the interface. During spreading of the initial PDMS dro- order are not easily picked up, while our approach allowed
plet, the particles were pushed together, similar to compression for straightforward comparison of the crystallinity of different
in a Langmuir trough. This resulted in close-packed mono- monolayers.
layers composed of the islands formed during deposition. Figure 3gi shows SEM images of the monolayer
A representative example of nanopatterned PDMS is shown embedded in PDMS. We found that PDMS did not disrupt
in figure 1c, in which the Bragg reflections can be clearly seen. the monolayer, as a similar order was observed in light
The crystallinity of the monolayers was assessed by microscopy and SEM. Figure 4 presents SEM images of the
scooping monolayers, which were not cast in PDMS, and resulting nanopatterned PDMS surfaces after particle removal.
imaging them with brightfield microscopy (figure 3ac). Colloidal monolayers were obtained for a range of particle
We confirmed the presence of a monolayer, rather than multi- sizes, which resulted in nanostructured PDMS with different
layers, by changing the focus and thus scanning through feature sizes (electronic supplementary material, figure S2).
the sample. The positions of all particles in the images were These results indicate that our method allows direct control
found using a particle-tracking algorithm, and this spatial over the topography of PDMS by tuning the properties of the
information was used to quantify the number of crystalline colloidal monolayer.
particles, i.e. particles with exactly six nearest neighbours
(see Experimental section and the electronic supplementary
material, S7). An exemplary result of this analysis is shown
3.2. Influence of subphase
Figure 3 clearly shows that the degree of crystallinity of the
in figure 3d f, where crystalline particles are shown in red
colloidal monolayer was strongly dependent on the pH of
(dark grey in printed version), and non-crystalline particles
the subphase. The sample at pH 3 (figure 3a) was largely
are shown in light grey. The normalized crystallinity per
disordered, while large domain sizes were obtained for
image was then calculated using
samples prepared at pH 10 (figure 3c). Samples prepared
crystalline particles at intermediate pH value ( pH 7) exhibited intermediate
normalized crystallinity : (3:1) order (figure 3b). We then systematically investigated the
total number of particles
degree of order of the samples as a function of the compo-
sition of the subphase for two types of particle dispersions,
This quantitative assessment of the crystallinity of the CP and CP-W (figure 5). The CP dispersion contained
monolayer proved insightful to study the effects of subphase residual styreneitaconic acid oligomers, whereas the CP-W
characteristics on the ordering of particles. Common, more dispersion was thoroughly washed with ultrapure water to
qualitative methods include visual inspection of micrographs remove these oligomers to a great extent. To study the pH
or iridescence of the monolayer [22,26,3739], observation of effect on monolayer formation, the pH was adjusted with
moire patterns upon thin film folding [40], diffraction pattern HCl and NaOH, while keeping the ionic strength constant.
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
(b) 6
(a)
rsif.royalsocietypublishing.org
10 m 10 m
10 m 100 m
Figure 4. SEM images of PDMS replicas of colloidal monolayers self-assembled onto water at (a) pH 3, (b) pH 7 and (c,d ) pH 10.
0.8
normalized crystallinity
0.6
0.4
0.2
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
pH pH pH
Figure 5. Degree of order as a function of subphase composition. (a) Normalized crystallinity as a function of pH for CP (triangle) and CP-W (diamond) at [salt]
0.1 M. Every point represents a different sample and standard deviations were obtained from crystallinity determination at 20 different positions within a sample.
(b) Normalized crystallinity as a function of pH for CP at [salt] 0.1 M. Solid symbols denote samples for which the pH of the subphase was constant during
processing. Open symbols denote samples for which the pH of the subphase was kept at pH 10 during self-assembly and adjusted afterwards. (c) Normalized
crystallinity as a function of pH for CP-W at [salt] 1023 M. Solid symbols denote samples for which the pH of the subphase was constant during processing.
Open symbols denote samples for which the pH of the subphase was kept at pH 10 during self-assembly and adjusted afterwards.
The pH was calculated from the known composition of the particles. The increased electrostatic repulsion counteracts
subphase and not determined experimentally, but we the attractive van der Waals and capillary forces. Consequen-
expect no strong deviations between calculated and actual tly, the particles remained more mobile at the interface,
values due to the low particle concentration in the overall resulting in more ordered monolayers [20,22,23]. Thus, at
system. The crystallinity was again determined by scooping pH 3, particles that approach each other closely (i.e. by capil-
monolayers and subsequent light microscopy imaging. lary forces) immediately feel the attractive van der Waals
Figure 5a shows that the crystallinity increased with forces that bring them in close contact. This process immobi-
increasing pH for both CP and CP-W. At higher pH, the lizes the particles and results in monolayers with relatively
charge density at the particle surface was increased. Itaconic low crystallinities. At pH 10, the electrostatic repulsion is
acid groups contain two carboxyl groups, with pKa1 3.85 greatly increased. This increases the energy barrier for a
and pKa2 5.44 [28]. The onset of increasing crystallinity close contact between particles and the particles have suffi-
indeed coincided with dissociation of the first carboxyl cient time and mobility to crystallize in a low free energy
group at pH higher than 4. The increased charge density position. Hence, monolayers with higher crystallinities are
resulted in larger electrostatic repulsion between the obtained [23].
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
It was observed that monolayers made from CP were (a) Milli-Q water (b) pH = 3, [salt] = 103 M 7
lessmobile and mechanically more stable than CP-W mono-
rsif.royalsocietypublishing.org
layers. More specifically, when part of the CP monolayer
was scooped with a coverslip, the hole left in the monolayer
remained square, whereas for CP-W the particles within the
monolayer rearranged after scooping, leaving an irregular-
1 m 1 m
shaped hole in the monolayer. We speculate that the surface
active styrene itaconic acid oligomers left in the CP disper- (c) pH = 10, [salt] = 103 M (d ) [NaCl] = 0.1 M
sions glue the particles together, in a similar fashion as
was observed by Ho et al. [44] for polyethylene oxide poly-
mers. To study the difference in mobility, CP and CP-W
samples were prepared at high pH, which resulted in well-
rsif.royalsocietypublishing.org
500 86 nm
10 S1
400
1
arb. units
300 135 nm
nm
S2
200 0.1
100 S3
0.01 173 nm
0
100 0.001
0 0.5 1.0 1.5 2.0 250 200 150 100 50 0 50 100 150
0.4 1.00
1.20
1.15
0 0.90
force (mN)
0.4 0.80
1.05
0.6 0.75
0.70 1.00
0.8
Figure 7. Different (nanopatterned) surface structures and adhesion thereof. (a d) (circle) Smooth PDMS; (diamond) sample 1; (triangle) sample 2; and (square)
sample 3. (a) Cross section from AFM image for smooth PDMS and S1 S3. (b) Histogram of height differences from an AFM picture for smooth PDMS and S1 S3.
(c) Representative force displacement curve of nanopatterned PDMS. Preload is the maximum on the y-axis (5 mN), and pull-off force is the minimum on the y-axis
(0.92 mN). Note the loading curve consists of only 14 data points due to the quick approach. (d ) Pull-off force of smooth PDMS and S1 S3 as a function of preload
with 95% confidence intervals. (e) Normalized pull-off force against normalized actual area for smooth PDMS and S1 S3 with 95% confidence intervals.
l
a
sample subphase crystallinity a a (nm) l (nm)
a
Determined by AFM.
approximated by hemispherical dimples separated by flat relatively large inter-dimple regions to obtain flat inter-
inter-dimple regions, with a as the dimple depth and l as dimple regions. As expected, upon decreasing the pH, the
the dimple wavelength. The wavelength l is defined as the dimple depth increased from sample 1 (S1; dark grey), to
diameter of the particles for a cross section through the sample 2 (S2; grey) and to sample 3 (S3; light grey) (see
centre of the particles. Actual average wavelengths of the also the electronic supplementary material, figure S4).
samples are a function of the degree of crystallinity and Second, histograms of the height distributions of the three
will deviate slightly from this value. Dimple depths were nanopatterned samples and smooth PDMS were determined
determined from the histogram in figure 7b. by AFM and show that the same trend in depth was also
As expected, the normalized crystallinity differed found over larger areas (5 5 mm). The histogram for
between the three samples and decreased with decreasing smooth PDMS (black) peaks around the normalized zero
pH value of the subphase: 54% for S1, 35% for S2 and 14% (see Experimental section for details) and extends slightly
for S3. Apart from the crystallinity, the dimple depth also to the right, indicating the presence of some small bumps
differed between the three samples, which was determined or dirt particles (figure 7b). The histograms of the nanopat-
by AFM. First, a cross section through the centre of a terned structures show a broad distribution and extend to
dimple of each of these samples was drawn, as shown in the left, indicating dimples. These data were used to
figure 7a. Profiles were created by selecting dimples with determine dimple depths.
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
Force displacement curves were obtained by moving a Substituting parameter values as obtained by Jin et al. [49] 9
smooth spherical probe (of 4.76 mm diameter) against the for PDMS with similar material properties, i.e. E* 1.5 MPa
and Wad 0.163 J m22, we obtain a maximum dimple depth
rsif.royalsocietypublishing.org
sample to a defined preload. Subsequently, the probe was
retracted with a constant velocity of 1 mm s21 until the final of 0.6 mm, which is a larger value than the depths of our
detachment event occurred giving the pull-off force, which samples. Thus, the dimples in this study are most likely to
is a measure of the adhesion performance. A representative be shallow enough to spontaneously jump to contact.
force displacement curve is given in figure 7c. At least Although theoretical considerations suggest that complete
five series of nine force displacement curves with nine contact formation may occur for our dimple dimensions, in
different preloads were measured on at least five different practice air can be trapped in the dimples and prevent complete
areas per sample. The example in the electronic supplemen- contact from being achieved. A difference between our
tary material, figure S5, shows that the pull-off force did dimpled structures and wrinkles or an array of pillars is that
not change within the series. This indicates that we did dimples are isolated and air can get locked in. However,
(a) (b) 10
15
rsif.royalsocietypublishing.org
10
friction (mN)
and the elastic penalty upon compression. These two effects PDMS resulted in pronounced stickslip motion, patterned
counteract each other. A deeper dimple results in more surfaces with vastly different patterns did not show stickslip
energy dissipation [53]. However, the larger the dimple instabilities and showed decreased friction [55,56]. Lorenz &
depth, the more elastic energy is stored due to surface defor- Persson [57] explained the disappearance of stickslip motion
mation to achieve complete contact. This stored elastic energy for patterned surfaces by the occurrence of sequential interfacial
is (partly) recovered and may help to break interfacial bonds. slip. They showed that surface patterns with millimetre-sized
As all three structured samples show enhancement of adhesion surface features will reduce the elastic coupling between the
compared with the flat control, the energy dissipation through different surface features, which can cause these features to
unstable detachment outweighs the elastic penalty, and the net move more independently of each other than surface patches
effect of one dimple is an increase in adhesion. on the flat homogeneous surface [57]. This reduces the tendency
In addition to the dimple depth, the dimple density will also for macroscopic stickslip motion. Whereas our dimples are
affect the adhesion. More specifically, a higher density may relatively small and shallow they still result in a heterogeneous
have a larger influence on adhesion. For our samples, the stress distribution and a decrease in lateral stiffness, as
dimple density can be assessed by the crystallinity. A higher explained by Rand & Crosby [58]. This decreases the friction.
(normalized) crystallinity stems from a more efficient packing
and thus a higher dimple density. Indeed, if we count all par-
ticles in figure 3df and compare S1 ( pH 10, normalized
crystallinity 0.54) and S2 (pH 5, normalized crystallinity 0.35),
4. Summary
In summary, we developed a scalable method to fabricate
a decrease in the number of particles of 6% is found. S3 (pH 3,
dimpled dry adhesives. Particles formed close-packed islands
normalized crystallinity 0.13) shows a decrease of 13%
at the air/water interface that transitioned to well-ordered
compared with S2.
monolayers by adding PDMS. Colloidal monolayers at the
In our experiments, both the dimple depth and the
air/water interface were used directly as a template for
dimple density are varied for each sample. We find
PDMS. The composition of the subphase was altered to tune
the largest increase in adhesion for the sample with interme-
the order and immersion depth of the monolayer, which was
diate dimple density and depth. The increase in adhesion per
translated into the PDMS nanopatterned surfaces. These sur-
dimple should therefore be higher with increasing depth in
faces showed an increase in adhesion, which is attributed to
this range of depths, but the decrease in the number of
an energy-dissipating mechanism during pull-off, and a
dimples with increasing depth counters this effect.
decrease in friction. These results show that interfacial templat-
ing of colloidal particles provides a promising route for creating
3.5. Friction functional surfaces of which the three-dimensional topology
The friction of the samples was measured by lowering the can be tuned without elaborate and energy-consuming litho-
probe onto the sample, applying a defined constant normal graphic methods. With the emergence of a wide variety of
load, and subsequently moving the probe in plane with new methods to tailor the size and shape of colloids, and
100 mm s21 for 20 s. The static friction of flat PDMS was too their organization at liquid interfaces, this opens a wealth of
high (more than 70 mN) to be measured with our set-up. For opportunities for lithography-free design of new adhesive and
the nanopatterned samples, the shear force first rose up to a low-frictional materials.
peak (static friction), and subsequently dropped to a nearly
constant value (dynamic friction), which indicates a sliding Acknowledgements. S.A. would like to thank Elma Vroege for her work on
monolayers as templates during her Bachelors thesis and Marcel Gies-
frictional mechanism (figure 8a) [55]. Figure 8b shows that, com- bers for his help with Auger experiments and SEM at such short notice.
pared with flat surfaces, friction was decreased by the patterns. Funding statement. M.K., J.A., J.S. and S.A. acknowledge the Nether-
This is in agreement with several previous publications on lands Organization for Scientific Research (NWO) for financial
patterned elastomers: whereas friction measurements on flat support.
Downloaded from http://rsif.royalsocietypublishing.org/ on May 5, 2016
References 11
rsif.royalsocietypublishing.org
1. Gorb S. 2001 Attachment devices of insect cuticle. 16. Davis CS, Martina D, Creton C, Lindner A, Crosby AJ. three dimensions. Opt. Expr. 17, 46854704.
New York, NY: Springer. 2012 Enhanced adhesion of elastic materials to (doi:10.1364/OE.17.004685)
2. Loskill P, Puthoff J, Wilkinson M, Mecke K, Jacobs K, small-scale wrinkles. Langmuir 28, 14 899 14 908. 30. Steinhardt PJ, Nelson DR, Ronchetti M. 1983 Bond-
Autumn K. 2013 Macroscale adhesion of gecko (doi:10.1021/la302314z) orientational order in liquids and glasses. Phys. Rev.
setae reflects nanoscale differences in subsurface 17. Lin P-C, Vajpayee S, Jagota A, Hui C-Y, Yang S. 2008 B 28, 784. (doi:10.1103/PhysRevB.28.784)
composition. J. R. Soc. Interface 10, 20120587. Mechanically tunable dry adhesive from wrinkled 31. Higler R, Appel J, Sprakel J. 2013 Substitutional
(doi:10.1098/rsif.2012.0587) elastomers. Soft Matter 4, 1830 1835. (doi:10. impurity-induced vitrification in microgel crystals. Soft
3. Kamperman M, Kroner E, del Campo A, McMeeking 1039/b802848f ) Matter 9, 53725379. (doi:10.1039/c3sm50471a)
RM, Arzt E. 2010 Functional adhesive surfaces with 18. Davis CS, Crosby AJ. 2011 Mechanics of wrinkled 32. Drechsler P, Federle W. 2006 Biomechanics of
45. Lyklema J, van Leeuwen HP, Vliet M, Cazabat A-M. to full contact. Soft Matter 7, 10 72810 736. Phys. Solids 55, 473 488. (doi:10.1016/j.jmps. 12
2005 Fundamentals of interface and colloid science. (doi:10.1039/c1sm06367g) 2006.09.007)
Degrandi-Contraires E, Beaumont A, Restagno F,
rsif.royalsocietypublishing.org
New York, NY: Academic Press. 50. 55. Rand CJ, Crosby AJ. 2009 Friction of soft elastomeric
46. Fuller K, Tabor D. 1975 The effect of surface Weil R, Poulard C, Leger L. 2013 CassieWenzel- wrinkled surfaces. J. Appl. Phys. 106, 064913
roughness on the adhesion of elastic solids. like transition in patterned soft elastomer adhesive 064914. (doi:10.1063/1.3226074)
Proc. R. Soc. Lond. A 345, 327 342. (doi:10.1098/ contacts. Europhys. Lett. 101, 14001. (doi:10.1209/ 56. Varenberg M, Gorb SN. 2009 Hexagonal
rspa.1975.0138) 0295-5075/101/14001) surface micropattern for dry and wet friction.
47. Persson B, Tosatti E. 2001 The effect of 51. McMeeking RM, Ma L, Arzt E. 2010 Bi-stable adhesion Adv. Mater. 21, 483 486. (doi:10.1002/adma.
surface roughness on the adhesion of elastic of a surface with a dimple. Adv. Eng. Mater. 12, 200802734)
solids. J. Chem. Phys. 115, 5597. (doi:10.1063/ 389397. (doi:10.1002/adem.201000091) 57. Lorenz B, Persson B. 2012 On the origin of why
1.1398300) 52. ChangaKim Y. 2008 Analysis of pressure-driven air static or breakloose friction is larger than kinetic
48. Briggs G, Briscoe B. 1977 The effect of surface bubble elimination in a microfluidic device. Lab Chip friction, and how to reduce it: the role of aging,