Biomimetic wall-shaped hierarchical microstructure for gecko-like attachment of

b Most biological hairy adhesive systems involved in locomotion rely on spatula-shaped terminal elements, whose operation has been actively studied during the last decade. However, though functional principles underlying their amazing performance are now well understood, due to technical di ﬃ culties in manufacturing the complex structure of hierarchical spatulate systems, a biomimetic surface structure featuring true shear-induced dynamic attachment still remains elusive. To try bridging this gap, a novel method of manufacturing gecko-like attachment surfaces is devised based on a laser-micromachining technology. This method overcomes the inherent disadvantages of photolithography techniques and opens wide perspectives for future production of gecko-like attachment systems. Advanced smart-performance surfaces featuring thin- ﬁ lm-based hierarchical shear-activated elements are fabricated and found capable of generating friction force of several tens of times the contact load, which makes a signi ﬁ cant step forward towards a true gecko-like adhesive.


Introduction
Hairy (brillar) attachment systems of insects, arachnids and reptiles have been intensively studied during the last decade, 1 aiming to reveal and possibly utilize functional principles underlying their amazing dynamical adhesive performance. These systems consist of arrays of hairs (setae) with two or more levels of hierarchy, which allow for a large contact area on almost any surface and hence feature high adhesion and friction derived from a combination of molecular interaction and capillary attractive forces. [2][3][4][5][6][7] The topmost hierarchical level of seta is responsible for the formation of intimate contact with the substrate and shows up as one or more terminal elements involving thin lms. [8][9][10][11][12] In general, thin-lm elements can be subdivided into two main groups characterized by different appearance and function. The thin-lm elements of one type resemble mushroom caps (by protruding circumferentially from the seta stem) and are used for passive long-term attachment, such as in pairing, whereas the thin-lm elements of the other type resemble spatulae (by protruding from the seta stem in one direction only) and are used for muscle-activated shearinduced short-term attachment required in locomotion. 13 Based on the studies of different animal groups, an interesting correlation between the geometrical properties of setal tips and animal weight was found: the heavier the animal, the smaller and more densely packed the tips. 14 This scaling effect was explained by introducing the principle of contact splitting, according to which splitting up the contact into ner subcontacts increases adhesion, 15 which, based on its beauty and simplicity, has drawn an attention of many researchers. However, numerous attempts at employing this principle in manufacturing usable materials were surprisingly unsuccessful: in most cases, simple arrays of micropillars [16][17][18][19][20][21][22] have not exhibited a stronger adhesion than at controls made of the same materials. Furthermore, it eventually appeared that the role of highly exible terminal thin-lm elements as compliant contacting surfaces is critical 23 and successful dry adhesive can hardly be constructed without these elements being involved. 24 Earlier failures with simple micropillars have led many groups to start experimenting with thin-lm terminal elements, and rst truly working biomimetic adhesive was reported 25 aer replicating mushroom-shaped attachment setae evolved in male beetles from the family Chrysomelidae. Mushroom-shaped structures were relatively easy to be fabricated due to their symmetry and thorough studies of their various properties and abilities have followed, performed rst experimentally 21,26-41 and then theoretically. [42][43][44][45] However, though mushroom-shaped microstructure is well adapted for static passive applications such as glass safety coverings 46 or medical patches, 47 it is not able to withstand shear load, detach with zero load and respond directionally, which makes it not suitable for active dynamic short-term attachment required in many cases.
Considering the fact that most biological hairy attachment systems involved in locomotion actually rely on spatula-shaped terminal elements, 13 the spatulate systems were also studied, though the results achieved are mainly related to theoretical rationalizations. Preceded by pioneer work on peeling of thin-lms, 48 the principles underlying the spectacular performance of biological spatula-shaped brillar attachment systems are continuously uncovered, 7,12,[49][50][51][52][53][54][55][56][57][58][59] opening the way to further advance. However, due to technical difficulties in manufacturing the complex hierarchical structure of adhesiondriven spatulate systems, there were only a few attempts to study them experimentally [60][61][62][63][64] and the progress made is limited. To try bridging this gap, here we report on a manufacturing method and a novel smart-performance surface design featuring thin-lm hierarchical shear-activated elements capable of generating friction force of several tens of times the contact load, which makes a step forward towards a true geckolike adhesive.

Results and discussion
Fabrication A set of problems that had to be faced in order to manufacture thin-lm-based hierarchical shear-activated surface elements and the solutions we suggest are summarized in Table 1. The main points are these.
(1) As long as the counter surface is smooth, splitting the attachment pad in parallel to the horizontal component of the peeling force does not enhance the attachment ability, 57 so there is no need in pillar-based projections that were invariably used so far. Instead, we propose wall-shaped projections, whose lengths (dimension in perpendicular to the horizontal component of the peeling force) are much larger than their widths and heights. In addition to easier fabrication due to smaller number of contact elements, this also allows more efficient use of available contact area and longer total peeling line length, which was found to govern biological brillar adhesion. 57 (2) As long as so elastomers are used, thin lms are involved and shear motion is necessary to actuate the system, there is no need in making surface projections slanted, as they are exible enough to build good contact without much elastic energy being stored. This allows making them essentially perpendicular to the contact plane, which signicantly simplies fabrication. Interestingly, even in relatively stiff keratin-based biological attachment systems, terminal spatulate plates are oriented in perpendicular to the contact plane, 61,65 which conrms our approach.
(3) It is well known that biological adhesive hairs become thinner towards their ends, 10,12,66 which suggests that thickness gradients have to be built into the articial systems as well. Photolithography methods used widely in fabrication of biomimetic adhesives are, however, hardly capable of imparting such feature into the design of surface projections. To this end, here we suggest using laser machining in preparing casting templates, which allows creating gradients by adjusting the laser beam geometry. In addition to (a) ability of cutting depressions with a depth-dependent width, other advantages of using laser machining instead of photolithography methods are (b) ability to machine non-at templates, such as rolls that can be used in continuous mass production, (c) practical lack of limitation in treated area, (d) ability to machine thin metal sheets that are much more robust in handling than brittle silicon wafers, and (e) ability to obtain smaller surface roughness of deep cut walls.
(4) Given that in an unloaded wall-shaped projection the working surface is perpendicular and not parallel to the contact plane, we suggest making the template depressions as deep as possible and using fast-polymerizing elastomers for creating structured surfaces. This allows fabricating wall-shaped hierarchical projections of different heights using the same template by adjusting the dwell time between mixing and pouring the polymer, which solidies before reaching the bottom of the template.
(5) Given that there is no need to ll the template completely, it can consist of through slots instead of blind depressions. Openings at the back side of the template simplify the fabrication process signicantly by preventing air trapping during the template lling, easing the cast release and facilitating the template cleaning.
Based on the above guidelines, we have manufactured two types of wall-shaped hierarchical microstructures using two different laser-machined templates ( Fig. 1 and Table 2, see Experimental section for details). The main difference between the two types of projections is in the form of the projection base, whose section resembles a crown in one case and a triangle in the other case. In fact, the crown base was not planned originally and appeared as a laser-cutting artefact during a rst attempt to produce a casting template. The problem was resolved by adjusting the process parameters in the following attempts. Fig. 1d and f-h demonstrates that this approach indeed allows fabricating wall-shaped hierarchical microstructures with bases of gradually changing thickness and thin-lm terminal elements of different heights. Fig. 2 presents the behavior of wall-shaped hierarchical microstructures loaded in normal and tangential directions. Interestingly, both types of microstructures were found to be easily detachable and non-sticky by default, which corresponds well to  60 This is explained by the following reason. The surface projections are perpendicular to the contact plane and, when loaded in a normal direction only, they buckle randomly (Fig. 2a). Deformed elastically, they store elastic energy that is recovered during the surface withdrawal as reaction contact forces that act against adhesion and lead to negligible externally applied load known as pull-off force (Fig. 2d). This effect becomes even more pronounced due to a non-simultaneous detachment of the wall-shaped projections, which do not share the same height as a result of casting into a template without a bottom.

Measurements
The picture changes completely when the normally loaded projections are subjected to shear. Initially randomly buckled projections are gradually sheared to take similar shape under the growing tangential load. They stretch, overturn and eventually all get oriented in the same direction ( Fig. 2b and c), which allows them acting in concert against the external tangential load. This unites contributions of single thin-lm microstructures and makes them capable of generating stable and unusually high friction force of up to several tens of times the normal load at the onset of sliding (Fig. 2e). Withdrawing the wall-shaped microstructures in normal direction on completion of tangential motion, we have learned that simple detachment that follows shearing stage leads to increase in peeling angle and, in accord with analysis of biological attachment systems, 7 to negligible detachment force. Thus, in order to study the directional performance of the wall-shaped microstructure, it is rst necessary to enable application of peeling force at constant angle, which is planned for the future.
Maximum friction force obtained at the onset of sliding was determined for all tested surfaces. These data are presented in Fig. 3, where the performance of both types of wall-shaped hierarchical microstructures is shown. It can be seen that the slope of the friction force vs. normal load curve, which is the "true" friction coefficient, 67 is similar in all cases except that of the triangular-based microstructure bearing medium aps, which demonstrated different slope at the normal loads of less than 10 mN (shown as lled triangular markers). Interestingly, the above generic slope of the friction force vs. normal load curve characterizes not only the microstructured surfaces but   Crown  80  90  80  6  750  150  Triangular  70  50  25  12  Sample-size  95  ---40  12  -----70  11 --also the smooth reference (ring lying on its at side, see Experimental section for details), whose frictional performance correlates well with that reported earlier for the samples having the same geometry and made of the same material. 68 This points out that knowing friction coefficient is not enough to evaluate friction force, as it can be offset signicantly by changing the surface geometry.
Comparing the frictional performance of triangular-based microstructures, we can suggest that there is an optimum ap height with which a maximum friction force is obtained. This is explained as follows. Based on the peel-zone model developed for pressure-sensitive adhesives but applicable also to a dry adhesion case, 54 we can conclude that if a thin-lm ap is not high enough, the real contact area is narrower than a possible active peel zone (dened by a balance of elastic, adhesive and frictional energies) to allow high attachment force. On the other hand, if the ap is too high, then, when in contact with the mating surface, it can interfere with the neighbor projections once they are close enough, which again handicaps overall attachment (most likely it was also the case of the crown-based microstructure that had the highest aps). Somewhere in between the two extreme cases, however, there is an optimum in the ap height, which should be dened by the peel zone width and the surface projections pitch. This assumption is yet to be further veried.  Coming back to the slope of the friction vs. load curve obtained with the medium aps of the triangular-based microstructure at the normal loads of less than 10 mN (see lled triangular markers in Fig. 3), we can propose the following possible reason for it being signicantly different from those obtained under other conditions. To allow for self-aligning required to bring at surfaces in complete contact, both specimens are xed on tense orthogonal threads 68 that are twisted when the surfaces coming in contact are not parallel, which necessitates a certain load. Most likely, in the discussed case the surfaces were initially not parallel, which resulted in that not all wall-shaped hierarchical projections yet came in contact at normal loads of less than 10 mN and hence led to low friction. Thus, the observed result is identied as a measurement artefact.
It is also worth trying to evaluate the progress of the present work, including the manufacturing method itself and the frictional performance of the surfaces made, in comparison to the previous studies on spatulate attachment microstructures. [60][61][62][63][64] In template preparation, the present technique seems to be easier as it is based on a straightforward two-step laser machining (rst, cutting blind trenches and, then, cutting through slots at the trench bottoms) of a commercially available metal sheet that can be used for a virtually innite number of casts. The photolithography techniques used previously are more complicated as they are based on several-step procedures including various materials and chemicals, and most of the templates are limited to about 10 casts. In casting the microstructured surfaces, the present technique also seems to be easier as it allows working in air at room temperature, as long as the photolithography-based techniques require casting under vacuum and heat curing.
Comparison of frictional performance is not so simple, as in most cases the test conditions differ signicantly. However, given that the largest ratio of friction force to normal load is obtained at light loads, the present microstructure performs seemingly better with the friction to load ratio of 50 (Fig. 2e) as opposed to that of 3 O 16 obtained with previously reported spatulate microstructures having architectures of wedge-shaped 60,61,64 and ap-shaped pillars. 62,63 Direct comparison is possible only with ref. 64, where a friction force of 400 mN was obtained with a 12.6 mm 2 area sample under the normal load of 25 mN. The present study reports a friction force of 200 mN obtained with a 3.14 mm 2 area sample under the same normal load, which gives a two-fold better result.

Conclusion
We have devised a novel method of manufacturing gecko-like hierarchical shear-activated attachment surfaces based on laser micromachining. This method overcomes the inherent disadvantages of photolithography techniques and opens wide perspectives for future production of low-cost gecko-like attachment systems. Advanced surfaces we have fabricated are found capable of generating friction force of several tens of times the contact load, while exhibiting optimum geometrical properties presumably determined by interrelations between the height and pitch of surface elements. This presents a further progress in comparison to the previously reported achievements and makes a signicant step forward towards a true gecko-like adhesive.

Specimens, preparations and conditions
Four types of hierarchical microstructured surfaces were cast from poly(vinylsiloxane) (PVS; Coltène Whaledent AG, Altstätten, Switzerland) using two negative templates produced by laser micro-machining (Oxford Lasers, Oxon, UK) from 0.15 mm-thick tungsten sheets to have through slots of different sizes. To allow easier cast release, the tungsten templates were oxidized in Ozone PSD-II Probe Cleaner (Novascan Technologies, Ames, IA) for 30 min. The height of individual surface projections was controlled by the dwell time of 30, 60 and 90 s between mixing and pouring the PVS onto the template, while the curing time of PVS is about 3 min when its two components are stored at room temperature before mixing. The cast thickness was controlled by using spacers between the template and a covering at surface. The cast release was done in a water bath aer a 30 min stay in ultrasonic cleaner. Microstructured side of the cast was used for contact surface of bio-inspired specimens, while smooth at side of the same cast was used for contact surface of reference specimens. Structured and reference smooth samples of 2 mm in diameter were punched out of the same 1 mm-height PVS cast (Young's modulus of about 3 MPa) 19 using disposable biopsy Uni-Punch (Premier Products Co., Plymouth Meeting, Pennsylvania). Based on that the real contact area of structured samples is much smaller than the apparent area of the sample (Fig. 2c), it was decided to have comparable real contact area in both structured and reference smooth samples. To this purpose, the center area of the smooth sample (disk of 2 mm in diameter and 1 mm in height) was removed to form a ring of outer diameter of 2 mm, inner diameter of 1 mm and height of about 0.2 mm. The central depression was rst cut with a 1 mm-in-diameter punch and then removed with a sharp knife. Contact area of structured and reference smooth samples had roughness average, R a , of 760 and 541 nm, respectively. The structured PVS samples were mounted in such a way as to orient the surface projections in perpendicular to the sliding direction. The tests were performed in contact with smooth (R a ¼ 5 nm) glass slide of 30 Â 5 Â 1 mm 3 in size. The temperature and relative humidity in the laboratory were 20-22 C and 40-50%, respectively. The pressure inside the scanning electron microscope (SEM) vacuum chamber was 300 Pa.

Equipment
Surface appearance of the specimens used was inspected with an optical stereomicroscope Leica M125 (Leica GmbH, Wetzlar, Germany) and imaged in a FEI Quanta 200 environmental SEM (FEI Co., Brno, Czech Republic). The tests were performed on a home-made tribometer 69 capable of operating inside the environmental SEM to enable charge-free imaging of non-conductive uncoated PVS specimens. The tribometer (Fig. 4) incorporates two main units used for driving and measuring purposes. The drive unit consists of three motorized translation stages used to load the contact by moving the glass specimen. The measurement unit consists of two load cells used to determine the forces acting on the PVS specimen. The load cells are xed in that way that normal force acts in the plane in which tangential load cell is not sensitive, and tangential force acts in the plane in which normal load cell is not sensitive, thus preventing cross-talk between force sensors (Fig. 4). To guarantee full contact and full the 'equal load sharing' principle 18 during force measurements in a at-on-at contact scheme essential in surface texture testing, a passive self-aligning system of specimen holders was used. 68 In order to examine the microstructure behavior in contact, the contact area was imaged with a monochrome digital camera UI-1240LE (IDS Imaging Development Systems, Germany) enhanced by high-magnication optics Zoom-12X (Navitar Inc., Rochester, New York).

Procedure
All specimens were tested in the following way. Adhesion tests were done by bringing the glass specimen in contact with the PVS specimen and then, aer applying a normal load of 80 mN, withdrawing the translation stage in the normal direction at a velocity of 100 mm s À1 while measuring the pull-off force. In friction tests, aer bringing the specimens in contact and applying a normal load chosen between 2 and 80 mN, the translation stage was moved in the lateral direction at a velocity of 100 mm s À1 for a distance of 1200 mm with the normal load being kept constant, while the friction force resisting the specimen motion was recorded. Immediately on completion of lateral motion, the translation stage was withdrawn in a normal direction at a velocity of 100 mm s À1 and the pull-off force affected by shearing was also measured.
During the experiments carried out inside SEM, the translation stages were moved incrementally and, following each incremental displacement, the contact projections were imaged with SEM to examine visually their gradual deformation. Each PVS specimen was tested at every load on a different region of the glass specimen and then replaced. Prior to experiments, the specimens were washed with deionized water and liquid soap, and then dried in blowing nitrogen.