1 Hydrodynamic function of shark skin and two

2 biomimetic applications
3

4
5
6 J. Oeffner and G. V. Lauder*
7
8
9
10 *Museum of Comparative Zoology, Harvard University, 26 Oxford Street, Cambridge, MA 02138,
11 USA
12
13
14 *Author for correspondence (email: [email protected])
15
16 Please address all correspondence to:
17 George V. Lauder
18 Museum of Comparative Zoology
19 Harvard University
20 26 Oxford Street
21 Cambridge, MA 02138, USA
22
23 [email protected]
24
25
26
27 Keywords: Shark skin, locomotion, riblet, drag reduction, foil, swimming, Fastskin
28
29 Running head: Function of shark skin
30

1
31 SUMMARY
32 It has long been suspected that the denticles on shark skin reduce hydrodynamic drag during
33 locomotion, and a number of man-made materials have been produced that purport to use shark
34 skin-like surface roughness to reduce drag during swimming. But no studies to date have tested
35 these claims of drag reduction under dynamic and controlled conditions in which the swimming
36 speed and hydrodynamics of shark skin and skin-like materials can be quantitatively compared with
37 that of controls lacking surface ornamentation or with surfaces in different orientations. We use a
38 flapping foil robotic device which allows accurate determination of the self-propelled swimming
39 speed of both rigid and flexible membrane-like foils made of shark skin and two biomimetic models
40 of shark skin to measure locomotor performance. We studied the self-propelled swimming speed of
41 real shark skin, a silicone riblet material with evenly spaced ridges, and Speedo “shark skin-like”
42 swimsuit fabric attached to both rigid flat plate foils and made into flexible membrane-like foils.
43 We found no consistent increase in swimming speed with Speedo fabric, a 7.2% increase with riblet
44 material, and shark skin membranes (but not rigid shark skin plates) showed a mean 12.3% increase
45 in swimming speed compared to the same skin foils after removing the denticles. Deformation of
46 the shark skin membrane is thus critical to the drag reducing effect of surface denticles. Digital
47 particle image velocimetry of the flow field surrounding moving shark skin foils shows that skin
48 denticles promote enhanced leading edge suction which may have contributed to the observed
49 increase in swimming speed. Shark skin denticles may thus enhance thrust, as well as reduce drag.
50
51

2
52

53 INTRODUCTION
54 In recent decades the skin of sharks has achieved a certain biomimetic status among both
55 science popularizers and in research circles for the notion that the specialized skin surface
56 structure could reduce drag and enhance the efficiency of locomotion. Manufactured body suits
57 have been loosely modeled on shark skin with various ridges and dents to induce surface
58 roughness that purportedly enhance swimming performance in humans, and researchers have
59 long suspected that the special surface structure of shark skin contributes to the efficiency of
60 locomotion (e.g., Applegate, 1967; Lang et al., 2008; and Reif, 1982, 1985, who provides a
61 comprehensive overview of shark skin structure; also see images in Castro, 2011).
62 A variety of “shark-inspired” engineered materials have also been produced to reduce drag
63 when applied to the surface of submerged bodies. For example, riblets are fine rib-like surface
64 geometries with sharp surface ridges which can be aligned either parallel or perpendicular to the
65 flow direction and may reduce drag. A diversity of riblet shapes and sizes have been
66 investigated experimentally and theoretically (Bechert and Bartenwerfer, 1989; Bechert et al.,
67 2000; Bechert et al., 1997; Büttner and Schulz, 2011; Koeltzsch et al., 2002; Luchini et al., 1991;
68 Luchini and Trombetta, 1995; Neumann and Dinkelacker, 1991) and drag reduction of stiff
69 bodies covered with riblet material has been shown to occur (Bechert et al., 1997; Bechert et al.,
70 1985; Dinkelacker et al., 1987). Experiments with an adjustable surface with longitudinal blade
71 ribs and slits revealed the highest stiff-body drag reduction of 9.9% with a groove depth of half
72 the size of lateral riblet spacing (Bechert et al., 1997). Scalloped riblets, somewhat similar to the
73 ridges in shark denticles, produce a maximal stiff body drag reduction of about 7% (Bechert et
74 al., 1985).
75 A silicone-replica of the skin of Carcharhinus brachyurus attached to a rigid flat plate
76 resulted in a drag reduction of 5.2 - 8.3% compared to smooth silicone on a flat plate (Han et al.,
77 2008). A hard plastic shark skin replica achieved a drag reduction of 3% (Bechert et al., 1985).
78 But these cases involved study of a rigid body covered with a biomimetic skin, which is not the
79 situation for a shark in vivo where body undulations may greatly alter the structure of surface
80 ornamentation and change flow characteristics over the skin. In addition, a variety of tests with
81 the Speedo FS II swimsuit “shark-like” material resulted in a 7.7% (Benjanuvatra et al., 2002)

1
 82 and 10-15% (Mollendorf et al., 2004) reduction in the stiff-body drag compared to normal
83 swimsuits under certain conditions, but other studies or tests showed no significant drag
84 reduction (Benjanuvatra et al., 2002; Toussaint et al., 2002).
85 Because sharks are self-propelled deforming bodies, thrust and drag forces are hard to
86 decouple (Anderson et al., 2001; Schultz and Webb, 2002; Tytell, 2007, Tytell et al., 2010),
87 which makes it difficult to isolate drag forces alone during normal freely-swimming locomotion
88 to assess the effect of surface ornamentation. In order to investigate the possible drag reducing
89 properties of surface ornamentation such as shark skin denticles or various biomimetic products
90 (or whether surface structures might possibly enhance thrust) it is necessary to use a study
91 system that permits (1) the use of self-propelling bodies possessing different surface
92 ornamentations where thrust and drag forces are naturally balanced throughout an undulatory
93 cycle, (2) accurate measurement of self-propelled swimming speed so that the swimming
94 performance of different surfaces can be compared statistically, (3) the imposition of different
95 motion programs so that the effect of moving the ornamented surfaces in different manners can
96 be assessed, and (4) various experimental manipulations of surface structure to test directly the
97 hypothesis that it is the surface ornamentation alone that causes drag reduction and hence
98 increased swimming speed.
99 In this study we use a robotic flapping foil device to test the effect of shark skin surface
100 ornamentation and two biomimetic surfaces on self-propelled swimming speed. The flapping
101 foil robotic device was developed for the study of fish-like self-propulsion in both rigid and
102 flexible foils, and allows accurate measurement of free swimming speeds, the production of
103 controlled motion programs to move foils under a variety of heave and pitch conditions (Lauder
104 et al., 2007, 2011a, 2011b), and quantification of flow over the foil surface using particle image
105 velocimetry. We make foils that are both rigid and flexible out of fresh shark skin, and also
106 study the propulsion of two manufactured shark skin mimics. We directly test the hypothesis in
107 each case that surface ornamentation produces an increase in swimming speed by comparing to a
108 control condition with reduced or absent ornamentation.
109

2
110 MATERIALS AND METHODS

111 Flapping foil materials

112 We produced foils to be mounted in the robotic flapping device with three different surface
113 materials: Speedo Fastskin® FS II fabric, riblet material, and real shark skin.
114 Flexible moving foils with Speedo Fastskin® FS II fabric (provided by Speedo
115 International Ltd., Nottingham, UK) were produced. Two rectangular fabric pieces (20 x 7 cm in
116 size) were bonded together using adhesive spray glue (Duro Spray Adhesive, Henkel, OH, USA)
117 and clamped in a metal sandwich bar holder (consisting of two rectangular metal rods of
118 0.15 x 1 x 28 cm, identical to the one shown in Fig. 3) vertically and horizontally, respectively.
119 Every foil is manufactured three times. In two foils the outside (biomimetic surface, see Fig. 6 B)
120 of the Speedo fabric faced the fluid with the dent structure parallel and perpendicular to the flow,
121 respectively. In the third foil the smoother underside (not biomimetic surface, see Fig. 6 A) of
122 the fabric faced the fluid.
123 A rubber riblet membrane (average thickness of 1 mm) is made of silicone (Elastosil M
124 4630, Wacker Silicone, Drawin Vertriebs GmbH, Germany) by pouring the silicone on a
125 lenticular foil resulting in longitudinal U-shaped riblet structures with a height of 87 µm and a
126 spacing of 340 µm (see Fig. 7). The rubber riblet material was glued (using adhesive spray glue,
127 Duro Spray Adhesive, Henkel, OH, USA)) onto a 19 cm long metal NACA 0012 foil with a
128 chord length of 6.85 cm. Every foil was manufactured three times. In two foils the outside
129 (biomimetic surface, see Fig. 7 A) of the riblet material faced the fluid with the dent structure
130 parallel (standard orientation of riblets) and perpendicular to the flow, respectively. In the third
131 foil the smooth underside (not biomimetic surface) of the fabric faced the fluid.
132 Both rigid and flexible shark skin foils were made with the skin of a male shortfin mako
133 (Isurus oxyrinchus) with a total body length of 190 cm and a male porbeagle (Lamna nasus) with
134 an body length of 155 cm. Both species are fast swimming sharks (according to the
135 classifications of Reif, 1985). Both individuals were obtained shortly after being caught by
136 fisherman off Boston, Massachusetts, and had not been frozen at the time we first removed the
137 skin. Skin panels were extracted using dissection instruments and cleaned with a water jet to
138 remove any surface material that had adhered as a result of capture and transport. Scraping the
139 underside of the skin with a single-edge razor blade mechanically removed attached muscle
140 fibres and generated a skin membrane that could be attached to either a rigid foil or another piece

3
141 of skin to make a bilaminar flexible skin membrane with the surface ornamentation facing out.
142 Rigid shark skin foils were made with skin pieces (6.5 x 17.5 cm in size with an average
143 thickness of 3 mm), at the longitudinal position of the dorsal fin, removed along the lateral side
144 of the shortfin mako, and were glued to both sides of a flat plate (3.1 x 6.8 x 19 cm in size) using
145 a thin layer of commercially available cyanoacrylate “Instant Krazy Glue” (Elmers Products
146 Inc., OH, US). The denticle ridges were orientated parallel to the chord length of the foil.
147 Semicircle-shaped hollow metal bars with a diameter of 8.17 cm (17.5 cm in length) were glued
148 to the long edges of the foil to obtain smooth leading and trailing edges. The “mako flat plate”
149 foil (Fig. 2) has a 1.5 cm broad uncovered top where the foil holder is attached. The symmetry of
150 the flat plate allows for conducting flapper experiments with two orientations by simply
151 reversing the foil orientation in the robotic flapper: in flow, where the denticle crown tips point
152 downstream (as in the live shark), and against flow with upstream-facing crown tips.
153 Three flexible “shark skin membrane” foils were produced: two with skin of the shortfin
154 mako (no.1 and 2) and one with skin of the porbeagle. For every foil, rectangular skin pieces
155 with a size of 6.4 x 9.2 cm were extracted from below the dorsal fin (10 cm above the midline) of
156 both lateral sides and bonded together. Skin membranes were clamped in two rectangular metal
157 rods (0.15 x 1 x 28 cm) so that the denticle ridges were oriented parallel to the chord length and
158 with the denticle crown peaks facing downstream (Fig. 3) – equivalent to the alignment in living
159 sharks. The leading edge was sealed with epoxy to form an even sharp leading edge and prevent
160 delamination of the glued membrane.
161 After finishing the self-propelled measurements with the intact foils, the denticles on each
162 shark skin foil were removed by carefully sanding the skin surface under a microscope with
163 wetted sandpaper (aluminum-oxide cloth) in order to design a control object (sanded foil) with
164 the properties of shark skin but greatly reduced surface ornamentation (discussed further below).
165 We tried a number of methods of removing denticles but none was as effective as careful sanding
166 under a microscope. We avoided chemical removal methods in order not to affect skin flexural
167 stiffness. Although this method did not remove every denticle and small nubs of the denticle
168 bases were left on the surface, we found that more aggressive denticle removal methods
169 damaged the skin surface. Sanding did not remove the epidermis of the underlying skin.

170 Surface imaging

171 In this study an Environmental Scanning Electron Microscope (ESEM - Zeiss EVO SEM 50,

4
172 Carl Zeiss SMT Inc., NJ, USA) was used in vacuum in order to take high-resolution images of
173 dried and coated foil samples. Dissected skin pieces were mechanically cleaned with a soft tooth
174 brush and placed into an ultrasonic cleaner for two minutes. Then, the skin was chemically
175 cleaned by placing it in 6% Sodium Hypochlorite (NaOCl) for 30 seconds to remove any mucus.
176 Both steps were repeated three times before a final cleaning in running water under pressure.
177 Wet skin samples were dried under mild pressure for 24 hours. All skin samples were extracted
178 0.5 cm downstream of the cut skin pieces for foil production (e.g., Figs. 2, 3). For biomimetic
179 materials, both the front surface and the underside plus a cross section of a piece each from the
180 riblet material and the Speedo fabric were visualized in the ESEM. These samples were coated
181 uniformly with Platinum/Palladium using a sputter coater (208 HR, Cressington Scientific
182 Instruments Ltd., UK).
183 Although the skin pieces used for foil construction here were removed from the mid-body
184 area to obtain single pieces that were large enough to be used for the foils, we also imaged
185 denticle structure from various locations around the body. The basic structure of the denticles in
186 the mid-body area with the three surface ridges (see Figure 6) was similar to that seen in the tail
187 and on the surfaces of the fins. Denticle morphology on the head was quite different than body
188 denticle structure.
189 To estimate the wetted surface area of the shark skin foils we measured the surface area of
190 single denticles (e.g., Fig. 4) and used the number of denticles counted within an area of 1 mm²
191 in the mako flat plate (Fig. 2) to provide an approximation of the total foil wetted surface area.

192 Foil manufacturing

193 For each of the foil materials described above, we took care to press the material onto the rigid
194 foil surface by hand and to smooth out any wrinkles. No clamps were used and strong pressure
195 was not applied to prevent damage to scales and membrane surface. The adhesive glues used
196 (details provided above for each foil) worked well even after many flapping cycles in the water,
197 and we did not observe separation of the two foil surfaces for any of the materials. Pieces of
198 shark skin removed and prepared as described above were blotted dry and gently spread onto
199 either the rigid foil, or another piece of skin to which glue had been applied to the inner surface.
200 Firm continuous pressure was all that was necessary to bond the two skin pieces together or to
201 attach the skin to the rigid foil surface.
202 The different foil types used different pitch axes that depended on the material being

5
203 studied. For the rigid foils (Fig. 2), the pitch axis was in the center of the foil. This allowed us
204 to pitch the foil surface and, when combined with heave movement, produce a motion that was
205 generally similar to that of the flexible foils. For the flexible foils (Fig. 3), the pitch axis was at
206 the leading edge to hold the skin firmly and prevent any distortion of the flow by bending of skin
207 in front of the pitch axis. All statistical comparisons of swimming speed were conducted among
208 trials with a single foil design.

209 Flapping foil robotic apparatus

210 Movement of the rigid and flexible foils are controlled using a robotic flapping device as
211 described in Lauder et al. (2007, 2011a, 2011b). This robotic device (see Fig. 1) was mounted on
212 a carriage attached to an air-bearing system to allow independent self-propulsion of an immersed
213 foil with little frictional loss (also see Tangorra et al., 2011). A desired foil motion was specified
214 using a custom LabVIEW program (LabVIEW 8, National Instruments, TX, USA). Under the
215 condition of self-propulsion at a steady speed, the stroke averaged thrust generated by the model
216 must equal the average drag force experienced by the flapping foil (Lauder et al., 2007). During
217 self-propulsion we measured the flow tank speed (Ueq) required for the foil to hold position at the
218 equilibrium position (Xeq) and determined the self-propelled swimming speed of different motion
219 programs for changes in heave (amplitude in cm), pitch (maximal angle in °) and cycle speed
220 (frequency in Hz). Amplitude values given here are +/- the mean midline position. We
221 determined an accurate Ueq using another custom LabVIEW program that tracked the x-position
222 of the flapping foil via a magnetic linear encoder and the propeller speed via an optical rotary
223 encoder to calculate an average Ueq for five slightly different flow speeds (see Lauder et al.,
224 2011a, b for further details). Nine replicate trials were completed at each motion program setting
225 for each foil.

226 Kinematics and hydrodynamic flow visualization

227 We took high-speed video sequences (using a FastCam 1024 PCI, Photron USA Inc., CA, USA)
228 during self-propelled speed experiments with the above mentioned foils to document foil shape
229 during swimming. Video data were acquired at 500 Hz and each frame had a resolution of 1024
230 x 1024 pixels. Additionally, we trained 1.5-year-old spiny dogfish (Squalus acanthias; body
231 length 27 cm) to swim steadily in the laboratory flow tank and recorded high-speed video
232 sequences from a ventral view at swim speeds of 1.25 and 1.75 BL s-1. Dogfish were obtained

6
233 from local commercial vendors in Woods Hole, Massachusetts. We determined in vivo body
234 curvature from these live shark videos and from the kinematic sequences of the mako membrane
235 no.2 foil by analyzing three image frames from individual tail beats on which the shark body or
236 the foil, respectively, showed the highest bending. Maximal body curvature was calculated from
237 coordinates for three points within 2 cm distance along the edge of the curved body or foil
238 respectively. These data were used to ensure that our experiments using the flexible shark-skin
239 foils were conducted using a motion program that replicated the in vivo curvatures of the lateral
240 body surface of live freely-swimming sharks.
241 Flow visualization was conducted for the mako flat plate and mako membrane no.2 foils
242 using digital particle image velocimetry (DPIV), as described in our past research (Drucker and
243 Lauder, 1999, 2002; Johansson and Lauder, 2004; Lauder et al, 2007). Briefly, water in a 600 l
244 recirculating flow tank is seeded with 6 g of hollow, silver-coated near neutrally buoyant glass
245 particles. A 10 W continuous-wave argon-ion laser (Innova 300 Series, Coherent Laser Group,
246 CA, USA) was focused onto a thin horizontal light sheet (1-2 mm thick) with a size of
247 approximately 40 x 25 cm, which cut the foil at the midpoint on a plane halfway between top and
248 bottom. Recorded images were analyzed with DPIV software (DaVis 7.2.2.272, LaVision
249 GmbH, Göttingen, Germany).
250 Average Reynolds numbers (Re) and Strouhal numbers (St) for each foil type self-
251 propelling are presented below in the Results section; in all cases values for these dimensionless
252 numbers are within the range of those found in studies of fish swimming in vivo (e.g., Flammang
253 et al., 2011; Lauder and Tytell, 2006; Triantafyllou and Triantafyllou, 1995).

254 Statistics
255 Mean values and standard errors were calculated for the nine trials of the self-propelled
256 swimming speed (SPS) measurements for each condition and are plotted in Figures 8 – 11 which
257 show the results of the swimming speed trials. A Kolmogorov-Smirnov test verified the normal
258 distribution of the nine trials in every case. Within each foil comparison set (e.g., shark skin with
259 denticles facing out versus the sanded condition), we used an independent two-sample Student’s
260 t-test with an equal sample sizes. Note that it is only appropriate to do corrected pairwise
261 comparisons among each particular experimental treatment (such treatment conditions are
262 plotted in the same color in Figures 8-11 to facilitate determination of which trials were
263 compared statistically). Thus we deliberately did not test for swimming speed differences among

7
264 foils moved with different motion programs. Statistical tests are performed using SPSS 15.0
265 (SPSS Inc., IL, USA).

266

267 RESULTS

268 Surface morphology

269 Environmental Scanning Electron Microscope (ESEM) images of surface denticle structure and
270 distribution for the mako flat plate is shown in Fig. 2, and for mako skin flexible membrane no.2
271 in Fig. 3. Denticle shape and size differs somewhat from top to the bottom of the foils. All
272 denticles all have three ridges with a spacing of 46 to 55 µm. Denticles on the mako flat plate
273 range from 130 to 180 µm in width and from 180 to 210 µm in length. The denticles on the very
274 top of the mako membrane no.2 (Fig. 3, A) have an oval shape, where the width (170-180 µm) is
275 longer than the length (130-140 µm). Fig. 4 shows a close-in view of a single denticle with three
276 ridges and a toothed crown from an area similar to that of Fig. 2D on the mako flat plate foil.
277 Sanding the shark skin foils to reduce surface denticles to small nubs was successful as seen in
278 Figure 5 where a part of the sanded mako membrane no.1 foil is shown. The denticle surface
279 was greatly reduced and only small relics of the denticle bases and small remaining stubs seen in
280 the upper left of Figure 5 that escaped complete the sanding process remained. From ESEM
281 images, we estimated that sanding the denticles (compare figures 3 and 5) reduced the wetted
282 surface area by over 70%.
283 The underside (Fig. 6, A) of the Speedo “shark skin” fabric shows a typical fabric surface
284 with parallel seams. The seams on the surface of the outside “biomimetic” surface (Fig. 6, B) are
285 slightly bigger and more significantly indented. Fig. 6, C shows the cross-sectional morphology
286 of two triangle-shaped dents in the outer surface which lie about 1.25 mm apart and form the
287 biomimetic “shark skin” surface ornamentation. These indentations in the fabric are part of the
288 biomimetic manufacturing process, and vary slightly in size.
289 The outside of the riblet material (Fig. 7, A) is covered with parallel tapering peaks (seen
290 as white lines). The cross section (Fig. 7, B) provides a clearer view of the riblet structures which
291 have a scalloped shape with a height of 87 µm and a spacing of 340 µm. The underside of the
292 material is smooth.

8
293 Self-propelled foil swimming speeds
294 The self-propelled swimming trials for flexible membranes made of the Speedo biomimetic
295 fabric (Fig. 6; results in Fig. 8) showed that the “shark skin” surface did not have a consistent
296 effect of allowing increased swimming speeds for a given motion program. For example,
297 moving the flexible Speedo foil at 2Hz with 2cm heave when the ridges were in the vertical
298 orientation did not result in any swimming speed change at all as compared to a foil with the
299 smoother inner fabric surface exposed to the water. When surface dents/ridge were oriented
300 perpendicular orientation swimming speeds of the foil slowed by an average of 5.2%. Adding a
301 20° pitch motion to the foil leading edge more than doubled overall swimming speeds but no
302 clear effect of the biomimetic surface could be detected (Fig. 8). Under 3Hz actuation, the foil
303 with the biomimetic surface on the outside actually swam slower than when the smoother inner
304 fabric surface was exposed (Fig. 8). Speedo biomimetic foils self-propelled at an average
305 Reynolds number of 27,000 and an average Strouhal number of 0.28 in these experiments.
306 Self-propelled swimming (SPS) trials for rigid foils with a surface of the riblet material
307 (Fig. 7; results in Fig. 9) showed a highly significant (P<0.001) increase in the self-propelled
308 swimming speed when the riblets were located on the outer surface of the foil compared to an
309 inside orientation. The biggest difference was present at a motion program of 2 Hz, 1 cm heave,
310 0° pitch, where parallel-oriented riblets had a mean SPS of 35.6±0.3 cm s-1 and the inside-
311 oriented riblets a mean SPS of 32.5±0.4 cm s-1, an 9.5% increase in self-propelled speed due to
312 the riblet surface (values are means ± one standard error). With the same motion program, the
313 comparison of the parallel and perpendicular ridge orientations showed no significant difference.
314 However, adding a pitch of 10° to the former motion program made the foil with perpendicular
315 ridge propel significantly slower (2.5 %) than the parallel foil, but both ridged foils self-
316 propelled at a faster speed than the foils with a smooth surface. Riblet foils self-propelled at an
317 average Reynolds number of 22,000 and an average Strouhal number of 0.12 in these
318 experiments.
319 Self-propelled speed trials for mako shark skin attached to the rigid flat plate (Fig. 2;
320 results in Fig. 10) showed significant increases in swimming speed for all three motion programs
321 when denticles were sanded off as compared to foils with intact denticles on the surface. Sanded
322 foils swam at an average of 13.4% faster than foils with intact denticles in their normal
323 orientation. When rigid mako skin foils were tested in the reverse orientation, only the 2 Hz, 1.5

9
324 cm heave, 10° pitch angle motion program showed a significant difference, and in this case the
325 foil with the reverse orientation of denticles swam 11% faster (Fig. 10). Rigid mako shark skin
326 foils self-propelled at an average Reynolds number of 24,000 and an average Strouhal number of
327 0.22 in these experiments.
328 Self-propelled speed trials for mako and porbeagle skin made into a flexible membrane
329 (Fig. 3; results in Fig. 11) showed that for all 8 test conditions, sanding the denticles reduced the
330 swimming speed by an average of 12.3%. Seven of the 8 comparisons were significantly
331 different, and only the 2 Hz, 2 cm heave, and 30° pitch motion program failed to be significantly
332 different. The most striking difference is obtained with the mako membrane no.1 at a motion
333 program of 2 Hz, 2 cm, 0° pitch (see Fig. 11, A), for which the intact foil (mean SPS: 19.7±0.15
334 cm s-1) swam almost 20% faster than the sanded foil (mean SPS: 16.3±0.13 cm s-1). Flexible
335 shark skin foils self-propelled at an average Reynolds number of 13,000 and an average Strouhal
336 number of 0.37 in these experiments.
337 We emphasize that this is the opposite result to that obtained with the rigid foils described
338 above: for flexible shark skin membranes, removing the denticles slows down swimming speed;
339 if shark skin is attached to a rigid foil, sanding the denticles can increase swimming speed.

340 Kinematics and hydrodynamics

341 Freely-swimming spiny dogfish (Squalus acanthias) displayed mean body curvatures of
342 0.14±0.01 cm-1 (mean ± 1 standard error) swimming at 1.25 body lengths per second (BL s-1)
343 (Re=90,000), and 0.20±0.01 cm-1 at 1.75 BL s-1 (Re=127,000). For comparison we measured the
344 curvatures of flexible mako membrane no.2 which showed mean curvature values of 0.17±0.01,
345 0.18±0.01 and 0.25±0.01 cm-1 for the three different motion programs at the relevant self-
346 propelled speeds. The swimming flexible shark skin membranes studied here thus achieved
347 curvature values comparable to those of freely-swimming live sharks at commonly observed
348 cruising speeds.
349 In order to better understand the mechanism of thrust generation by flexible shark skin
350 foils, and the hydrodynamic basis of the performance difference between the normal and sanded
351 flexible shark skin membranes, we quantified fluid flow over the flexible foil surface during self-
352 propulsion using particle image velocimetry. Sample analyses are shown in Figures 12 and 13.
353 The flexible shark skin membrane foils moved with a motion program of 2Hz, +/- 2 cm
354 heave, and 10° pitch, showed a clear leading edge vortex (LEV) that remained attached to the

10
355 foil for most of the flapping cycle (Fig. 12). This LEV moved posteriorly down the foil from its
356 initial position near the leading edge (see Fig. 12: 300 – 400 ms time frames), and was then shed
357 into the wake before reforming at the leading edge as the flapping cycle resumed. Plots of flow
358 velocity in the x-direction (parallel to free-stream flow, Vx) versus the distance from the foil
359 surface are shown for six frames (0, 100, 200, 300, 400, 500 ms) representing 100% of a motion
360 cycle in Figure 12. The attached LEV is clearly seen by the negative velocities (portions of the
361 curve to the left of the zero line in Fig. 12) that occur near the foil surface where flow is reversed
362 relative to free stream velocities (e.g., Fig. 12: 100 ms time). Along the transect away from the
363 foil surface flow velocities change sign as the LEV core is traversed, and at a distance of
364 approximately 3 cm from the foil surface the free stream velocity has been reached.
365 Comparisons of the flow pattern at the foil leading edge between shark skin foils with
366 denticles on the surface and those with the denticles sanded off show substantial differences in
367 LEV location for the flexible membrane shark skin foils, but not for the flat plate foils,
368 corresponding well to the differences in self-propelled speed reported above for flat plate versus
369 membrane shark skin foils (Figs. 10, 11). Figure 13A shows the analysis of LEV vorticity for
370 the intact and sanded mako flat plate foil at a motion program of 2 Hz, 2 cm, 10°. This plot
371 shows that the peak vorticity, which reflects the vortex core location, appears at almost the same
372 distance from the foil edge in the two rigid foils. The intact foil has a peak vorticity of 13.6 s-1 at
373 a distance of 11.3 mm from the foil edge, while the sanded foil peak vorticity of 11.4 s-1 occurs
374 at a distance of 11.9 mm from the sanded foil edge.
375 The analogous plot for the intact and the sanded flexible mako membrane no.2 foil at a
376 motion program of 2 Hz, 2 cm, 10° is shown in Figure 13B. Here, the intact foil shows a peak
377 vorticity of 9.5 s-1, which occurs at a distance of 6.3 mm to the foil surface. In contrast, the
378 sanded foil has a maximal vorticity of 11.1 s-1 at 14.5 mm distance from the foil surface.
379 Sanding the denticles on the flexible foil membrane leads to a displacement of the leading edge
380 vortex core a distance of nearly a centimeter (8.2 mm) further away from the foil surface.
381

382 DISCUSSION
383 In this paper we have compared the swimming performance of both rigid and flexible shark skin
384 foils under both intact and sanded conditions, with the aim of quantifying possible locomotor

11
385 benefits of the surface denticles on the skin. In addition, we analyzed the swimming
386 performance of two biomimetic shark skin surfaces, a ribbed rubber material and the Speedo
387 Fastskin fabric which possesses surface indentations.
388 Our most noteworthy results were: (1) that the shark denticles had no beneficial
389 locomotor effect on the moving rigid shark skin foils, and in fact the unmodified rigid foils swam
390 more slowly than those on which the denticles were removed for two motion programs (Fig. 10),
391 (2) that surface denticles did improve swimming performance significantly (by an average of
392 12.3%) on flexible shark skin membrane foils compared to those in which the denticles had been
393 removed, and (3) that biomimetic surface indentations and riblets can enhance swimming
394 performance under certain motion programs, but not for other types of foil movement.
395 These data emphasize both the utility of using a highly-controlled robotic system to test
396 for changes in swimming performance, and the importance of flexibility in locomotor dynamics:
397 studying flexible shark skin membranes proved essential to demonstrating a significant increase
398 in locomotor performance due to surface ornamentation. In addition, as we discuss below, shark
399 skin denticles may enhance thrust in addition to causing a reduction in drag.

400 Flapping foils and shark swimming

401 How closely did swimming by our flexible shark skin membranes match the locomotor
402 conditions seen in live sharks? The shark skin foils in this study self-propelled at average
403 Strouhal numbers between 0.28 and 0.38 which is well within the range of the Strouhal numbers
404 used by live sharks. Many fishes, including sharks, swim at Strouhal numbers between 0.21 and
405 0.41 (Flammang et al., 2011; Lauder and Tytell, 2006; Triantafyllou and Triantafyllou, 1995).
406 The Reynolds numbers during flexible shark skin foil self-propulsion were also in the range of
407 swimming by smaller sharks which typically cruise at swimming speeds of 0.5 to 1.0 lengths sec-
1
408 , 13,000 to 24,000, although certainly large open water sharks swim at much higher Reynolds
409 numbers during fast swimming or during periods of burst accelerations. Since our foils were self-
410 propelling they swam at their natural Strouhal and Reynolds numbers under the driven motion
411 program, and we could not alter these parameters experimentally without removing the self-
412 propelled condition.
413 Surface curvatures achieved by the flexible shark skin foils could also have a potentially
414 important influence on locomotor performance as surface denticles might be expected to bristle
415 as the skin is bent and thus alter flow near the skin (e.g., Lang et al., 2008). We compared the

12
416 measured surface curvatures of our flexible shark skin foils to those we measured in live sharks
417 swimming in our laboratory flow tank, and the values of 0.17-0.25 cm-1 for foils accord well
418 with measured maximal mid-body values from live spiny dogfish swimming at 1.0 lengths sec-1
419 (0.14-0.20 cm-1). Body curvature values vary considerably in swimming sharks, depending on
420 location and swimming speed, from 0 to 0.3 cm-1, so these curvature measurements show that the
421 shark skin membranes when self-propelling bend to a similar extent to the skin of a live shark
422 during unrestrained locomotion.

423 Swimming of biomimetic foils

424 Self-propelled speed results from the Speedo material foils under conditions with the biomimetic
425 surface oriented toward the water versus located inside away from the water show that
426 swimming performance depends on the motion program used and the orientation of the surface
427 ridges (Fig. 8). For two of the three motion programs used, the biomimetic surface actually
428 reduced swimming performance compared to flexible Speedo foils with the underside located
429 toward the water surface. The orientation of surface ridges can also have a significant effect on
430 swimming performance, and placing the surface ridges in a perpendicular orientation to
431 oncoming flow reduced swimming performance for all three motion programs.
432 The proposed drag-reduction properties of Speedo Fastskin surface ridges is thus called into
433 question when the locomotion of this material made into flexible foils is compared to controls
434 with no surface indentations. While it is theoretically possible that both skin friction reduction
435 and pressure drag reduction could be increased by the surface of this material (Benjanuvatra et
436 al., 2002), the complexities of surface deformation in human swimmers and the three-
437 dimensional flows that occur over the body during movement suggest that there is no consistent
438 flow direction over the surface ornamentation on body suits made of this material. Our ESEM
439 images (Fig. 6) of the Speedo fabric also evoke doubts about whether the swimsuit surface
440 functions as shark skin-like riblets, as the surface indentations bear little resemblance to either
441 shark skin or engineered riblet materials (e.g., Fig. 7). Manufactured riblet materials have sharp
442 longitudinal rib structures (Bechert and Bartenwerfer, 1989) which can cause alterations in
443 boundary layer flow (Fig. 7). A cross-cut of the Speedo material shows dents instead of ribs,
444 with large distances between the dents. If we treat these dents as riblets, the height to spacing
445 ratio would be h/s ~ 0.1. A two dimensional riblet surface with h/s = 0.1 resulted in a maximal
446 passive drag reduction of 0.5% (Bechert et al., 2000), supporting our results of either no drag

13
447 reduction or drag enhancement with impaired swimming performance resulting from the Speedo
448 material.
449 Our studies of self-propelled speeds achieved with the engineered riblet material (Fig. 7)
450 show that improvements in swimming speed through drag reduction can occur depending on the
451 way in which the material is moved (Fig. 9). We also altered the orientation of the riblets and
452 compared parallel and perpendicular orientations. Placing the riblet surface on the inside so that
453 the foil had a smooth outer surface reduced swimming performance, suggesting that the riblets
454 were effective in reducing drag: when riblets are exposed to the flow in a parallel orientation
455 they improved swimming speeds by up to 9.5%. But both parallel and perpendicular orientations
456 improved swimming performance, suggesting that the surface roughness was more important
457 than the precise orientation of the riblets.
458 Our ESEM images revealed a height to spacing ratio of 0.25 for this riblet material (Fig. 7).
459 Bechert et al. (2000) studied a system with adjustable longitudinal slit riblets which yielded a
460 motionless-drag reduction (riblet foils were held still in moving flow) of about 5% at a ratio of
461 h/s = 0.2. Our data confirm previous studies that riblets can reduce not only the motionless-drag
462 (as shown by Bechert and Bartenwerfer, 1989; Bechert et al., 1997; Luchini and Trombetta,
463 1995), but also the drag-in-motion, when foils with riblet surfaces self-propel.

464 Swimming of shark-skin foils

465 One mechanism of reducing drag during locomotion is to reduce skin friction drag, and the
466 magnitude of skin friction drag force depends on the wetted surface area (Hoerner, 1965). We
467 estimate that sanding the denticles reduced the wetted surface area of the flexible foils by 70%.
468 If this were the only factor in understanding shark skin locomotor function, then the sanded foils
469 would be exposed to a skin friction drag force which is much less than the force acting on the
470 intact foils, and sanded foils would self-propel at higher speeds than the foils with denticles
471 intact.
472 However, flexible shark skin foils actually showed a substantial improvement in swimming
473 performance of an average of 12.3% (with a maximal improvement of almost 20%) as compared
474 to the same foils with the surface denticles sanded off. Orientating the denticle crown tips
475 against the flow stream, opposite to the natural alignment, still did not reduce swimming
476 performance to the degree produced by sanding the foil surface (Fig. 10), but did reduce
477 swimming performance compared to the natural orientation for the 2Hz, 1.5 cm heave, 0° pitch

14
478 motion program.
479 Interestingly, the same treatments applied to a rigid foil did not have the same effect:
480 sanding the surface of shark skin applied to a rigid foil increases the self-propelled speed, the
481 opposite effect seen in flexible foils. These results show the substantial effect that flexibility of
482 the swimming surface can have on the results of foil surface tests, and indicates that future
483 testing of shark skin drag-reduction effects should include flexible surfaces that are freely-
484 swimming and approximate the bending seen in vivo by swimming sharks: other conditions may
485 not show a drag-reduction effect.

486 Hydrodynamic function of shark skin

487 We were able to detect significant differences in the pattern of water flow over the swimming
488 shark skin flexible foils for normal versus sanded surfaces (Figs. 12, 13), and these
489 hydrodynamic differences can help explain why flexible foils with the sanded surface swam
490 more slowly compared to those with the intact surface denticles. Particle image velocimetry
491 showed the presence of a leading edge vortex (LEV) on the swimming foils that is located at a
492 greater distance from the foil surface after sanding the denticles than in the intact foil with
493 normal denticles (Fig. 13B). The presence of denticles on the surface thus alters the flow
494 environment near the flexing foil surface in such a manner that the LEV adheres more closely to
495 the foil surface than after the denticles are sanded off.
496 Vortices are regions of low pressure with a pressure minimum at the vortex core (Rossi,
497 2000). The low pressure zone of the leading edge core will affect suction forces on the foil
498 surface, and this pressure force acts normal to a surface in fluids (Sigloch, 2005). This pressure
499 vector can be divided into two resultant forces (normal to each other), one acting along the free-
500 stream direction, while the other acts orthogonal to this path. Appropriate foil orientations where
501 the inclined surface with an attached LEV is angled upstream (Fig. 12: 0 – 100 ms) will result in
502 enhanced propulsion through leading edge suction as one component of the low pressure vector
503 will draw the foil upstream. At these positive angles of attack, the force parallel to the flow
504 points upstream, so it represents a thrust force. The lower the pressure on the foil surface and the
505 closer the vortex core is to the foil surface, the higher the thrust force. We predict that the LEV-
506 derived suction (thrust) force is reduced after sanding the denticles, which alter flow near the
507 membrane surface and cause the LEV to move almost 1 cm further away from the leading edge
508 surface of the moving foil. This would reduce leading edge suction and hence thrust.

15
509 This result suggests that one important effect of the skin denticles is to enhance thrust, and
510 not simply to reduce drag. The overwhelming emphasis of the existing literature on shark skin
511 has been on drag reduction, but denticles alter vortex location, and especially on the tail surface
512 where flow separation and vortex formation have been demonstrated (Wilga and Lauder, 2002,
513 2004) could increase thrust. Similar effects of denticles could occur on pectoral fins also, where
514 vortices are generated during maneuvering (Wilga and Lauder, 2000) and denticles that enhance
515 vortex attachment to the fin surface would increase lift and hence maneuvering forces.
516 The precise nature of flow modification on the foil surface when denticles are present is
517 not known at present, and studying this will require close views (on the order of 1-2 mm2 field of
518 view) of the flows on the surface of moving shark skin foils. This is certainly a challenging
519 proposition in vivo, but will likely reveal the specifics of flow modification due to surface
520 denticles on the skin as compared to a sanded condition. Indeed, the question of what flow
521 modification occurs in the boundary layer region in the presence of natural shark skin denticles
522 during self-propulsion remains unanswered at present, and remains a key area for future study.
523
524

525 ACKNOWLEDGEMENTS
526 Many thanks to Vern Baker and Erik Anderson for assistance with the flapper
527 programming and experiments, and to all members of the Lauder lab and the MCZ Fish
528 Department (Karsten Hartel and Andy Williston) for their assistance with this project. Special
529 thanks to Dr. Brooke Flammang for videos of spiny dogfish. This work was supported by
530 National Science Foundation grants IBN0316675 and EFRI-0938043.
531

16
532 REFERENCES
533 Anderson, E. J., McGillis, W. and Grosenbaugh, M. A. (2001). The boundary layer of
534 swimming fish. J. Exp. Biol. 204, 81-102.
535 Applegate, S. (1967). A survey of shark hard parts. In Sharks, Skates and Rays, (ed. W. C.
536 Hamlett) pp. 37–67. Baltimore: Johns Hopkins Univ. Press.
537 Bechert, D. and Bartenwerfer, M. (1989). The viscous flow on surfaces with longitudinal ribs.
538 J. Fluid Mech. 206, 105-129.
539 Bechert, D. W., Bruse, M. and Hage, W. (2000). Experiments with three-dimensional riblets as
540 an idealized model of shark skin. Exp. Fluids 28, 403-412.
541 Bechert, D. W., Bruse, M., Hage, W., Van Der Hoeven, J. and Hoppe, G. (1997).
542 Experiments on drag-reducing surfaces and their optimization with an adjustable geometry.
543 J. Fluid Mech. 338, 59-87.
544 Bechert, D. W., Hoppe, G. and Reif, W. E. (1985). On the drag reduction of the shark skin. In
545 AIAA Shear Flow Control Conference, vol. AIAA 85-0546, pp. 1-18. Boulder, Colorado.
546 Benjanuvatra, N., Dawson, G. and Blanksby, B. (2002). Comparison of buoyancy, passive and
547 net active drag forces between Fastskin TM and standard swimsuits. J. Science Medicine
548 Sport 5, 115-123.
549 Büttner, C. C. and Schulz, U. (2011). Shark skin inspired riblet structures as aerodynamically
550 optimized high temperature coatings for blades of aeroengines. Smart Mater. Struct. 20,
551 094016.
552 Castro, J. I. (2011). The Sharks of North America. Oxford: Oxford Univ. Press.
553 Dinkelacker, A., Nitschke-Kowsky, P. and Reif, W. (1987). On the possibility of drag
554 reduction with the help of longitudinal ridges in the walls. In Proceedings of the IUTAM
555 Symposium on Turbulence Management and Relaminarization,, pp. 109–120. Bangalore,
556 India: Springer Verlag.
557 Drucker, E. G. and Lauder, G. V. (1999). Locomotor forces on a swimming fish: three-
558 dimensional vortex wake dynamics quantified using digital particle image velocimetry. J.
559 Exp. Biol. 202, 2393-2412.
560 Drucker, E. G. and Lauder, G. V. (2002). Experimental hydrodynamics of fish locomotion:
561 functional insights from wake visualization. Int. Comp. Biol. 42, 243-257.

17
562 Flammang, B. E., Lauder, G. V., Troolin, D. R. and Strand, T. (2011). Volumetric imaging
563 of shark tail hydrodynamics reveals a three-dimensional dual-ring vortex wake structure.
564 Proceedings of the Royal Society B 278: doi:10.1098/rspb.2011.0489 (in press, available on-
565 line).
566 Han, X., Zhang, D., Li, X. and Li, Y. (2008). Bio-replicated forming of the biomimetic drag-
567 reducing surfaces in large area based on shark skin. Chinese Sci. Bull. 53, 1587-1592.
568 Hoerner, S. F. (1965). Fluid-dynamic drag: practical information on aerodynamic drag and
569 hydrodynamic resistance. Midland Park, NJ.
570 Johansson, L. C. and Lauder, G. V. (2004). Hydrodynamics of surface swimming in leopard
571 frogs (Rana pipiens). J. Exp. Biol. 207, 3945-3958.
572 Koeltzsch, K., Dinkelacker, A. and Grundmann, R. (2002). Flow over convergent and
573 divergent wall riblets. Exp. Fluids 33, 346-350.
574 Lang, A. W., Motta, P., Hidalgo, P. and Westcott, M. (2008). Bristled shark skin: a
575 microgeometry for boundary layer control? Bioinsp. Biomimet. 3, 046005.
576 Lauder, G. V., Anderson, E. J., Tangorra, J. and Madden, P. G. (2007). Fish biorobotics:
577 kinematics and hydrodynamics of self-propulsion. J. Exp. Biol. 210, 2767-80.
578 Lauder, G., Madden, P. G. A., Tangorra, J., Anderson, E. and Baker, T. V. (2011a).
579 Bioinspiration from fish for smart material design and function. Smart Mater. Struct. 20::
580 doi:10.1088/0964-1726/20/9/094014.
581 Lauder, G., Tangorra, J., Lim, J., Shelton, R., Witt, C. and Anderson, E. J. (2011b). Robotic
582 models for studying undulatory locomotion in fishes. Marine Tech. Soc. J. 45:41-55.
583 Lauder, G. V. and Tytell, E. D. (2006). Hydrodynamics of undulatory propulsion. In Fish
584 Biomechanics, Vol 23 (Eds. R. E. Shadwick and G. V. Lauder, eds.), pp. 425-468. San
585 Diego: Academic Press.
586 Luchini, P., Manzo, F. and Pozzi, A. (1991). Resistance of a grooved surface to parallel flow
587 and cross-flow. J. Fluid Mech. 228, 87-109.
588 Luchini, P. and Trombetta, G. (1995). Effects of riblets upon flow stability. Flow Turb.
589 Combust. 54, 313-321.
590 Mollendorf, J., Termin, A., Openheim, E. and Pendergast, D. (2004). Effect of swim suit
591 design on passive drag. Med. Sci. Sports Exer. 36, 1029-1035.
592 Neumann, D. and Dinkelacker, A. (1991). Drag measurements on V-grooved surfaces on a

18
593 body of revolution in axial flow. Appl. Sci. Res. 48, 105-114.
594 Reif, W. E. (1982). Morphogenesis and function of the squamation in sharks. 1. Comparative
595 functional morphology of shark scales, and ecology of scales. Neues Jah. Geol. Palaeon.
596 Abh. 164, 172-183.
597 Reif, W. E. (1985). Squamation and ecology of sharks. Stuttgart: Senckenbergische
598 Naturforschende Gesellschaft.
599 Rossi, M. (2000). Of vortices and vortical layers: an overview. In Vortex Structure and
600 Dynamics, (Ed A. Maurel and P. Petitjeans), pp. 40-123. Berlin: Springer.
601 Schultz, W. W. and Webb, P. W. (2002). Power requirements for swimming: do new methods
602 resolve old questions? Int. Comp. Biol. 42, 1018-1025.
603 Sigloch, H. (2005). Technische Fluidmechanik. Berlin: Springer.
604 Tangorra, J., Phelan, C., Esposito, C. and Lauder, G. (2011). Use of biorobotic models of
605 highly deformable fins for studying the mechanics and control of fin forces in fishes. Int.
606 Comp. Biol. 51:176-189.
607 Toussaint, H., Truijens, M., Elzinga, M., Van De Ven, A., De Best, H., Snabel, B. and De
608 Groot, G. (2002). Effect of a Fastskin bodysuit on drag during front crawl swimming.
609 Internat. Soc. Biomech. Sports 1, 1-10.
610 Triantafyllou, M. S. and Triantafyllou, G. S. (1995). An efficient swimming machine. Sci.
611 Amer. 272, 64-70.
612 Tytell, E. D. (2007). Do trout swim better than eels? Challenges for estimating performance
613 based on the wake of self-propelled bodies. Exp. Fluids 43, 701-712.
614 Tytell, E. D., Borazjani, I., Sotiropoulos, F., Baker, T. V., Anderson, E. J. and Lauder, G.
615 V. (2010). Disentangling the functional roles of morphology and motion in the swimming of
616 fish. Int. Comp. Biol. 50, 1140-1154.
617 Wilga, C. D. and Lauder, G. V. (2000). Three-dimensional kinematics and wake structure of
618 the pectoral fins during locomotion in leopard sharks Triakis semifasciata. J. Exp. Biol. 203,
619 2261-2278.
620 Wilga, C. D. and Lauder, G. V. (2002). Function of the heterocercal tail in sharks: quantitative
621 wake dynamics during steady horizontal swimming and vertical maneuvering. J. Exp. Biol.
622 205, 2365-2374.
623 Wilga, C. D. and Lauder, G. V. (2004). Hydrodynamic function of the shark's tail. Nature 430,

19
624 850.
625

20
626 FIGURE CAPTIONS
627
628 Fig. 1. Robotic flapping foil apparatus used to test the hydrodynamic function of shark skin and
629 biomimetic models. Both rigid (f) and flexible foils with attached riblet material or shark skin
630 are clamped in the flapper shaft (fs) and immersed in the flow tank. Robot motion is driven by a
631 heave motor (hm) and a pitch motor (pm), mounted on air bearings (ab), so that the equilibrium
632 position is only dependent on the cables (c). There is no cable effect as all tests are conducted at
633 the equilibrium position (see Lauder et al., 2007, 2011a, b) for details of this testing apparatus.
634 Tuning the flume speed to match flapper thrust enables one to find the self-propelled speed (see
635 text). Bumpers (b) limit large deflections. Note that the robot can drive two separate foils, but
636 only one is used during the experiments here. Blue arrow indicates the flow direction. The rigid
637 foil shown here is 19 cm in height, 6.85 cm in chord length (Fig. 2), but the membrane foils
638 tested were a different shape (see text and Fig. 3 for description).
639
640 Fig. 2. Flat plate foil (with rounded leading and trailing edges) covered with skin from the lateral
641 midline area of a male shortfin mako shark (left). Dark skin color indicates skin from the lateral
642 shark surface, while whitish color indicates more ventral skin. The foil is 7.22 cm in chord length
643 (width) and 19 cm in height (whereof 17.5 cm is covered with shark skin). Distribution of the
644 skin structure on the surface of the mako flat plate foil (right). Environmental Scanning Electron
645 Microscope (ESEM) images from parts of a top (A), a top middle (B), a middle (C), a middle
646 bottom (D) and a bottom (E) areas. Images were taken from skin pieces extracted 0.5 cm
647 downstream of the right foil edge at each location. Scale bars 200 µm. The leading edge of the
648 denticles is on the left and the natural water flow pattern would thus be left to right in this figure.
649
650 Fig. 3. Flexible shark skin foil - mako membrane no.1 (left). Bonded skin pieces (6.4 cm in
651 length and 9.2 cm in height) of both lateral sides from 10 cm above the midline of a male
652 shortfin mako shark are clamped 2.5 cm above the lower end in a sandwich bar holder. Right:
653 Distribution of skin structure on the surface of this mako membrane foil. ESEM images from
654 parts of an upper (A), a middle (B) and a lower (C) area. Image samples were taken from skin
655 pieces extracted 0.5 cm downstream of foil edge of this foil at each location. Blue arrow
656 indicates direction of water flow during testing. Scale bars 100 µm.

21
657
658 Fig. 4. Close view ESEM image of denticles from the surface of the mid-body region in a
659 bonnethead shark (Sphyrna tiburo) to show details of typical denticle structure with the three
660 surface ridges and three posteriorly-pointing prongs. Such denticle structure is common on the
661 body, fins, and tail, although denticles of this species on the head have a different morphology.
662 Scale bar is 50 µm.
663
664 Fig. 5. ESEM image of a part of the mako membrane no.1 foil (Fig. 3) after the process of
665 sanding. The full denticles are almost completely removed with only small stubbs remaining
666 (compare to the intact denticle surface shown in Fig. 3 above). These stubbs could not be
667 removed without damaging the underlying collagen surface framework since denticles are
668 embedded in the skin. Scale bar 200 µm.
669
670 Fig. 6. ESEM images of Speedo Fastskin® FS II fabric. A: Surface image of the underside (non-
671 biomimetic) surface of the fabric. B: Surface image of the outside (biomimetic surface) of the
672 Speedo fabric at the position of V-shaped printing. C: Image of a cross-section of the Speedo
673 fabric, showing the dents on the biomimetic side (red arrows, d) in the fabric that make up the
674 “ribbed” surface. Scale bars 500 µm.
675
676 Fig. 7. ESEM images of the biomimetic riblet silicone material. A: Front side of riblet surface
677 with clearly visible height peaks (white lines). B: Image of a cross section showing the riblet
678 structures with a height (h) of 87 µm and a spacing (s) of 340 µm. Scale bars 200 µm.
679
680 Fig. 8. Histogram of the mean self-propelled speed results (mean from N=9 trials for each test)
681 for Speedo membrane foils. Error bars are +/- 1 standard error. The motion program settings
682 below each group of similarly-colored bars show the programmed foil movement with frequency
683 in Hz, amplitude in cm and pitch in degrees (°). Speedo fabric was tested on two different foil
684 orientations (vertical and horizontal – see Materials and Methods), and with the biomimetic
685 surface ridges parallel to the free stream flow, perpendicular to free stream flow, and “inside” or
686 reverse orientation with the biomimetic surface on the inside glued to the foil and the non-
687 biomimetic surface exposed to the water. Within each group of similarly-colored bars, bars with

22
688 # symbols are not significantly different from each other (P>0.05). All other comparisons are
689 significant at between P<0.05 and P>0.0001.
690
691 Fig. 9. Histogram of the mean self-propelled speed (mean from nine trials for each test) for the
692 silicone riblet material applied to a NACA 0012 foil surface. Error bars are +/- 1 standard error.
693 The motion program settings below each group of similarly-colored bars show the programmed
694 foil movement with frequency in Hz, amplitude in cm and pitch in degrees (°). Silicone riblet
695 material was tested with the biomimetic surface ridges parallel to the free stream flow,
696 perpendicular to free stream flow, and “inside” or reverse orientation with the riblet surface on
697 the inside glued to the foil and the non-biomimetic (smooth) surface exposed to the water.
698 Within each group of similarly-colored bars, bars with # symbols are not significantly different
699 from each other (P>0.05). All other comparisons are significant at between P<0.05 and P>0.000.
700 On average the foils with the “inside” orientation propel at a 7.2 % slower SPS than the foils
701 with parallel-oriented ridges.
702
703 Fig. 10. Bar diagram of the mean self-propelled speed results (mean from nine trials for each
704 test) for the mako shark skin attached to a rigid flat plate. Error bars are +/- 1 standard error. The
705 motion program settings below each group of similarly-colored bars show the programmed foil
706 movement with frequency in Hz, amplitude in cm and pitch in degrees (°). Mako shark skin on
707 the flat plate was tested with denticle surface oriented parallel to the free stream flow in the same
708 direction as on a living shark (in flow), oriented opposite to the in vivo shark denticle orientation
709 (against flow), and sanded where most of the denticle surface has been removed (Fig. 5). Within
710 each group of similarly-colored bars, bars with # symbols are not significantly different from
711 each other (P>0.05). All other comparisons are significant at between P<0.05 and P>0.0001. On
712 average the sanded rigid foils propel at a 13.4 % higher speed than the foils with denticles, with a
713 maximum difference of 18.03 % (P<0.001) at a motion program of 2 Hz, 1.5 cm, 10°.
714
715 Fig. 11. Bar diagram of the mean self-propelled speed results (mean from nine trials for each bar)
716 for the flexible moving shark skin membranes. A: mako membrane no.1, B: mako membrane
717 no.2 and C: porbeagle membrane. Error bars are +/- 1 standard error. The motion program
718 settings below each group of similarly-colored bars show the programmed foil movement with

23
719 frequency in Hz, amplitude in cm and pitch in degrees (°). Shark skin membranes were tested
720 intact with denticles oriented as in live sharks, and sanded where most of the denticle surface has
721 been removed (Fig. 5). All paired comparisons within similar motion programs were significant
722 at between P<0.001 and P>0.0001. The sanded foils swim at a 12.3 % average lower speed than
723 the intact foils.
724
725 Fig. 12. Time series of flow velocities over a whole motion cycle of the normal mako shark skin
726 membrane no.2 foil, swimming at its average self-propelled speed of 0.2 m s-1 (motion program:
727 2 Hz, 2 cm heave, 0° pitch). For the cycle period of 500 ms, images (above) and plots
728 (underneath) of 0, 100, 200, 300, 400 and 500 ms are shown. A cycle begins and ends at 0 ms
729 and 500 ms, respectively. Yellow arrows in DPIV images show velocity vectors. Colored
730 contours indicate velocity in x-direction (Vx) of downstream (red) and upstream (green) moving
731 fluid. Foil shape at each time is overlaid on each image. Black scale bar is 2 cm. Orange scale
732 vector is 1m s-1 and indicates flow direction from left to right. Values for Vx are taken along the
733 blue lines and plotted against the distance from the foil edge in the Vx-plots. The blue line starts
734 3-4 mm above the lower foil edge. In this area, Vx values approximate zero, so the actual foil
735 edge begins at the point where the Vx-plots shows the first slope. Note that negative velocities,
736 where the flow travels upstream, occur near the foil edge. White areas above the foil membrane
737 indicate areas that were in shadow and so no vectors were calculated in these regions.
738
739 Fig. 13. Images of streamlines showing flow patterns of an identical position in the motion cycle
740 of the intact (i, cyan, upper panel) and the sanded (s, light purple, lower panel) from the mako
741 shark skin attached to the rigid flat plate (A) with a motion program of 2 Hz, 2 cm heave, 10°
742 pitch and the mako membrane no.2 flexible foil (B) with a motion program of 2 Hz, 2 cm heave,
743 30° pitch. Values for the vorticity near the intact (i) and the sanded (s) foils are taken along the
744 red line transects and plotted against the distance from the foil edge on the right. Note that the
745 position of maximal vorticity is much farther away from the flexible foil surface on the sanded
746 foil than for the intact foil with denticles. Yellow scale bars = 1 cm; flow direction from left to
747 right.
748

24