Pacific Science (1985), vol. 39, no. 1 © 1985 by the University of Hawaii Press. All rights reserved

Design of a Small Cantilevered Sheet: The Sail of Velella velella1

LISBETH FRANCIS 2

ABSTRACT: The upright sail of the sailing hydrozoan Velella velella is sup­ ported by a very thin cantilevered sheet of colorless and transparent chitinous material. The skeletal material is a layered fibrous composite that is similar structurally to arthropod exoskeleton; but the appearance and mechanical pro­ perties (breaking stress, breaking strain, and stiffness or Young's modulus) are more similar to vertebrate hyaline cartilage. Since the homologous perisarc of some sessile hydroid species is both stiffer and stronger, the Velella skeletal material probably has not been selected evolutionarily for extreme strength or stiffness. Several specific design features make this thin cantilevered sheet of relatively floppy material a suitable support for Velella's permanent sail. The sail sheet is thicker than the rest ofthe skeleton, and is further reinforced by two overlapping patterns ofraised ridges. The sheet is triangular-to-semicircular, and this taper­ ing shape provides a larger cross section of material at the base to resist the greater bending moment there. A three-dimensional curve at the insertion line between sail and float provides more flexural stiffness, further reducing the tendency for the sheet to fold at the base. Consequently, the sail bends smoothly and progressively under an increasing load and quickly returns to the upright position when unloaded, rather than curling or kinking at the bottom. This, plus some tilting of the whole animal, may reduce stress on the sail in heavy gusting winds.

THE BY-THE-WIND SAILOR, Velella velella (Fig­ metazoans (Glaessner and Wade 1966, Cham­ ure la), is a conspicuous surface-dwelling hy­ berlain 1971, Yochelson, Sturmer, and Stan­ drozoan ofthe open ocean. Large flotillas are ley 1983). often reported by ships in both major oceans Since Totton (1954) reclassified these and from both hemispheres, from equatorial animals, creating the hydrozoan order Chon­ waters to latitudes as high as 55° (Bieri 1959, drophora, Velella velella and its near relative Savilov 1961). The animal is large: specimens Porpita porpita are described as free-floating, 5 to 6 cm long are not uncommon among the solitary (Mackie 1959, Fields and Mackie large numbers blown ashore on the exposed 1971) hydroid polyps. The large central beaches ofcentral California in the late winter mouth is surrounded by budding medusae and spring (pers. obs.); and Savilov (1961) within a ring of tentacles (Figure Ib, from reports specimens up to 13 cm long. The flesh Agassiz 1833). Extending beyond the edge of is bright blue, and the skeleton is complex, the oral disc is a thin, skirtlike flap of tissue, beautifully transparent, and colorless. Im­ the mantle, that adheres to the surface of the print fossils of obviously similar skeletons water and floats out around the animal like a from the Precambrian show that animals of blue halo. While the sessile hydroids typically this kind were among the earliest of earth's have an aboral stalk supported by a tubular, external skeleton (the perisarc), the pelagic chondrophorans lack a stalk and secrete the 1 Manuscript accepted 27 September 1984. 2 Bates College, Biology Department, Lewiston, Maine aboral skeleton within an invaginated pocket 04240. of ectoderm (Leloup 1929). A B

o F -

FIGURE 1. A whole specimen and a naked skeleton of Velella velella shown from several angles. Watercolor illustrations drawn by Agassiz (1833) show the whole animal (a) from a position broadside to the sail, (b) from underneath with the central mouth surrounded by budding medusae, a ring of tentacles, and the peripheral mantle, which clings to the water's surface when the animal is alive, and (c) from above, showing the shape of the float. Photographs ofa 5.6 cm long skeleton from Winter Harbor, B. C., provide (d) a lateral view showing the conspicuous raised ridges on the sail, (e) the oral (underwater) surface, and (f) another lateral view oriented at 90° to d, which clearly shows the three-dimensional cupping of the hull/float and the curvature of the sail. The Sail of Velella velella-FRANcIS 3

The chondrophoran skeleton is a gas-filled examination and tensile testing was quite float which provides shape and buoyancy for easy; the flesh was gently torn, allowing the the soft, fleshy parts ofthe animal. Thin sheets skeleton to slide free. Before cutting samples, of a chitin-containing, composite material I tracedthe outline ofthe skeletons as a record (Rudall 1955, Rudall cited in Fields and of the size ofeach individual. Pieces cut from Mackie 1971) form the walls of the float, the flat, sheetlike sail support were kept re­ which is subdivided into a series ofconcentric frigerated in seawater or in various other tubes (Figure Ie). The Porpita skeleton is rela­ solutions until they could be examined or tively flat and circular, while that of Velella is tested. To prevent desiccation during me­ cupped and rhomboidal, with a flat keellike chanical tests, wet specimens were dampened projection that extends above the water's sur­ with a few drops ofseawater or other aqueous face, where it acts as a sail (Figures Id, I). holding solution. Fossil skeletons of several kinds have been Near Friday Harbor, Washington, speci­ described, including some that are radially mens of the sessile hydroid Tubularia marina symmetrical and some that are bilateral, some were collected from docks and pilings and by that may have a sail sheet and some that do diving. These animals were kept in tanks sup­ not (Caster 1942, Glaessner and Wade 1966, plied with running seawater, and sections of Chamberlain 1971, Glaessner 1971, Wade the tubular perisarc were examined micro­ 1971, Yochelson, Sturmer, and Stanley 1983). scopically and used in tensile tests. Observers report that Velella velella orients its sail nearly broadside to the wind, and that Determining Dry Weight and Density it sails at angles ofup to 45° off the downwind direction (Mackie 1962). Four specimens of Velella that had been held frozen for 6 months were thawed and the Since Velella's skeletal sail sheet cannot be folded or furled during rough weather, I won­ skeletons freed offlesh. Large pieces ofskeletal material were cut from the sail support sheets, dered how that apparently vulnerable struc­ weighed, dried in a sealed container with a ture survives the rigors of the open ocean. Is chemical desiccant for 15 hours, and then re­ the skeletal material especially stiff and strong, or does the design avoid stress concentrations weighed to determine wet weight, dry weight, and water content of the material. that could cause kinking? To address these questions, I investigated the mechanics of the Toestimate the density ofthe Velella skeletal skeletal material and the structure and design material, small sections ofsail skeletal material of the cantilevered sheet that supports the from thawed specimens were placed in dextrose solutions whose density ranged from 1.09 to Velella sail. 1.28. When the specimen and solutions were free of bubbles, the density of the skeletal material could be estimated by noting the MATERIALS AND METHODS solutions in which the specimensjust began to sink or float. Collection and Handling ofSpecimens Microscopic Examination During April, May, and June of 1981, I collected and froze live specimens of Velella Whole specimens, hand-cut sections, and velella from among a large number that came sections cut with a freezing microtome were ashore near Santa Cruz, California. Several examined using compound and dissecting hundred were kept sealed in plastic bags and microscopes equipped with crossed polaroid held frozen as a large mass in a household filters and a first order red interference com­ freezer. Six additional specimens were collected pensator (Red I). The presence or absence of alive in July of 1983 near Winter Harbor, birefringence and the sign ofthat birefringence British Columbia, and kept refrigerated in were used as indications of macromolecular seawater until they were used. orientation in the material (Wood 1964, Frey­ Preparing the skeletons for microscopic Wyssling 1953). 4 PACIFIC SCIENCE, Volume 39, January 1985

I v

TUbUI8ri~••••••: 20

0° ••••••~ ... N 'e z t 2 UJ ... /' If) t 10 ~ ./ If) srl"a W u. a: I- If) .:: __----v EXTENSION _ .04 .08 .12 FIGURE 2. Sample force/extension curves for Velella skeletal material (solid line) and for a distal segment of STRAIN perisarc from a specimen ofthe hydroid Tubularia marina FIGURE 3. Sample stress/strain curves for fresh and dry (dotted line). As force was increased, the specimens specimens of Velella skeletal material and for a fresh stretched until they tore. Some specimens ofeach species sample of perisarc from Tubularia marina, showing broke suddenly, as this Velella specimen did, producing a stretching of the samples before they began to tear. (The sudden drop in the curve with little or no evidence of shape of these curves is different from that of the force/ terminal yielding. The tailing off at the end of that curve extension curves shown in Figure 2 because data from the was caused by some raggedness in the tear. Others, like terminal, descending portions of those curves was not the T. marina specimen shown here, yielded first, produc­ included here.) Differences in the slopes of these three ing some unevenness in the final section of the curve curves reflect large differences in material stiffness. before final failure.

fixed grip ofthe tensometer measured increas­ ing force, and a linearly variable differential Mechanical Testing transfonper attached to the movable grip To measure the tensile properties of the measured extension. Connected to an x/y materials, I stretched whole sections of the recorder, the electronic output from these Tubularia perisarc, and 5mm-wide, rectangu­ two instruments produced a continuous lar strips ofthe Velella sail support in a racklike force/extension curve (Figure 2). Machine device called a minitensometer. Before testing, compliance (deformation ofthe machine itself I measured the thickness of all samples and under stress) was measured by recording force the width of the perisarc samples using a and displacement with the grips attached microscope equipped with an ocular mi­ directly to each other. This was subtracted crometer. These data allowed calculation from total experimental deformation to give ofcross-sectional areas for the samples. Then the deformation of the specimen. the sample was mounted in the grips of the Increasing stress (force/cross-sectional area) tensometer. For wet sail material, the grips and natural strain (l n [instantaneous length/ were padded with damp strips ofpaper towel­ initial length]) were calculated from the ing to prevent slippage. Dry pieces ofsail were force/extension curves (Figure 3). In doing attached to each grip with a piece of tape those calculations, I made the following as­ before being clamped, and a drop of cyano­ sumptions: (1) that the sheet of material was acrylate glue was used to help secure the tubular of uniform thickness, (2) that there was no perisarc. After a sample was firmly mounted, change in sample volume during testing, (3) I measured the length of the section exposed that the specimen was deformed uniformly between the grips with vernier calipers. along its length, and consequently (4) that the By turning a hand crank, I stretched the changing cross section ofthe sample could be specimen at a rate of approximately 0.05 mm calculated by dividing the initial (calculated) per second. A force transducer attached to the volume by the instantaneous length (the initial The Sail of Velella velella-FRANcIS 5

length plus the measured displacement, minus mesoglea swell in distilled water and shrink in displacement due to machine compliance). concentrated saline solutions, becoming stif­ Since the specimens actually varied somewhat fer or less stiff, respectively, as they do so in thickness along their lengths, the calculated (Wainwright et al. 1976; Gosline 1971). In cross-sectional areas are only estimates. As a addition to measuring the stiffness ofsamples further consequence of this, the assumption held under different osmotic conditions, that the sample is distorted uniformly during I examined handcut cross sections of the testing must also be false. This means that material under a dissecting microscope before the calculated figures for Young's modulus and after changing the osmotic concentration (hereafter referred to simply as "stiffness" ofthe medium. Using an ocular micrometer, I or "material stiffness"), breaking stress measured the thickness ofthe sections, first in (hereafter called "strength"), and breaking seawater (29 ppt), and then at regular inter­ strain (called "extensibility" here) are only vals for several hours after placing them into order-of-magnitude estimates, and that it will distilled water, or a hypertonic solution of be possible to detect only relatively large dif­ sodium chloride (150 ppt or 300 ppt). ferences in the material properties of samples receiving different experimental treatments. Bending ofthe Sail under Load To explore the nature of the Velella material, and to determine whether handling The float of a freshly thawed Velella skele­ procedures might effect the mechanical pro­ ton was pinned to a mound of modeling clay perties, I cut and tested multiple samples from on a small sheet of lead. The skeleton, thus each sail sheet. To determine the effects of secured and with the sail exposed and upright, sample orientation, vertical and horizontal was placed in the bottom of a flow tank filled samples from the same five specimens were with flowing seawater (Vogel and LaBarbera tested and compared. To determine the effects 1978). As the water velocity in the tank was of freezing on the tensile properties of the slowly increased, it was possible to observe the material, two strips were cut from each of six bending of the sail under an increasing load. specimens. One from each specimen was tested immediately. The others were frozen in Modeling the Effects ofSail Shape on a standard household freezer, then thawed Bending and refrozen a total of five times before test­ ing. To further test the effects of handling, To explore the effects of sail shape and the several additional strips were cut from four of importance of the three-dimensional curve at the previous six specimens. One from each the base of the Velella sail, I worked with was sealed in a plastic bag, and aged in the paper models. Sheets of different shapes and freezer for several months before it was tested. sizes were cut from ordinary notebook paper One was held in distilled water for an hour. and taped to a flexible plastic ruler. The ruler One was held in an isotonic solution of mag­ provided support at the base of the sheet, nesium sulphate for an hour. One was dried making it possible to change the shape of the flat between two paper towels weighted with a "insertion line" simply by bending the ruler. book, and tested after four to six hours, when By blowing air against these models I could it was fairly dry. These data were then com­ mimic loading ofthe sail in gusting winds and pared to determine whether the treatments compare the bending and righting behaviors had substantially affected the tensile pro­ of various models. perties of the material.

Swelling and Shrinking ofthe Velella Skeletal RESULTS Material in Hypotonic and Hypertonic Solution General Observations Hydrostatically stiffened materials like The skeletal material of Velella looks quite vertebrate hyaline cartilage and anemone different from that of Tubularia species. The 6 PACIFIC SCIENCE, Volume 39, January 1985 perisarc of Tubularia marina varies from whit­ cedarwood oil. This large, variable compo­ ish or transparent at the distal tip to light tan nent is form birefringence (caused by oriented and somewhat opaque at the base, while the discontinuities between two materials with skeleton of Velella is colorless and quite trans­ different refractive indices-presumably ma­ parent throughout. The sail sheet of Velella is trix and chitin fibers in this case) rather than much thicker than the perisarc wall of Tubu­ birefringence intrinsic to the macromolecules laria species (commonly O.2mm for the sail themselves. support sheet of Velella, as compared with Regardless of how the section is cut (hori­ 0.05 mm for the distal perisarc wall of Tubu­ zontally, vertically, radially, or otherwise), laria marina). Nonetheless, the Velella ma­ cross sections ofthe Velella sail support sheet terial feels softer and more flexible when it is always show retardation by the Red I com­ handled. pensator when the long axis of the section is Although nonliving, both the skeleton of oriented parallel to the slow direction of the Velella and the perisarc of Tubularia are in filter. This indicates preferred orientation par­ contact with the living tissues of the animals, allel to the plane of the section. Random ori­ which may thicken and extend the skeletons entation ofchitin in thin layers parallel to the by adding layers ofmaterial. Cross sections of surface ofthe sheet would give this result, and the Tubularia perisarc are obviously layered; is therefore the choice of parsimony. and since large sails are thicker than small Cross sections of the Tubularia perisarc sails (data used to calculate specimen cross are also much more birefringent than are sections), the Velella skeleton is probably also flattened pieces of the whole tube. Again, thickened by accretion. this suggests preferred orientation of fibers through the thickness ofthe walls ofthe peri­ sarc tube, but little or no preferred orientation Dry Weight and Density ofthe Velella parallel to the surface ofthe tube. The polarity Skeletal Material of this birefringence is the same as for the Velella skeleton, indicating again that the The average water content of the Velella chitin fibers could be randomly arranged skeletal material (estimated by drying the within thin layers that are parallel to the sur­ specimens, and calculated using the arcsine face of the tube. transformation) was 87 percent (s.d. = 0.3 percent). The density of the Velella skeletal material, estimated using the gravimetric Mechanical Properties: Response in Tension method, was 1.28. Typical force/extension curves for Velella skeleton and for Tubularia marina perisarc are shown in Figure 2. The initial steepening of Birefringence the force/extension curve is characteristic of Mounted in seawater and viewed with fiber/matrix composites, and may indicate crossed polaroid filters through the broad flat any of several things: (1) the sample might surface, the sail support sheet of Velella shows simply be pulling taut and straightening be­ little or no birefringence, indicating that there tween the grips; (2) there may be some stretch­ is no net preferred macromolecular orien­ ing ofthe amorphous matrix before tension is tation in the plane of the sheet. Any cross transmitted fully to embedded chitin fibers; section of the sheet mounted in seawater is and (3) fibers may be realigning as the material strongly birefringent, indicating that there is is pulled. The middle section of the curve is preferred orientation of the macromolecules quite linear. Sometimes the material breaks through the thickness of the sheet. Birefring­ sharply (curve for the Velella specimen, Fig­ ence decreases progressively as cross sections ure 2) and sometimes it yields first, producing are dehydrated by transfer from seawater a flattened and uneven section in the curve to increasingly more concentrated alcohol (Tubularia specimen, Figure 2). Most speci­ solutions, then into toluene and finally to mens tore raggedly, producing a jaggedly de- The Sail of Velella velella-FRANcIS 7

TABLE I TENSILE PROPERTIES OF Velella SKELETON AND Tubularia PERISARC: EFFECTS OF SAMPLE ORIENTATION, HYDRATION, IONIC AND OSMOTIC STATES, FREEZE/THAw, AND SAMPLE AGE

E 0", S, N MATERIAL SOURCE STIFFNESS STRENGTH EXTENSIBILITY SAMPLE AND TREATMENT (MNm-2 ) (MNm-2 ) (%) SIZE

I. Velella velella, sail support sheet

Santa Cruz, CA freeze/thaw, horizontal strip *25 ± 10 0.4 ± 0.2 2.5 ± 0.05 5

SC, f/t vertical strip 15 ± 8 0.6 ± 0.2 4.0 ± 0.04 5

Winter Harbor, BC fresh sample 15 ± 6 3±2 19.5 ± 1.9 6

WH, f/t, fresh 17 ± 5 3±2 18.9 ± 1.6 6

WH, f/t, aged 27 ± II 4±2 9.9 ± 3.0 4

WH, f/t, distilled water 30 ± 14 7±3 9.8 ± 9.5 4

WH, f/t, 0.58M MgSO 24 ± II 3 ± I 1l.l±0.4 4

WH, f/t, dry 400 ± 157 5±3 9.3 ± 3.4 4

II. Tubularia marina, untanned perisarc

Friday Harbor, WA fresh 140 ± 19 20 ±4 14.6 ± 0.1 5

• Values given are means, plus or minus their standard deviations. The arcsine transformation was used. for extensibility calculations. NOTE: Abbreviations used are as follows: SC = Santa Cruz, California; WH = Winter Harbor, British Columbia; f/t = freeze/thaw; MN m-2 = meganewtons per square meter; 0", = maximum tensile stress at breaking = force/cross- sectional area; E = stiffness = u/E; s, = extensibility = {In(final length/initial length)} x 100. scending terminal portion (Ve/ella specimen, tions, some were consistently stiffer than the Figure 2). rest. Eight ofthe eleven specimens tested had The material responds similarly when the an average elastic modulus of1.2 x 107 N m-2 sheet is stretched in any direction. There are (s.d. = 0.22). The other three specimens were no large differences in the tensile properties of consistently about twice as stiff (E [stiffness] = 7 2 7 strips cut straight across the sail (at right 2.4 x 10 N m- ; s.d. = 0.61 x 10 ). angles to the water line), as compared with In addition to individual differences in stiff­ strips cut vertically (Table 1); nor are samples ness, there were also populational differences from the middle obviously different from peri­ in strength (Table 1, Figure 4). Those col­ pheral samples. This again indicates a lack of lected at Winter Harbor in 1983 were consis­ preferred orientation parallel to the plane of tently stronger than those collected at Santa the sheet. Cruz in 1981 (i.e., they broke at higher stress Not all specimens showed the same tensile and strain levels). properties (Figure 4). Within both popula- The 1983 Winter Harbor specimens also 8 PACIFIC SCIENCE, Volume 39, January 1985

2 burn" (freeze drying), because they also were ~ significantly stiffer than fresh samples from Ie the same skeletons (Table 1). The tensile prop­ z ~ erties were not affected by the osmotic char­ acteristics of the holding solution or by the f/) charge of ions in solution. The material re­ f/) w sponded similarly in tension when freshly re­ a: l- moved from the animal, when held in cold f/) seawater for up to a week, and when held in distilled water or in an isotonic solution of magnesium sulphate either for a few minutes STRAIN or for several hours. The distal tip of the perisarc of Tubularia FIGURE 4. Sample stress/strain curves for skeletal marina is stiffer and stronger than the wet material from four specimens of Velella, showing stretch­ ing of the samples until they began to tear, and demon­ skeletal material of Velella, and both mate­ strating differences in tensile properties within and be­ rials break at about the same strain levels tween populations. Samples from the 1983 population (Table 1 and Figure 3). collected at Winter Harbor, B. C. (w) broke at· higher strain levels than did samples from the 1981 population collected at Santa Cruz, Calif. (s). A few ofthe specimens Swelling and Shrinking ofVelella Skeletal from each population were about twice as stiff as the rest Material in Response to Desiccation and (e.g., the two steeper curves). Osmotic Variation had more conspicuous radial and concentric The Velella skeletal material shrinks drama­ ridge patterns than did the 1981 specimens tically when it is dried; however, in contrast from Santa Cruz; many of the 1983 sail sup­ to vertebrate hyaline cartilage, the Velella port sheets had jagged margins (Figure 1d), material neither swells nor shrinks in solutions while the 1981 specimens had smooth mar­ with osmotic concentrations that vary con­ gins. However, the differences in strength and siderably from those of natural seawater. extensibility were probably not due simply to sculpturing differences. Samples from rela­ Bending ofthe Sail: Behavior ofthe Specimen tively flat sections of highly sculptured sails in a Flow Tank showed the same properties as samples from more sculptured areas of the same sails. When the sail was loaded in drag under­ None of these differences within and be­ water at low water velocities, only the tip of tween populations were associated with indivi­ the sail bent. As the velocity was increased, the dual size. Some larger and some smaller speci­ whole sheet inclined at an increasingly steep mens were examined from both populations: angle. ofthe three stiffer specimens, one was among After several hours in the tank at one veloc­ the largest, one was among the smallest, and ity, the sail did not stretch or bend greatly one was in the middle ofthe size range ofthose beyond the initial response. When the force tested (ranges for maximum sail lengths: was removed by turning offthe flow, the speci­ Santa Cruz specimens, 3.0 to 5.7cm; Winter men quickly returned to the original position. Harbor specimens, 3.9 to 5.7 cm; less stiff Apparently the skeletal material does not specimens from both populations, 3.9 to creep appreciably under a small, steady load. 5.7 cm; stiffer specimens, 3.0 to 5.7 cm). Of the treatments tested, only drying had a Effect ofSheet Shape on Bending: Behavior detectable effect on the mechanical properties ofModels ofthe material. Samples that were deliberately dried before testing were much stiffer than wet A rectangular sheet of paper that is sup­ specimens (Table 1). Samples kept in a freezer ported along the straight, bottom edge will for 3 months apparently suffered "freezer tend to collapse by curling over near the base The Sail of Velella velella-FRANcIS 9

the curved sheet kinks readily when the same force is applied from the opposite direction, against the convex face (Figures 5b-d). A slight S-shaped curve at the base increases the flexural stiffness equally for loading from either direction (Figure 5e). A

DISCUSSION Although the skeletal material of Velella velella has been identified as a chitinous com­ posite (RudallI955; Rudall cited in Fields and Mackie 1971), the mechanical properties of the material, including macromolecular ori­ entation and the functional morphology of the skeleton itself, apparently have not been c investigated. In the discussion below, I com­ pare the properties ofthe Velella material with homologous and analogous materials, discuss the probable microstructure of the material, and present an analysis of miniature sail design. I am presently studying the sailing dy­ namics of Velella and the behavior of living animals. Effects of body design and behavior E on orientation and movement will be dis­ FIGURE 5. Bending ofpapermodels in a wind. Sheets of cussed at a later date. notebook paper were taped to a flexible plastic ruler and their behavior was observed as puffs of air (indicated by arrows) were blown at right angles to the center of one Material Properties and Submicroscopic surface. A flat rectangular sheet (a) tends to kink at the Structure base even without wind. A flat triangular sheet (b) bends more gradually along its length. A triangular sheet with a In the realm of animal skeletal materials, simply curved base remains upright at relatively high there is nothing very remarkable about the velocities if the wind blows from inside the curve (c); but tensile properties of the Velella skeletal above some critical velocity, wind from the opposite direction causes kinking (d). An S-shaped curve at the material. It is neither particularly strong nor base ofthe sheet (e) supports it equally well no matter how unusually extensible. This is probably not the wind blows. surprising, considering its function. A force strong enough to stretch or tear the skeleton would probably capsize the animal or tear the of the sheet (Figure 5a). In contrast, a flat flesh free instead. triangular sheet that is held upright, supported What does seem surprising is that the ma­ only at the bottom edge, tends to remain terial is also not very stiff. Itis about as stiffas upright. Blow gently against one face of the vertebrate hyaline cartilage, but less stiff than triangular sheet and it bends slightly (Figure insect cuticle by more than two orders ofmag­ 5b), inclining more steeply in a more vigorous nitude (Table 2). In fact, as the name "chon­ wind and recoiling when the force is removed. drophoran" would suggest, the Velella skele­ When the wind blows against the concave ton looks and acts remarkable like vertebrate side of a simply curved sheet, the sheet re­ hyaline cartilage. The two materials have mains upright under much heavier loads than about the same tensile stiffness, strength, and will a flat sheet of the same dimensions; but extensibility (Table 2); however, changes in to PACIFIC SCIENCE, Volume 39, January 1985

TABLE 2

COMPARATIVE DATA: THE TENSILE PROPERTIES OF SOME PLIABLE MATERIALS

DENSITY EXTENSIBILITY STRENGTH STIFFNESS DATA MATERIAL (Kgm-3) (% increase) (Nm-2 ) (Nm-2 ) SOURCE·

Organic rubbers 1.3 X 103 >200 3 x lOs .6-4 x lOs t

Anemone mesoglea > 200 3-4 3 x 104 •

Vertebrate hyaline 3 7 cartilage 1.1 x 10 9-31 2-5 x lOs 1.3 X 10 t, t

Wet Velella 3 7 skeleton 1.3 x 10 2-20 4-30 x lOs 2 X 10 §

Dried Velella 6 8 skeleton 9 5 x 10 4 X 10 §

7 8 Tubularia perisarc 14 2 x 10 1.4 X 10 §

3 7 9 Locust cuticle 1.2 x 10 9.6 X 10 2 X 10 t

*Sources ofdata: (*) Alexander 1968; (t) Currey 1970; mYamada 1973; (§) new data reported here. A blank space indicates lack of information. the osinotic concentration or the ionic compo­ alignment. Structurally, fibers arranged paral­ sition ofthe medium do not measurably affect lel to the surface of a sheet can also provide either the dimensions or the stiffness of the much greater reinforcement, both in tension Velella skeletal material. This suggests that and in shear, than fibers arranged at right the material is not stiffened hydrostatically, as angles to the plane ofthe sheet (Wainwright et is hyaline cartilage. al. 1976). Arthropod exoskeleton, nematode Indirect evidence from polarizing micro­ cuticle, and fish skin are also fiber/matrix scopy suggests that the chitin fibers in the sheets that are birefringent in cross section Velella skeleton may be randomly arranged (Wainwright et al. 1976). within thin layers, with relatively few fibers running through the thickness of the sheet. Evolution ofthe Velella Skeleton However, this is a weak deduction which should be confirmed or denied by fixing, stain­ Cnidarians may have anticipated the ar­ ing, and sectioning the material and examin­ thropods in developing a layered chitinous ing it using transmission electron microscopy. exoskeleton. Most, if not all, natural fiber/matrix sheets Members of all three cnidarian classes are are birefringent in cross section, because the known to secrete cuticular chitinous struc­ fibers tend to be oriented parallel to the tures (Rudall 1955, Neville 1975, Muzzarelli sheet's surface. Developmentally, this may be 1977). Examples include the perisarc of the because these structures are secreted as thin hydrozoans, Tubularia and Velella (Rud­ layers; relatively long fibers within a thin sheet all 1955), the calcareous skeleton of the are almost certain to show preferred orienta­ (anthozoan) coral Pocillopora damicornis tion parallel to the surface of the sheet. For (Wainwright 1962), and the cuticle of poly­ example, noodles poured onto a flat surface poid generation of the scyphozoan jellyfish show this alignment, and Wainwright (1963) Aurelia aurita (Chapman, 1968). argues that the mechanical interactions of Electron micrographs of the Aurelia polyp fibers and crystals during the formation of cuticle (Chapman :1968), show a parabolic ceramic composites should produce a similar pattern that is typical of the helicoidal ar- The Sail of Velella velella-FRANCIS II rangement of fibers found in the arthropod cuticle (Richards 1951, Neville 1975). Since we do not yet know the details of fiber ar­ rangement in the layered skeletons of Tubu­ laria, Velella, or the fossil chondrophorans, it is possible that there, too, the arrangement is 1 helicoidal. Like the arthropods, modern sessile hy­ droids produce long, thin support structures by enclosing parts ofthe body in an external, 2 tubular sheath; and like the arthropods, some hydroids construct flexible joints that allow them to bend the stiffcuticular tube (Murdock 3 1976). However, these animals also retain the typical cnidarian commitment to feeding with nematocysts, which requires exposing the soft tentacles to contact rather thancompletely en­ casing the body in armor as the arthropods do. While soft flesh completely enfolds the elaborate chondrophoran skeleton, the deri­ vation ofthat complex endoskeleton from the tubular perisarc of sessile hydroids is not dif­ ficult to imagine (Figure 6). Since it is secreted inside a fleshy pocket instead ofon the outside ofthe stalk, the chondrophoran skeleton must be inside-out relative to the perisarc tube (Fields and Mackie 1971). Ifone end ofa tube were flattened, it would form a double-layered sheet, like the sail support. (In fact, when the sail sheet of Velella is cut into very thin strips, . 3 it delaminates readily into two equal layers, indicating that it probably is formed as two FIGURE 6. Suggested derivation of the Velella skeleton adjoining sheets.) By abruptly belling out from an inverted, tubular perisarc. To form the sail, the below the sail, the tube could form the upper tube is flattened at the top (arrows at 1) and the oppressed surface ofthe float, and by rapidly pinching in faces are joined. An abrupt bulge in the middle (2) is immediately constricted again at the base (3), producing a again, it could form the lower surface of that flattened baIloon, which is the first, centralchamber ofthe float. As the animal grows, the sail support gas-fiIled float. sheet is easily enlarged by the addition of material at the edges. Enlarging the float re­ quires adding gas-filled tubes at the periphery. the wet Velella skeleton by an order ofmagni­ Since the Velella skeleton is secreted and tude (Table 1), we must conclude that natural molded by the ectoderm, it would presumably selection has not favored maximal stiffness for take the shape ofthe enfolding flesh, so drastic this material. distortions ofthe kind proposed here might be Apparently it is possible, and perhaps quite simply accomplished. even advantageous, to construct a thin, free­ standing sheet of relatively floppy material. Design ofCantilevered Sheets: The Velella The question, then, is how such a structure can be stiff enough to stand upright. A flat Sail Sheet sheet of very floppy material would simply Since the dried skeleton of Velella and the crumple under its own weight, while a flat wet perisarc of Tubularia are both stiffer than sheet of slightly stiffer stuff would tend to 12 PACIFIC SCIENCE, Volume 39, January 1985

FIGURE 7. Three successively more realistic models showing the shape ofthe joint between the sail and the low, cone­ shaped float of Velel/a. (a) A flat sail intersects the upper surface ofa cone, producing an inverted, V-shaped insertion. (b) The base of the sail is often S-shaped rather than straight, so the arms of the inverted "v" are curved in opposite directions along the surface ofthe cone. (c) The sail surface is not vertical, but slopes oppositely at the two ends. Both the steepness of the cone and the curvature of the sail are exaggerated here.

collapse by bending near the bottom (Sal­ the addition of superficial thickened ridges. vadori and Heller 1963, Figure 5a). The degree of sculpturing varies considerably In engineering shorthand, the sail is a can­ among specimens, but the patternis quite con­ tilevered, sheetlike beam that is loaded in sistent, and strikingly reminiscent of leaf ve­ bending (Wainwright et al. 1976). The bend­ nation patterns, (Figure Id). A set of radial ing resistance (flexural stiffness) of a beam is ridges that fans out from the middle of the the product of the material stiffness (E) and bottom edge could be described as palmate the second moment of area (I), which is a venation. Another set runs parallel to the free complex morphological term describing the edge of the sail and looks like concentric distribution of material in the beam's cross growth lines. The concentric set plus the cen­ section (Wainwright et al. 1976). Given that tral vertical rib ofthe radial set form an oppo­ the Velella skeletal material is not very stiff, sitely branching, pinnate venation pattern. the resistance of the sail to bending must be For the same net increase in weight, this build­ largely due to the shape of the beam's cross up ofmaterial as a pattern ofthickened ridges section, described by the second moment of places material farther from the axis of bend­ area. ing than would uniform thickening of the An obvious way to stiffen a beam by in­ whole sheet. In addition, this specific arrange­ creasing the second moment ofarea is to make ment of the ridges provides disproportionate that beam stouter. A thick sheet is stiffer in reinforcement vertically, at the midline. The bending than a thin sheet ofthe same material. palmate set converges in the middle at the However, doubling the thickness of a flat bottom of the sail, and the pinnate set pro­ sheet more than doubles its flexural stiffness, duces a series of V-shaped reinforcements because material contributes to the I compo­ along the central rib, where the shape of nent of flexural stiffness in proportion to the the pointed sail tip is repeated as a series of square ofits distance from the axis ofbending thickened growth lines (Figure ld). ofthe beam. The axis ofbending, in this case, The second moment of area may also be is the plane at the center ofthe sheet. Not only increased by changing the shape of a beam is there more material to bend in a thicker without adding material. Many specimens of sheet than there is in a thinner one, but the Velella have a slight, S-shaped curve at the additional material is all farther from the axis bottom of the sheet (Figures Ie, 5e, 7b, 7e), ofbending than that in the thinner sheet. The which increases the flexural stiffness equally sail support of Velella is about twice as thick for loading from either direction. as the walls ofthe float, and four times as thick Under some circumstances, a somewhat as the perisarc of Tubularia marina. flexible sheet may be more advantageous than But a thicker sheet is heavier and more a maximally stiffone. Ifit is brittle or deforms expensive to make than a thin sheet. For a plastically, a stiff sheet may crack or kink relatively smaller increase in weight, the sail permanently when bent. A flexible structure support of Velella is stiffened further by that is tough and springy may bend smoothly The Sail of Velella velella-FRANcIS 13

when loaded, and recoil when the force is is not a sharp right angle, but a smoothly removed. Many biological structures respond sloping curve (Figure Ie). The impression this to transitory forces in this accommodating, gives is that the sail is twisted oppositely at stress-reducing fashion. Leaves, branches, either end where the concave face ofthe sheet and whole trees sway in the wind (Vogel becomes the upper surface ofthe float (Figure ) 1984). Skins and arteries stretch and recoil as 7e). When loaded in drag by wind blowing they are repeatedly loaded and unloaded against one face, the sail angles backward, and (Wainwright et al. 1976). By bending in high tension along that face can be transferred to winds, the sail of Velella presents a lower pro­ the upper surface at one end of the float. The file, thereby reducing the drag forces acting on opposite, leeward face may then act as a leaf it. To be advantageous, this bending must be spring, absorbing and storing energy as the smooth and reversible rather than sharp and sail is bent backward. Bubbles of gas that catastrophic. extend into the base of the sail (Figure Id) Because the sail projects upward and at may also be compressed, reversibly storing right angles to the surface ofthe float, it could energy and providing an outward pressure be quite vulnerable to failure by kinking at the that would counteract any tendency for kink­ base. Two design features reduce that risk: (1) ing when the sail is bent. Bending of the sail the shape ofthe sail itselfand (2) the design of and tilting of the whole animal provides a the joint between the sail and float. purely passive mechanism for reducing stress The pattern of bending under an increasing and moderating sailing speed in the face of load depends not only on the characteristics of unpredictably variable wind velocities. the material and the thickness ofthe sheet, but also on the shape ofthe cantilevered sheet and the way that it is supported at the base. Under SUMMARY uniform loading, a square or rectangular sheet tends to bend suddenly near the base when The skeletal material of Velella is a layered the load reaches a critical level (Figure Sa), chitinous composite that is neither especially because the momentarm is longest at the base, stiff nor particularly strong. Various design providing the greatest bending moment there. features increase the flexural stiffness of the This may explain the advantage to Velella of cantilevered sheet that supports the thin, flat having a triangular to semicircular sail rather sail ofthe animal: the sheet is thicker than the than a rectangular one. The tapered shape pro­ rest ofthe skeleton and decorated with a pat­ vides an increasingly larger cross section of tern of raised ridges. However, a somewhat material toward the bottom of the sail, where flexible sail is probably better design, in this it resists the increasing bending moment. case, than a maximally stiffone. The sail shape The insertion line between sail and float is a and sculpturing, and the shape ofthe insertion three-dimensional curve (Figures 1, 7e) which line between sail and float, all help to prevent provides an additional increase in the second kinking as the sail bends and recoils, absorb­ moment ofarea and better resistance to bend­ ing and reducing the drag forces acting on it ing at the base of the sail. From the side, the under what must commonly be very caprici­ insertion line is an inverted "V", where the sail ous wind conditions. attaches to the low, cone-shaped, upper sur­ Unlike many other natural cantilevered face ofthe float (Figures Id, 7a). Itis therefore sheets, the skeletal sail sheet is relatively impossible for the sail simply to fold over at simple, both structurally and functionally: (1) the base by bending straight along the inser­ it is a solid structure, made entirely ofa single tion line. Many specimens also show a slight S­ kind of composite material, without conduc­ shaped curve at the base of the sail (Figure tive elements, and without included living 7b). This again increases the second moment tissue of any kind, and (2) its only known ofarea, providing greater bending resistance, function is support. This makes it a good especially at the base ofthe sail. Furthermore, subject for the analysis of design in natural the joint between the float and the sail support cantilevered sheets. 14 PACIFIC SCIENCE, Volume 39, January 1985

ACKNOWLEDGMENTS GLAESSNER M. F. 1971. The genus Cono­ medusites Glaessner and Wade and the di­ I thank: S. Vogel and M. LaBarbera for versification ofthe Cnidaria. Paleont. Zeits. guidance and encouragement; K. P. Irons, S. 45:7-17. Smiley, R. L. Miller, and S. F. Norton for col­ GLAESSNER, M. F., and M. WADE. 1966. The lecting specimens of Velella and Tubularia; and Late Precambrian fossils from Ediacara, the visitors, staff, and faculty of the Friday South Australia. Paleontology 9: 599-628. Harbor Laboratories, and the Duke University GOSLINE. J. M. 1971. Connective tissue me­ Zoology Department for valuable ideas, in­ chanics ofMetridium senile: II. visco-elastic formation, and advice. The work was largely properties and macromolecular model. done at the University of Washington's Fri­ J. Exp. BioI. 55: 775-795. day Harbor labs with electronic equipment LELOUP, E. 1929. Recherches sur l'anatomie et generously supplied by M. LaBarbera. S. A. Ie developpement de Velella spirans Forsk. Wainwright and B. A. Best kindly read and Arch. BioI. Paris 39: 397-478. commented on the manuscript, and Lisa J. MACKIE, G. O. 1959. The evolution of the Croner drew Figures 5 and 7. Chondrophora (Siphonophora-Disconan­ thae): New evidence from behavioral studies. Trans. Roy. Soc. Canada: 53, ser. 3, sect. 5: 7-20. LITERATURE CITED ---. 1962. Factors affecting the distri­ bution of Vellela velella (Chondrophora). AGASSIZ, A. 1833. Exploration of the surface Internationale Revue der Gesamten Hy­ fauna of the Gulf Stream III, Part I. The drobiologie 47: 26-32. Porpitidae and Velellidae. Mem. Mus. MURDOCK, G. R. 1976. Hydroid skeletons and Compo Zool. Harvard 8(2): 1-16. fluid flow. In G. O. Mackie, ed. Coelente­ ALEXANDER, R. McN. 1968. Animal mechan­ rate ecology and behavior. Plenum, New ics. Univ. ofWash. Press, Seattle. 346 pp. York. BIERI, R. 1959. Dimorphism and size distri­ MUZZARELLI, R. A. A. 1977. Chitin. Per­ bution in Velella and Physalia. Nature gamon Press, New York. 309 pp. 184: 1333. NEVILLE, A. C. 1975. Biology of Arthropod CASTER, K. E. 1942. Two siphonophores from cuticle. In D. S. Farner, ed. Zoophysiology the Paleozoic. Palaeontographica Amer. and ecology, Volume 4/5. Springer-Verlag, 3(14).34 pp. New York and Ithaca. New York. 448 pp. CHAMBERLAIN, C. K. 1971. A "by-the-wind­ RICHARDS, G. 1951. The integument of Arth­ sailor" (Velellidae) from the Pennsylvanian ropods. Univ. ofMinnesota Press, St. Paul. Flysch of Oklahoma. J. Paleont. 45: 411 pp. 724-728. RUDALL, K. M. 1955. The distribution of CHAPMAN, D. M. 1968. Structure, histo­ collagen and chitin. Symp. Soc. Exp. BioI. chemistry and formation of the podocyst 9:49-70. and cuticle of Aurelia aurita. J. Mar. BioI. SALVADORI, M., and R. HELLER. 1963. Struc­ Ass. U. K. 48: 187-208. ture in architecture. Prentice-Hall Inter­ CURREY, J. D. 1970. Animal skeletons. Studies national Series in Architecture, M. Salva­ in biology, no. 22. Edward Arnold, Lon­ dori, ed. Prentice-Hall, Englewood Cliffs, don. 52 pp. New Jersey. 370 pp. FIELDS, W. G., and G. O. MACKIE. 1971. Evo­ SAVILOV, A. I. 1961. The distribution of the lution ofthe Chondrophora: Evidence from ecological forms of the by-the-wind sailor, behavioral studies on Velella. J. Fish. Res. Velella lata Ch. and Eys., and the Portu­ Bd. Canada 28: 1595-1602. guese man-of-war, Physalia utriculus (La FREY-WYSSLING, A. 1953. Submicroscopic Martiniere) Esch., in the North Pacific. morphology of protoplasm. Elsevier Pub. Trudy Inst. Okeanol. Akad. Nauk SSSR Co., New York.411 pp. 45: 223-239. The Sail of Velella velella-FRANcIS 15

TOTTON, A. K. 1954. Siphonophores of the WAINWRIGHT, S. A., W. D. BIGGS, J. D. Cur­ Indian Ocean. Discovery Rep. 27: 1-162. rey, and J. M. GOSLINE. 1976. Mechanical VOGEL, S. 1984. Drag and flexibility in sessile design in organisms. Edward Arnold, Lon­ organisms. Amer. Zool. 24: 37-44. don. 423 pp. VOGEL, S., and M. LABARBERA. 1978. Simple WOOD, E. A. 1964. Crystals and light, an flow tanks for research and teaching. Bio­ introduction to optical crystallography, science 28: 638-643. 2nd rev. ed. Dover, New York. WADE, M. 1971. Bilateral Precambrian chon­ YAMADA, H. 1973. Strength of biological drophores from the Ediacara fauna, South materials. Krieger, New York. 297 pp. Australia. Proc. R. Soc. Vic. 84: 183-188. YOCHELSON, E. L., W. STURMER, and G. D. WAINWRIGHT, S. A. 1962. An anthozoan chi­ STANLEY. 1983. Plectodiscus discoideus tin. Experientia 18: 1-3. (Rauff): A redescription of a chondro­ ---.1963. Skeletal organization in the co­ phoran from the Early Devonian Hunsruck ral, Pocillopora damicornis. Quart. J. Micr. slate, West Germany. Palaont. Z. 57: Sci. 104: 169-183. 39-68. .