Interfibril hydrogen bonding improves the strain-rate response of natural armour

Abstract

Fish scales are laminated composites that consist of plies of unidirectional collagen fibrils with twisted-plywood stacking arrangement. Owing to their composition, the toughness of scales is dependent on the intermolecular bonding within and between the collagen fibrils. Adjusting the extent of this bonding with an appropriate stimulus has implications for the design of next-generation bioinspired flexible armours. In this investigation, scales were exposed to environments of water or a polar solvent (i.e. ethanol) to influence the extent of intermolecular bonding, and their mechanical behaviour was evaluated in uniaxial tension and transverse puncture. Results showed that the resistance to failure of the scales increased with loading rate in both tension and puncture and that the polar solvent treatment increased both the strength and toughness through interpeptide bonding; the largest increase occurred in the puncture resistance of scales from the tail region (a factor of nearly 7×). The increase in strength and damage tolerance with stronger intermolecular bonding is uncommon for structural materials and is a unique characteristic of the low mineral content. Scales from regions of the body with higher mineral content underwent less strengthening, which is most likely the result of interference posed by the mineral crystals to intermolecular bonding. Overall, the results showed that flexible bioinspired composite materials for puncture resistance should enrol constituents and complementary processing that capitalize on interfibril bonds.

1. Introduction

Fish scales are an interesting structural material that serves as a dermal armour [1–6]. Along with other dermal armours, including those of the alligator, turtle, etc. (e.g. [7–12]), fish scales have been identified as a source of inspiration in the design and development of new concepts for protection [13–17].

The scales of modern fish have evolved into four primary groups, including cosmoid, elasmoid, ganoid and placoid scales [18,19]. The primary building blocks of the scales in each of these groups are type I collagen fibres and apatite [20–23]. In contrast to the high flexibility of elasmoid scales, the other three are relatively stiff owing to their larger thickness and the presence of bony regions with higher mineral content [24]. The flexibility of the last three scale assemblies is achieved by articulation between the scales. By contrast, elasmoid scales are much thinner and stacked or over-layered with respect to one another, which enables flexing of the scales themselves as needed for locomotion. They are more common on fish with greater speed and are clearly multi-functional in light of their required flexibility and puncture resistance [2,25]. In fact, elasmoid scales are inspiring the designs of protective ‘skins’ with imbricated scale-like structures [16].

The microstructure of elasmoid scales consists of two primary layers, including a thin outermost highly mineralized layer and the secondary interior layer, which occupies the bulk of the scale thickness as shown in figure 1a. The outermost limiting layer (LL) is a highly mineralized matrix of calcium-deficient apatite crystals that is reinforced by a sparse dispersion of collagen fibres. The underlying elasmodine layer consists of plies of mineralized unidirectional type I collagen fibrils that are organized into a laminate. The elasmodine layer actually consists of two separate layers [26], both consisting of discrete plies of unidirectional collagen fibres with a specific stacking sequence [27]. The internal elasmodine (IE) and the external elasmodine (EE) layers are only differentiated by the higher mineral content of the latter [21]. Both the EE and IE have substantially lower mineral content than the outermost LL [28]. The plies of the entire elasmodine layer are arranged in the form of a twisted-plywood structure, with rotation between plies ranging between 60° and 90° depending on the fish [29]. For the carp, the average angle of rotation between plies is 75°, in comparison with close to 90° rotation for the arapaima [27].

Figure 1.

Details of the scales and the effects of exposure to ethanol. (a) Carp scale cross-sections for representative scales from the head and tail regions. The three principal layers are highlighted including the limiting layer (LL), external elasmodine (EE) and internal elasmodine (IE). (b) Three-dimensional arrangement of the tropocollagen triple helix of glycine, proline and hydroxyproline. The development of inter- and intrafibrillar hydrogen bonds stimulated by exposure to ethanol are highlighted as dashed yellow lines. (Online version in colour.)

The development of bioinspired structural materials generally involves understanding the microstructure of existing natural systems and then mimicking that structure with an appropriate set of synthetic material analogues and methods of processing [30]. Nevertheless, biological structural materials can also serve as a vehicle for exploring new approaches to amplify strength and toughness of synthetic materials [31,32] or to explore ways to activate property changes in response to a stimulus. For instance, in a recent study by Murcia et al. [28], the stiffness and strength of elasmoid scales were increased by exposure to a polar solvent. The change in mechanical behaviour was nearly instantaneous and apparently resulted from interfibril hydrogen bonding of the collagen enabled by the displacement of water molecules by ethanol (a weaker hydrogen-bond-forming solvent). The removal of water allows hydrogen bonds to develop between the peptide chains [33]. Although hydrogen bonds are rather weak because of their lower energy when compared with covalent bonds, a high number of interpeptide bonds can initiate substantial changes to the mechanical behaviour of single fibrils [34] and dense fibril systems [35].

In contrast to the increases in tensile strength and stiffness of fish scales with exposure to ethanol, Murcia et al. [28] reported that the change in the mode III tear resistance to failure was minimal; this type of loading corresponds to what is called ‘tearing mode’, in which the direction of shear is oriented out of plane with regard to the plane of the structure. The different behaviours in tension and out-of-plane shear were interpreted to result from the interference of interfibril bonding and mineral to the reorientation of the collagen fibrils in mode III tear failures. However, Nalla et al. [36,37] showed that tooth dentin, a relatively highly mineralized tissue, underwent a significant increase in its resistance to crack growth under mode I loading after exposure to polar solvents, citing the same mechanism. Thus, both the nature of loading and relative mineral content appear to be important to the effects of interfibril bonding on the structural behaviour of collagen-based tissues.

The design of fish scales has evolved to be resistant to puncture, which occurs through dynamic loading. Although the resistance of scales to indentation and puncture has received considerable attention (e.g. [4,14,38–42]), the importance of dynamic loading has not [20]. In addition, the contribution of interpeptide bonding to the strain-rate characteristics of this interesting structural material has not been reported. Therefore, in this investigation, the mechanical behaviour of elasmoid fish scales was evaluated under uniaxial tension and transverse puncture over a large range of strain rates. The primary objectives were to: (i) evaluate the effects of changes in interpeptide bonding achieved by application of different solvents on the toughness of the scales in each loading orientation and (ii) assess the relative importance of this bonding on the mechanical behaviour as a function of loading rate. The roles of mineral reinforcement and extent of interpeptide bonding on the resistance to failure are examined and the importance of the findings to bioinspired designs of armours is discussed.

2. Material and methods

Scales of the Cyprinus carpio (i.e. the common freshwater carp) were obtained by extraction from across the body of several fish. These fish were marketed as East Asian carp, and no additional information was available from the distributor. Each of the fish was selected to ensure that the weight, number of scales along the lateral nerve line and microstructure of the scales were consistent. The scales were obtained nearly equidistant between the ventral and dorsal aspects of the body (i.e. adjacent to the midline) and from three regions including near the head, mid-length (beneath the dorsal fin and regarded as ‘middle’) and near the tail (figure 1a). To be consistent in the collection for all fish, these regions were defined at distances from the gill plate of approximately 15%, 40% and 80% of the total number of scales counted along the lateral line over the fish length after previous work [43]. All of the scales were less than 1 mm thick and possessed a diameter that varied according to the anatomical position but generally between 1 and 2 cm. After extraction, the scales were stored in Hanks' balanced salt solution (HBSS) at 3–5°C for less than one month.

To support an evaluation of the tensile properties, conventional dog-bone-shaped tensile specimens were sectioned from the scales using a dedicated punch and stamping process [21]. A single specimen was stamped from the centre region of each scale where the thickness is most uniform. The specimens possessed a gauge section length and width of 5.5 mm and 1.5 mm, respectively. Owing to anisotropic behaviour in scales from the head region [26], all of the specimens were obtained with alignment parallel to the fish length, for consistency.

After sectioning, the specimens were placed in a bath of HBSS or an ethanol bath at 3–5°C for a minimum of 24 h. The effectiveness of polar solvents to form hydrogen bonds in collagen is often described in terms of the Hansen solubility parameter for hydrogen bonding (δ_h). Polar solvents with high δ_h values preferentially form hydrogen bonds with collagen peptides, thereby preventing the molecules from developing interpeptide bonds. Water has one of the highest reported δ_h values for polar solvents 37.3 (J cm⁻³)^1/2. In the absence of water, or when the matrix is immersed in solvents with δ_h below 19 (J cm⁻³)^1/2, which is the value postulated for collagen in air [33], interpeptide hydrogen bonding occurs (figure 1b). Thus, exposure of the scales to ethanol with δ_h of 19 (J cm⁻³)^1/2 is perceived to cause an increase in the stiffness and strength of the scales due to the displacement of water molecules between the tropocollagen triple helix and development of interpeptide hydrogen bonds [44].

Tensile testing of the hydrated fish scale specimens was performed to failure at room temperature. The loading was performed under displacement control using a commercial universal testing machine (ElectroPuls E1000; Instron, Norwood, MA, USA) equipped with a load cell having a full-scale range of 250 N and load precision of 0.01%. Stroke rates were selected to obtain strain rates from 10⁻⁴ to 10² s⁻¹ in decades, which were estimated according to the specimen gauge length. All specimens were tested to failure. Five specimens were evaluated for each strain rate and anatomical region (head, middle and tail), which resulted in 7 rates × 3 regions × 5 specimens = 105 total tests for each liquid environment. Results of the experiments performed on the scales tested in HBSS were also reported in Ghods et al. [43] to explore the strain-rate dependence of the mechanical behaviour in HBSS. An additional 105 experiments were performed on the scales subjected to the ethanol treatment.

The elastic modulus (E), strength (S) and modulus of toughness (MOT) were determined from results of the tension tests at each rate using the engineering stress–strain definitions. The elastic modulus was determined using the tangent method for strains less than 1% and the strength was defined by the maximum stress realized by the sample. The MOT was calculated by integrating the area under the stress–strain curves as a function of strain until failure.

Puncture tests were conducted on specimens from the three anatomical regions using a sharp stainless steel punch with a tip radius of curvature of 40 µm and included an angle of about 30°. Loading was conducted using a universal testing machine (model ELF 3300; BOSE, Eden Prairie, MN, USA) on scales clamped within a dedicated fixture at rates that ranged from 0.05 to 50 mm s⁻¹. The scales were unsupported and clamped around the periphery with the exposed portion of 12.7 mm. With scales having an average thickness of approximately 0.5 mm, and ignoring the out-of-plane deflection, the loading rates corresponded to effective strain rates of approximately 10⁻¹ to 10³ s⁻¹; the actual strain rates for the puncture tests varied according to the scale thickness. In all testing, the scales were arranged with the LL oriented towards the punch. The load and displacement signals were acquired at a rate of 4 kHz. Similar to the evaluation performed using tensile loading, five specimens were evaluated for each strain rate and anatomical region (head, middle and tail), which resulted in 4 rates × 3 regions × 5 specimens = 60 tests for each liquid environment. Consistent with the tensile testing, scales were evaluated after 24 h treatment in HBSS or in ethanol. Results of the experiments performed with HBSS storage were reported in Ghods et al. [43] to evaluate the strain-rate dependence of the impact behaviour. An additional 60 puncture experiments were performed on the scales subjected to the ethanol treatment, for a total of 120 experiments overall.

The scale performance under puncture loading was evaluated in terms of the stiffness at the onset of loading, the load to initiate and complete puncture, as well as the corresponding work to puncture. The measures of work were obtained by integrating the load-line displacements to the displacement of interest. All data obtained from the tensile and puncture testing were analysed in terms of the rate of loading and solution and are freely available for public access. The properties were compared using an analysis of variance (ANOVA) with p ≤ 0.05 used to define significant differences.

Both the tensile and puncture testing were performed in air and on specimens that were pre-soaked in the appropriate solvent, either HBSS or ethanol. Each of these two liquids was volatile enough that the scales could begin to undergo dehydration during the testing that was conducted at the lowest loading rates. Therefore, the specimens were treated throughout the experiments with a pipette and the appropriate solution (HBSS or ethanol) to prevent the potential effects of dehydration associated with free convection. Droplets of the solution were applied immediately after the specimens were mounted in the grips and sequentially during the testing. This process maintained the treatment condition.

The microstructure of scales within each region of the evaluation was examined using scanning electron microscopy (model JSM-6010PLUS/LA; JEOL, Peabody, MA, USA). The samples were sputtered with Au/Pd and observed in secondary electron imaging mode. In addition, selected scales were evaluated after puncture testing using micro-computed tomography (microCT) scanning with a North Star Imaging model X5000. For this effort, the scales were treated with a very light coating of diluted ethylene glycol after Ryou et al. [45] to maintain the moisture content within the scale and to minimize any distortion related to dehydration during scanning. MicroCT scanning was performed on scales from the head region and that were subjected to puncture at a loading rate of 50 mm s⁻¹. The scans were performed using 50 kV and 1000 µA, 1 frame s⁻¹ step scan mode with 10 ms delay, 900 projections and 2-frame averaging. These conditions were identified as most appropriate for evaluating the scales via preliminary studies. Based on the fixture arrangement of the scales, X-ray source and detector position, a geometric zoom of approximately 50× and a resolution of 3 µm pixel⁻¹ was achieved. Radiographs were treated using the systems imaging software (efX-CT, version 1.6; North Star Imaging, Rogers, MN, USA) to obtain images of relevance for display.

3. Results

Typical stress–strain responses resulting from the uniaxial tension tests are shown in figure 2. A single response obtained for a scale from the middle region evaluated after ethanol treatment and at a rate of 10¹ s⁻¹ is shown in figure 2a. The definitions of the elastic modulus (E), strength (S) and MOT are shown, and were used to quantitatively characterize the tensile responses of the scale samples. Representative responses obtained at low (10⁻³ s⁻¹), intermediate (10⁻¹ s⁻¹) and high (10¹ s⁻¹) strain rates are shown in figure 2b–d, respectively. These graphs include an example of responses from scales of the three regions and under both water and ethanol treatment conditions. A comparison of the responses in this figure shows that there is an increase in the resistance to failure with increasing loading rate and that ethanol treatment results in greater strength and toughness.

Figure 2.

Representative stress–strain responses for scales from the three representative regions at three different strain rates. (a) Response for a scale from the head region and definition of the measured properties. E, S and MOT represent the elastic modulus, strength and modulus of toughness, respectively. Strain rates of (b) 10⁻³ s⁻¹, (c) 10⁻¹ s⁻¹ and (d) 10¹ s⁻¹. (Online version in colour.)

The contribution of loading rate to the elastic modulus, strength, strain to failure and MOT of the scales is shown in figure 3a–d, respectively. The results in each graph represent the average of responses for scales from the three regions, namely the head, middle and tail, as well as within both water and ethanol environments. In general, there is an increase in all four mechanical properties of the scales with loading rate, and in both liquid environments. In water, the elastic modulus underwent the largest increase in magnitude with strain rate (up to 3×); the smallest increase with loading rate was in the strain to fracture (less than 1.5×), as shown in figure 3c. After ethanol treatment, the elastic modulus underwent the largest increase with strain rate (by nearly 6×; p < 0.001) and the strain to fracture was least affected by the ethanol condition. Both the tensile strength and MOT of the scales subjected to ethanol increased by approximately 3× (p < 0.05) over the range of strain rate.

Figure 3.

Strain-rate dependence in the tensile responses of the scales for each of the three regions. (a) Elastic modulus, (b) ultimate tensile strength, (c) strain at failure and (d) modulus of toughness. (Online version in colour.)

The strain-rate sensitivity of the scales was estimated from the measures of MOT for each of the three regions. A power law response was adopted to characterize the responses, in which the toughness (T) was defined according to $T=B{\dot{e}}^m$, where m and B represent the strain-rate sensitivity exponent and coefficient, respectively. In tension, the values of m for the head, middle and tail regions evaluated in water were approximately 0.06, 0.07 and 0.08, respectively. Ethanol treatment caused an increase in those values with m for the head, middle and tail regions of 0.09, 0.10 and 0.13, respectively. That represents between a 30% and 50% increase in the strain-rate sensitivity, with the largest occurring in scales from the tail region.

A representative load versus displacement response for transverse puncture loading of a scale to failure is shown in figure 4a. This response was obtained for a scale from the middle region evaluated after ethanol treatment and at a loading rate of 5 mm s⁻¹. There are three events highlighted in this figure, including (i) where the punch initiates puncture and penetrates the interface of the LL and elasmodine, (ii) where the maximum load occurs during the puncture process, and (iii) where the punch completely penetrates the scale beyond its tapered neck. The failure process is characterized by the apparent stiffness at the onset of loading, the load and work to initiate puncture defined at point (i), the maximum load to failure (P_max) defined at point (ii) and the work to failure defined at point (iii). These quantities were used for comparing the mechanical behaviour of the scale samples. Representative puncture responses for selected low (0.5 mm s⁻¹), medium (5 mm s⁻¹) and high (50 mm s⁻¹) loading rates are shown in figure 4b,c; these three rates correspond to strain rates of 10⁻¹, 10¹ and 10³ s⁻¹. The graphs include typical responses for the three regions of evaluation and within both the water and ethanol environments; evaluations in ethanol (e) are marked accordingly. In general, the resistance to failure increased with increasing loading rate in both liquid environments. However, in comparison with the responses in tension, there was no change in the displacement to puncture with loading rate as evident from the responses in figure 4.

Figure 4.

Representative puncture responses for the scales from the three regions at three different ramp rates. (a) Response for puncture of an ethanol-treated middle scale at 5 mm s⁻¹. (i) Initial puncture, (ii) maximum load and (iii) full puncture. (b) 0.5 mm s⁻¹, (c) 5 mm s⁻¹ and (d) 50 mm s⁻¹. (Online version in colour.)

The influence of the loading rate on the puncture resistance of the scales and importance of ethanol treatment are shown in figure 5. Specifically, the cumulative responses for the stiffness, as well as the load and work to failure, are presented in figure 5a–c, respectively, over the range in loading rate evaluated. There is an increase in each of the properties with increasing loading rate, and it occurs in both the water and ethanol treatment conditions. Note that results for the puncture displacement to failure are not shown owing to the absence of distinct trends in the data. Akin to evaluating the tensile responses, the strain rate sensitivity exponents (m) were estimated for the work to failure. In the hydrated condition, the values of m for the head, middle and tail were 0.13, 0.11 and 0.11, respectively. After ethanol treatment, the values of m for the head, middle and tail regions were 0.06, 0.04 and 0.06, respectively. Exposure to ethanol increased the resistance to puncture but resulted in between 40% and 70% decrease in the strain-rate sensitivity.

Figure 5.

Strain-rate dependence in the puncture responses of the scales for each of the three regions. (a) Stiffness, (b) load at full puncture, (c) work at full puncture. (Online version in colour.)

To further explore contributions of loading rate and ethanol treatment to the resistance to puncture, the properties of the scales obtained at the lowest loading rate (0.05 mm s⁻¹) and in water were used to normalize the responses obtained at the higher rates and after ethanol testing. The results of this analysis are shown in figure 6. Specifically, the load and work to the initiation of puncture, i.e. defined as the initiation of penetration of the punch (point (i) in figure 4a), are shown in figure 6a,b, respectively. Following this normalization scheme, the results for the head, middle and tail obtained for loading at 0.05 mm s⁻¹ in water are unity. The properties at the higher loading rates and after ethanol treatment are greater than 1, indicating that those conditions caused an increase. According to figure 6a, the scales underwent an increase in the resistance to puncture with loading rate. Although ethanol exposure was important at the lowest rate, it was not important to the load to initiate puncture at the higher rates of loading. There was also an increase in the work to initiate puncture with loading rate as evident in figure 6b. Of note, ethanol exposure caused the largest increase in work to puncture, with up to 7× greater required work to puncture than in the hydrated condition; the increase was significant (p < 0.01). The normalized load and work to failure, i.e. defined as the complete penetration of the punch and point (iii) in figure 4a, are shown in figure 6c,d, respectively. In general, there is an increase in both the load and work to puncture with loading rate for the three anatomic regions of the scales. The largest increase in these properties occurred under ethanol exposure. In the load to puncture, all three regions exhibited similar increases as a result of the ethanol. This is not consistent with the work to puncture, which increased the most in the scales of the tail region.

Figure 6.

Comparison of the rate dependence of puncture loading metrics for hydrated versus ethanol (e) treated scales. All values are normalized to hydrated scales at 0.05 mm s⁻¹. (a,b) The normalized initial puncture load and work, respectively, to point (i) in figure 4a. (c,d) The normalized full puncture load and work, respectively, to point (iii) in figure 4c. (Online version in colour.)

After completion of the puncture testing, representative scales were evaluated by microCT scanning to investigate contributions to the structural behaviour in the water and ethanol conditions. Representative microCT scans obtained from transverse puncture loading are shown in figure 7. Scans of scales punctured at 50 mm s⁻¹ in water and ethanol are shown in figure 7a,b, respectively. These are cross-section views at the centreline of the puncture. Similarly, microCT images representing ‘in-plane’ views of punctured scales evaluated in water and ethanol are shown in figure 7c,d, respectively; the depth of this cross-sectional slice is within the IE region of the scales. Note that the bright regions within these figures are the combined LL and EE, which occurs as a result of their higher mineral content relative to the IE. The cross-section views show that the LL and EE layers remain largely intact, without delamination or separation between the LL and EE (figure 7a,b). There is evidence of delamination between individual plies of the IE, particularly for the hydrated scale in figure 7a. In figure 7c, the delamination extends radially outward and approximately 2× the puncture diameter. The disruption in plies of the ethanol-treated scales is much different from the hydrated scales; the delamination appears to have occurred profusely throughout the entire IE thickness (figure 7b), and is not confined within specific interfaces. In addition, the extent of radiating delamination from the centre of puncture outwards is much less in the ethanol-treated scale (figure 7d) than in the hydrated scale.

Figure 7.

MicroCT scans of selected representative scales after transverse puncture loading at 50 mm s⁻¹. (a,b) Cross-sections at the centre of puncture for hydrated and ethanol treated, respectively. (c,d) In-plane views within the internal elasmodine (near the exit of penetration) for hydrated and ethanol treated, respectively.

4. Discussion

Elasmoid fish scales are composite laminates consisting of an assembly of unidirectional plies of type I collagen fibrils that are arranged with a twisted-plywood stacking sequence. Although these materials were not designed by engineers, they are clearly advanced in light of their hierarchical microstructures and performance [8]. In this investigation, elasmoid scales of the carp were exposed to a polar solvent to explore the importance of hydrogen bonds within the tropocollagen molecules to the strength and toughness of this dermal armour material. The apparent importance of interface mechanics on the mechanical behaviour of natural materials was recently highlighted by Barthelat et al. [46]. The present findings provide a unique contribution by showing the potential for tuning the interface properties through molecular bonds.

While the influence of loading rate on the elastic behaviour of scales was investigated by Lin et al. [20], other aspects have received limited attention. Ghods et al. [43] recently showed that the layered microstructure of fish scales is more effective at resisting failure by transverse impact than in-plane tension. Furthermore, the laminated structure of the scales exhibits spatial variations in structure and mechanical behaviour across the body. Scales from the head region achieve the greatest resistance to puncture, which results from the larger percentage of mineralized layers of the elasmodine in comparison with the middle and tail regions. After exposure to ethanol, scales from the head region were also the most resistant to puncture, and over the entire range of strain rates (figure 5). Nevertheless, the greatest increase in the load and work to failure after exposure to ethanol was achieved by scales from the tail region (figure 6).

As evident from the stress–strain responses in figure 2, the scales undergo a form of strain hardening with deformation. Chintapalli et al. [15,47] commented on this characteristic of biological composites, and that structural materials in Nature rely on strain hardening to maintain their functionality and tolerance to defects and damage. In most mineralized structural materials, the strain hardening response is restricted to the interfaces between highly mineralized components that are mediated by connective proteins. However, apart from the LL, the composition of elasmoid scales is dominated by the collagenous proteins of the fibrils, as well as the milieu of connective organic elements at the interface of the fibrils and the lamina. This increases the capacity for strain hardening. Strain hardening can also result from the inherent resistance of the individual lamina to rotation and sliding about the interfaces in response to the orientation of mode I loading [6]; mode I corresponds to tensile loading that causes opening of cracks that are oriented perpendicular to the direction of loading. Scales from the tail region underwent the greatest extent of strengthening overall with inelastic strain, and in both the water and ethanol conditions. Owing to the lower relative mineral content of scales from the tail region [28], the intra- and interfibrillar mineral appears to temper the effectiveness of the strengthening mechanisms in this highly organic system, independent of the loading rate.

The extent of strain hardening in the tensile response of the scales was dependent on the rate of loading and environment (figure 2). Both the strength and toughness of the scales increased with strain rate. Exposing the scales to ethanol increased the extent of strain hardening, which was further amplified with increasing strain rate. Indeed, after exposure to ethanol the strain rate sensitivity exponents for the tensile strength increased by 30–50% with respect to those values for the scales in water. Mechanistically, this is interpreted to result from the increased resistance to interfacial sliding posed by interpeptide hydrogen bonds that formed between the tropocollagen molecules of the fibrils. In the hydrated condition, the water molecules serve as a molecular lubricant that facilitates sliding. When replaced with ethanol, the ethanol molecules are retained within and about these interpeptide sites but they do not contribute to bonding. That causes an increase in the resistance to interfacial sliding between the fibrils owing to the interfibrillar bonds (figure 1b) and without reduction in strain to failure owing to the retention of a liquid lubricant. Air drying could also enable hydrogen bonds to develop but in the absence of this interpeptide and interfibrillar lubrication.

Contributions from molecular-scale and ply-level mechanisms to the strain hardening response of the scales during inelastic deformation are rather easy to conceive. But what quality or mechanism is responsible for the strain-rate sensitivity in the scales? The principal constituent is collagen, which undergoes an increase in resistance to deformation with increasing strain rate [48–50]. The next contribution is potentially at the interface of the individual plies. The Bouligand-type structure of scales allows the lamina to undergo rotation according to the loading orientation [51] and reorientation along the tensile axis. That process promotes interfacial sliding and stretching of the connective proteins that exhaust mechanical energy [6]. These proteins are also rate sensitive. Therefore, it is expected that the macroscopic response is attributed largely to the strain-rate sensitivity of the proteins.

The strain-rate sensitivity exponents for tension and puncture were nearly equivalent (m ≈ 0.1) in water. However, the ethanol treatment caused an interesting change. Under tension, the strain-rate sensitivity increased with ethanol treatment by at least 30%. The mechanisms of inelastic deformation are considered to be consistent in both environments. This suggests that the interpeptide hydrogen bonds stimulated by ethanol treatment are more rate sensitive than the hydraulic processes associated with the displacement of free and bound water during deformation. In other words, the rate sensitivity of deformation in the tropocollagen molecules is increased with interpeptide hydrogen bonds. These mechanisms contribute to the global deformation that occurs during tensile loading of the samples.

Contrary to the increase in rate sensitivity in tension, the ethanol treatment caused a decrease in m for the puncture responses by approximately 50%. According to a review of the dynamic behaviour of layered biological materials [52], energy dissipation occurs principally via mechanisms operating at the micro-scale, including delamination and microbuckling of the lamella. The dissipated energy increases with the extent of new surface area generated. After exposure to ethanol, the microCT scans (figure 7d) showed the delaminated zone was smaller, which must result from the increased intermolecular attraction due to the hydrogen bonding, and the large stress gradient under puncture loading. This resulted in suppression of the zone contributing to energy dissipation overall, and the relative contribution of interfacial processes to the rate sensitivity. In addition, the LLs play an important role in the puncture resistance, which is influenced less by the polar solvent owing to the low volume fraction of collagen. Admittedly, this interpretation is speculative, and further work is required to understand the contributing mechanisms more clearly.

One interesting aspect of the responses to the tension was that the strain to failure increased with strain rate until reaching a plateau at approximately 10¹ s⁻¹, and then it decreased significantly with a higher rate (figure 3c). In water, the strength and toughness also decreased with strain rate greater than 10¹ s⁻¹. Bone undergoes a ductile to brittle transition at moderate to high strain rates that are very similar to the range noted in the scales [53]. While exposure to ethanol had almost no influence on the strain to failure, it promoted strengthening and increased toughness over the entire range of strain rate (figure 3b,d). Most mineralized tissues undergo reductions in strength and toughness with dehydration in air [54–58] and even in ethanol [36,59]. Johnson et al. [60] presented results on the strain-rate characteristics of the bighorn sheep horn keratin in both hydrated and dehydrated conditions. Although the strength increased with strain rate in both tension and compression, there was a decrease in toughness, which appeared to be due to the reduction in strain to failure with dehydration. By contrast, Torres et al. [61] reported on the impact resistance of scales from Arapaima gigas under Charpy testing in both hydrated and dehydrated conditions. In that study, the energy to fracture of the dehydrated scales exceeded that of the hydrated condition by 2× or more. Thus, interpeptide hydrogen bonding appears to be effective at resisting the embrittlement that occurs with increasing strain rate, but it appears limited to systems with lower mineral content. As evident from the comparison of responses of scales from the head to tail and the reduced effectiveness of the polar solvent on scales from the head, the mineral interferes with the molecular bonding essential for this mechanism of toughening.

The dynamic response of engineered composites to axial and transverse impact loading is of substantial importance in a variety of applications. In these materials, an increase in stiffness with strain rate is common, but generally limited to a factor of 2×, which is substantially less than that of the scales (figure 3a). For carbon/epoxy composites, the strain-rate sensitivity is largely based on the resin properties, as expected, but is also dependent on the stacking sequence [62,63]. And while an increase in stiffness with strain rate is common, the change in toughness is dependent on the constituents. An excellent review of the strain-rate effects on the mechanical properties of composite materials was presented by Jacob et al. [64]. Efforts concerning the impact resistance of engineered composites are largely focused on the extent of delamination, not the strain-rate sensitivity. In an evaluation of various textile composites for impact resistance, Flanagan et al. [65] highlighted that stacking/blending layers with unique material characteristics (e.g. Spectra/aramid) improved the energy absorption as well as the material structural integrity at the interfaces. One aspect of the scale responses that was not considered in detail was the role of the highly mineralized LL, as well as the interface between the EE and IE to the puncture responses. Clearly, this is an area that requires further consideration as there are additional inspirational qualities of scales to learn from.

The findings of this investigation advance our understanding of the mechanical behaviour of this unique class of composite materials and highlight the capacity for capitalizing on interfibril bonds to render changes in rate sensitivity and toughening. Nevertheless, there are some recognized limitations to the investigation that are important to consider. For instance, although the armour of fish involves an assembly of imbricated scales and skin, the experiments were focused on the properties of individual scales, and not multiple scales. Previous work under quasi-static loading showed that stacking of elasmoid scales amplifies the overall puncture resistance [14] and is important to the strain-stiffening response of scaled skins [66]. Stacked scales or the skins may exhibit different behaviours from the individual scales. Similarly, the puncture loading was performed on scales restrained by peripheral clamping, and not on a compliant backing that simulates the fish body. Although the work of [14] did not find contributions from the compliant substrate backing to the puncture resistance of fish scales, it could be important to the type of damage incurred to the fish (contusion rather than puncture). An additional concern is that experiments were limited to the scales of the carp. This is important for two reasons, including (i) the differences in the number of mineralized plies, which would change the EE ratio, and (ii) the difference in lamination and relative ply orientations [27]. Rotation of the individual lamina in response to the orientation of tensile stress is a primary contribution to the toughness of scales [6,51]. As such, differences in the stacking sequence of plies and their orientations could contribute to the strain-rate response of these materials.

An increase in resistance to failure was observed after exposure to ethanol, which is speculated to originate from hydrogen bonds that form primarily within the tropocollagen molecules of the collagen fibrils. This increased degree of bonding would act to suppress delamination of the fibrils, as well as between the fibrils and between plies. If stretching and sliding of the lamina relative to one another is an important contribution to the global deformation [51], then molecular bonds that resist that process should be beneficial. Indeed, in the microCT images of the punctured scales subjected to ethanol treatment, there was evidence that gross delamination was suppressed (figure 7) with respect to those evaluated in water. Nevertheless, the increase in hydrogen bonding was not quantified, it was only inferred from the increase in stiffness of the scales (figure 2) and its potential contribution based on the differences in the solubility parameter of the liquid treatments. According to Shen et al. [67], there exists a complex milieu of ionic and hydrogen bonding interactions that exist between molecules of the tropocollagen within a fibril. Thus, in order to make definitive statements about the contributions of interpeptide bonding in the absence of water, direct measurements that can distinguish among them would be valuable. Of equal relevance, proteoglycans reside within this collagenous system and undoubtedly play a role in the mechanics of deformation of this tissue [68]. Dehydration via application of the polar solvent could alter their structure and their contribution to the connective arrangement between the collagen fibrils. That could also affect the constitutive behaviour of the dehydrated scales, in addition to the interpeptide hydrogen bonds.

The effectiveness of polar solvents to form hydrogen bonds in collagen is often described in terms of the Hansen solubility parameter for hydrogen bonding. Polar solvents with high values, like water, preferentially form hydrogen bonds with collagen peptides, thereby preventing the molecules from developing interpeptide bonds. When the water is displaced by a polar solvent with a lower solubility parameter than that of collagen in air (e.g. ethanol), interpeptide hydrogen bonding occurs. Therefore, it would be possible to evaluate the importance of this bonding on the dynamic response of these fibrous composites through the use of polar solvents with different solubility parameters. Of related interest, the solubility parameter for hydrogen bonding of collagen in air (interpeptide bonding) is equivalent to that for ethanol. That implies that the same extent of interfibrillar bonding would occur for the scales in air and scales in ethanol (with adequate time for displacement of the water molecules). The difference in these two conditions is the liquid molecules residing within the interpeptide sites of the tropocollagen that would serve as an intermolecular lubricant. It may be possible to explore the importance of this lubricant on the toughness of these fibrous flexible composites through careful choice of experiments. That exercise is reserved for future study.

5. Conclusion

An experimental investigation of the dynamic loading responses of elasmoid fish scales was conducted involving in-plane axial tension and transverse puncture loading to failure. Scales were obtained from near the head, middle and tail regions of multiple fish, and then evaluated as fully hydrated (in water) or after exposure to a polar solvent (ethanol). There was a significant increase in the resistance to failure with loading rate, and the largest resistance to failure overall resulted from the scales exposed to ethanol. The strain-rate sensitivity exponent in tension increased by up to 50% after exposure to ethanol, but decreased in transverse puncture loading by as much as 70%. The largest strength and toughness in tension, as well as the largest load and work to puncture, were obtained by scales from the head region of the fish, which is due to the greater proportion of mineralized layers in those scales than in the middle and tail regions. However, the largest increases in the resistance to failure with exposure to polar solvent (up to 7× increase in the work to puncture; p < 0.01) were in scales of the tail. This spatial variation is interpreted to result from the lower mineral content in the scales of the tail and the consequent lower interference to interpeptide bonding. The increased strength and damage tolerance with intermolecular bonding are quite unconventional for structural materials and are a unique characteristic of those with low mineral content. These results highlight the importance of weaker interfibril bonding on the mechanical behaviour of these natural laminated composites and the potential for pursuing bioinspired material designs with substantially greater impact resistance based on this concept.

Data accessibility

Research data related to this submission will be made available upon request.

Authors' contributions

D.A. designed the study, analysed results and helped write the paper. S.G. performed the experiments, analysed results and helped write the paper. C.S. performed experiments and helped write the paper. S.M. performed experiments and helped edit the paper. E.A.O. analysed the data and edited the paper.

Competing interests

We declare we have no competing interests.

Funding

D.A. is grateful for the financial support from the National Natural Science Foundation of China under grant no. 11872240 and for a seed grant from the University of Washington. E.A.O. acknowledges support from RutaN of Colombia through contract 080C-2015.