Elasmoid fish scales as a natural fibre composite: microscopic heterogeneities in structure, mineral distribution, and mechanical properties

Abstract

The elasmoid scales in teleost fish serve as exemplary models for natural fibre composites with integrated flexibility and protection. Yet, limited research has been focused on the potential structural, chemical, and mechanical heterogeneity within individual scales. This study presents systematic characterizations of the elasmoid scales from black drum fish (Pogonias cromis) at different zones within individual scales as a natural fibre composite, focusing on the microscopic structural heterogeneities and corresponding mechanical effects. The focus field at the centre of the scales exhibits a classical tri-layered collagen-based composite design, consisting of the mineralized outermost limiting layer, external elasmodine layer in the middle, and the unmineralized internal elasmodine layer. In comparison, the rostral field at the anterior end of the scales exhibits a two-layered design: the mineralized outermost limiting layer exhibits radii sections on the outer surface, and the inner elasmodine layer consists of collagen fibre-based sublayers with alternating mineralization levels. Chemical and nanoindentation analysis suggests a close correlation between the mineralization levels and the local nanomechanical properties. Comparative finite element modelling shows that the rostral-field scales achieve increased flexibility under both concave and convex bending. Moreover, the evolving geometries of isolated Mandle’s corpuscles in the internal elasmodine layer, transitioning from irregular shapes to faceted octahedrons, suggest the mechanisms of mineral growth and space-filling to thicken the mineralized layers in scales during growth, which enhances the bonding strength between the adjacent collagen fibre layers. This work offers new insights into the structural variations in individual elasmoid scales, providing strategies for bioinspired fibre composite designs with local-adapted functional requirements.

1. Introduction

Biological materials, which exhibit complex and hierarchical architectures, are increasingly serving as models and sources of inspiration for scientists and engineers [1–3]. The natural protection given by biological structural materials is the result of millions of years of evolution. As one of the representative biological materials, fish scales can be used as a model for designing new protective materials and structures with good combined mechanical properties, including flexibility, strength, and toughness [4–6].

There are four major types of fish scales: (i) ganoid, (ii) placoid, (iii) cycloid, and (iv) ctenoid [7–9]. The ganoid scales have a dense bony foundation with an enamel-like ganoine surface, which are found in a few freshwater fish, such as alligator gars [10] and Senegal bichir [11]. The placoid scales are commonly found in cartilaginous fish (e.g., sharks, rays, and skates), which show tooth-like dermal denticles with riblets and longitudinal grooves for hydrodynamic functions [7]. In comparison, both cycloid and ctenoid scales belong to the collagen-based elasmoid scales, commonly found in teleosts [12,13]. The elasmoid scales have concentric rings (circuli) on the outer surface, symbolizing the annual growth of the fish similar to tree rings [12]; the cycloid scales have these circuli with no interruptions, while the ctenoid scales have comb-like structures at the posterior margins. Among the different types of fish scales, the elasmoid scales represent the most abundant type in living fish. This collagen-based composite is often mineralized (primarily by hydroxyapatite (HAP)) in the outer layers to resist penetration and hence to enhance protection [14]. Specifically, elasmoid scales can be considered as laminated composites with multilayered cross-plywood structures, including the limiting layer (LL, mineralized), external elasmodine layer (EEL, partially mineralized), and internal elasmodine layer (IEL, unmineralized) [15–17]. Some studies also defined the combined LL and EEL as the external mineralized layer and considered the elasmoid fish scale as a bilayer composite. Together with the overlapping assembly, fish scales offer an excellent model system for developing fibre composite materials and flexible armour systems [18].

There have been many studies working on the structure–mechanical relationship in the elasmoid fish scales, and tension has been the most common mechanical test to determine the strength and stiffness (Young’s modulus). In general, the tensile response of elasmoid scales was linear at the initial stage, followed by plastic yielding before breakage, where the yielding is a direct result of the sliding and dissociation of collagen fibres [19]. Additional tensile studies of fish scales investigated the influence of scale locations (over the fish body), mineral contents, fibre orientations, and hydration conditions [12]. For instance, Garrano et al. analysed the tensile properties of Cyprinus carpio fish scales from several anatomical regions (head, body, and tail) as a function of moisture content in scales [20]. The strength and elastic modulus of the head scales were nearly double the values of tail scales when hydrated [20]; in contrast, no discernible variation was observed on the dehydrated scales [12]. Zhu et al. analysed the arrangement of collagen fibrils and tensile strength of striped bass (Morone saxatilis) scales [21]. The tensile strength and Young’s moduli vary between 30 and 50 MPa and between 600 and 850 MPa, respectively, along 0°, 45° and 90° from the longitudinal axis of the fish [21].

With the highest mineral level, the LL of the fish scale exhibits strong stiffness and hardness but undesirable brittleness [17,22]. Therefore, the exceptional fracture toughness of the scale composite is primarily due to the elasmodine layers (EEL and IEL) [18,21]. Dastjerdi et al. designed a customized fracture test configuration to probe the different fracture behaviours of fish scales, including fresh intact scales, collagen-only scales (by removing LL and EEL), and mineralized-only scales (by removing IEL) [23]. Specifically, small steel plates were clamped to the notched sample to transfer uniform loading and control crack growth. The results indicated that mineralized-only scales exhibited four times lower toughness than the collagen-only scales [23]. Moreover, the fibre orientations in the sublayers had a major influence on the directions of crack propagation, where the toughening mechanisms in the collagen layers involved massive delamination and crack bridging by the connecting fibres/ligaments [23].

The fish scale assembly achieves great combinations of protection and flexibility, which provide promising inspiration for multifunctional designs. For example, Rudykh et al. proposed a design map of scaled systems by investigating the design trade-off between flexibility and penetration resistance [24]. The design map was constructed based on the experimental results of 3D-printed models with an assembly of hard scales embedded in a soft matrix. The microstructural configurations were varied by changing the volume fraction and inclination angle of the slanted scales. The scale armour had the stiffest reaction at a soft/hard phase ratio of 1:1, and the indentation force decreased as the inclination angle increased [24]. Martini and Barthelat designed the slanted scale system by glueing the engraved alumina plates onto the pre-strained polyurethane strip, which enhanced the puncture resistance and maintained the torsional and flexural flexibility [4]. In general, while the fish scale-inspired composites can resist penetration, they are also flexible due to the various deformation mechanisms, such as inter-plate matrix shear, plate rotation, and plate bending [24,25].

Previous studies on elasmoid fish scales focused primarily on the structure–mechanical properties of the entire scale as a layered composite with gradient properties. However, there is a lack of systematic characterization of the scales at the material level. The primary objective of this work is to characterize the structural, chemical, and mechanical properties at different zones of the individual scales from black drum fish (P. cromis) scales exemplifying the elasmoid scales. Specifically, three zones were selected along the longitudinal direction of the scales (see the three zones and corresponding abbreviations in figure 1a–c), including the focus field (F0, centre of the circuli), the far end of the rostral field (R2, anterior end of the fish scale), and the middle between the focus and rostral end (R1). Corresponding comparative analysis includes structural characterizations based on scanning electron microscopy (SEM) and micro-computed tomography (μCT), chemical characterization using energy-dispersive X-ray spectroscopy (EDS), and multi-scale mechanical characterizations via nanoindentation, comparative tension tests, and finite element modelling (FEM). The overall aim is to provide a multi-scale characterization of the structure–property correlation of fish scales at different zones and to provide a mechanical understanding of the structural variations in an individual scale for the effective bioinspired design of flexible armours.

Figure 1.

Structural characterization at different regions of black drum fish scales. (a,b) Coordinate definition of (a) the fish body and (b) individual scales, where the longitudinal (L) and transversal (T) directions are defined along the anterior–posterior and ventral–dorsal axes, respectively, and ‘N’ denotes the normal direction. (c) SEM image of the selected region on a fish scale exhibiting the surface morphologies at rostral (R2, R1), focus (F0), and caudal fields, where R2 denotes a region close to the anterior end of the scale, F0 represents the focus field, and R1 resides at the middle between the R2 and F0 zones. (d–f) SEM images on the local surface morphologies of the (d) R2, (e) R1, and (f) F0 zones, respectively, where the zoom-in images highlight the microscopic circuli features and the nanoscopic protrusion on each circulus. (g–i) Vertical cross-sections of μCT reconstruction at (g) R2, (h) R1, and (i) F0 zones on the transversal T–N planes of the fish scale, exhibiting different multilayered structures including the limiting layer (LL) and external and internal elasmodine layers (EEL + IEL), as well as the fusiform-like radii sections (RS) in the rostral field and the distributed mineralized Mandl’s corpuscles (MC) in IEL.

2. Material and methods

2.1. Materials and sample preparation

The black drum fish (P. cromis) scales were obtained from Fruge Seafood Companies (TX, USA). These scales were shipped in fresh condition and stored in frozen condition. After thawing and sonicating in deionized water, the selected scales were dried in air with heavy weights on top to minimize morphological changes. These scales were then cut into narrow strips along the longitudinal (anterior-to-posterior) and transversal (ventral-to-dorsal) directions (figure 1a,b) to minimize epoxy used for embedding (Epo-Fix, Electron Microscopy Sciences) and thus reduce the influence of exothermic curing of epoxy. Those epoxy-embedded samples were used for elemental and nanoindentation characterizations. Sequential polishing was applied to the scale cross-sections using an automatic polishing machine (MultiPrep TM System, Allied HighTech Products, Inc.) with diamond lapping films (15, 9, 6, 3 and 1 μm). The final procedure of surface finish involved cloth polishing with 40 nm colloidal silica suspension. Later, the samples were sonicated again in deionized water to remove the remaining colloidal particles.

2.2. Scanning electron microscopy and energy-dispersive X-ray spectroscopy

The scale samples were dried and then coated with a 10 nm Pd/Pt layer to reduce the charging effect before imaging. SEM images were acquired with an FEI Quanta 600 FEG environmental SEM with an acceleration voltage of 15 kV and a working distance of ~12 mm. The backscattered electron images were acquired with a Bruker QUANTAX 400 XFlash 4010 (10 mm²) silicon drift detector. In addition, the SEM–EDS data were also obtained using the same scope with an acceleration voltage set at 20 kV. The elemental composition was quantified using Bruker Esprit 2.1 software with PB Linemarker-ZAF correction. Areal maps and line profiles were collected to reveal the distribution of the selected elements (figure 2), including the major elements of carbon (C), phosphorus (P), and calcium (Ca), as well as the trace elements of sulfur (S), chlorine (Cl), and oxygen (O).

Figure 2.

Chemical characterizations at different regions of black drum fish scales. (a–f) EDS elemental areal maps at the focus region, including major elements (C, P and Ca) and trace elements (S, Cl and O), respectively. (g–i) EDS line profiles of the selected major elements across the thickness of the fish scales at (g) R2, (h) R1, and (i) F0 zones, respectively. The line profiles of the element distributions are normalized by the maximum atomic percentage (at.%).

2.3. Synchrotron-based micro-computed tomography

The μCT measurements were conducted using synchrotron X-rays at beamline 2-BM of the Advanced Photon Source, Argonne National Laboratory (Chicago, IL). The beam energy used was 27.4 keV. The camera imaging resolution was 1.30 μm per pixel and a field of view (FOV) of 2560 × 1280 pixel², corresponding to ∼3.33 mm in width and ∼1.66 mm in height. During the μCT scans, the stage was rotated 180° to capture 1500 projection images for 3D reconstruction. Tomographic reconstructions were performed using TomoPy, segmented using the open-sourced machine-learning-based software Ilastik and rendered with Avizo (Thermo Fisher Scientific, USA). Especially, Mandle’s corpuscles (MC) were further labelled using the Random-Walk Distance Transformation module from AmiraZIBEdition (Thermo Fisher Scientific, Zuse Institute, Berlin, Germany), which quantified the spatial and geometrical information of the isolated particles (figure 3; electronic supplementary material, figures S8–S10).

Figure 3.

Synchrotron-based μCT at different regions of black drum fish scales. (a–c) 3D reconstruction of the fish scale at (a) R2, (b) R1, and (c) F0 zones, where the mineralized layers and particles are coloured with local surface mean curvatures (−0.1 to 0.1 μm⁻¹) and the collagen layers are shown in transparent mode to exhibit the distributed MC particles in IEL. (d–f) Horizontal slices of the fish scale at (d) R2, (e) R1 and (f) F0 zones, where the insets highlight the different morphologies of mineralized particles. The white arrows in (d) mark the perpendicular orientations of particle alignment in adjacent sublayers at the R2 zone. (g) 3D morphologies of the separated MC particles in the F0 zone, which can be generally classified into three categories: (i) small particles with irregular features, (ii) medium particles with oval or faceted shapes formed by particle attachment, and (iii) large particles exhibiting octahedron faceted shapes. (h,i) Quantification of the separated MC particles, including (h) 3D volumes (<5 × 10³ μm³ only) and (i) particle lengths and in-plane distribution.

2.4. Nanoindentation

The property gradients on the polished fish scales were measured using an instrumented nanoindentation system (NanoTest Vantage platform 4, Micro-Materials, UK). Line mapping over the scale thickness was conducted using a Berkovich tip to examine hardness (H) and reduced modulus (E_r). The spacing between adjacent indents was 2 μm, and the maximum load was 2 mN with a loading profile of loading (15 s), holding (10 s), and unloading (15 s). Thermal drifting was monitored for 30 s at the unloading stage when the load was unloaded to 10% of the maximum load (0.2 mN). The system was set to equilibrate for at least 3 h prior to performing nanoindentation. For each sample, three lines were drawn on the rostral (R2 and R1) and focus (F0) fields of the scale cross-sections for statistical purposes under dry conditions (relative room humidity level at ~30%; figure 4). Nanoindentation areal mapping was conducted on a selected radii section (RS) of a transversal cross-section (T–N plane) near the anterior end of the scale with spacing of 10 × 10 μm² per indent (R2 field; electronic supplementary material, figure S11). For comparative purposes, line-mapping nanoindentation was also conducted and analysed under 90% relative humidity conditions (electronic supplementary material, figures S12 and S13).

Figure 4.

Nanoindentation properties along the thickness of fish scales at (a) R2, (b) R1 and (c) F0 zones, including reduced modulus (E_r) and hardness (H) under dry conditions (room humidity).

2.5. Macroscopic uniaxial tension

The dogbone-shaped tensile samples were cut along the longitudinal and transversal directions using customized steel cutting dies (WhiskyTime, Etsy, Inc.) from large fish scales (electronic supplementary material, figures S14 and S15). The gauge area was 40 mm² with dimensions of 4 mm in width and 10 mm in length. The average thickness of the samples was ca 0.4–0.5 mm, and the measurement of the individual scales was used for the following analysis of the strength approximation. Demineralized scale samples were also prepared to investigate the mechanical contribution of the mineral phase. The demineralization protocol involved acid treatment by immersing the scale samples in 0.4 mol l⁻¹ hydrochloric acid for 90 min (electronic supplementary material, figure S15a) [26]. The acid-treated scales were then immersed in artificial seawater solution for more than 24 h before thickness measurements and tensile tests. The demineralization treatment was confirmed by SEM imaging, where the flattened surface circuli indicated the complete removal of minerals (electronic supplementary material, figure S1g). In total, 40 scale samples were prepared with 10 samples for each direction (longitudinal versus transversal) and condition (intact versus demineralized). The uniaxial tensile tests were carried out on an Instron 6800 mechanical testing machine, with a load cell of 2 kN. The mechanical tests were performed at room temperature at a displacement rate of 0.2 mm min⁻¹, and the hydration of the samples was maintained by water spraying during the tests. The tensile results were presented in terms of engineering stress–strain curves, and the tensile properties were quantified including strength (maximum stress), modulus (linear fitting of the initial deformation), and energy (integration of the stress–strain profiles; figure 5).

Figure 5.

Comparative tensile tests of intact and demineralized black drum fish scales. (a,b) Stress–strain curves of the intact and demineralized scales along the (a) longitudinal and (b) transversal directions. (c) Comparative tensile results of different samples, including strength, modulus, and failure energy. The asterisk denotes a significant difference at a confidence level of 95%. (d) Top-view SEM image of the fractured scales, revealing fibrous sheet delamination from the mineralized layer. (e,f) SEM images of fractured collagen sheets with MC particles anchoring (e) in-plane fibres and (f) the adjacent fibre sheets with orthogonal fibre orientations, where the yellow rhombuses mark the MC particles and yellow dashed lines indicate the fibre orientations. (g) Schematics of crack arrest and deflection (red curves) in IEL, including (i) in-plane crack deflection along the curved fibres around the MC particles and (ii) out-of-plane crack deflection with MC particles binding adjacent sheets with orthogonal fibres. (h) Bottom-view SEM image of the failed intact scale, revealing the bridging mechanisms by collagen fibres in the IEL.

2.6. Finite element simulation

Based on the characteristics of rostral- and focus-field structures of black drum scales, models for FE simulation were generated using Abaqus/CAE 6.14 (figure 6; electronic supplementary material, figures S16 and S17). The dimensions of real-scale beam models with multilayer structures and various external surfaces were designed based on the optical and SEM images in R2, R1, and F0 regions of the scale (figure 6a): (i) the R2 field consisted of alternating collagen and mineralized layers (9 sublayers with decreasing thickness) with surface RS units; (ii) the R1 field consisted of RS units in the LL, mineralized EEL, and collagen IEL; and (iii) the F0 field consisted of mineralized LL and EEL with collagen IEL. The width of the beam models was 3.38 mm (10 RS units), and the thickness (or height) of the beam models corresponds to the measurements from the multilayered scale samples (i.e., 0.31 mm of the R2 and R1 models, and 0.35 mm of the F0 models; figure 6a). 3D models were configured by extruding along the out-of-plane direction with a depth of 0.5 mm (electronic supplementary material, figure S16). For direct comparison, the comparative models C2, C1, and C0 with simple multilayered structures were generated with the same thickness in alternating mineralized and collagen layers (figure 6b). The 8-node linear brick elements (C3D8R) with decreased integration and hourglass control were used in FE beam simulations. Each beam model had 1 million quadrilateral mesh elements. The material property inputs in the FEM were obtained from the tensile tests. The strength (σ_c) and modulus (E_c) of the collagen layers were estimated from the tensile strength of demineralized scales, and the properties of the mineralized phase were determined based on the rule of mixture in composite materials.

Figure 6.

FE modelling results of fish scales under load-controlled three-point bending. (a,b) Geometrical models for FE modelling, including (a) the real-scale models at R2, R1, and F0 zones and (b) the comparative multilayered C2, C1, and C0 models consisting of alternative layers of mineralized and collagen layers. (c,d,f,g) Distribution of von Mises stress in fish scales at R2, R1 and F0 zones under (c,d) concave bending (loading from outside) and (f,g) convex bending (loading from inside). (d,g) Zoom-in regions of the stress distribution at the contact region of the R2 zone under (d) concave and (g) convex bending. (e,h) Displacement–force profiles of the FE beam models under (e) concave and (h) convex bending.

$${\displaystyle E_{\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}}=V_{\mathrm{c}}{E}_{\mathrm{c}}+V_{\mathrm{m}}{E}_{\mathrm{m}},}$$ (2.1)

where E_scale is the tensile modulus of intact scale, E_m is the modulus of the mineral phase, and V_c ~ 0.4 and V_m ~ 0.6 are the volume fractions of the collagen and mineral phases, respectively. It must be noted that the rule of mixture used here may not precisely predict the property correlations in fish scales, but the approximated modulus E_m here should be sufficient for comparative analysis of the multilayered designs in the following FE simulations. The strength of the mineralized phase σ_m was estimated as the strength of the intact scale σ_scale, where the maximum stress corresponded to the breakage failure of the mineralized phase. The Poisson’s ratios of the collage fibres and HA were approximated as ca 0.3 [27,28]. In FEM, the three-point bending load was applied to the central point until a maximum force of 100 N. Only half of the beam models were generated by taking advantage of the structural and loading symmetry, and the lateral edge of the beam models was fixed to the bottom or the top edge for the concave or convex bending, respectively (electronic supplementary material, figure S16). The FEM was used to analyse the deformation capability of fish scales in different fields via von Mises stress distribution and displacement–load correlations.

3. Results

3.1. Microstructure

The surface and cross-sectional features were characterized systematically using SEM and μCT images. First, the coordination and orientation systems of the fish scales are defined as shown in figure 1a,b. Based on the fish body anatomy, the anterior–posterior orientation is defined as the longitudinal (L) direction, the ventral-to-dorsal orientation is defined as the transversal (T) direction, and the normal (N) direction defines the orientation from the interior (proximal to body) to the exterior (distal to body) of the scale. On individual scales, the surface characteristics can be divided into five fields (figure 1b) [12], including the rostral field close to the anterior end (ca 1/2 of the scale length), the focus field at the centre (ca 1/3 of the scale length and ca 1/2 of the scale width), the caudal field at the posterior end (ca 1/6 of the scale length), and two lateral fields close to the ventral and dorsal sides. The rostral, focus, and lateral fields are embedded underneath the overlapping scales, while only the caudal field is exposed.

The surface morphologies of the black drum fish scales reveal intricate surface characteristics at different fields (figure 1c–f; electronic supplementary material, figure S1). First, the focus field is the initial part for scale growth, exhibiting labyrinthic patterns of surface ridges with microscopic hierarchical features on those labyrinthic patterns (figure 1c,f). Ring-like circuli ridges exhibit parallel patterns surrounding the central focus region, which are the developed ‘growth rings’ interrupted by the caudal field (figure 1c). On the rostral field of the scales, groove-like radii along the radial directions disrupt circuli ridges, where the primary radii extend from the focus field to the anterior end, and secondary radii initiate within the rostral field (figure 1c). In contrast, the circuli ridges are continuous with no interruptions in the lateral fields (electronic supplementary material, figure S1b). The surface morphology difference between the circuli ridges at the rostral and lateral fields is that hierarchical protrusions on the ridges are only available at the rostral field (figure 1d,e) but not the lateral field (electronic supplementary material, figure S1c). On the caudal field of the scales, the microstructures showed overlapping arrangements of the individual tubercles (ctenii) with sharpened tails, probably for hydrodynamic functions (electronic supplementary material, figure S1d). In comparison, smaller scales exhibited similar structural features and field divisions to the larger scales; yet the focus field is much smaller and regular (electronic supplementary material, figure S2).

In this study, the primary focus is to compare the structure designs at different zones of the fish scale. Three zones of interest were selected on the scale (figure 1c), including the rostral field close to the anterior end (R2), the focus field (F0), and the approximate middle between the anterior end and focus field (R1). As noted earlier, the surface circuli represent the growth rings of the scales with early-formed ridges close to the focus field and newly formed ones near the anterior end and lateral edges; that is, the R2 zone is the most recently formed region and F0 zone is the earliest formed region.

Apart from the difference in surface morphologies between the rostral field (R2 and R1) and the focus field (F0), the multilayered features on the transversal cross-sections exhibit more interesting variations (figure 1g–i; electronic supplementary material, figures S3–S5). In general, the fish scale exhibits multilayered structures, which can be generally divided into three layers, including a limiting layer (LL, highest mineralization) and an elasmodine layer further divided into external and internal elasmodine layers based on the mineralization difference, as shown by the SEM images (electronic supplementary material, figure S3) and μ-CT reconstructions (figure 1g–i; electronic supplementary material, figure S4). In the outermost LL, fusiform-like radii sections (RS) are observed on the transversal cross-sections (T–N planes) of both the R2 and R1 zones, where the edges of the RS units correspond to the radii intercepting the circuli ridges (figure 1g,h). In the most developed F0 zone, the surface profile in the F0 zone has irregular morphologies, corresponding to the labyrinthic patterns in the focus field (electronic supplementary material, figures S3d and S4d). The external elasmodine layer (EEL) consists of collagen fibrous sheets with higher mineralization, and the internal elasmodine layer (IEL) is typically not mineralized. The collagen cross-ply layers exhibit orthogonal fibre bundles in the alternating layers and decreasing thickness towards the scale interior, as revealed by the high-resolution SEM images of the polished and dried scale samples (electronic supplementary material, figure S5), which should correlate with the circumferential and radial fibres of the individual scale (figure 7a,b). In the newly formed R2 zone, no defined separation can be found between the external and internal elasmodine layers, where the micro- and nano-sized mineralized particles, named Mandl’s corpuscles (MC) [29], are distributed with alternating high and low concentrations (figure 1g; electronic supplementary material, figures S3b and S4b). Interestingly, it seems that the sublayers with fibre orientations parallel to the L direction have high concentrations of mineralized particles compared with the alternating in-plane fibre layer (fibres parallel to T direction), which can be evidenced by the electron density variations from the backscattered SEM image (electronic supplementary material, figure S3b). At the macroscopic level within the individual scale, the sublayers with radial fibres should have a higher mineralization level than the circumferential fibres. Later, the R1 zone is developed with more defined boundaries between the EEL and IEL (figure 1h; electronic supplementary material, figures S3c and S4c), where the MC particles in the EEL grow in size and channels to densify the mineralized phases and additional isolated MC particles float in the IEL. In the F0 zone, the mineralized particles become densely packed in the EEL, and isolated MC particles exhibit rhombus-shaped cross-sections and aggregates of multiple particles (figure 1i).

Figure 7.

Fibre arrangement and mineralized particles exposed on the fractured fish scales. (a–c) SEM images of fractured fish scales, including (a) delamination failure exposing fibrous multilayers under the surface circuli and (b,c) orthogonal fibre arrangements. The yellow dashed lines trace the circumferential alignment of collagen fibres (surrounding the focus region), and the white dashed arrows point to the radial alignment of the fibres. (d–f) SEM images of fractured fish scales exposing (d,e) the aggregated clusters of rhombus-shaped mineralized MC particles with major axes along the local fibre orientation and (f) corrugated sheets due to the local mineralization in the fibrous IEL.

3.2. Chemical analysis

The elasmoid scales are composed of HAP and type-I collagen fibre sheets [30], which can be confirmed by the chemical characterizations (figure 2; electronic supplementary material, figure S6). Based on the areal mapping at the focus field, the detected elements of interest include phosphate (P) and calcium (Ca) concentrated in the LL and EEL of the bulk scale, with highest concentrations in the LL (figure 2b,c). In comparison, no Ca or P is found in the IEL (except for the MC particles) due to the low mineralization content. Additional trace elements of sulfur (S) and chloride (Cl) are found with generally uniform distributions across the scale thickness (figure 2d,e).

Comparative EDS line profiles were collected at R2, R1, and F0 zones to correlate with the multilayers across the scale thickness (figure 2g–i). In the R2 zone, P and Ca trends show a local increase with electron-dense regions in the multilayers; except for the LL, the higher mineralized alternating layers correspond to those with fibre orientations (along L direction) perpendicular to the plane of view (T–N plane; figure 2g). The elasmodine layer with high P and Ca exhibits similar levels of mineralization, which are slightly lower than the LL. In the R1 zone, higher P and Ca levels reside in LL and EEL (in total ~60% of scale thickness; figure 2h), showing a similar trend to that in the R2 zone. A sharp drop of mineralization is observed across the boundary of EEL and IEL, with oscillating C levels in the collagen layers in IEL correlating with alternating fibre orientations. In the F0 zone, the general oscillating trends of the chemical elements P, Ca, and C agree with the trends in the R1 zone, with alternating high and low mineralization in the EEL and no mineralization in IEL (figure 2i).

3.3. Mandl’s corpuscles

According to the above structural and chemical investigations, the LL and EEL constitute the majority of the mineralization in the black drum fish scales. Yet, isolated mineral particles and aggregated assembly are present in the IEL. These MC particles can be revealed by SEM imaging on the scales after peeling off LL and EEL layers (figure 7).

On the one hand, the peeling failure of the black drum fish scale induces delamination between the sublayers, which reveals the alternating circumferential and radial fibre orientations in adjacent layers of the bulk scale (figure 7a), forming orthogonal fibre alignment at the local scale (figure 7b). The fibre bundles act to bridge the cracks in the adjacent layers, which prevent premature failure (figure 7c). In comparison, the fractured surface near the outer circumference of the rostral and caudal fields reveals U-shaped fibre bundles pointing outward along radial directions (electronic supplementary material, figure S7a–d). It is proposed that these U-shaped bundles persist along the entire outer circumference of the scales, forming local cross-ply structures in adjacent fibrous sheets (electronic supplementary material, figure S7e). On the other hand, the mineral phases are shown as the regions with higher electron density, where the isolated MC particles fall further away from the aggregates than the connected mineralization front (figure 7d). The mineral aggregates can be generally considered as a cluster of rhombus-shaped MC particles with their major axes (ca. 50–100 μm in dimensions) along and perpendicular to the local fibre orientations (figure 7e). The corrugated surface of the fibrous sheets after delamination failure suggests that these particles are not 2D features but 3D volumes, where volume expansion of these particles during mineralization deforms fibrous sheets in the IEL (figure 7f).

Detailed analysis of the distribution and 3D geometries of these MC particles was conducted using the synchrotron μCT results (figure 3; electronic supplementary material, figures S8–S10). First, from the 3D rendering of the selected zones in fish scales, the distribution of mineral phases exhibits clear differences. (i) The newly formed R2 zone exhibits layered distribution in the elasmodine layer with the innermost layer of mineralization merging by the micro-sized mineral particles (figure 3a; electronic supplementary material, figure S8a,d). Here, it should be emphasized that the alternating mineralization does not originate from the local fibre orientations but truly correlated with the different mineral densities, as evidenced by the contrast between adjacent sublayers bordered by the white dashed line in figure 3d. (ii) The R1 zone shows defined EEL with higher mineralization content, yet the mineralization is not fully densified but with remaining sites for further filling. The individual MC particles are sparsely distributed with ellipse-shaped particles (figure 3b; electronic supplementary material, figure S8b,e). (iii) The developed F0 zone shows fully mineralized EEL, and additional faceted octahedron MC particles are enlarged in IEL (>50 μm in length) and aggregated to form clusters (figure 3c; electronic supplementary material, figure S8c,f). From the surface curvature mapping, the MC particles exhibited mean curvatures below zero (figure 3a–c; electronic supplementary material, figure S8d–f). Especially, the MC particles exhibited morphological evolution at different stages of mineralization, starting from ellipse particles (sizes of 1–10 μm) in the R2 zone (figure 3d) to enlarged ellipse (sizes of 20–30 μm) with some faceted boundaries in the R1 zone (figure 3e), and finally more regular faceted MC particles (major axes of ca 50 μm) with edge merging (figure 3f). The 3D reconstruction of the isolated MC particles confirmed the 2D observation from the reconstructed slices, where the small irregular particles grow in size probably by layered deposition (to form medium-sized ellipses) and by particle attachment (to form irregular aggregates) and later develop into larger faceted particles with roughened edges (figure 3g).

Detailed characterization of the MC particles in a scale sample at the F0 zone was conducted to characterize the dimensions and orientations (electronic supplementary material, figures S9 and S10). After segmentation and 3D rendering, the geometrical parameters were characterized by length (l), width (w), thickness (t), and the 3D volumes of the individual MC particles (electronic supplementary material, figure S10a). Excluding the larger particles (>5 × 10³ μm³ probably due to particle merging), the remaining smaller particles were used for size distribution and orientation analysis. The counts of MC particles decrease exponentially as the particle size increases (figure 3h; electronic supplementary material, figure S10c). For geometrical characterizations, while the thickness-to-width (t/w) ratio gives a normal distribution near 1.0, the length-to-width (l/w) ratio gives a skewed distribution with a maximum at 2 (electronic supplementary material, figure S10d,e). Therefore, the individual MC particles exhibit equiaxed dimensions in the thickness–width plane and elongated morphology along the length axes. For orientation analysis, spherical coordinates with paired angles (θ, φ) were used to quantify the orientations of the length axes of the particles, where θ was the polar angle characterizing the out-of-plane orientations and φ was the azimuthal angle characterizing the in-plane orientations (electronic supplementary material, figure S10b). First, the polar angle θ exhibits a centred distribution between ~78° and 86° (electronic supplementary material, figure S10f), suggesting the in-plane distribution of the MC particles (with a slight offset tilting angle of ~8° against the analysed plane). Second, the azimuthal angle φ, as well as the length l and l/w ratio, does not show preferred in-plane orientation in the range of 0–360° (figure 3i; electronic supplementary material, figure S10g,h), confirming the coalignment between the long axis of the MC particles and the local fibre orientations (completing a 360° rotation along the circumferential and radial directions in the fish scales).

3.4. Microscopic mechanical behaviour: nanoindentation

The mineral phases have a close correlation with the mechanical behaviours of the fish scales at the material level. To probe the mechanical gradients across scale thickness, comparative nanoindentation line mapping was conducted at different zones (R2, R1, and F0) of the black drum fish scales (figure 4).

In general, close correlations were found between the property profiles and the distribution of mineralized phases along the scale thickness, with the highest reduced modulus (E_r) and hardness (H) in the LL and lowest in the collagen-based IEL (electronic supplementary material, table S1). First, the property trends exhibit obvious oscillations in the R2 zone (figure 4a), where the cyclic undulation correlates with the high and low distribution of the mineralization levels as revealed by the EDS chemical mappings in figure 2g. The modulus and hardness show an obvious decrease from LL (H = 1.2 ± 0.6 GPa, E_r = 31.7 ± 11.2 GPa) to the elasmodine layer, which oscillates between H ~ 0.16–2.25 GPa (average of 0.8 GPa) and E_r ~ 9.2–36.0 GPa (average of 20.9 GPa). In comparison, the property profiles are quite similar in the R1 and F0 zones (figure 4b,c), with obvious oscillations in the EEL of the R1 zone (due to incomplete mineralization) but not in the EEL of the F0 zone (due to densified mineralization). Especially, the properties in the EEL (H = 0.43 ± 0.25 GPa, E_r = 13.54 ± 5.11 GPa) and IEL (H = 0.38 ± 0.23 GPa, E_r = 7.31 ± 1.94 GPa) show a stepwise decrease in the F0 zone. Considering that nanoindentation line mappings only measure the property gradients across the scale thickness, an additional area map was conducted on a highly mineralized RS unit of the R2 zone, which verified the uniform properties in the sublayers with the same fibre orientations (electronic supplementary material, figure S11). Also, within the RS feature, E_r results did not vary significantly with fibre orientations in the adjacent sublayers (E_r ~ 30–40 GPa), but the H map observed higher values in the sublayer with out-of-plane fibres (H ~ 1 GPa) than the in-plane fibres (H ~ 0.5 GPa), agreeing with the observation that the sublayers with out-of-plane fibres have higher electron density (electronic supplementary material, figure S11a). Considering that in the fully developed F0 field with fully densified mineralization in EEL, the hardness profile is almost constant (figure 4c), suggesting an ignorable difference in the mechanical contribution in the sublayers with in-plane and out-of-plane fibres. Therefore, the property variations observed in the R2 and R1 zones should mostly correlate with the mineralization levels rather than fibre orientations.

In addition, comparative nanoindentation was also conducted under 90% relative air humidity conditions in the R1 and F0 zones of the fish scales (electronic supplementary material, figure S12). The property profiles inherited the general trends with higher properties in the mineralized layers. At the R0 zone, while the LL layer maintained a similar hardness (H ~ 0.13–3.75 GPa, average of 0.97 GPa) and modulus (E_r ~ 3.8–82.6 GPa, average of 23.7 GPa) under increased humidity compared with the dry condition, the EEL (H = 0.18 ± 0.10 GPa, E_r = 6.29 ± 2.50 GPa) and IEL (H = 0.06 ± 0.04 GPa, E_r = 1.35 ± 0.78 GPa) showed a significant decrease due to the softening effect of the collagen-dominated constituents (electronic supplementary material, figure S13 and table S1).

3.5. Macroscopic mechanical behaviour: tension

The mineral contents and the cross-ply arrangements of the collagen fibres also have a close correlation to the mechanical behaviours of the fish scales. Therefore, comparative tensile tests were conducted to investigate the influence of testing orientations (longitudinal versus transversal) and mineralization (intact versus demineralized). Figure 5a,b plots the stress–strain curves for tensile tests of dogbone samples along the longitudinal and transversal samples, while electronic supplementary material, figures S14 and S15 compare the tensile tests of intact and demineralized samples, respectively.

First, the stress–strain curves of intact scales under tension along the longitudinal and transversal directions did not show obvious differences, both exhibiting classical polymer material behaviours before cracking of the mineralized layers. Cracks in LL initiated at the grooves of the surface circuli (regarded as pre-existing microscopic notches) parallel to the loading direction, and the cracked LL induced following delamination between the LL and the underlying elasmodine layers (figure 5d). From the post-failure samples, failure usually occurred near the clamped end due to the decreased thickness towards the scale circumference (electronic supplementary material, figure S14d,e). The mineralized MC particles also act as bonding agents for the adjacent fibre bundles in individual layers (figure 5e) and adjacent cross-ply layers (figure 5f; electronic supplementary material, figure S14f). This structural feature, by ‘glueing’ fibre bundles together locally, is able to resist and/or deflect crack propagation around the MC particles in 3D (figure 5g). In addition, the non-mineralized IEL layer showed bridging fibres connecting the crack openings (figure 5h). The highly mineralized LL, however, exhibited brittle fracture exposing failure surfaces resembling typical ‘conchoidal fractures’ observed in biominerals (electronic supplementary material, figure S14g). The analysed properties showed similar strength (σ_scale ~ 50 MPa) and failure energy (W_scale ~ 5.4–6 × 10⁶ J m⁻³) along the longitudinal and transversal directions under tensile tests, while only the tensile modulus measured statistically higher values along the transversal direction (figure 5c).

Second, for demineralized scale samples, no directional dependence was reported for the samples along the longitudinal and transversal directions regarding the tensile stress–strain curves and analysed properties (electronic supplementary material, figure S15). The mechanical equivalence along these two directions could correlate with the cross-ply arrangement of the fibrous sheets in fish scales; with the circumferential and radial alignment of the collagen fibres at the macroscopic scale, we expect the demineralized scales to exhibit in-plane mechanical isotropy along any random orientations (electronic supplementary material, figure S7). From the SEM images of the post-facture samples, failure also occurred near the clapping ends due to the reduced thickness there, where the fibre bundles exhibit bridging mechanisms and wrinkle after tensile failure (figure 5h; electronic supplementary material, figure S15d–f). Based on the comparison of the analysed properties, the demineralized samples exhibited sharp decreases in the tensile strength and modulus, suggesting the strengthening and stiffening effects of the mineral phases (figure 5c). However, the demineralized scales exhibited ~60% increased strain at maximum stress.

3.6. Macroscopic mechanical behaviour: computational analysis

Previous structural, chemical, and nanomechanical characterizations revealed the intrinsic differences of different zones at the material level. In this section, to understand the mechanical differences of fish scales at the macroscopic level, FEM analysis was conducted at different zones. The multilayered structures with alternating high and low mineralization, as well as the different features in the LL, were replicated and standardized based on the SEM and μCT images (figure 1g–i; electronic supplementary material, figure S3) of the transversal cross-sections (on the T–N planes) of the scales at R2, R1 and F0 zones, respectively (real-scale models in figure 6a). In addition, comparative multilayered beams without LL features (RS in the R2 and R1 zones and triangular corrugations in the F0 zone) were generated to investigate the influence of these surface features (C2, C1 and C0 models in figure 6b). Three-point bending was applied to the centre of the scale-inspired models by applying load-controlled displacement to generate concave and convex bending, corresponding to bending from the scale exterior (figure 6c–e; electronic supplementary material, figure S17a,b) and interior (figure 6f–h; electronic supplementary material, figure S17c,d), respectively.

The specific questions to be answered include (i) What are the mechanical adaptations of different zones in individual fish scales? (ii) How do the mechanical designs of fish scales with specified multilayered structures and surface structures compare with simple multilayered structures? (iii) What is the bending asymmetry under concave and convex bending? In the following paragraphs, the von Mises stress distribution and the displacement–load curves are used for the comparison between the structural adaptations between the real-scale models and later with the simple multilayered models. The results for the concave and convex bending cases are introduced separately, which are compared later to discuss the mechanical asymmetry under these two loadings.

First, under concave bending, the bottom side of the scales (interior) is under tension while the scale exterior is under compression. The stress distribution and the displacement–load curves exhibit clear differences (figure 6c–e; electronic supplementary material, figure S17a,b). Upon the maximum load of 100 N, the R2 zone exhibited the most significant displacement (0.092 mm) than the R1 and F0 zones, and the bottom surface of the R2 scale had the highest von Mises stress of 12.24 MPa due to easy tension of the collagen sublayers in the elasmodine layer. The mineralized RS units ensured increased resistance against contact loading, while the grooves between RS units enabled spaces for compressive deformation (figure 6d). In comparison, the R1 and F0 scales with tri-layer structure (mineralized EEL and unmineralized IEL) and LL features showed lower deformability, and the maximum stress occurs near the grooves of the surface features (RS in R1 scale and ridges in F0 scale) due to the local stress concentration (figure 6c). The comparison between R2 and R1 models suggests that multilayered collagen fibrous sheets with alternating high and low mineralization have the advantage of increasing deformability under concave bending. The comparative multilayered models (C2, C1 and C0) show much smaller deformation displacements and von Mises stresses (electronic supplementary material, figure S17a,b), indicating that the surface features with grooves allowed compressive deformation without cracking the mineralized LL and thus increased deformability of the entire scale. The maximum von Mises stress in the multilayered models all occurred underneath the contact point, which followed C0 > C2 > C1 although their differences were negligible.

Second, under convex bending, the general stress distribution is reversed with the scale interior under compression and the scale exterior under tension (figure 6f–h; electronic supplementary material, figure S17c,d). The highest von Mises stress is found at the grooves of these surface features in all three real-scale models, where the local stress concentration is induced by the tensile stress on the convex side (figure 6f,g). The comparison between the three real-scale models suggests the order of flexibility R2 > R1 > F0 under convex bending (figure 6h), similar to the concave bending case. Such structural–chemical–mechanical adaptations at different regions of the scales could correlate with the functional requirements, where the newly formed R2 zone at the anterior end inserted into tissue shows the best flexibility, and the developed F0 zone of the scale overlaps to resist predator loading requiring higher stiffness and strength. The comparative multilayered models also exhibit much smaller stress and displacement compared with the real-scale model under convex bending, confirming the increased flexibility by the surface structures regardless of the loading directions.

In addition, regarding the concave–convex bending asymmetry, the simple multilayered models show ignorable differences in the displacement curves (electronic supplementary material, figure S17b,d). Yet, it is likely due to reduced flexibility by the flat mineralized layers reduces the deformability significantly and limits the deformation only within the elastic regime upon the maximum load of 100 N (displacement < 2 × 10⁻³ mm compared with beam thickness of 0.3 mm). In those cases, the displacement is mainly induced by the local deformation in the collagen layers rather than the entire beam models. Even in the R1 and F0 of the real-scale models, the similar maximum displacement does not indicate obvious concave–convex bending asymmetry (figure 6e,h). This is mainly attributed to the elastic deformation in the beam models under both concave and convex bending with maximum stress below the yielding point of collagen of ~7 MPa (table 1); the deformation in the two scale models might not be significant enough upon 100 N to show the bending asymmetry. In comparison, the R2 zone exhibits approximately two times higher displacement under concave bending (0.92 mm) than under convex bending (0.35 mm). Such a difference could be mainly attributed to the material gradients. The key assumption for composite beams under bending is the linear distribution of bending strain over the section, which induces generally higher strains in the collagen layers. For the R2 scales, more portions of the collagen layers are under tensile stress when applied to concave bending, leading to more significant deformation.

Table 1.

Tensile results of elasmoid scales under different conditions.

species	scale type	conditions	directions	strength (MPa)	modulus (GPa)
Pogonias cromis (this work)	ctenoid (intact)	wet	0°	48.9 ± 17.9	0.57 ± 0.12
			90°	49.6 ± 20.3	0.85 ± 0.38
	ctenoid (demineralized)	wet	0°	7.3 ± 2.4	0.04 ± 0.01
			90°	7.0 ± 2.9	0.04 ± 0.01
Arapaima gigas [16,31]	cycloid	dry	—	46.7–53.85	1.2–1.38
		wet	—	22.26–25.2	0.1–0.83
Pagrus major [19]	cycloid	wet	—	93 ± 1.8	2.2 ± 0.3
Megalops atlanticus [32]	ctenoid	wet	0°	22.6 ± 5.1	0.22 ± 0.03
Morone saxatilis [21,33]	ctenoid	wet	0°	~32a	~0.85a
		wet	45°	~48a	~0.62a
		wet	90°	~52a	~0.8a
Cyprinus carpio [20]	cycloid	wet	0°	33.4–34.0	0.45–0.59
Latimeria chalumnae [34]	elasmoid (cycloid)	—	0°	~50	~0.21
		—	90°	~50	~0.25

^a Note that some values in the table are estimated from the stress–strain curve.

4. Discussion

4.1. Mechanical comparison with different biological scale systems

Elasmoid fish scales consist of collagen and calcium-based minerals (primarily calcium phosphate and calcite carbonate) [30]. The mineral-to-collagen weight ratio varies significantly in different species [7], e.g., 1:1.2 in red seabream [19] and 1:2 in the scutes of boxfish [35]. Thus, as a combination of mineralized segments and soft matrix, scaled skin gives a combination of flexibility and protection. Also, the overlapping pattern of scales minimizes the drag during swimming through wave regulation about the body [36] and even contributes to effective undulatory locomotion through scale interlock and release [37]. Table 1 compares the tensile results of elasmoid fish scales, where the 0° and 90° samples are defined as cut along and perpendicular to the long axis of the fish to study the in-plane anisotropy [21,38]. Regardless of the large variations in shapes, sizes, and overlapping arrangements of the elasmoid scales, their tensile response and properties are rather similar. Yet, the strength and modulus of the P. major scale are roughly twice as high as those of the other elasmoid scales, which could be attributed to the higher mineral-to-collagen ratio in the scale thickness [19]. In addition, hydration has a significant effect on the properties of elasmoid scales, where the highly mineralized (HAP) fibres contributed to greatly increased stiffness and modulus under dry conditions [16,31], especially in the partially mineralized EEL as suggested by the comparative nanoindentation results (electronic supplementary material, figure S14).

It should be noted that some fish scales exhibit anisotropic responses when loaded along different directions, which should correlate with the cross-ply orientations in the collagen fibrous matrix [39]. For example, the cross-ply orientations follow orthogonal directions (with 90° rotation) between consecutive layers in the scales of sea bream (P. major) [19], striped bass (M. saxatilis) [33], rainbow fish (Poecilia reticulata) [40], and black drum (P. cromis) in this study (figure 7a–c). In comparison, Bouligand patterns with plywood-like layers completing 180° rotations across thickness direction were also reported, where the inter-ply angles could be irregular (e.g. A. gigas [38] and jewelfish (Hemichromis bimaculatus) [40]) or regular (e.g., 60° inter-ply angles in Atlantic tarpon (Megalops atlanticus) scales [32] and 75° in tarpon (Megalops atlanticus) and carp (C. carpio) scales [39]). In addition, the coelacanth (L. chalumnae) scales represent an interesting exception, which consist of two sets of Bouligand structures perpendicular and interpenetrating with each other [34,41].

As teleost fish scales are among the toughest biological materials known (15–18 kJ m⁻² for M. saxatilis) [23], the orthogonally arranged collagen fibre lamellae provide increased toughness through different responses to the crack front in adjacent layers, e.g., fracture, separation, and deflection, all contributing to stress delocalization [42]. Under tension, lamellae reorientation towards the tensile axis was observed under small angle X-ray scattering [43] and modelled considering the elastic stretching, strain-rate sensitivity, and inter-fibrillar sliding [38]. In situ tensile tests on notched samples also revealed lamellae delamination as an energy-absorbing mechanism [38] and crack bridging as a prevailing toughening mechanism [8].

There are other models of flexible armour in nature, such as osteoderms from reptiles (alligators and turtles) and mammals (armadillos) [8], keratinous armour in pangolin [8], chitin-based crustacean exoskeletons [44], and carapace scutes from boxfish [7,35], all contributing to balanced penetration resistance, flexibility, lightweight, and mobility. These armour designs have different material constituents, structural designs, and specific adaptations to functional requirements. The alligator osteoderms contain bony denser structures on the outer surface with irregular dorsal ridges and layered collagen lamellae with 100–500 μm pores inside [8]. In comparison, the armadillo osteoderms have a dark-brown keratin layer coating on top and Sharpey’s fibres (non-mineralized collagen fibres) connecting the adjacent hexagonal tiles [45], while the leatherback turtle has a skin covering with jagged edges of the tiles [46]. The sutures between the individual osteoderm tiles bite with adjacent tiles for interlocking strengthening of the entire armoured system, but such a mechanism also limits the motion of the tiles to ~±15°. The hexagonal scutes of boxfish carapace consist of a bilayered composite similar to fish scales, with an outer high-mineralized surface plate reinforced by raised struts and an inner compliant collagen base [35]. However, the sutured scutes do not have Sharpey’s fibres in between, which does not have a strengthening contribution to the overall carapace but allows for accommodation of the changing pressures in the ocean habitat and growth without moulting [35]. In comparison, the entire crustacean exoskeletons are developed as one piece with chitin Bouligand sheets partially mineralized by calcium carbonate for integrated protection, but the crustaceans replace the entire exoskeletons via periodic moulting [44,47].

Compared with the aforementioned armoured systems, the elasmoid scales have individual scales inserted into the skin forming a slanted overlapping pattern, not only enabling higher puncture resistance than the single-layered scale units [14,18] but also preventing scale tilting effectively under asymmetric puncture loading [48,49]. Previous theoretical analysis by Vernerey and Barthelat correlated the bending curvature of scaled skin with the rotation and bending of individual scales, demonstrating balanced control of the bending capability by the scale arrangement parameters, including scale length, spacing between scales, and total body length [50]. Systematic tests on different printed scale patterns give a wide spectrum on the Ashby chart, indicating an optimized choice for desired protection–compliance requirements, suggesting the advantage of having a large design space with overlapping arrangements of elasmoid scales [48].

4.2. Adaptive function of different microstructures in the individual fish scales

Previous studies on the mechanical significance of scaled armour focused primarily on the effects of scale tilting angles, mineral-to-collagen ratios, and scale dimensions. In the present study, we focused on the structural and mechanical heterogeneity within individual scales, which may suggest location-dependent material adaptations within scales.

On the outer surface of the fish scales, the circuli ridges surround the focus field. The structural difference has been reported on the arapaima scales between the exposed and covered regions, where the exposed regions have thicker but less regular mineral ridges [15] leading to enhanced fracture resistance against predatory tooth (Pygocentrus natteri) during penetration experiments [8,46]. Based on the comparative FE results of the scales with surface ridges (R2, R1, and F0 in figure 6) and simplified multilayers (C2, C1, and C0 in electronic supplementary material, figure S17), the corrugations on the mineralized outer surface (including the radii sections and circuli with groves in between) have the advantage of enabling increased flexion in scales without cracking the minerals [15]. The comparison between the rostral-field models (R2 and R1) and the focus-field F0 model suggests that thicker radii sections were more effective in increasing bending flexibility than the simple circuli ridges (figure 6e,h). In addition to the general circuli features, the caudal field of the ctenoid scales aligns the surface roughness with the water current direction for hydrodynamic advantage [51].

Despite the differences in surface features in individual scales across species [8], the cross-sections of scales usually follow a common design motif: a mineralized outer region and a non-mineralized collagen inner region (figure 1g–i) [15,16]. This layered system in elasmoid scales enables piercing resistance under predatory bites [25,52]: with the mineralized outer layers resisting high local stress (while the surface grooves enable deformation without cracking), and the tough inner layers distributing the stress over a larger area. In addition, our results suggest the presence of local structural variations within the scales at different zones. Especially, the R2 zone near the anterior end of the scale exhibits (i) alternating multilayers (thickness of 180 μm) of high and low mineral densities and (ii) fusiform-like radii sections on the exterior surface with radii grooves (thickness of 130 μm). In comparison, R1 region close to the centre of scales, the external elasmodine layer (thickness of 120 μm) is almost fully mineralized and the internal elasmodine layer (thickness of 120 μm) is non-mineralized. The comparison between the R2 and R1 models suggests that alternating layers with high and low mineralization provide increased flexibility (figure 6e) even with similar mineralization (ca. 67% versus 61% of mineralized layer thickness in the R2 and R1 models; figure 6a). The flexibility at different zones follows R2 > R1 > F0 (figure 6e), indicating increased flexibility towards the proximal edge (anterior end) of the scales. Our results suggest that this gradient in flexibility results from the microstructural architecture and different mineralization levels. The greater deformability of the rostral-field scale (inserted into tissue) may facilitate flexible body movement during locomotion.

Additional mechanical significance resulting from the local structural heterogeneities within scales can be inferred from the FE results, including the out-of-plane asymmetry (concave bending versus convex bending) and in-plane anisotropy (by bending on T–N and L–N planes). On the one hand, the comparison between concave and convex bending suggests higher flexibility under concave bending, especially in the R2 zone, which may also suggest adaptive significance. For instance, predatory loadings induce concave bending with great deformability, which delays tissue penetration/failure during the puncture loading and increases survivability. On the other hand, scales exhibit different structures on the T–N and L–N planes (i.e., transversal and longitudinal cross-sections, respectively). In the rostral field, R1, for example, the scales exhibit RS units in LL with a defined boundary between EEL and IEL on the T–N plane (figure 1h, modelled by the R1 model in figure 6a) but show surface radii grooves with triangular ridges (electronic supplementary material, figure S4c, approximated by the F0 model in figure 6a), leading to different bending response on these two planes. Since the FE results suggest the deformability that R1 > F0 (figure 6e), we can interpret that the bulk rostral field of the individual scales should have better deformability on the T–N plane (captured by the R1 model) than the L–N plane (captured by the R0 model). In fish body locomotion, however, higher flexibility is required along the anterior–posterior bending, corresponding to the bending in the L–N plane of individual scales; the discrepancy between the bending properties of individual scales and the functional requirement of the total fish body is compensated by the overlapping scales, where the scaled skin acquires higher flexibility along the anterior–posterior direction [50].

4.3. Mineralization in elasmoid fish scales

The structural and mechanical anisotropy at different zones of the scales is closely associated with the mineralization levels. In this work, we revealed that Mandl’s corpuscles may act as the precursors of mineralization in EEL during the scale mineralization process.

The growth of teleost fish scales continuously extends from the focus to the surrounding rostral and lateral fields (electronic supplementary material, figure S7). Therefore, the mineralization in the newly formed rostral field is lower than that in the developed focus field, which is exemplified by the larger relative thickness in the highly mineralized layers (LL + EEL) in the focus field than that in the rostral field (figure 2g–i). The growth of the scales in thickness and size initiates by fibroblasts forming collagen fibrous sheets beneath the IEL, and the fibre orientations are controlled by the surface ridges and invaginations of the fibroblasts [53]. The mineralization in the fish scale is rather complicated and involves different mechanisms in different structural layers [53,54]. (i) The mineralization of the fish scales initiates by depositing needle-like or flaky crystallites in the EEL (at the outer surface), which is accompanied by the formation of calcification loci in cellular vesicles during the formation of this layer. (ii) The subsequential mineralization of EEL inward is deposited without the mediation of those vesicles but via contact mineralization from previously mineralized regions. (iii) The mineralization in the outermost LL follows a different mechanism from the mineralization in the underlying EEL, involving the secretion of a collagen-free organic matrix. (iv) In the unmineralized IEL, however, random nucleation of the Mandl’s corpuscles and isolated calcifications of collagen fibrils are observed in the absence of vesicles and without contacting the pre-existing calcified tissue. Especially, it would be expected that residual stress is developed within the IEL due to the local contraction of the fibrils after mineralization [55], which induces tensile residual stress in the unmineralized fibrils and thus increases the penetration resistance against predatory loading.

Here, our μCT characterizations at different regions suggest that Mandl’s corpuscles could act as the precursors before full mineralization (figure 3). Specifically, in the newly formed R2 zones, the mineralized particles are small and exist more dominantly in the sublayers with longitudinal (radial) fibrils (electronic supplementary material, figures S3 and S4b), which could correspond to the initial mineralization of EEL by calcification induced by matrix vesicles. Later, the continuous mineralization in the EEL is followed by subsequential contact mineralization, as shown in electronic supplementary material, figure S4c,d. The isolated Mandl’s corpuscles exhibit random distribution and orientations within the scale (electronic supplementary material, figure S10), suggesting the nature of random nucleation. As the scale thickness increases, the mineralization boundary line between EEL and IEL should move inward by fusing the isolated Mandl’s corpuscles. The morphological changes in these particles at R1 and R2 regions could help explain the space-filling mineralization at the front: first, the high density of Mandl’s corpuscles nucleate not far away from the EEL–IEL boundary, where the small particles located below the large ones suggesting delayed nucleation and growth (figure 3e; electronic supplementary material, figure S8b,e). These mineralized particles may also have the function of bonding the adjacent fibrous laminates against delamination, yet further fracture analysis is needed to verify these assumptions (figure 7d–f). The particle morphologies change prominently from irregular geometries (figure 3e) to larger faceted crystallites (figure 3f) and then aggregates of multiple particles (figure 7e). It must be emphasized that the developed crystallites in the fibrils exhibit elongated octahedron shapes with the major axis along the fibril orientations. Yet, the mineral phase in elasmoid scales is HAP (space group P6₃/m) with a hexagonal crystal system, different from the developed octahedron geometry; the correspondence between the crystallography and the particle shapes requires further detailed analysis. Later, these aggregates of Mandl’s corpuscles, possibly induced by layered deposition of crystallites, fuse and fill the non-mineralized fibrils. Therefore, it may be inferred that Mandl’s corpuscles are the ‘precursor phases’ at the frontline of mineralization between the highly mineralized EEL and unmineralized IEL.

5. Conclusion

In summary, black drum fish scales represent a typical example of elasmoid scales with gradient mineralization towards scale interior deposited in collagen-based fibre laminates (i.e., a highest mineralized outer layer, a highly mineralized external elasmodine layer in the middle, and a non-mineralized internal elasmodine layer). At the material level, the gradient distribution of the mineral phases is directly associated with the stepwise decrease and local oscillation in the nanoindentation modulus and hardness towards the scale interior. At the macroscopic scale of the entire scale, the tensile properties of black drum scales are similar along the longitudinal and transversal directions, suggesting with the in-plane isotropy of the orthogonal cross-ply structures with circumferential and radial fibres. The deformation mechanisms in the fish scales involve the crack initiation at the circuli grooves in the limiting layer, delamination failure between the collagen fibrous laminates, and fibre bridging by the orthogonal fibre bundles across the cracked laminates.

Apart from the systematic structural–chemical–mechanical investigations of fish scales, this study highlights the structural variations at different zones within the individual scale. The newly formed anterior end of the scales consists of an elasmodine layer with alternating high and low mineralization and a limiting layer with fusiform radii sections (segmented by the surface radii interrupting the continuous circuli). Further growth leads to space-filling mineralization in the external elasmodine layer, resulting in a well-defined boundary from the unmineralized internal elasmodine layer underneath. The radii diminished at the fully developed focus field, leading to irregular ridges on the exterior surfaces. These different structural features along the scale length enable different bending resistance, which induces the greatest deformability at the anterior end of the scale to accommodate the flexible movement of the fish body in contrast to the stiffest focus field against puncture loading.

The isolated Mandl’s corpuscles at different zones map the mineralization evolution in the internal elasmodine layer. These particles nucleate randomly and are isolated in collagen fibrils, which apparently do not involve the cellular matrix vesicles or contact mineralization with previously calcified tissue. The particles are initially deposited as small irregular geometries and later develop into faceted octahedrons before coalescence, which drives the mineralization front towards the scale interior.

The insights of the structural variations within individual fish scales of black drum revealed here suggest that the fish scales do not have a uniform structure; instead, intricate microstructural control and local material heterogeneities are present, which may contribute to the animals’ multi-purpose requirement, such as locomotion and protection. The results of this work could also provide inspiration for designing novel bioinspired scaled armour materials.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

The data used in this work are available within the paper and in the electronic supplementary material [56].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

Y.T.: data curation, formal analysis, investigation, methodology, visualization, writing—original draft; Z.J.: investigation, methodology; Z.D.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing; L.L.: conceptualization, formal analysis, funding acquisition, methodology, project administration, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

L.L. gratefully acknowledges partial funding support by the United States-Israel Binational Science Foundation (BSF-2016341), National Science Foundation (DMR-1942865), and the Air Force Office of Scientific Research (FA9550-19-1-0033).