Introduction

After four decades of intensive research, one must conclude that the bone tissue engineering community stills faces still technical challenges of designing and manufacturing biodegradable scaffolds to fill the Form, Fixation, Function and Formation needs proposed by Hollister and Murphy.[1] Today, the field which might be defined as scaffold-guided bone regeneration (SGBR) does focus the development on composite scaffolds and offers a new approach to improve the treatment of large bone defects resulting from trauma, infection, or tumour removal,[2] which has been a challenge for patients and clinicians.[3] A typical case is a traumatic fracture in the tibial region, where bone is not surrounded by protective muscular tissues and fragmentation of the bone tissue is common.[4] Removing bone fragments is often the optimal method to control infectious propagation[5] eventually causing a large segmental bone defect in a load-bearing region that will complicate the healing.[6] Over the past 30 years, researchers have developed biodegradable composite scaffolds, such as 3D-printed polycaprolactone with ceramic filler material, with osteoconductive capacity and bone regeneration observed in pre-clinical and clinical studies.[7,8,9,10] These composites offer solutions to many significant challenges in the field, providing advantages in material processability, structural versatility, bioactivity, biomimicry, degradation kinetics and material properties. Biodegradable composites with weak interfaces could be particularly intriguing for addressing SGBR in large segmental defects that require load bearing, as they possess the ability to dissipate energy.

Indeed, one way to increase both strength and stiffness of composite scaffolds is to increase their ceramic content, but this strategy is usually detrimental for the toughness, since these materials eventually exhibit an undesirable brittleness.[11,12] Yet, examples of mineralized biological materials in Nature, such as mollusc shell, teeth, crab exoskeleton, or natural bone, teach us that it is possible to simultaneously optimize the mechanical properties in materials with high ceramic content. Indeed, these materials are made of brittle mineral blocks that are layered, assembled and combined with thin and soft polymeric interfaces that are capable of deflecting propagating cracks.[13,14] For example, nacre combines 95% of brittle and stiff mineral tablets (calcium carbonate) embedded in 5% of soft polymeric matrix (conchiolin). This unique architecture promotes many toughening mechanisms and dissipates 3000 times more energy than the plain mineral.[15,16] Enamel is structured with 95% of parallel crystals of mineral arranged in rods with soft interfaces (proteins) in between, conferring an increased overall toughness.[17,18] Finally, natural bone is composed of calcium phosphate-based mineral (60–70%) with an organic matrix made of proteins, mostly collagen (30–40%), responsible of many toughening mechanisms occurring at each of the hierarchical levels of its architecture.[19,20] In most cases, the toughening mechanisms are crack deflections. In fracture mechanics, a propagating crack follows a path that minimizes the energy. Brittle materials, such as ceramics, dissipate little energy since fractures propagate straightforwardly. Ceramics are a class of materials that are strong, stiff but not tough. Although counter-intuitive, mineralized biological materials teach us that the presence of soft, compliant interfaces confers to ceramic the capability to deflect cracks and increase the toughness of the material. When a crack is deflected from its original path and trapped into a longer and more tortuous course, new surfaces are debonded and energy is dissipated until, eventually, the crack stops. Interestingly, the propagation of crack in an interface, perpendicular to the initial direction, increases the toughness but does not affect the strength of the overall material (the cross-section remaining constant). In addition, a crack propagating straightforwardly in a brittle ceramic has a stress intensity factor at a crack tip that increases, which ease the propagation (further explanations are provided in “Engineering new toughened ceramics” section). On the opposite, if the crack extends in an interface and parallel to the applied stress, the stress intensity factor at the crack tip decreases with the immediate effect of dissipating more energy.[21] Then, the frictional forces between the layers, and eventually crack bridging, twisting or debonding are all toughening mechanisms congruous to crack deflection and contributing to the toughness.[22] Finally, when the crack kinks out of the interface, the crack re-initiation in the surrounding matrix will participate in increasing the overall strength.[23] The advantage of crack deflection in ceramics is therefore multiple. Crack deflection is a powerful toughening mechanism capable of increasing the total dissipated energy (and so the toughness) by several orders of magnitude,[14,15] as long as a series of geometrical and mechanical conditions on the mineral matrix and the interface are met.

This review aims to expose the role of the soft interfaces in the structure of natural bone, the associated toughening mechanisms and the resulting increased toughness. Then, the governing equations of the general case of cracks propagating in multi-layered materials provide insights and considerations for the design of toughened ceramic–polymer multi-layered biomaterials. This review will depict from these equations a series of guidelines for the design of future bone graft materials with high ceramic content and superior mechanical properties and directly inspired from natural bone.

Bone, a tough mineralized biological material

The simultaneous optimization of strength, stiffness and toughness of natural bone comes from its complex structure and composition acquired through million years of evolution. Bone exhibits a high ceramic content (calcium phosphate-based mineral, accounting for 60–70% of its mass) responsible of its high stiffness, whilst the structural protein matrix (mostly collagen, accounting for 30–40% of its mass) adds toughness.[19,20] A small percentage of non-collagenous proteins (1–3%) plays an active role on the overall toughness by being involved in many toughening mechanisms, such as osteopontin or osteocalcin that serves as binders between minerals.[24] Besides its composition, bone exhibits a unique hierarchical architecture with specific building blocks at each level. Each family of building blocks is paired with thin layers of ductile materials, called soft interfaces. Each structural level, with its building block and associated soft interface, is depicted in Fig. 1. The contrasting mechanical properties between the interfaces and the building blocks generates a variety of toughening mechanisms at each of the seven hierarchical levels, all contributing to the overall toughness of the whole bone.[14,19,25]

Figure 1
figure 1

The building blocks of the seven hierarchical levels of natural bone and their soft interfaces: (a) whole bone. (b) Cancellous and cortical bone. (c) Osteon structure. (d) Cross-ply structures in lamellae. (e) Isolated mineralized collagen fibre. (f) Mineralized collagen fibrils connected by non-collagenous proteins. (g) Collagen molecules connected by hydrogen bonds. Created with BioRender.com.

Full size image

A hierarchical architecture

At the first level is the organ bone [Fig. 1(a)] composed of several tissues (periosteum, endosteum, bone marrow,…). From an engineering point of view, the material bone can be divided into a porous inner part, the cancellous bone and a denser outer shell, the cortical bone [Fig. 1(b)]. This neat separation is the second level of the architecture. Cancellous bone is composed of interconnected struts, the trabeculae, directed along the principal biomechanical stresses of the organ bone. This porous microstructure affects the propagation of an incoming crack as it constantly deflects its directional path, dissipating more energy than a straight-forward propagation.[26] Cortical bone is made of osteons (200–300 µm) which are the third level of the architecture [Fig. 1(c)]. Osteons consist of concentric layers of mineralized collagen, the lamellae, wrapped into a thin layer with lower collagen content and increased mineralization, the cement line.[27] Cement lines, which are much weaker and brittle than the bone matrix, have the ability to deflect and to slow the microcracks propagating in the cortical bone. They also accumulate damages, therefore contributing to the overall toughness of bone.[28,29] The lamellae of osteons are the fourth level of the architecture [Fig. 1(d)]. These layers are composed of mineralized collagen fibres arranged in a variety of patterns, such as woven, plywood-like or parallel structures. They are separated by non-collagenous proteins, disordered collagen and minerals, all together constituting softer interlamellar interfaces capable of deviating microcracks.[14] The fifth hierarchical level is the mineralized collagen fibres [Fig. 1(e)], themselves an array of mineralized collagen fibrils, at the sixth level [Fig. 1(f)]. The fibrils contribute to the toughness by deforming elastically and inelastically, sliding on each other or closing the crack wake (a mechanism called bridging).[30,31] Fibrils are held together by a thin 1–2 nm layer of weakly bonded non-collagenous proteins, such as osteopontin or osteocalcin [Fig. 1(f)]. These proteins form sacrificial molecular complexes that are more compliant than the fibrils and are associated with crack deflection, pullout or crack bridging and can rapidly reform after failure.[32,33] Finally, at the last level, fibrils are divided into collagen molecules combined with plate-shaped minerals (carbonated apatite). Collagen molecules can stretch or slip over minerals and are connected by weak hydrogen bonds [Fig. 1(g)]. Similarly to the inter-fibrillar bonds, hydrogen bonds can break and continuously reform, contributing to the toughness at the nanoscale.[34,35]

Role of soft interfaces on toughness

The range of the mechanical properties resulting from the multiphasic morphology of bone is wide but remarkably high for a biological material. A tensile strength of about 150 MPa (cortical bone) and a Young’s modulus of 10 GPa are commonly reported for cortical bone.[19,20,25] These properties are combined with a high fracture toughness, ranging from 3 to 20 MPa m1/2, resulting from the sum of the toughening mechanisms activated by the soft interfaces.[15,19,33] The alteration of the soft interfaces at a specific hierarchical level can have dramatic effects on the overall mechanical properties Many examples have been identified at the micro- and nano-scales: The deformations of the inter-fibrillar interfaces are responsible of 60% of the deformation of bone[36]; because of the action of cement lines, the toughness of cortical bone is five times higher in the transverse direction (the direction where cracks split osteons) than the longitudinal direction[37]; inhibiting the adhesive action of osteopontin significantly decreases the overall toughness[38]; the increase of stiffer covalent bonds in the inter-molecular interfaces (seventh level) and inter-fibrillar interfaces (sixth level) is responsible of the decline of the mechanical properties with age.[39]

Therefore, the role of soft interfaces in the bone toughness has been clearly evidenced, and the dissimilarity between the mechanical properties of the matrix and those of the interfaces is the key to efficiently promote crack deflection and increase toughness. Indeed, amongst the variety of toughening mechanisms occurring in bone (such as crack deflection, debonding, pullout of collagen fibrils, pullout of osteons, crack bridging, crack twisting[13,14,40]), crack deflections have been identified by experimental models as the main contributor to toughness.[19,41] For instance, crack bridging in bone microstructure, mainly due to collagen fibrils and collagen fibres stretching, contribute in an increase of 1–2 MPa m1/2 of the toughness, whilst the crack deflections at soft interfaces, such as osteon cement lines, can contribute as much as 3–20 MPa m1/2 to the fracture toughness.[19] This result suggests that future engineered ceramics must consider architectures with interfaces to promote crack deflections. The next section reviews the equations and the conditions to deflect a crack at an interface with the perspective of defining criteria on the mechanical properties and geometry of an ideal interface.

Engineering new toughened ceramics

Strength-based criterion

In fracture mechanics, a unique principle governs crack propagation: a propagating crack seeks to a path of least resistance. In this section, we are considering a multi-layered material made of alternating layers of matrix and soft interface. For simplification purposes, we suppose that each material (interface and matrix) is elastic, has same properties in all directions (isotropic) and is in mixed-mode loading (loaded in all directions): opening mode (mode I), in-plane shear (mode II) and out-of-plane shear (mode III). Then, ahead of the crack tip can be written the stress field σij in polar coordinates [Fig. 2(a)] that is function of the mechanical properties and distance from the tip:

$${\sigma }_{i,j}\left(r,\theta \right)=\frac{{K}_{\text{I}}}{\sqrt{2\pi r}}{f}_{i,j}^{\text{I}}\left(\theta \right)+\frac{{K}_{\text{II}}}{\sqrt{2\pi r}}{f}_{i,j}^{\text{II}}\left(\theta \right)+\frac{{K}_{\text{III}}}{\sqrt{2\pi r}}{f}_{i,j}^{\text{III}}\left(\theta \right),$$
(1)

where θ and r are the angle and distance from the tip; KI, KII and KIII are the stress intensity factors for the mode I, II and III and depend on the mechanical properties, microstructure of the material and fIij, fIIij and fIIIij are factors that depend on the angle of propagation.[42] The propagation of the crack is determined by the value of the stress at the crack tip, e.g. by \(\underset{r\to 0}{\text{lim}}{\sigma }_{i,j}(r,\theta )\). The interpretation of this mechanism is that crack propagation can be seen as the crack was “scanning” a small area ahead of the tip and selecting the path with the highest stress field, and more particularly, the path with the highest Hoop stress σθθ. If the crack branches by an angle θ, σθθ must verify[43]:

Figure 2
figure 2

Successive steps of the propagation of a crack in a multi-layered material. (a) Crack propagating straight-forward in the stiff ceramic matrix. The stress field σij as function of θ and r is shown. (b) Crack breaking through the soft interface and (c) a first case is a small kink extending out of the soft interface, back in the stiff matrix. (d) A second case is a 90° deviation of the propagation path, along the interface–matrix border. (c) and (d) Illustrated by microscopy examples of crack propagation in an alumina/lanthanum phosphate composite ceramic. Adapted with permission from Reference 44.

Full size image
$$\left\{\begin{array}{c}{\sigma }_{\theta \theta }=\frac{{K}_{\text{I}}}{\sqrt{2\pi r}}{\text{cos}\left(\frac{\theta }{2}\right)}^{3}+\frac{{K}_{\text{II}}}{\sqrt{2\pi r}}\left[-3\text{sin}\left(\frac{\theta }{2}\right){\text{cos}\left(\frac{\theta }{2}\right)}^{2}\right],\\ \frac{\partial {\sigma }_{\theta \theta }}{\partial \theta }=0.\end{array}\right.$$
(2)

which can be solved for θ in

$$\text{tan}\left(\frac{\theta }{2}\right)=\frac{1}{4}\left(\frac{{K}_{\text{I}}}{{K}_{\text{II}}}-\sqrt{{\left(\frac{{K}_{\text{I}}}{{K}_{\text{II}}}\right)}^{2}+8}\right).$$
(3)

Equation 3 gives immediately that for pure loading mode I (\(K_{{\text{I}}} \gg K_{{{\text{II}}}}\)), θ = 0 and no branching is possible. On the opposite, pure in-plane shear (\(K_{{{\text{II}}}} \gg K_{{\text{I}}}\)) or mixed-mode loading will provide non-null values of θ, showing that the crack is prone to branch in the material. In a scenario where the crack tip crosses an interface made of a material with dissimilar properties (e.g. a softer material with a lower toughness and thus a lower KI, as discussed later on Fig. 2), the ratio \(\frac{{K}_{\text{I}}}{{K}_{\text{II}}}\) may be drastically modified, therefore affecting the direction of branching in the interface. A soft interface with dissimilar mechanical properties affects the stress field ahead of the crack tip and the direction of propagation.

This scenario is depicted in Fig. 2: after the crack crosses the interface and reaches the matrix, the stress field could be higher in the interface after a 90° deviation than straightforward in the matrix. The conditions on the mechanical properties of the interface to allow such scenario have been proposed in two approaches, strength based and energy based, which led to establish criteria for crack deflection at interfaces.

In a simple analysis of the stress field, Cook et al.[45] revealed that the maximum normal stress ahead of the crack was about five times greater than the maximum normal stress perpendicular to the crack tip and suggested that an interface with a strength σi five times smaller than the strength of the matrix σm was sufficient to deflect a crack at the interface–matrix border: \(\frac{{\sigma_{{\text{i}}} }}{{\sigma_{{\text{i}}} }} < \frac{1}{5}\). With the extension of this concept on the stress field around a sharp crack tip in a elastically homogeneous system, Gupta et al. revised the criterion in 1992[46] to

$$\frac{{\sigma }_{\text{i}}}{{\sigma }_{\text{m}}}<\frac{1}{3.5}.$$
(4)

In natural bone, push-out tests on individual osteons have measured the shear strength of the cement lines enveloping the osteons and demonstrated that the cement lines and bone matrix were verifying the criterion established in (2): The shear strength of the cement lines was measured at 8 MPa,[47,48] which was nine times smaller than that of the surrounding bone matrix (73 MPa).[49]

However, the disadvantage of this approach is that crack deflection becomes much less likely if the interface is more compliant than the matrix.[23] A criterion based on the strength only is therefore problematic as it does not describe the variety of situations observed in mineralized biological materials. Indeed, compliant interfaces successfully deflect cracks from a stiff matrix like conchiolin and calcium carbonate in nacre.

Energy-based criterion

He & Hutchinson proposed a new approach in 1989[50] involving the energy release rate at the crack tip, G. In elastic fracture mechanics, G is the energy that is dissipated because of the creation of new surfaces due to the advance of the crack, in Joules per unit area. Since the crack seeks to a path of least resistance, the ratio of the energy release rate ahead of a straight-ahead crack advance, G0, to the energy release rate of a 90° kinked crack tip, G90, must be considered [Fig. 2(c), (d)]. To ensure crack deflection, this ratio must be lower than the ratio of the toughnesses in the two directions: Γm, the toughness of the matrix ahead of the crack tip and Γi, the toughness of the interface in the 90° direction. The criterion is then

$$\frac{{G}_{0}}{{G}_{90}}<\frac{{\Gamma }_{\text{m}}}{{\Gamma }_{\text{i}}}.$$
(5)

Or

$$\frac{{G}_{90}}{{G}_{0}}<\frac{{\Gamma }_{\text{i}}}{{\Gamma }_{\text{m}}}.$$
(6)

Although this criterion now explains how ductile interfaces are more likely to provoke crack deflection [with a lower Γi, (4) is more likely to be verified], its disadvantage is the difficulty to estimate the ratio \(\frac{{G_{{{90}}} }}{{G_{{0}} }}\). Γm and Γi are, however, properties of the materials that can measured during specific fracture tests. In 1994, He et al.[51] found an expression for the ratio \(\frac{{G_{{{90}}} }}{{G_{{0}} }}\) and showed that it was highly dependent on the Dundurs parameter \(\alpha = \frac{{E_{{\text{i}}} - E_{{\text{m}}} }}{{E_{{\text{i}}} + E_{{\text{m}}} }}\), where Ei and Em are, respectively, the plane strain tensile moduli of the interface and the matrix. The parameter α can therefore be interpreted as the elastic mismatch between the interface and the matrix. Beside the newly expressed ratio \(\frac{{G_{{{90}}} }}{{G_{{0}} }}\), numerical investigations on the effect of the interface–matrix elastic mismatch have shown that \(\frac{{G_{{{90}}} }}{{G_{{0}} }}\) reaches a minimum of 0.25 (Fig. 3). Finally, as long as the ratio of toughnesses \(\frac{{\Gamma_{{\text{i}}} }}{{\Gamma_{{\text{m}}} }}\) is below a critical threshold of 0.25, the criterion expressed in (4) will be satisfied for all α and crack deflection will occur. The energy criterion to ensure crack deflection is then

Figure 3
figure 3

Graph of the ratio of the energy release rates \(\frac{{G_{{{90}}} }}{{G_{{0}} }}\) as function of the elastic mismatch α. The ratio reaches a low point at approximately 0.25 for α slightly negative. Adapted with permission from Reference 51.

Full size image
$$\frac{{\Gamma }_{\text{i}}}{{\Gamma }_{\text{m}}}<\frac{1}{4}.$$
(7)

A simplified version of the energy-based approach therefore consists in selecting an interface that is at least a four times lower toughness than the matrix, and crack deflection will be initiated at the interface–matrix border. Then, as the deflected crack grows in the interface, G90 increases proportionally with the length of propagation and finally reaches a steady state (when the crack propagation length is large compared to the thickness of the interface) that is independent of the mode of loading, crack length or properties of substrate.[52] Consequently, once crack deflection has been initiated, the crack will not automatically kink out of the interface and a local increase of G0 (for instance, a flaw or defect in the matrix) is required to draw the crack out of the interface and initiate cracking in the next matrix layer.[53] Paradoxically, imperfect ceramics with flaws are beneficial for the toughness as they ensure that cracks cannot propagate in the interface until catastrophic delamination. The existence of an optimal critical flaw size that maximizes the toughness and that is function of the ratio of the toughnesses was demonstrated by Kovar et al.[53] Indeed, large flaws draw cracks out of interfaces early, with little energy dissipated, but small flaws promote longer travelling distance of crack in interfaces and increase the risk of delamination. Interestingly, after a delamination crack kinks out of the interface, the re-initiation of matrix cracking will then increase the overall strength of the material.[23]

The criterion (5) is also verified by the cement lines of osteons, which exhibit a toughness of 1–2 MPa m1/2, approximately 10 times lower than that of bone matrix.[41] Another example is that natural bone, and more generally mineralized biological materials are much tougher in hydrated conditions. This can be explained by the hydration that plasticizes the organic matrix, lowers its toughness, which results in better crack deflection capabilities, and higher overall toughness.[13] In nacre, although hydration slightly decreases its strength and Young’s modulus, it increases the toughness by a factor 3.5.[54]

Guidelines for the design of tough multi-layered composite scaffolds

The previous analysis highlighted that dissimilar mechanical properties between the matrix and the interface are essential to promote crack deflection. Both strength and toughness govern the capability of a layered material to deflect cracks, as demonstrated by the criterions (2) and (5). However, considering the strength only in the choice of the interface may not be systematically sufficient since the strength-based approach failed to include many observations from biological materials. Parmigiani and Thouless have numerically investigated in 2005[23] this strength–toughness duality and have finally shown that the dominating regime (toughness-controlled fracture or strength-controlled fracture) was influenced by the fracture-length-scale parameter \(\frac{E\Gamma }{{\sigma^{2} L}}\), a nondimensional parameter that is function of the modulus E of the material, the work of fracture Γ of the material (a version of the fracture toughness that is based on the overall dissipated energy and calculated by measuring the area under the traction–separation curve), the strength σ of the material and L, a characteristic length of the material. At large values of fracture-length scale (> 0.01), the failure mechanism is controlled by the strength criterion. At smaller values (< 0.001), crack deflection becomes governed by the toughness criterion. In between, the regime controlling fracture is influenced by both toughness and strength, and the simultaneous verification of both criterions guarantees the crack deflection capability of the composite. In the case of natural bone, it is reasonable to estimate that the fracture-length-scale parameter is in the intermediate range (0.001–0.1), meaning that both strength and toughness are governing the fracture mechanisms. For brittle ceramics used for bone tissue engineering (calcium phosphates or calcium sulphates), however, an estimation of the fracture-length-scale parameter indicates a range lower than 0.001. Therefore, a toughness of the interface four times lower than that of the calcium cement matrix is decisive in the design of these multi-layered ceramics.

Lastly, numerical and experimental investigations on multi-layered calcium cements for bone tissue engineering have highlighted the influence of geometrical parameters on the toughness of the material. For instance, a higher number of layers[55,56,57] or a lower volume fraction of interface[21,44,57] (e.g. thin interface layers) were beneficial to promote toughening mechanisms, similarly to what is currently observed in biological materials. Indeed, more layers provide more opportunities to deflect an incoming crack, as shown by the successive peaks in stress–strain curves [Fig. 4(a)], corresponding to the successive failures of the layers, delaying the complete failure of the sample.[21] Thinner interfaces, provided that they are thick enough to initiate crack deflection and propagation,[21,58] promote the re-initiation of the crack back in the matrix,[23] thus avoiding an undesirable delamination. The work also emphasized the benefit of a less stiff interface on crack deflection, already noticed by He et al. whilst exploring the effect of the elastic mismatch[51] and observable on Fig. 3. Finally, experimental results have shown the importance of the adhesion between layers that requires to be optimized to promote crack deflection but not catastrophic delamination.[55]

Figure 4
figure 4

(a) Comparison of the stress–strain curves of monolithic and multi-layered silicon carbide, thus showing the successive failures of the layers and the increased area under the curve. Adapted with permission from Reference 21. (b) Different patterns of crack propagation in a silicon nitride ceramic multi-layered from soft (top) to tough (bottom) interfaces. Propagation lengths decrease as the toughness of the interface increases and reaches that of the matrix layers. Adapted with permission from Reference 53. (c) Work of fracture of Al2O3/LaPO4 composite versus interface thickness. The toughness of the multi-layered ceramic decreases as the thickness of the interface increases, as thicker interfaces increase the risk of longer propagation lengths and delamination. Adapted with permission from Reference 44.

Full size image

Examples of tough engineered ceramics

First designs of tough multi-layered ceramics have been reported in 1990 by Clegg et al.[21,58] In these works, thin sheets of silicon carbide were combined with graphite as soft interface. Both materials had very close flexural moduli and flexural strength, but because of the dissimilar toughnesses verifying the criterion (5), the average work of fracture of the composite was 4625 J m−2 with values as high as 6700 J m−2. The average value represented a 75-fold increase compared to plain silicon carbide (62 J m−2). Figure 4(a) displays the flexural force–deflection curve of the multi-layered silicon carbide where the multiple crack bifurcations result in a progressive and delayed failure. The work of fracture (area under the stress–strain curve) of the composite is significantly increased compared to plain silicon carbide that fails suddenly and straight-forwardly, resulting in a curve with very little work of fracture.

Another impressive design was descripted by Kovar et al.,[53] where silicon nitride was multi-layered with silicon nitride containing boron nitride as soft interfaces. The fabrication technique allowed to tailor the composition of the interface from 0 to 80 vol% of silicon nitride, and the interface toughness from 30 to 80 J m−2. As depicted on Fig. 4(b), no crack deflection was observed and very little energy was absorbed for specimens with high interfacial toughness (> 80 J m−2). Crack deflection occurred for specimen with intermediate interfacial toughness (50–80 J m−2), but with short propagation lengths, and extensive delamination was observed for specimens with the lowest interfacial toughness (30–50 J m−2). The resulting toughness of the composite increased as the toughness of interface decreased. The study demonstrated that the energy dissipated by the toughened ceramic was not only influenced by its capability to deflect cracks but also by the crack path in the interfaces and the capability to maintain an interfacial propagation.

A last example of toughened ceramic is a work from 2007 by Tomaszewski et al.,[44] in which alumina (Al2O3) was layered with lanthanum phosphate (LaPO4) as soft interphase. Whilst the flexural strength of the composite (356 MPa) could be explained by the “rule of mixture” (the flexural strength of Al2O3 was 432 MPa and that of LaPO4 was 103 MPa), the work of fracture of the composite was superior to its components by a factor 30: the work of fracture of Al2O3 and LaPO4 were, respectively, 36.2 and 4.7 J m−2 [satisfying the criterion (5)] and that of the composite was 1165.1 J m−2. In addition, the study revealed the impact of the layer thickness and the reduction in toughness as the interface layer increases [Fig. 4(c)], as discussed previously. Indeed, as predicted by Parmigiani and Thouless,[23] thicker interfaces increase the risk of catastrophic delamination.

Manufacturing processes

Multi-layered composite ceramics can be manufactured using a variety of methods,[59] but some are more adapted to follow the guidelines previously mentioned. After selecting the appropriate materials for ceramic and interface, the manufacturing process must ensure a high control on the thickness of each layer, in order to provide a number of layers as large as possible and interfaces as thin as possible. First, the moulding of ceramic sheets can be performed via tape casting of an aqueous ceramic suspension on a smooth substrate[55] or by rolling technique, where the ceramic mixture is laminated at an optimal speed. These techniques allow fabricating ceramic sheets as thin as 200 µm,[58,59] which can then be sintered if necessary. Interestingly, tape casting could be particularly advantageous in the manufacturing of patient-specific bone implants (e.g. with particular geometry). The process consists in using 3D-printed substrates with the desired geometry for the casting of the ceramic suspension, but the concept needs to be further developed and experimented.[55] Once fabricated, the ceramic sheets can be assembled by sequential stacking and pressing. The interfacial material can be applied on the ceramic sheets by dip coating or “painting” before pressing,[59] since these techniques allow the formation of thin interfaces and guarantee a higher toughness.[57]

Another way to fabricate multi-layered scaffolds for bone regeneration is extrusion-based additive manufacturing, a 3D printing technique allowing to print a variety of materials with ideal physico-chemical properties for SGBR.[60] The technique was successfully experimented for instance on polymers (such as polyether ether ketone,[61] polycaprolactone,[62,63] polylactic acid[64]), ceramics (such as β-tricalcium phosphate,[65] wollastonite,[66] hydroxyapatite,[67]) or their composites.[68,69,70,71,72] It is however crucial to achieve a stable and homogeneous composite blend with proper ceramic particles distribution and dispersion throughout the polymer matrix in order to fully realize the benefits of such blends (Fig. 5). Successful dispersing not only mechanically reinforces the resulting composite but also enhances its bioactivity by enabling the material to release osteoinductive ions. These benefits are further amplified as the dispersion resolution is increased and the particle size homogeneity is improved. The understanding of the dispersion mechanisms (Fig. 5) opens the door to 3D-printed composites with high ceramic content (up to 40%).[73] Extrusion-based 3D printing technique appears particularly tailored to fabricate versatile composites combining challenging multi-layered structure and customized geometry for specific bones.[60]

Figure 5
figure 5

Reproduced with permission from Reference 73.

Recent comprehension of the dispersion mechanism of ceramic (here, calcium phosphate, CaP) particles, driven by the presence of positive and negative charges at the surface within a printed polymeric matrix, have led to increase the ceramic content up to 40%. Section I shows a representative SEM image and illustrates the particles position within the nozzle of a 3D printer. Section II represents the particle distribution after the composite has been extruded through the nozzle, with representative SEM image.

Full size image

Conclusion

Mineralized biological materials exhibit a multi-scale structure in which soft interfaces play a central role by channelling deformations and deflecting cracks, thus increasing the overall toughness. Mimicking these strategies in mineral-based synthetic materials is a promising approach for bone tissue engineering where ceramics, an important class of material used in this field, exhibit a lack of mechanical properties. The field of fracture mechanics has identified the mechanical and geometrical conditions that governs crack propagation at interfaces. The understanding of these mechanism has led to a series of general design guidelines for toughened composite scaffolds: (i) a multi-layered structure must be preferred, with calcium-based ceramics as matrix layers, because of their advantageous physico-chemical properties for bone regeneration. (ii) A ductile phase, like a polymer, with a toughness at least four times smaller than the matrix must be selected as interfacial layers. (iii) The fabrication technique of the composite must allow a number of layers as large as possible and an interface thickness as small as possible. The application of these guidelines could lead to new medical bone graft technologies with improved mechanical properties. Such materials could have significant impacts on restoring functionality and quality of life in patients requiring reconstruction of structural bone defects with load-bearing property, due to a variety of pathologies.

Despite the potential improvements promised by those guidelines, quantitative predictions of the overall toughness enhancement remain delicate, given that the increase in toughness is the sum of the contributions of each deflected crack. Indeed, the path of a crack was modelled here in its ideal version, but the distribution of imperfections and flaws in the material confer unpredictability to the real path of a crack. However, when the properties of interfaces promote crack deflection, the material will eventually dissipate more energy, a statistical consequence illustrated in the works of Clegg, Kovar of Tomaszewski. Further work using statistical tools could provide not only qualitative but also quantitative prediction on toughness enhancement and refine the conditions required by an interface to promote toughening mechanisms.

The development of multi-layered scaffolds for segmental bone defect repair shows promises, but further investigation is needed to move from bench to bedside. One key area requiring attention is the adhesion between layers in a multi-layered matrix, as it significantly impacts the overall mechanical behaviour. Currently, empirical methods are used to determine interfacial adhesion, but greater control is needed to prevent delamination. Whilst extrusion-based 3D printing is a promising manufacturing method, it has limitations, such as being dependent on the melt viscosity of the composite material. As a result, polymers with low ceramic content have been predominantly used, despite the preference for layers with higher ceramic content. Our group has recently reported on the fused deposition modelling of a composite scaffold with a Voronoi design, using a material combination of 60% polylactic acid and 40% ceramic. This emerging technique could accelerate the development of toughened biodegradable composites for SGBR. Another critical area is the sterilization of the implant, as most studies are conducted on raw materials rather than final prototypes. To advance the treatment of large segmental bone defects, pre-clinical studies in well-characterized and validated large animal models are imperative, as these treatments require bones with load-bearing properties. Lastly, cost reduction is a significant barrier. Designing patient-specific implants with complex architectures requires expensive medical grade materials and engineering time for the design.[60] However, a recent investigation has shown that routine clinical implementation of SGBR can be facilitated and accelerated using custom-made software.[74] This emerging technique could pave the way to automated and standardized scaffold designs with significantly lower costs.