## Introduction

In millions of years, evolution developed materials with self-healing properties. These natural materials have the capability, to partially or completely restore in shape or function after the event of damage1. In nature there are many familiar examples of self-healing that we take for granted, such as spontaneous healing of ruptured skin or mending of broken bones. Healthy human bone is a material that is permanently in change2. Cells, osteoclasts and osteoblasts, respond to biomechanical stimuli and degrade or build bone accordingly, by using ions, such as calcium and phosphate that are present in the blood plasma as well as stored in already existing bones and teeth3. Due to these actions, a broken bone is able to mend.

In the recent 20 years, intense research on the development of self-healing materials was performed1,4,5. Hereby, several strategies are to be distinguished. First of all, a distinction is made between autonomic healing, where no external trigger or action is required and non-autonomic healing, that presupposes external triggers. Furthermore, a differentiation is made between intrinsic healing, based on the inherent properties of the material, and extrinsic healing using external agents supplied in, for example, capsule systems6. Since polymers show chain mobility even at comparatively low temperatures, and a large number of chemical modification options are available, a multitude of self-healing mechanisms are found for this by far most intensively studied material class4,5,7,8,9,10,11,12,13. Nevertheless, pathways for metals, ceramics and concretes were developed, as well12,14,15,16. Many healing methods require external triggers, with temperature regimes being the most utilized5. In terms of extrinsic healing with capsule systems, the repeatability of the healing process is restricted1. These factors can limit the fields of potential applications. However, applications with poor accessibility, for example in the construction industry or especially in clinical practice, particularly depend on self-healing materials, since here the replacement or even the detection of defects can be problematical. For clinical practice, numerous self-healing polymeric biomaterials and hydrogels were investigated and developed in the past 10 years9,13,17. For brittle materials like cements or concrete the stabilization of the defect is essential for the healing strategy and can be achieved by different pathways18. In the case of concretes utilized for construction some systems with self-healing capability were developed. Hereby, different strategies were investigated with re-hydration, capsule systems or biological processes being most prominent12. For the latter, bacteria are added to the cement system, which produce mineralization products closing the crack19. Defect stabilization and healing itself is typically based on the formation of salts in the crack, blocking of cracks with impurities or particles resulting from water infiltration or crack spalling, further hydration of unreacted cement or finally the expansion of cementitious matrix in the crack flanks15. Self-healing with apatite materials was shown for apatite coatings and by utilizing phosphate-containing hydrogel in Portland cement paste20,21.

To design a self-healing calcium phosphate cement (CPC) we suggest a mechanism with inspiration from self-healing mechanisms used for construction materials, utilizing crack stabilization and closing by damage tolerance and subsequent the formation of precipitate in the crack. Although both cements in the construction industry and CPCs, have to struggle with their brittle behavior and their poor accessibility in application in terms of self-healing, there are some very important differences in the challenge of designing a self-healing material. Besides others, the size of the components and workpiece differ and the choice of materials in vivo is limited to biocompatible ones. In general, CPCs result from the setting reaction of a calcium phosphate (CaP) powder and a liquid phase. Depending on the pH of the reaction either apatite or brushite is formed. However, in contrast to polymeric biomaterials or hydrogels this implant material, in clinical application since 1994 and under intense research for roughly 35 years by now, has significant advantages22,23,24,25. Due to the paste-like state of CPC between mixing and setting, the cement can be injected and is therefore of particular interest in minimal invasive surgery25. Beyond, the paste is capable to fill complex geometries and to ensure an intimate bone-implant-contact. Setting itself is an autonomous process, at moderate temperatures and without significant volume changes, which occurs in vivo26. Thus, under physiological conditions the resulting precipitate, shows high chemical and structural similarity to mammalian hard tissue, that is composed of biological apatite27. Finally, setting reactions at physiological temperatures allows the incorporation of drugs such as antibiotics, anti-inflammatory agents or growth factors in order to stimulate certain biological reactions28. Beyond these advantages, however, CPCs are suffering from their inherent brittleness, which restricts their application29. To overcome this drawback, damage tolerant systems were developed using, e.g. fiber reinforcement30,31,32,33,34. Among others, carbon fibers (C-fiber) proved to be effective35,36,37. In a previous study we investigated the influence of chemical fiber treatment on the mechanical performance of CPC35. The bending strength of CPC was increased from 9 MPa to 19 MPa by adding 1 wt% untreated C-fibers. A chemical pretreatment of the C-fibers with aqua regia and calcium chloride lead to a further increase to 30 MPa. Additionally, the opening of occurring cracks was stabilized and fatal crack propagation was prevented, resulting in enhanced Work of fracture (WOF5). In addition to the tensile strength of the C-fibers, the surface chemistry of the fibers also influences the setting reaction of the CPC and thus the resulting matrix structure and strength. In the case of CPC with modified fibers, the fiber surface treatment increases the matrix fiber bond, leading to adhered matrix material that remains on the fibers surface after fracture. These remaining matrix particles impede the fiber pull-out, since they cause higher friction. Consequently, higher strength and WOF5 is observed for the composite35.

In the present study, we go even further and present a self-setting, damage tolerant CPC with in vitro self-healing capability. For this purpose, an apatite cement with C-fiber reinforcement was damaged in a controlled manner and healed in simulated body fluid (SBF). Since SBF mimics the composition of the inorganic part of human blood plasma, it can reproduce the in vivo formation of apatite on bioactive surfaces and is therefore used to investigate the in vitro bioactivity38,39,40. In this context, a surface is considered to be bioactive, if it facilitates the heterogeneous nucleation and growth of apatite41,42. Combining the in vitro ability of SBF to form apatite with a damage-tolerant bioactive material, the aim of the present study is to design a cement-based implant material with an intrinsic self-healing capacity, where cracks are instantly healed by a biomimetic mineralization process. Beyond, also an extrinsic healing pathway will be investigated, using H2PO4- supplying capsules, which trigger the precipitation of CaP. Such self-healing CPC are of potential interest for load-bearing orthopedic applications, since occurring cracks might be filled with new bone instead of leading to catastrophic in vivo failure.

## Materials and Methods

The experimental procedure for sample preparation and self-healing tests is schematically illustrated in Fig. 1. As indicated on the left, the CPC samples were prepared as a composite of C-fibers embedded in an α-tricalcium phosphate (α-TCP) matrix. Partly sodium dihydrogen phosphate (NaH2PO4) capsules were added to enable extrinsic healing experiments. Besides other tests, self-healing was examined regarding to the regime presented on the right in Fig. 1. In the following chapters, the single parts of the experimental procedure are described in detail.

### Cement preparation

α-TCP was prepared by sintering calcium hydrogen phosphate (CaHPO4, Mallinckrodt-Baker, Germany) and calcium carbonate (CaCO3, Merck, Germany) in a molar ratio of 2:1 for 5 h at 1400 °C followed by quenching to room temperature. The sintered cake was crushed and passed through a 125 μm sieve followed by ball milling at 200 rpm for 4 h43. A solution of 1 M trisodium citrate (Na3C6H5O7, Carl Roth, Deutschland) and 2.5 wt% disodium hydrogen phosphate (Na2HPO4, Carl Roth, Deutschland), prepared with demineralized water, served as liquid phase in the cement preparation.

C-fibers (Nippon Graphite Fiber Corporation, Japan) with a diameter of d = 7 µm were cut to a length of 1 cm before desizing twice in boiling isopropanol (Carl Roth, Germany). Subsequently, the fibers were dried in a freeze dryer and processed as they are or stirred for 40 min in aqua regia (HCl, Carl Roth, Germany and HNO3, Carl Roth, Germany), washed with water and stored in 1 M calcium chloride (CaCl2, VWR Chemicals, Belgium) solution for 1 day, as reported in detail previously35. After the modification the C-fibers were freeze-dried.

To prepare capsules NaH2PO4·2 H2O (VWR Chemicals, Belgium) salt was coated with Degacryl (PMMA, 15 kDa, Evonik Industries AG, Germany) dissolved in acetone using a fluidized bed coating (Mini Glatt 12104, Glatt, Germany). Particle size and surface ratio were estimated by analyzing light microscopic images of NaH2PO4 capsules and cement fracture surfaces using the software ImageJ. Light microscopy (Axio Vision, Carl-Zeiss, Germany) was  performed in light field using z-stack.

α-TP, 1 wt% C-fibers, and optionally 2 wt% NaH2PO4, according to the compositions listed in Table 1, were mixed in dry state by hand and combined with the liquid phase at a powder-to-liquid ratio (PLR) of 3 g∙mL−1. The mixed CPC pastes were transferred into silicon molds and stored at high humidity and 37 °C for 4 h. Then they were stored to demineralized water for final setting at 37 °C for 7 days. Hereby, 40 g α-TCP and the corresponding amounts of additives were used to obtain 32 samples, 16 of which were each tested with similar parameters (e.g. 16 reference in water and 16 immersion in SBF for specific time). After setting in demineralized water for 7 days, CPCs were ground plane-parallel using abrasive paper up to a grit of 1000.

Applying the Gilmore needle test, the initial (needle diameter 2.12 mm, weight 113 g) and final (needle diameter 1.06 mm, weight 454 g) setting time was estimated using plate specimens with a diameter of 20 mm.

### Mechanical properties and self-healing capacity

The mechanical properties of CPC were characterized using a universal testing machine (Z020, Zwick, Germany). In three-point bending tests specimens with dimensions of length l = 30 mm, height h = 3–4 mm and breadth b = 6 mm were tested applying a support span of L =  20 mm, a loading rate of 1 mm·min−1, and a preload of 0.1 N. The bending stress σb was calculated according to Eq. 1, with the force F. The strain εb was calculated according to Eq. 2 with the  deflection s. Additionally, the Work of fracture up to a strain of 5% WOF5 was calculated according to Eq. 3. Hereby the force F was integrated over the deflection s from s0 = 0 to the deflection s5, where εb equals 5%. Although samples show much higher strain at failure, the residual strength at higher strain strongly depends on fiber orientation, causing the error  being vast. Since 5% strain is closer to the actual strain of bone44 this limit was chosen for a better comparability also concerning clinical applications.

$${\sigma }_{b}=\frac{3FL}{2b{h}^{2}}$$
(1)
$${\varepsilon }_{b}=\frac{6sh}{{L}^{2}}\cdot 100 \%$$
(2)
$$WOF5=\frac{{\int }_{{s}_{0}}^{{s}_{5}}Fds}{bh}$$
(3)

For the test of the self-healing capacity the steps I to V, shown in Fig. 1 were applied. Hereby each test was performed with n = 16 specimens. (I) Samples were prepared according to the cement preparation procedure and dried after preparation. (II) Cement bars were sputtered with gold, which is bio-inert, to ensures that a mineralization is directed to freshly formed defects. (III) Subsequently, the specimens were damaged in a controlled manner by measuring the bending stress up to an elongation of 1% in both directions A and B one after the other by turning the sample. 1% strain was chosen, since the strain at maximum bending strength is in this range and a reduction of strength upon second loading without healing is measurable. (IV) Samples were stored for 7 days in either SBF or demineralized water at 37 °C. SBF solutions with ionic concentrations almost equal to the inorganic part of human blood plasma were prepared according to Müller et al.38 Hereby the immersion volume was 200 mL. (V) After rinsing with demineralized water and drying, the bending stress up to a strain of 1% was measured in both directions in the same order as previously. The degree of self-healing SHD is defined as the sum of strengths in both directions A and B after storage σbA* and σbB*, divided by the sum of strengths before healing σbA and σbB, Eq. 4.

$$SHD=\frac{({\sigma }_{bA}^{\ast }+\,{\sigma }_{bB}^{\ast })}{({\sigma }_{bA}+\,{\sigma }_{bB})}\cdot 100{\rm{ \% }}$$
(4)

Healing kinetics were measured mechanically, applying the above described regime for healing periods of 1, 3, 5 and 7 days, respectively. Beyond, the composition of SBF was measured everyday by taking 5 mL from the solution. 5 mL of 20 vol% HNO3 were added and the solution was passed through a syringe filter with a pore size of 0.45 µm. The concentration of calcium and phosphor was quantified using a simultaneous radial ICP-OES spectrometer 725ES (Agilent, Waldbronn, Germany) with a CCD-detector. For these tests the immersion volume was raised to 250 mL without further refilling. On the basis of the results of the kinetic measurements, cyclic healing in SBF was performed using a healing period of 5 days for three cycles by repeating steps III to V. Scanning electron microscopy (SEM, Sigma VP, Carl-Zeiss, Germany) on reinforced and healed CPC was performed after 7 days of (I) setting or (V) healing and subsequent drying.

### Statistical analysis

The self-healing capacity in terms of SHD was analyzed for significant differences between groups by one-way ANOVA using both the comparison with the respective reference in water and the differences between the healed samples. For all tests the level of significance was set at p < 0.05.

## Results

### Cement characteristics and mechanical properties of CPC

In this study, the self-healing of cements reinforced with C-fibers and to which additional NaH2PO4 capsules were optionally added was investigated. Whereas the properties of C-fibers reinforced CPC were already published in a previous study35, the characteristics of the capsules and their influence are reported in the following. The capsules consist of a NaH2PO4 salt, which was coated with the purple stained polymer Degacryl, which is based on PMMA. The capsules are evenly distributed in the cement and have a surface ratio of 7% ± 3% at the fracture surface (Fig. 2a). The size distribution of the irregularly shaped NaH2PO4 capsules displays a broad distribution with a d50 of 40 µm (Fig. 2b).

The X-ray diffraction in Fig. 3 shows, that in all systems a cement matrix consisting of calcium deficient hydroxyapatite (CDHA) is formed in a setting process from the raw powder α-TCP and an aqueous solution. The diffraction patterns show the presence of two phases, Hydroxyapatite (HAp) and α-TCP. The Rietveld analyses presented in Table 2 revealed that for CPC without additives the content of α-TCP is highest at 13%, while the addition of both fibers and NaH2PO4 capsules leads to a reduction of the α-TCP content to 9% and 5%, respectively.

Figure 4a illustrates the resulting morphology of the CPC matrix without further additives. It consists of CDHA crystals, which form a porous network. The initial and final setting time for pure CPC was estimated as ti = 25 min and tf = 105 min and shows upon fiber addition a significant reduction35. This effect was even more pronounced, when PO43- capsules were added in the PO4-Cm-CPC system, where the setting times are reduced to ti = 13 min and tf = 21 min. Beyond, the addition of capsules led to a smaller crystal size (Fig. 4b).

During the mixture, some of the capsules tear apart and release H2PO4-, causing a higher local saturation and consequently an accelerated nucleation. Thus, the crystal size is reduced, also resulting in altered mechanical behavior (Table 3). The incorporation of capsules increased the bending strength by 100% when untreated fibers were used and by 50% when chemically pretreated fibers were used. Besides the increase of strength, all reinforced cements show enhanced damage tolerance in comparison to pure CPC. Hereby, the WOF5 was enhanced from 0.02 kJ·m−2 for pure CPC to 1.9 kJ·m−2 in the case of Cm-CPC35. Upon the addition of PO4-supply, WOF5 remained in the same range with WOF5 = 1.7 kJ·m−2 for both, PO4-C-CPC and PO4-Cm-CPC.

The damage tolerance of the reinforced systems is caused by changes in the cracking behavior (Fig. 5). Whereas for CPC only one single crack with a fatal crack opening propagation is observed, the fiber reinforcement leads to multiple cracking and a crack stabilization with crack openings below 10 µm. Thus, the damage tolerance can be assumed for all fiber reinforced CPC-systems shown in this study.

### Self-healing capacity

The self-healing capacity of cements was proven by the mineralization-induced closure of cracks and a recovery of their mechanical properties, respectively. The comparison of the stress-strain curves for CPCs prior and after immersion shows self-healing for all samples in SBF and for CPCs with capsules also in water. Since the bending tests were applied only up to a strain of 1% and in both directions, the individual measurements are not significant for the mechanical properties of the material, but the relation of strength from first measurement to second measurements reflects the self-healing capability of the CPCs. Exemplarily, stress-strain curves for Cm-CPC at first and second loading stored for 7 days in water and SBF, respectively, are shown in Fig. 6. Hereby, not only the maximum strength, but also the similar slope in the case of samples stored in SBF implicates the self-healing of CPCs. However, while individual maxima are visible at the first load, the stress increases more continuously with strain at the second load.

Figure 7a shows the SEM image of a Cm-CPC after self-healing, reflecting the damaging-healing-damaging process. In composites consisting of a CDHA matrix (light grey) reinforced with chemically treated C-fibers, defects were introduced by straining. These defects healed during exposure to SBF by a mineralization within the crack. The mineralization product can be   distinguished by its darker grey color. In the detailed view of the healed crack a porous plate-like morphology of the precipitate can be observed (Fig. 7b). It can also be seen that the crack propagating from the left splits into two, both of which are mineralized. When the strength was examined again after self-healing, the healed crack is reopened. Contrary to the previous course, this time, however, the crack did not split, but ran exclusively along the upper line. A similar cracking pattern can be observed in the case of PO4-Cm-CPC shown in Fig. 7c,d. For this particular crack, which was formed during the first loading, the closure occurred during immersion in SBF. At the second loading, the crack was reopened, with the crack running through the mineralized crack filling. Figure 7e quantifies the self-healing capacity of C-CPC, PO4-C-CPC, Cm-CPC and PO4-Cm-CPC. In general, both plots show a similar course. However, as mentioned earlier, but not obvious in the boxplots due to the normalization in the formula of the self-healing degree SHD, the absolute strength of the cements strongly differs. Damaged C-CPCs did not recover when stored in water but reached only 60–80% of their original strength (Fig. 7e). On the contrary, when analogous C-CPCs were stored in SBF an average SHD of 102% was observed. A similar increase in SHD was also detected for Cm-CPC (Fig. 7e). Here, the reference in water shows about 63% in the second test of the mechanical properties, while the healing in SBF causes an average restoration of SHD of 106%. The addition of capsules changed the healing behavior particularly in water, where SHD reached 110% and 98% for untreated and modified fibers, respectively (Fig. 7e).

Figure 8 illustrates the healing kinetics and repeatability for Cm-CPC and PO4-Cm-CPC. Measurements of the mechanical recovery in dependence of time revealed the beginning of healing at day one with statistical significance (p < 0.05) (Fig. 8a,b). Hereby, only negligible differences were observed in the average SHD over the time period of 7 days (p > 0.05). Analyses of the amount of precipitate from the media are shown in the Figure 8c and reveal a precipitation over the whole time period. Slight differences between Cm-CPC and PO4-Cm-CPC are noted by the increased concentration of PO43- in solution for PO4-Cm-CPC. Fig. 8e,f demonstrate the repeatability of self-healing for Cm-CPC and PO4-Cm-CPC. Self-healing was proven for Cm-CPC over three cycles without significant efficiency loss (p > 0.05), whereas a significant decrease in SHD from cycle 1 to 2 was observed for PO4-Cm-CPC (p < 0.05), whereas to cycle 3 no further decrease was detectable (p < 0.05).

Figure 9 shows a crack in a Cm-CPC during closure via mineralization. An overview of the crack is given in Fig. 9a, which shows mineralization and crack bridging at several points. In the detail view in Fig. 9b one particular spot is magnified, exposing a sheared α-TCP grain with mineralization at its surface. The mineralization product, similar to the previously shown results, is of porous nature and consists of needles.