Microstructure, stability and biocompatibility of hydroxyapatite – titania nanocomposites formed by two step sintering process

2.1 Synthesis of HA-based composites

Commercial HA (<200 nm; 99.99%; Sigma-Aldrich) calcined at 900 °C in air for one hour (HA900) and commercial TiH₂ (TH) (100–150 μm; water atomized; >98%; Merck) powders, were mixed in HA900:TH = 3:1 (wt.%) ratio to prepare low porosity TiO₂-HA composites. Unilateral cold compaction at 150 MPa was performed for the HA900 and TH powder mixture (HAT0) to obtain cylindrical pellets of 10 mm diameter and 4–6 mm in height. Nano-sized HA-based composites were obtained by two-step sintering (TSS) method (Gingu et al., 2011). During the first step of TSS method, the sample is heated up to T₁ of 900 °C for τ1 = 1–5 min to reach a critical relative bulk density of about 70% and to avoid unwanted HA900 dehydroxylation which begins at T_onset = 900 °C (Sofronia et al., 2014). In the second step, the HAT0 sample is rapidly cooled down and held at a lower temperature (T₂ = 800 °C) for two different times (τ2), 5 (HAT5) and respectively 10 h (HAT10).

2.2 Structural characterization

Powder X-ray diffraction (XRD) patterns were recorded using a Panalytical XPERT-PRO diffractometer with Bragg-Brentano configuration, operating with a PIXcel detector using Cu Kα radiation (wavelength of 1.5406 Å). Data were acquired in the range 15–80° (2θ) using a step size of 0.0262° (2θ) at room temperature. The sample was rotated on a monocrystalline silicon support in order to reduce preferential orientation and increase the quality of the pattern. Quantitative phase analysis was performed with Quanto program (Altomare et al., 2001) by Rietveld method. XRD analysis was carried out to identify different phases present in the starting powders. To measure the instrumental line broadening a known silicon powder standard was used. The mean crystallite size (d) was calculated using Scherrer equation:

2.3 Surface and morphology characterization

Specific surface area (S, m² g⁻¹) of the samples was measured by the Brunauer-Emmett-Teller (BET) nitrogen adsorption method in a Nova 2200e Quantachrome surface area analyzer. Samples were dried and degassed at 150 °C over the night and analyzed using a multipoint N₂ adsorption/desorption method.

The sample morphology was examined by scanning electron microscopy (SEM) using a high-resolution FEI Quanta 3D FEG (Dual Beam) microscope. Microstructural observations on polished surfaces of the sintered powder compacts (chemically etched by immersion in 2 vol.% HF solution for 2 min) were made to study the metal oxide-hydroxyapatite-interconnect diffusion couples and the degree of sintering.

2.4 Spectroscopic analysis

2.5 Thermal analysis

For the thermal characterization a simultaneous high temperature thermogravimetry (TG) coupled with differential scanning calorimetry (DSC) technique was employed. Thermal properties (enthalpy and temperature of transformations and mass change) of the HA-based samples were measured by a TG-DSC Setaram Setsys Evolution 17 analyzer, in the temperature range from 40 to 1400 °C, with a scanning rate of 10 °C/min in alumina crucibles under Ar flow. Sample mass for simultaneous TG-DSC measurements was about 20 mg. The thermoanalyzer calibration was previously reported (Marinescu et al., 2011). Error of TG measurement is ±0.154%.

Linear shrinkage was measured using a thermomechanical analyzer (TMA) Setaram Setsys Evolution 17, from 40 to 1400 °C at a heating rate of 5 °C/min in an Ar flow. Load is 0.049 N. For each analyzed sample a blank curve obtained under the same conditions as those employed to test the samples was used for TMA signal treatment. Average coefficient of thermal expansion (CTE) was calculated.

2.6 Tribological characterization

The wear tests have been performed on the TRB 01-02541 tribometer (CSM Instruments SA Switzerland) with a linear reciprocating module. The balls were made of DIN 100Cr6 tool steel, 6 mm diameter, Ra < 3.2 μm; HRc 60–64; density > 7.6 g/cm³. The disk, as a static counterpiece, is the studied biocomposite. The friction parameters are sliding linear velocity = 1 cm/s up to 50 cm/s; RT = 23 °C, room humidity = 30 percent; the normal load is 2 N. The important parameters determined by the tribometer are friction coefficient, μ, using balloon-disk friction couple, and wear rate. The wear rate of bioceramics was determined from the following relation Wr = V/(P_NL), where V is the worn volume, P_N is normal load and L is sliding distance. The worn track cross section of samples was measured from cross-sectional profiles of the wear track.

2.7 Cell viability assay. Cell morphology

Cell viability was determined by the MTT assay (Craciunescu et al., 2014) using mouse fibroblast cell line NCTC, clone L929. Briefly, 2.5 × 104 cells were seeded in 24-well culture plates, in Minimum Essential Medium (MEM) supplemented with 10% FBS and 1% antibiotics (penicillin, streptomycin and neomycin). After 24 h of treatment, the cells were incubated with 0.25 mg/mL MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) for 3 h at 37 °C. Then, the medium was removed and isopropyl alcohol was added to each well for formazan crystal solubilization. Optical density was measured at 570 nm using a microplate reader (Mithras LB 940, Berthold Technologies). Cell viability was expressed as a percentage of control cells considered to be 100% viable. Cell morphology was analyzed by Giemsa staining. The cells were fixed with cold methanol, stained with Giemsa for 15 min and observed on an inverted Zeiss Axio Observer D1 microscope (Carl-Zeiss, Germany).

3.1 Characterization of the samples

The XRD Patterns, SEM micrographs, pore size distribution and adsorption–desorption isotherms of the HA900, HAT5, HAT10 samples and reference starting HAT0 powder are shown in Figs. 1–3, respectively and Tables 1 and 2. XRD patterns of the compacted HAT5 and HAT10 powders showed typical patterns for hexagonal HA, rutile TiO₂ and orthorhombic perovskite CaTiO₃ (Fig. 1, Table 1).

Close

Figure 1

X-ray diffraction patterns of the HA900, HAT5, HAT10 samples and reference starting HAT0 powder.

Export to PPT

Close

Figure 2

(a) Pore size distribution and (b) adsorption–desorption isotherms for HAT5, HAT10 samples and reference starting HAT0 powder.

Export to PPT

Close

Figure 3

SEM imagines of the (a) TH, (b) HA900, (c) reference starting HAT0, (d) HAT5 and (e) HAT10 samples.

Export to PPT

HA900 characterization was already reported (Sofronia et al., 2014) and is discussed here for comparison. Technical specifications of commercial TiH₂ indicated a composition of TiH_1.9 (Gingu et al., 2011). Both compacted powders present a majority pore size distribution within the mesopores domain (Fig. 2, Table 2). Sintering for a longer time (10 h) leads to a decrease in the maximum domain of the pore size distribution and an increase in the distribution in the microporous domain.

Fig. 3(a)–(e) shows the morphology of the investigated samples obtained by SEM. As-received titanium hydride powder is highly inhomogeneous (Fig. 3a) presenting small and large aggregates (size values ranging from 170 nm to 3 µm). HAT0 microstructure (Fig. 3c) comprises both HA900 bidispersive round-shaped particles (Fig. 3b), with size values between 59 and 531 nm (Sofronia et al., 2014) and TH angular-shaped particles (size values ranging from 170 nm to 3 µm). Small cracks are observed on the HAT5 composite surface (Fig. 3d), while few regular pores (red circles) can be clearly observed for the HAT10 sample (Fig. 3e).

The increase in adsorption for HAT5 sample (Fig. 2b) is in agreement with cracks observed in SEM imagines. It is interesting to note that the titanium atoms replacing phosphorus atoms having different radius, with the TiO₂ formation in composite, lead to an abnormally crystal lattice and therefore, preclude the grain growth when sintered. Consequently, a small grain size of the composite is obtained.

3.2 Spectroscopic characterization

Functional groups of the HAT0 mixture heated up to different temperatures in the DSC analyzer and of HA900 sample for comparison have been investigated by FT-IR as shown in Fig. 4a.

Close

Figure 4

(a) FTIR spectra of HA900 and HAT0 powder heated at different temperatures; (b) Raman spectra of the HAT0, HAT5 and HAT10 samples.

Export to PPT

The characteristic PO₄³⁻ absorption bands of HA at 564, 570, 578, 589, 602, 962, 1032 and 1090 cm⁻¹ were observed in FTIR spectrum of HAT0 sample (Fig. 4a) with a broad band at 632 cm⁻¹ attributed to OH⁻ groups (Sofronia et al., 2014). The band assigned to CO₃²⁻ groups was observed in HAT0 and HAT0 heated at 700 °C spectra at 876 cm⁻¹ and 874 cm⁻¹, respectively.

Raman spectroscopy is a sensitive technique in investigation of HA, as well as phosphate and hydroxyl substitution by carbonate ions in the so called B- and A-substituted HA (Antonakos et al., 2007). Raman spectra of the three samples HAT0, HAT5 and HAT10 along with the TH are illustrated in Fig. 4b. Results of the band deconvolution with six-band model over 800–1200 cm⁻¹ and 3400–3700 cm⁻¹ domains are listed in Table 3.

Heterogeneous structure of the mechanical homogenized HAT0, e.g. the mixture of the commercial hydroxyapatite and titanium hydrate powders, is proved by the two types of spectra named HAT_1 and HAT_2 corresponding to hydroxyapatite and rutile, respectively (Fig. 4b). The latter one is formed by local heating in air under laser action. Thus, all the Raman modes of the PO₄³⁻, i.e. symmetric P—O stretch (ν₁ at 959 cm⁻¹), symmetric bending, ν₂, within 400–450 cm⁻¹, asymmetric stretching modes, ν₃, over 1028–1076 cm⁻¹ and asymmetric bending, ν₄, between 579 and 610 cm⁻¹ (Antonakos et al., 2007) are present in the HAT_1 spectrum. Lack of 1103 cm⁻¹ band confirms the absence of the A-type apatite (Antonakos et al., 2007) when CO₃²⁻ ions substitute OH⁻. Given the fact that uncarbonated HA shows the ν₁(CO₃²⁻) and ν₁(PO₄³⁻) bands with similar full width at half maximum (FWHM) (Zhou and Lee, 2011), bigger FWHM for the ν₁(CO₃²⁻) band (11 cm⁻¹ in comparison with 7.14 cm⁻¹ for the ν₁(PO₄³⁻) band in Table 3) indicates the presence of B-substituted apatite in the HAT0 sample. Typically, crystallinity and/or nanosize of the apatite materials is given by the full width at half maximum (FWHM) of the main phosphate band ν₁ (Antonakos et al., 2007). Carbonate content of an apatite material is derived from the area ratio of the ν₁(CO₃²⁻) and ν₁(PO₄³⁻) bands (Awonusi et al., 2007), e.g. area ratio of 1069 cm⁻¹ (A₁) and 959 cm⁻¹ (A₂) bands in the case of the HAT0 sample (Table 3). However, ν₁(CO₃²⁻) modes are overlapping with the triply degenerated ν₃(PO₄³⁻) modes (Tsiourvas et al., 2011). Also, the ν₁(PO₄³⁻) mode for biological apatite in bone, dentin and enamel was also reported (Pasteris et al., 2004) to be located at 959 cm⁻¹. Degree of hydroxylation and hence ionic order in the apatite unit cell is also estimated by the area ratio of ν(OH⁻) and ν₁(PO₄³⁻) bands (Pasteris et al., 2004) (area ratio of the 3568 (A₃) and 959 cm⁻¹ bands for the HAT_1 sample).

The second type of spectrum for HAT0 sample, HAT_2, is dominated by TiO₂ as rutile (Porto et al., 1967). Other phases present in the HAT_2 spectrum are amorphous calcium phosphate (ACP) and/or nanostructurated HA (Dorozhkin, 2009; Sadat-Shojai et al., 2013) by its wide band at 955 cm⁻¹. In accordance with the Raman modes reported by Porto et al. (1967) for rutile, e.g. 143 cm⁻¹ (B_1g), 235, 447 (E_g), 612 (A_1g) and 826 cm⁻¹ (B_2g), all the bands of the HAT_2 spectrum are downshifted very likely due to the reduced size of the rutile particles obtained by TH laser heating.

Low frequency Raman spectra for sintered HAT5 and HAT10 samples (see Fig. 4b) point out the presence of CaTiO₃ (Moreira et al., 2009) while ν₂(PO₄³⁻), ν₄(PO₄³⁻) and Ca₃—OH (Dorozhkin, 2009) (at about 318 cm⁻¹ in case of the HAT_1 spectrum in Fig. 4b) are highly masked by the strong Raman scattering of rutile. Literature data (Hirata et al., 1996; Moreira et al., 2009) reported Ca vibrations in Ca—TiO₃ (134 cm⁻¹), O—Ti—O bending modes (181, 224, 244, 287 and 339 cm⁻¹), Ti—O₆ torsional (464 and 495 cm⁻¹) and symmetric stretching Ti—O mode at 669 cm⁻¹ as Raman active modes for the orthorhombic CaTiO₃. Narrower ν₁(PO₄³⁻) band for HAT5 and HAT10 samples, i.e. a consequence of increased crystallite size and thus increased crystal perfection (Pasteris et al., 2004), is accompanied by the enhancement of the hydroxylation degree (Table 3).

3.3 Reaction mechanism between HA and TiH₂ at heating

The specific roles as well as mutual relation between the TH and HA900 at sintering are investigated by TG-DSC (Fig. 5a–c, Table 4) in order to study the reaction mechanism of the TiO₂-HA formation, mechanism which has been scarcely studied. TH decomposition heat of 73.3 kJ mol⁻¹ obtained in this work is within the values reported in the literature, 67 kJ mol⁻¹ (Fromm and Hörz, 1980) and 136 kJ mol⁻¹ (Herbst, 2002).

Close

Figure 5

(a) DSC; (b) TG curves of the HA900, TH, HAT0 and HAT5 and HAT10 samples; (c) DSC, TG and dTG curves for HAT5 and HAT10; (d) TMA curves for HAT5 and HAT10.

Export to PPT

Progress of the thermal effects of HAT0 at heating in Ar atmosphere further described (TG-DSC) has been correlated with FTIR analysis at different steps of HAT0 heating (Fig. 4). The HA effect on hydride decomposition is the increase in the onset temperature (T_onset = 485 °C, Fig. 5a) at which gas expansion occurs and maximum rate of hydride decomposition is shifted to a higher temperature, 608 °C comparing to 575 °C for TH (inset Fig. 5a) as well. The first stage of mass variation (Δm₁, Table 4) corresponds to the evolution of adsorbed water on HA surface (physically bound water) without any contribution from TH. Starting with the second stage of mass variation, thermal effects due to both components of the composite are overlapped. Consequently, water desorption in the ΔT₂ temperature range takes place along with Ti oxidation in the presence of the water physically desorbed from HA. Comparatively, for TH sample alone, Ti oxidation in the presence of the oxygen incorporated in Ti network is observed (Fig. 5b). Dominant effect in the ΔT₃ range is given by TH decomposition; besides it, lattice water evolution and carbonate removal from the hydroxyapatite take place (Sofronia et al., 2014). The Ti oxidation process predominates in the ΔT₄ temperature range. The blunt and smooth peak at 876 cm⁻¹ characteristic to CO₃²⁻ group is still observed in the FTIR spectrum of the sample heated at 700 °C, meaning that decarbonation is still ongoing; decarbonation is overlapping with Ti oxidation up to 900 °C as proved by the missing 876 cm⁻¹ band from FTIR spectrum. Above 900 °C, HAT0 sample presents the widening of the bands in the 550–650 cm⁻¹ range and also a slope between 650 and 800 cm⁻¹ that can be assigned to overlapping of PO₄³⁻ groups with those of Ti—O bond bands of TiO₂ formed by oxidation (Huang et al., 2010; Illekova et al., 2011). Titanium oxidation (Eq. (2)) is evidenced by the ascending trend of DSC curve with a broad exothermic effect (with a maximum at 1135 °C):

The presence of Ti in the composite leads to a decrease in the band intensity at 962 cm⁻¹ corresponding to PO₄³⁻ group (Ribeiro et al., 2006), (Fig. 4a). At the same time with TiO₂ formation at increasing temperature, a redistribution of the PO₄³⁻ stretching (950–1150 cm⁻¹) and vibration (550–620 cm⁻¹) band intensities can be found slightly shifted regarding the HA900 bands alone (Carrodeguas and De Aza, 2011) (Fig. 4a). This indicates that titanium atoms did enter into the structure of HA and replaced phosphorus atoms, which resulted in the change of HA crystal lattices. The last stage of mass loss (ΔT₅) corresponds to HA decomposition into phosphates with a pronounced endothermic effect on the DSC curve. HA decomposition is ongoing at 1100 °C, the bands corresponding to HA900 being split in FTIR spectrum, the spectrum recorded at 1280 °C is specific to β-TCP, while at 1400 °C, α-TCP spectrum (Carrodeguas and De Aza, 2011) is identified (Fig 4a).

The above-detailed analysis of the thermal behavior of HAT0 mixture led to obtaining the reaction path of the biocomposites formation. Phase compositions of the composites showed the presence of HA, rutile (TiO₂) and CaTiO₃ as determined by XRD, and typical results are shown in Fig. 1. Formation of a small quantity of CaTiO₃ at HAT0 sintering indicates that the reaction between HA and TiO₂ formed during the thermal treatment takes place as follows:

However, the XRD pattern of both composites sintered at 800 °C showed no diffraction peaks characteristic of β-TCP, due to the further reaction of β-TCP (Tanaka et al., 2000):

However, bone minerals are essentially calcium deficient hydroxyapatite (Ca-dHA) with Ca/P ratio of about 1.5, which is structurally similar to stoichiometric HA (Ca/P = 1.67) (Siddharthan et al., 2005).

3.4 Sintering time effect on TiO₂ – HA composites. diffusion mechanism between HA and Ti

The higher porosity of HA900 compared to HAT10 indicates the important role of titania particles in lowering surface roughness and densifying the composite. Also, the reduction in surface activity by halving of the surface area (Table 2) decreases the extent of self-agglomeration of HA particles, thus promoting the achievement of composites with homogeneous microstructure. This is relevant when designing materials for load-bearing applications which need improved mechanical properties (Farnoush et al., 2012; Sprio et al., 2013).

SEM images of the etched samples (Fig. 6) show that the appropriate holding time increase in the second sintering stage leads to homogenization of the particle size and pore distribution. Therefore, pore size is raging between 350 nm and 1.2 µm and 200–600 nm for HAT5 and HAT10 respectively. However, individual particles of the components were still visible after sintering, more obvious for HAT5 sample, meaning that CaTiO₃ formed in small amounts influences a little the contact area between the two main phases (ceramic/metal oxide). However, some initial bridges between the particles (sintering necks) occurred for HAT10. In this case, interrupting the sintering process in an early phase provides a way of achieving a well-defined and homogenous porosity and also avoids HA decomposition into TCP.

Close

Figure 6

SEM micrographs of HAT5 and HAT10 composites after being sintered, polished, and etched in HF 2% for 2 min.

Export to PPT

The sum of open and closed porosity is referred to as the total porosity. Total porosity obtained by BET analysis of HAT5 is lower than HAT10 (Table 2); after etching the samples (Fig. 6) it can be observed that HAT10 sample is more porous than HAT5. It means that HAT10 composite contains a high fraction of closed pores giving rise to denser ceramics suitable for biomedical implants able to bear high loads.

The comparison among HAT5 and HAT10 indicates that a predominant mass loss for HAT10 composite (75% of total mass loss) takes place in the 900–1400 °C temperature range and it corresponds to HA dehydroxylation, with a maximum rate at 988 °C (dTG curve, Fig 5c). On the other hand, HAT5 sample looses 55% of total mass up to 900 °C. This behavior relies on different carbonation degrees versus sintering time of HAT5 and HAT10. Thus, increasing sintering time to 10 h causes an enhancement of the hydroxylation degree on the expense of carbonate content (Table 3). Conversely, despite less hydroxylated HAT5, faster dehydroxylation process and hence enhanced HA decomposition are recorded in this case (Fig. 5c). The displacements of the ceramics presented in Fig. 5d support this behavior, HAT5 sample being less stable above ∼800 °C. The CTE obtained in the stability temperature range is 10.68 · 10⁻⁶ C⁻¹ for HAT5 (40–800 °C) and 11.46 · 10⁻⁶ C⁻¹ for HAT10 (40–1000 °C) proving an extended thermal stability of the HAT10 composite, which is useful in the application of HA-based ceramics. CTE is an important factor in assessing the capacity of a ceramic to resist fracture during clinical use of a restoration (Lopes et al., 2009). Decomposition of HA alters physical, chemical and mechanical properties and hence performances of implant material in a living body (Huang et al., 2010). Restraining HA decomposition is therefore an important outcome for both scientific and biomedical purposes.

The purpose of the present work was the evaluation of TiH₂ - HA interactions as extensively as possible in the TSS sintering conditions for a better understanding of the obtained microstructural and mechanical features. In order to assess optimum TSS parameters and considering the main application of these biocomposites, namely engineered tissue implants, the interface between metallic/ceramic implant and biocomposite graft needs to be studied from the wear behavior point of view. In our study, the metallic/ceramic implant is represented by the metallic/sapphire ball of the bal-on-disk friction couple and counterpiece, respectively the disc, is represented by the biocomposite samples (HAT5 and HAT10). In the case of a τ2: T2 ratio lower than that corresponding to the kinetic window, sintering did not occur for the HAT5 sample. The compacted powder processed for τ2 = 600 min (HAT10) led to a sintered biocomposite (Table 5) with a relative density of 59%.

A titanium oxide layer is formed on the particle surface during sintering which delays gas release and therefore it was important to find the optimal time for sintering as one sufficiently long for a total hydride decomposition and, at the same time, for full densification without cracks. During sintering, Ti atoms move and interdiffuse on the surface while oxygen atoms migrate toward HAT0 mixture bulk. The result of these simultaneous processes is titanium oxide in either amorphous or crystalline forms (Arifin et al., 2014). The presence of Ca, P and Ti atoms demonstrated that interdiffusion occurred in the HAT10 composite leading to a homogenous microstructure (Fig. 7). In this case, a composite of stable microstructure with matching mechanical and thermophysical characteristics of the constitutive phases was obtained.

Close

Figure 7

Electron image (SEM) and correlated XR maps (EDS) from the HAT10 composite.

Export to PPT

It seems then possible to conclude that use of a non-conventional sinterization process at high temperatures for short spans as TSS method could be suitable for obtaining homogenous small grain bodies of TiO₂-HA. By avoiding HA decomposition with TCP formation, using TSS sintering method is advantageous due to the fact that TCP presence may result in excessively fast resorption in vivo which may be detrimental to the establishment and maintenance of a proper bone regeneration process. One quantitative assay (MTT assay) (Fig. 8a) and one qualitative assay (Giemsa staining) (Fig. 8b–e) were used to investigate the cytotoxicity of HA-based composites.

Close

Figure 8

(a) The effect of treatment with HA-based samples on cell viability of L929 fibroblast cells assessed by MTT. Light micrographs of L929 fibroblasts cells (b) untreated and treated cells with 50 mg/mL of (c) HA900, (d) HAT5 and (e) HAT10 for 24 h using Giemsa staining (10×).

Export to PPT

Our data showed that the percentage of cell viability was above 80% (non-cytotoxic) for all HA-based composites and all tested concentrations (12.5, 25 and 50 mg/mL) after 24 h of treatment (Fig. 8a). At the highest tested concentration (50 mg/mL), viabilities of the cells treated with HA and HAT10 were 102.9% and 97.1% respectively, whereas for HAT5 composite cell viability was 85.5%. Furthermore, Giemsa staining evidenced that cells treated with HA-based composites maintained their normal fibroblastic phenotype, similar to the untreated cells (Fig. 8b–e). Overall, cell viability results correlated with morphological observations indicated a good biocompatibility of HA-based composites at all tested concentrations.