Regulating micro-structure and biomineralization of electrospun PVP-based hybridized carbon nanofibers containing bioglass nanoparticles via aging time

Abstract

The hybrids of bioactive glass-ceramic (BG) decorated carbon nanofibers (CNFs) have drawn wide interest as bone repairing materials. Herein, hybridized CNFs were produced from electrospinning the mixture solution of polyvinylpyrrolidone (PVP) and BG sol–gel precursors, followed by preoxidation and carbonization. Choosing 45S-type BG (mol%: 46.10% SiO₂–41.48% CaO–12.42% P₂O₅) as the model, the interaction of BG precursors with PVP and the micro-structural evolution of CNF/BG composites were systematically evaluated in relation to aging times (1–7 days) of BG precursor solution. With aging time prolonging, BG precursors underwent morphological changes from small sol clusters with a loosely and randomly branched structure at 1 day, to a fully developed Si-network structure at 3 days, and finally to a dense Si-network at 7 days. On one hand, it showed continuous increase in network linking degree. On the other hand, the gel particles underwent the process of size increase and subsequently decrease. This directly influenced the miscibility between BG precursors and PVP in solution, and the surface morphology of CNF/BG composites. At short aging times, the sol–gel solution of BG precursors mixed uniformly with PVP and the resulting BG nanoparticles were less likely to migrate toward the fiber surface. With the aging time prolonging, the phase separation between BG precursors and PVP facilitated more BG nanoparticles to form on the fiber surface. Calcium ions were found to be able to interact with carbonyl groups in PVP, while phosphorus element would be lost gradually depending on aging time, which led the CaSiO₃ formed in the final CNF/BG changing from a weak to strong crystal state along with longer aging time. By soaking in simulated body fluid, it was found that the CNF/BG composites prepared from BG precursor sol–gel solution with 7 day aging demonstrated the fastest apatite deposition, which was ascribed to those abundant BG nanoparticles on the fiber surface having exposed numerous nucleation sites for the apatite deposition. Promisingly, CNF/BG composites developed from electrospinning and carbonization of PVP/BG sol–gel mixtures were envisioned good choices for bone repairing.

---

1. Introduction

Bioactive glass-ceramics (BG) are promising materials for both hard and soft tissue regeneration due to their ability in enhancing cell proliferation and differentiation, as well as in binding to living tissues spontaneously.^1–3 As scaffolding materials, the processing techniques of porous BG substrates usually include the dry method, colloidal processing, direct casting methods, tape casting, solid freeform fabrication, and so on.^4–6 These porous structures are different from the nanofibrous network of natural extracellular matrixes (ECM).⁷ BG nanofibers were reported in recent years by using the sol–gel/electrospinning technique.^8,9 However, the nanofibrous BG was too brittle to hold a three dimensional scaffolding form.

For hard tissue repairing, carbon nanofibers (CNFs), which had high mechanical strength, have demonstrated favorable advantages in selectively promoting the proliferation and osteogenic differentiation of osteoblasts and bone marrow stromal cells (BMSCs).^10,11 The enhancements are thought mainly benefiting from their conductive and magnetic properties, which resemble the electrophysiological features of natural bone tissue.^12,13 On the other hand, the CNFs can mimic the nanofibrous network of natural ECM, which also contributes positively to biological behaviors.^14,15 Thus, the marriage of BG and CNFs will deliver promising hybridized materials for bone repairing.

Electrospinning technique offers opportunities to incorporate ceramic nanoparticles into CNFs.^16,17 In our previous works,^18–23 BG or beta-tricalcium phosphate (beta-TCP) have been successfully incorporated into CNFs by electrospinning the mixtures of polyacrylonitrile (PAN) and sol–gel solutions of BG or beta-TCP precursors, followed by preoxidation and carbonization. With the increase of surface roughness and the release of silicon and calcium ions, these hybridized CNFs not only accelerated the growth of hydroxyapatite (HA) from simulate body fluids (SBF), but also promoted the proliferation and osteogenic differentiation of osteoblasts or BMCSs in comparison with pure CNFs. The bioactivity and osteocompatibility of these hybridized CNFs, however, relied heavily on preparation parameters such as the sol–gel process of ceramic precursors in addition to carbonization temperature and time, because the micro-structure and the morphology of resulting hybridized CNFs changed accordingly with the parameters. For instance, Hosseini et al.²⁴ reported the aging time and temperature of precursor solution would significantly affect the content of calcium phosphate impurity phases in preparing the sol–gel-derived HA. Our previous study²⁵ also demonstrated that aging time had significant effect on micro-structural evolution of BG sol–gels and subsequent crystallization characteristics of BG component. Ma et al.²⁶ studied the crystallization induced by the sintering process. The results revealed that crystallization occurred at sintering temperature above 900 °C, and glass-ceramics with pseudowollastonite and wollastonite formed. With the further increase in sintering temperature, pseudowollastonite would like to transform into wollastonite.

PAN is a common and excellent starting material in producing electrospun CNFs,^27,28 and the resulting CNFs demonstrated high-yield, high strength and semicrystalline structure. Reasonably, that was why PAN was the first choice in preparing hybridized CNF/BG in our previous works.^18–23 However, this brings some uncertainties in obtaining reproducible CNF/BG. Coalescence and migration of BG nanoparticles were observed in CNF/BG during carbonization. The procedure was significantly affected by carbonization temperature and time, during which, the micro-structural transformation of CNFs resulting from denitrogenation and dehygrogenation played an important role in pushing BG nanoparticles toward fiber surface.²⁰ Due to the compact structure of resulting CNFs derived from PAN, parts of BG nanoparticles were entrapped inside the fibers, which would influence the bioactivity of CNF/BG because those entrapped nanoparticles could not contact with SBF or cell culture medium well.

Polyvinylpyrrolidone (PVP) is another choice in producing CNFs via the electrospinning method. Compared to PAN, its carbon residue is lower, therefore, the resulting CNFs display soft amorphous carbon state and larger surface areas.²⁹ It is speculated that the coalescence and migration of BG nanoparticles in PVP-based CNFs would be easier, and thus has more opportunity of being exposed to biological media to improve the bioactivity of CNF/BG. To this end, different from our previous works,^18–23,25 PVP was chosen as the new carbon source in the present study. And additionally, type 45S BG, composed of 46.10 mol% of SiO₂, 41.48 mol% of CaO and 12.42 mol% of P₂O₅, was set as the model BG component because of its well-known high bioactivity.³⁰ The hybridized CNF/BG was prepared via electrospinning mixtures of PVP and BG precursor sol–gel solution, followed by carbonization. The micro-structural evolution and biomineralization behavior of resulting CNF/BG composites were systematically evaluated by modulating the aging times of BG precursor solution. Comprehensive characterizations including sol–gel particle size and chemical structure in BG precursor solution with aging time, morphology and crystalline structure of resulting CNF/BG, especially the interaction between BG sol–gel solutions with PVP, were performed on purpose to correlate the state of BG sol–gel with the micro-structural evolution of PVP-based CNF/BG. Finally, the ability of apatite deposition from SBF was compared between CNF/BG with different morphological and crystalline features to evaluate their possible application in bone repairing.

2. Experimental

2.1. Materials

PVP (M_w = 1 300 000 g mol⁻¹) was purchased from Aladdin Industrial Co. (China). BG precursors including triethyl phosphate (TEP), calcium nitrate tetrahydrate (CN) and tetraethoxysilane (TEOS) were purchased from Aldrich (USA) and used directly. DMF (99.5%) was bought from Tianjin Fine Chemical Co. (China). Salts for preparation of SBF and other reagents required for the experiments were of analytically pure grade and obtained from Beijing Fine Chemical Co. (China).

2.2. Preparation of CNF/BG composites

The experimental illustration was performed similarly to the process in our previous reports.^18–23 Briefly, 0.966 mL as-hydrolyzed TEP and 0.962 g CN were dissolved together. Then, 1.002 mL TEOS was added into the clear solution to obtain 45S BG precursor solution with the molar ratio of Ca/P set as 1.67. The obtained sol–gel precursor solution was stirred continuously at 30 °C in a water bath for different aging times (1, 3 or 7 days). Afterwards, the as-prepared sol–gel solution was mixed into PVP solution by dissolving 0.982 g PVP in 12.305 mL DMF. The homogeneous solutions were then electrospun by using an electrostatic spinning apparatus with a rolling rod as the collector. The rod rotation speed was set at 800–900 rpm. Other electrospinning parameters were set as: 15 kV (voltage), 0.5 mL h⁻¹ (flow rate) and 15 cm (receiving distance). After that, the obtained membrane was stabilized at 270 °C for 0.5 h in air, and carbonized at 1000 °C for 3 h in N₂ atmosphere to obtain the final CNF/BG composites. For clarity, CNF/BG composites produced from the precursor solution aged for 1 day, 3 days and 7 days were named in terms of CNF/BG-1d, CNF/BG-3d and CNF/BG-7d, accordingly. For comparison, PAN-based CNF/BG composites with precursor aging time of 3 days were prepared under the same parameters (1000 °C/3 h/N₂).

2.3. In vitro test for bioactivity in SBF

Biomineralization in vitro was usually applied to evaluate bone binding ability of orthopedic substrates. Therefore, CNF/BG composites were soaked in SBF and the apatite depositions on CNF/BG composites were characterized. Referring to ISO standard 23317:2007, the SBF was prepared by precisely weighing and dissolving the reagents as follows: NaCl (11.9925 g), NaHCO₃ (0.5295 g), KCl (0.3360 g), K₂HPO₄·3H₂O (0.3420 g), MgCl₂·6H₂O (0.4575 g), CaCl₂ (0.4168 g), Na₂SO₄ (0.1065 g). The solution pH was adjusted to 7.4 at 37 ± 0.2 °C with the buffer solution (0.05 mol L⁻¹) composed of tris(hydroxymethyl)aminomethane (Tris) and hydrochloric acid. CNF/BG composite sheets were cut into rectangular pieces (2 cm × 4 cm) and immersed in the SBF at 37 ± 0.2 °C for 1–7 days. At each predetermined time point, samples were retrieved and gently rinsed with distilled water, and then dried at 80 °C for 24 h.

2.4. Characterizations

Sol–gel transformation of BG precursor solution was monitored by laser particle size analyzer (LPSA, 2000, UK) and nuclear magnetic resonance analysis (NMR, Bruker, Switzerland). Morphology of CNF/BG composite was observed by scanning electron microscope (SEM, S-250, UK) at accelerating voltages of 15–20 kV and transmission electron microscope (TEM, JEOL 2000 EX, Japan) at an accelerating voltage of 200 kV. Before SEM observation, the sample surface was coated with a thin layer of gold alloy (E5600, Polaron, USA). TEM imaging was done by amplitude and phase contrast, and images were acquired using a Gatan Orius SC600 high resolution camera. For TEM observation, the CNF/BG composites were dispersed in ethanol with the help of ultrasonication (60 W/1 min) and mounted onto 800-mesh copper grids. Mapping of BG elements (Si, Ca and P) was performed under the same parameters of TEM observation with an exposure time of 180 s. Crystalline micro-structures were investigated by high resolution TEM (HR-TEM, Hitachi H-800, Japan) at an operating voltage of 200 kV and an X-ray diffractometer (XRD, Rigaku D/max 2500 VB2+/PC, Japan) operating at 40 kV and 200 mA. The practical contents of BG components in final CNF/BG composites were determined by thermogravimetric analysis (TGA, Q500, TA, USA) at a heating rate of 10 °C min⁻¹ from room temperature to 700 °C in air atmosphere. The chemical composition of BG component in CNF/BG was determined by energy dispersive spectroscopy (EDS). Element analysis was conducted on X-ray photoelectron spectroscope (XPS, ESCALAB 250, Thermo Scientific) using monochromatised AlK X-ray source at a constant analyzer. Chemical structure was analyzed using Fourier transform infra-red spectroscope (FTIR, Nexus 670, Nicolet, USA).

3. Results and discussion

3.1. Aging of BG precursor solution

It has been well known that TEOS will hydrolyze when the system contains water.³¹ The reactions can be the formation of Si–OH followed by condensation to form Si–O–Si. Theoretically, the BG precursors will transform into highly crosslinked Si-network structure after critical aging time due to the continuous condensation. ²⁹Si NMR analysis is a powerful tool for investigating structural changes of Qⁿ on silica-based glasses quantitatively.³² Qⁿ (n ranging from 0 to 4) stands for silicon atoms in silicate network that are connected to other silicon atoms by n bridging oxygen (oxygen atoms covalently bond to two silicon atoms Si–O–Si) bonds, which leads to the structure of Si(OSi)_n(OR)_4−n.³³ Fig. 1(a) shows the ²⁹Si NMR spectra of 45S precursor solution after being aged for 1, 3 and 7 days. The chemical shifts around 109, 101 and 92 ppm were assigned to Q⁴, Q³ and Q², respectively.³⁴ Using the deconvolution procedure (Fig. 1(b)), the integral area percentages of different structural units were calculated and shown in Fig. 1(c). With aging proceeding, Q⁴ species increased along with the decreases in both Q³ and Q² species, suggesting the gradual formation of three-dimensional cross-linked Si-network structure in the BG precursor sol–gel solution.

Fig. 1 Sol–gel transformation in BG precursor solutions with aging time (1, 3, 7 days) characterized by (a–c) ²⁹Si NMR spectra and (d) particle size analysis: (a) as-obtained, (b) after deconvolution, (c) calculated characteristic parameters of different Si signals from (b), and (d) particle size and size distribution of precursor sol–gel particles.

To analyze the evolution of sol–gel particle size, sol–gel solutions obtained at selected aging time points were dispersed in ethanol and evaluated by LPSA. The size distribution of each sample was provided in Fig. 1(d), and the statically averaged particle size was illustrated above each curve. The size of precursor particles here was the dynamic dimension, which could be considered as the measurement of the movement territory of particles in the precursor solution.³⁵ As shown in Fig. 1(d), the particle size reached the maximum value at the aging time of 3 days, and subsequently decreased with longer aging time. These variations were supposed relating to the morphological and structural evolution of sol–gel precursor particles with aging times. Initially, the silicate monomers was hydrolyzed and condensed to form oligomers. Subsequently, these oligomers were linked together to make up well-developed silicate network. It explained the increase in particle size from 1 to 3 days of aging. As those inner silicon hydroxyls condensed with each other under longer aging time, liquid would be expelled out from the highly crosslinked Si-network structure, leading to the decrease in particle size from 3 to 7 days of aging.

3.2. Morphological evolution of CNF/BG composites

Fig. 2(a₁–a₃) show the morphological evolution of CNF/BG with aging times by SEM. Spherical nanoparticles with average size of 150 nm appeared on the surfaces of all resulting CNF/BG composites. Their amounts on fiber surface increased apparently with aging time prolonging. Referring to our previous studies,^20–22 these nanoparticles were suggested BG component. In TEM images (Fig. 2(b₁–b₃)), it was confirmed that the fibers held numerous nanoparticles on surface depending on aging times. However, there was no hint of nanoparticles existing inside the fibers in all cases. It was interesting to find that the morphology of PVP-based CNF/BG was different from that of PAN-based CNF/BG^18–23 (Fig. 2(a₄ and b₄)). The fiber diameter of PVP-based CNF/BG was apparently larger than that of PAN-based CNF/BG. The reason was the PVP used in the study having the M_w of 1 300 000, which was higher than that of PAN (M_w = 100 000). The BG nanoparticles were much larger in the former case (150 nm) than in the latter case (∼21 nm). Besides, the BG nanoparticles had come out of the fibers in PVP-based CNF/BG, while those were mainly inlaid on fiber surfaces or inside the fibers in PAN-based CNF/BG, even they were carbonized under the same parameters (1000 °C/3 h/N₂).

Fig. 2 (a₁–a₄) SEM images and (b₁–b₄) TEM images of PVP based CNF/BG-1d, CNF/BG-3d, CNF/BG-7d and PAN based CNF/BG-3d. The insets in SEM images are the average size and size distribution of BG nanoparticles.

During the thermal treatment, BG precursors transformed into BG nanoparticles and some nanoparticles were gradually expelled from the CNFs toward fiber surface by thermodynamical driving force and densification of CNF matrix as described in our previous work.²⁰ The CNFs resulting from PAN are usually made up of dense and rigid semicrystalline carbon species, which might prevent fast diffusion of BG nanoparticles.³⁶ However, PVP is an amorphous polymer and possesses high T_g due to the presence of the rigid pyrrolidone group. Under carbonization, PVP likely transformed into soft amorphous carbon due to the depolymerization of linear polymeric chains.³⁶ The structure of soft amorphous carbon was not so dense that the migration of BG nanoparticles toward fiber surface would be easier and faster, therefore, the expelled BG nanoparticles were also in larger size than those in PAN-based CNF/BG.

The phenomenon that more BG nanoparticles were detected on fiber surface with longer aging time, could be explained from two aspects, i.e. the sol–gel state and its compatibility with PVP. When the aging time was only 1 day, the BG precursor sol–gel particles were small and the Si network was barely formed. Then the BG precursor oligomers could disperse well between PVP macromolecules, and which entrapped quite a few of BG component inside resulting CNFs after the carbonization. With the aging time increasing, the silicate network developed, which would worsen the compatibility between BG and PVP. As the result, more BG nanoparticles had migrated toward fiber surface by the thermodynamical driving force. The well-developed dense silicate network at further longer aging time, undoubtedly, would lead much more BG nanoparticles to fiber surface during carbonization.

Although the CNF/BG composites displayed different amounts of BG nanoparticles on fiber surfaces, the loading amounts of BG components in all CNF/BG composites were measured to be around 75 wt% by TGA analysis (Fig. 3) due to the same fabrication process, and the values were consistent to the initial feeding ratios. It suggested that abundant BG component was still mixed with carbon in fiber form. After those nanoparticles on fiber surface were being stripped off with high power ultrasonication (120 W/10 min), the smooth fibers were conducted element mapping to show the presence of Si and Ca elements distributing uniformly along CNF (Fig. 4). The intensities of Si and Ca elements decreased with longer aging time, which were negatively correlated to the increasing amounts of BG nanoparticles on fiber surfaces.

Fig. 3 TGA curves of CNF/BG-1d, CNF/BG-3d and CNF/BG-7d composites.

Fig. 4 Element (Si, Ca and P) mapping results of CNF/BG-1d, CNF/BG-3d and CNF/BG-7d after BG nanoparticles on the surface being stripped off under high power ultrasonication.

To consider the data in both Fig. 3 and 4 together, it was easy to deduce that there were still abundant BG component in the CNFs, which distributed homogeneously along the fibers. It was quite different from those entrapped BG nanoparticles in PAN-based CNF/BG composite (Fig. 2(a₄ and b₄)). Thus, the BG precursor sol–gel should have some interaction with PVP in the electrospinning solution to prevent significant phase separation between them. FTIR and XPS were reported as the effective techniques to detect the possible interaction between BG and PVP.³⁷ As the FTIR spectrum shown in Fig. 5(a₁), the electrospun pure PVP sample showed the absorption bands of carbonyl stretching vibration (1660 cm⁻¹) and C–N vibration (1291 cm⁻¹).^36,38 While in the spectrum of as-electrospun PVP/BG precursor composites, the absorption band of carbonyl group in PVP shifted to 1646 cm⁻¹. In literature,³⁹ a similar shift in carbonyl absorption band was reported for 2-pyrrolidone when alkali metal ions were added. The authors ascribed the shift to the interaction between the carbonyl oxygen of 2-pyrrolidone and the metal ions. Also, shift in carbonyl absorption band in FTIR study was reported for systems containing both PVP and metal cations.³⁷ Then, it was suggested that the band shift of carbonyl group in the case of as-electrospun PVP/BG precursor composites resulted from the interaction of Ca²⁺ ions with the carbonyl groups. After oxidation, the release of pyrrolidone rings from PVP backbone originated the new absorption band at 1702 cm⁻¹, which was assigned to the stretching vibration of carbonyl in detached pyrrolidone ring (Fig. 5(b₁)).⁴⁰ While the FTIR spectrum of electrospun PVP/BG almost remained unchanged after the oxidation, indicating the strong interaction between Ca²⁺ ions and carbonyl groups able to stabilize pyrrolidone rings from being cut off during oxidation. In corresponding XPS spectra (Fig. 5(a₂ and b₂)), the binding energies of O1s in as-electrospun and oxidized PVP samples were peaked at ∼530.5 eV and 531.3 eV, respectively. While in the presence of BG precursors, the binding energies of O1s shifted to higher values (533.1 eV and 531.9 eV) for as-electrospun and oxidized PVP/BG samples. The increased binding energy of O1s also confirmed the interaction of Ca²⁺ ions with carbonyl groups in PVP. Thus, this interaction between BG and PVP helped to maintain the homogeneous hybridized morphology of CNF/BG.

Fig. 5 (a₁ and b₁) FTIR spectra and (a₂ and b₂) high resolution XPS spectra of O1s peak for PVP (black) and CNF/BG-7d (red) after (left column) electrospinning and (right column) oxidation.

3.3. Structural evolution of CNF/BG composites

From Fig. 4, noticeably, the P element in resulting CNF/BG was only able to be detected when the aging time of BG precursor sol–gel solution was 1 day. For other two aging time points, the P signal was too weak to be detected. EDS analysis was then applied to determine the chemical compositions of those nanoparticles on fiber surface. As shown in Fig. 6(a₂–c₂), P element also gradually decreased with aging time prolonging. It was wondered that this phenomenon was unique in preparing CNF/BG, or common in preparing BG-based materials by using sol–gel technique. To make this point clear, BG precursor sol–gel solutions from different aging times were similarly sintered (thermally treated at 1000 °C for 3 h) and analyzed with EDS. From Fig. 6(a₁–c₁), however, the atomic percentages of P element in the samples only decreased slightly with longer aging time. The difference between the change of P element in CNF/BG and in pure BG made us realize that the electrospinning operation might be an unnegligible factor to cause the loss of P element. During the process of electrospinning, high voltage (usually above 20 kV) was applied on the ejected liquid drop to elongate it into nano-scaled fibers, during which, some incompletely hydrolyzed TEP was liable to be vaporized along with the solvent evaporation.⁴¹

Fig. 6 EDS spectra of (left column) pure BG aged for (a₁) 1, (b₁) 3 and (c₁) 7 days without electrospinning and (right column) BG nanoparticles on (a₂) CNF/BG-1d, (b₂) CNF/BG-3d and (c₂) CNF/BG-7d after carbonization.

While for phenomenon that the loss in P element displayed dependence on aging time, it was thought relating to the micro-structural evolution of BG. P element is of high volatility at high temperature (200–300 °C).⁴² Thus, more P element would be lost during the following preoxidation (270 °C) and carbonization (1000 °C) treatment for electrospun PVP/BG fibers. If the silicate network could provide a kind of stabilizing or shielding effect on PO₄³⁻ ions, the loss of P element during the heating treatment would be reduced. At short aging time, the silicate network was loosely packed, the entrance of Ca²⁺ and its counter ion PO₄³⁻ into the network was relatively easy. When the silicate network got denser with longer aging time, the resistance for Ca²⁺ and PO₄³⁻ ions into the silicate network increased. Thus the former case could provide stronger stabilizing effect to avoid the volatilization of P element upon heating than the latter case.

To characterize the chemical and crystalline structure of final products, these CNF/BG composites were conducted analysis with FTIR and XRD. After the carbonization, the CNFs resulting from electrospun pure PVP nanofibers displayed only a broad weak being ascribed to (0 0 2) reflection of graphitic carbon in XRD pattern (Fig. 7(a)), and a wide peak at 1174 cm⁻¹ being assigned to C–O–C vibration in FTIR spectrum (Fig. 7(b)). These facts revealed that the PVP-based CNFs were predominantly composed of amorphous carbon with some oxygen residue. Characteristic peaks at 2θ = 26°, 27.6°, 31.8°, 36.6° and 45.8° were identified in all the CNF/BG samples, which matched well with those of CaSiO₃ (JCPDS 01-0720) in JADE software. The diffraction intensities became stronger with longer aging time, suggesting the crystallinity of BG component being enhanced. FTIR spectra shown in Fig. 7(b) confirmed the presence of Si–O–Si groups in all CNF/BG composites by bands of 1000 cm⁻¹ (asymmetric stretching of Si–O–Si) and 460–480 cm⁻¹ (bending or rocking of Si–O–Si).⁴³ The two peaks at 720 cm⁻¹ and 940 cm⁻¹, according to literature data,⁴⁴ were ascribed to the presence of pseudowollastonite structure. Those BG nanoparticles on fiber surface were further checked with HR-TEM. As shown in Fig. 7(c₁–c₃), clear lattice fringes of d-spacing of 0.248 nm were identified for all the samples, which were assigned to single crystallized calcium silicate corresponding to the peaks at 36.6° with the d-spacing of 0.245 nm in JADE software. In comparison, the lattice fringes in longer aged samples became more apparent, suggesting the higher crystallinity. With all these facts shown in Fig. 7, it was inferred that the BG component in CNF/BG was in a form of crystalline calcium silicate with possible domains of pseudowollastonite structure, and the crystallinity increased if the aging time of BG precursor sol–gel solution was longer.

Fig. 7 (a) XRD patterns and (b) FTIR spectra of CNF, CNF/BG-1d, CNF/BG-3d and CNF/BG-7d, and (c₁–c₃) TEM images of BG nanoparticles on various CNF/BG.

Based on all the aforementioned results, the mechanisms of morphological and micro-structural evolution of BG components in PVP-based CNF/BG composites were proposed, and illustrated in Fig. 8 to show their dependence on the heat treatment and aging time. When the BG precursor solution was mixed with PVP/DMF solution, the formed Si-network particles dispersed in the solution between PVP macromolecules. Ca²⁺ ions bonded to electronegative carbonyl groups due to their cationic feature, while PO₄³⁻ ions were supposed mainly as counter ions around Ca²⁺ ions. When this solution was submitted to electrospinning, the resulting PVP/BG precursor composite fibers would present relatively homogeneous distribution of silicate gel particles, Ca²⁺ and PO₄³⁻ ions along with PVP macromolecular chains. At this stage, few of Ca²⁺ ions had entered into the silicate network. Upon oxidation, theoretically, pyrrolidone rings were cut off from the PVP backbone and conjugated C=C bonds would form along the polymer chain. However, due to the stabilization of bonded Ca²⁺ ions, the detachment of pyrrolidone ring was not so significant in the case of electrospun PVP/BG fibers at the oxidation temperature (270 °C). In the following carbonization, PVP transformed into soft amorphous CNFs as the result of polymer depolymerization and cyclization.^29,36 At the same time, the debonded Ca²⁺ ions migrated into the silicate network with some PO₄³⁻ ions.⁴⁵ Accordingly, CaSiO₃ formed and some of them were driven to fiber surface because of the thermodynamical driving force. During this whole procedure, apparently, the miscibility between the PVP solution and BG precursor sol–gel solution decreased as the aging time of the sol–gel solution increasing due to the denser silicate network forming at longer aging time. Therefore, numbers of CaSiO₃ particles found on fiber surface increased in the order of CNF/BG-1d, followed by CNF/BG-3d and CNF/BG-7d. The loss of P element became more significant when the silicate network got denser, because fewer PO₄³⁻ ions had entered into the network and those free PO₄³⁻ ions were liable to volatilize under high temperature. As the facts and analysis shown, aging time of BG precursor solution could be applied to adjust the morphology and the composition of final CNF/BG composites, which would be practical in developing new bone repairing materials.

Fig. 8 Schematic illustration of morphological and micro-structural evolution of BG components in PVP-based CNF/BG composite with aging time during each heating step. The graphs in left and right columns show the evolution in chemical structure of PVP and micro-structure of BG component, respectively. The middle column shows the suggested interaction between BG and PVP during each step.

3.4. Apatite formation on CNF/BG composites

Biomineralization is a common method for evaluating bone binding ability of orthopedic materials, which is achieved by soaking materials in SBF and subsequently monitoring the apatite formation. Fig. 9 shows SEM images of different CNF/BG composites having been soaked in SBF for various times. Previously, the ability of CNF/BG composite membrane to maintain integrated forms was ever identified by being folded for several times.²² From Fig. 9, noticeably, it could be confirmed that the fibers remained continuous form without visible breakage during the soaking, which indicated the CNF/BG composite membrane had enough strength to serve as the substrate for biomineralization testing. Within 1 day, the surface of CNF/BG-1d became smooth with no visible BG nanoparticles in comparison with the original composite fibers (Fig. 9(a₁)). On the surfaces of CNF/BG-3d and CNF/BG-7d, the spherical BG nanoparticles had been replaced by deformed coverings (Fig. 9(b₁ and c₁)). As the biomineralization proceeding to the third day, fibers in all the samples could be seen smooth but with randomly deposited mineralites (Fig. 9(a₂–c₂)). The depositions were more on CNF/BG-7d than on the other two substrates. With the continuous mineralite deposition at longer soaking time, the morphology of the three biomineralized CNF/BG substrates became similar (Fig. 9(a₃–c₃)). Fibers were fully covered by mineralites after being soaked in SBF for 7 days (Fig. 9(a₄–c₄)). A distinct difference between the samples was found that the size of spherical depositions on fibers increased in the order of CNF/BG-1d < CNF/BG-3d < CNF/BG-7d, indicating faster growth of mineralites on CNF/BG composites prepared from BG precursor sol–gel solution with longer aging time. The spherical mineralites had displayed the typical flaky-like structure of HA on CNF/BG-7d after 7 days of biomineralization. Reasonably, it could be inferred that it was the dissolution of BG nanoparticles inducing the mineralite deposition from SBF. In comparison with CNF/BG-1d and CNF/BG-3d, CNF/BG-7d could induce the fastest apatite deposition and growth because it had the maximum BG nanoparticles on fiber surface, which were exposed to SBF to provide the most nucleation sites for mineralite with the dissolution of BG nanoparticles.

Fig. 9 SEM images obtained after different CNF/BG composites (from left to right) being soaked in SBF for (a₁–c₁) 1 day, (a₂–c₂) 3 days, (a₃–c₃) 5 days and (a₄–c₄) 7 days (from up to down) at 37 °C. The insets are the corresponding high-magnification SEM images.

The biomineralized CNF/BG composites were further characterized by conducting XRD and FTIR analysis. As shown in Fig. 10(a₁–c₁), the mineralite depositions on all CNF/BG composites at all time points, except the case of CNF/BG-1d for 1 day of SBF soaking, were identified crystalline HA as the two characteristic diffraction peaks at 2θ = 26° and 32° shown.⁴⁶ They were assigned to (0 0 2) and (2 1 1) reflections of crystalline HA. In the case of CNF/BG-1d for 1 day of SBF soaking, the small diffraction peak at 2θ = 28.5° was identified as the crystal plane of (0 4 1) of dicalcium phosphate dehydrate (DCPD), which was a thermodynamically instable intermediate usually forming at the early stage of biomineralization from SBF.²⁹ Compared to CNF/BG-3d and CNF/BG-7d, the initial mineralite deposition on CNF/BG-1d could be seen slower because it had less BG nanoparticles on fiber surface, while the growth of crystalline HA was quite rapid after the nucleation. From FTIR spectra (Fig. 10(a₂–c₂)), a peak at 1450 cm⁻¹ was found and attributed to the presence of CO₃²⁻,⁴⁷ in addition to those absorption bands (558, 605, 884 and 1038 cm⁻¹) assigned to PO₄³⁻ or HPO₄²⁻ groups.⁴⁸ In combining the results of XRD, FTIR and SEM together, it revealed that the formed apatite on CNF/BG composites was a kind of carbonated HA, which was the common product from SBF soaking.⁴⁹

Fig. 10 (a₁–c₁) XRD patterns and (a₂–c₂) FTIR spectra obtained for (a₁ and a₂) CNF/B-1d, (b₁ and b₂) CNF/BG-3d and (c₁ and c₂) CNF/BG-7d after soaking in SBF at 37 °C for various time points.

In summary, the CNF/BG composites were able to induce heterogeneous nucleation and growth of apatite from SBF, indicating the substrates having potentials to bind bone tissue. However, the nucleation depended on the morphology of CNF/BG composites. If more BG nanoparticles were exposed to SBF, faster nucleation and the following faster growth in crystalline HA were obtained. The whole process was basically relating to the preparation of CNF/BG composites, in which, aging time of BG precursor sol–gel solution was identified a critical factor influencing morphology and micro-structure of resulting CNF/BG composites.

4. Conclusions

PVP, which can form soft amorphous carbon after carbonization, was chosen as the starting material to produce hybridized CNF/BG in this study. It was identified that the BG precursors had some interaction with PVP via calcium ions bonding to carbonyl groups, and this interaction helped to maintain the hybrid structure during electrospinning and pre-oxidation. When BG precursor sol–gel solutions with different aging times were used for the electrospinning, loss of P element was found and ascribed to its volatilization with solvent evaporation during electrospinning and the following heating treatments. Due to the worsened compatibility between well-developed dense Si-networks and PVP, as well as the thermodynamical driving force under carbonization, more BG nanoparticles had been expelled from resulting amorphous CNFs and grew on fiber surface as the aging time of BG precursor sol–gel solution increasing. This made more BG nanoparticles be exposed to SBF, which induced faster apatite deposition. Benefiting from this feature, accordingly, the PVP-based CNF/BG could be a potential material for bone repairing.

Acknowledgements

The authors are very pleased to acknowledge financial support from National Basic Research Program of China (2012CB933904), National Natural Science Foundation of China (51373016 and 51473016), the Research Fund for the Doctoral Program of Higher Education (No. 20110010120014), and the Fundamental Research Funds for the Central Universities (No. ZY1106).

Notes and references

1. R. Ma, S. C. Tang, H. L. Tan, J. Qian, W. T. Lin, Y. G. Wang, C. S. Liu, J. Wei and T. T. Tang, ACS Appl. Mater. Interfaces, 2014, 6, 12214–12225.
2. A. J. Salinas, P. Esbrit and M. Vallet-Regí, Biomater. Sci., 2013, 1, 40–51.
3. O. D. Schneider, F. Weber, T. J. Brunner, S. Loher, M. Ehrbar, P. R. Schmidlin and W. J. Stark, Acta Biomater., 2009, 5, 1775–1784.
4. C. Galassi, J. Eur. Ceram. Soc., 2006, 26, 2951–2958.
5. R. B. Zhang, B. Y. Han, D. N. Fang and Y. S. Wang, Ceram. Int., 2015, 41, 11517–11522.
6. Q. Q. Guo, H. M. Xiang, X. Sun, X. H. Wang and Y. C. Zhou, J. Eur. Ceram. Soc., 2015, 35, 3411–3418.
7. P. Premnath, B. Tan and K. Venkatakrishnan, BioRes. Open Access, 2012, 5, 231–238.
8. A. M. Deliormanlı, Ceram. Int., 2015, 41, 417–425.
9. Y. Y. Lia, B. B. Lia, G. Xua, Z. S. Ahmad, Z. H. Ren, Y. Dong, X. Lia, W. J. Wang and G. R. Han, Colloids Surf., B, 2014, 122, 785–791.
10. R. L. Price, M. C. Waid, K. M. Haberstroh and T. J. Webster, Biomaterials, 2003, 24, 1877–1887.
11. Y. Z. Wan, G. F. Zuo, F. Yu, Y. Huang, K. J. Ren and H. L. Luo, Surf. Coat. Technol., 2011, 205, 2938–2946.
12. P. A. Tran, L. J. Zhang and T. J. Webster, Adv. Drug Delivery Rev., 2009, 61, 1097–1114.
13. I. Rajzer, R. Kwiatkowski, W. Piekarczyk, W. Biniaś and J. Janicki, Mater. Sci. Eng., C, 2012, 32, 2562–2569.
14. X. M. Li, R. R. Cui, W. Liu, L. W. Sun, B. Yu, Y. B. Fan, Q. L. Feng, F. Z. Cui and F. Watari, J. Nanomater., 2013, 2013, 1–16.
15. H. Q. Cao, T. Liu and S. Y. Chew, Adv. Drug Delivery Rev., 2009, 61, 1055–1064.
16. N. Bhardwaj and S. C. Kundu, Biotechnol. Adv., 2010, 28, 325–347.
17. D. Li and Y. Xi, Adv. Mater., 2004, 16, 1151–1170.
18. H. Y. Liu, Q. Cai, P. F. Lian, Z. Fang, S. Duan, X. P. Yang, X. L. Deng and S. Ryu, Mater. Lett., 2010, 64, 725–728.
19. H. Y. Liu, Q. Cai, P. F. Lian, Z. Fang, S. Duan, S. Ryu, X. P. Yang and X. L. Deng, Carbon, 2010, 48, 2266–2272.
20. X. L. Jia, T. H. Tang, D. Cheng, L. J. Guo, C. H. Zhang, Q. Cai and X. P. Yang, RSC Adv., 2014, 4, 64299–64309.
21. B. Han, X. H. Zheng, H. Y. Liu, X. L. Deng, Q. Cai, X. L. Jia, X. P. Yang, Y. Wei and G. Li, J. Biomater. Sci., Polym. Ed., 2014, 25, 341–353.
22. Q. Yang, G. Sui, Y. Z. Shi, S. Duan, J. Q. Bao, Q. Cai and X. P. Yang, Carbon, 2013, 56, 288–295.
23. C. H. Zhang, D. Cheng, T. H. Tang, X. L. Jia, Q. Cai and X. P. Yang, J. Mater. Chem. B, 2015, 3, 5300–5309.
24. H. E. Hosseini, M. R. Housaindokht and M. Chahkandi, Mater. Chem. Phys., 2007, 106, 310–316.
25. X. L. Jia, T. H. Tang, D. Cheng, C. H. Zhang, R. Zhang, Q. Cai and X. P. Yang, Colloids Surf., B, 2015, 136, 585–593.
26. J. Ma, C. Z. Chen, D. G. Wang, X. G. Meng and J. Z. Shi, Ceram. Int., 2010, 36, 1911–1916.
27. K. L. Elias, R. L. Price and T. J. Webster, Biomaterials, 2002, 23, 3279–3287.
28. S. Kobayashi and W. Kawai, Composites, Part A, 2007, 38, 114–123.
29. P. Q. Wang, D. Zhang, F. Y. Ma, Y. Qu, Q. N. Chen, S. H. Xie and J. Y. Li, Nanoscale, 2012, 4, 7199–7204.
30. L. L. Hench, J. Biomed. Mater. Res., 1998, 15, 511–518.
31. Y. Peng, N. W. Turner and D. W. Britt, Colloids Surf., B, 2014, 119, 6–13.
32. C. Vaid, S. Murugavel, C. Das and S. Asokan, Microporous Mesoporous Mater., 2014, 186, 46–56.
33. F. He, H. L. Zhao, X. H. Qu, C. J. Zhang and W. H. Qiu, J. Mater. Process. Technol., 2009, 209, 1621–1626.
34. E. Pomakhina, D. Deneele, A. C. Gaillot, M. Paris and G. Ouvrard, Cem. Concr. Res., 2012, 42, 626–632.
35. C. S. Johnson and D. A. Gabriel, Laser light scattering, Dover Publications, New York, 1994.
36. J. W. Shin, W. H. Ryu, K. S. Park and I. D. Kim, ACS Nano, 2013, 7, 7330–7341.
37. M. H. Liu, X. P. Yan, H. F. Liu and W. Y. Yu, React. Funct. Polym., 2000, 44, 55–64.
38. P. Barmpalexis, I. Koutsidis, E. Karavas, D. Louka, S. A. Papadimitriou and D. N. Bikiaris, Eur. J. Pharm. Biopharm., 2013, 85, 1219–1231.
39. J. L. Wuepper and A. I. Popov, J. Am. Chem. Soc., 1969, 91, 4352–4356.
40. K. Nakanishi, Infrared Absorption Spectroscopy, Nankido, Tokyo, 1962.
41. S. Zhang, Y. Q. Huang, X. P. Yang, F. Mei, Q. Ma, G. Q. Chen, S. K. Ryu and X. L. Deng, J. Biomed. Mater. Res., Part A, 2009, 90, 671–679.
42. M. Anastasescu, M. Gartner, A. Ghita, L. Predoana, L. Todan, M. Zaharescu, C. Vasiliu, C. Grigorescu and C. Negrila, J. Sol-Gel Sci. Technol., 2006, 40, 325–333.
43. A. Meiszterics, L. Rosta, H. Peterlik, J. Rohonczy, S. Kubuki, P. Henits and K. Sinko, J. Phys. Chem. A, 2010, 114, 10403–10411.
44. M. Rokita, W. Mozgawa and A. Adamczyk, J. Mol. Struct., 2014, 1070, 125–130.
45. S. Lin, C. Ionescu, K. J. Pike, M. E. Smith and J. R. Jones, J. Mater. Chem., 2009, 22, 1276–1282.
46. H. Wang, S. C. Zhao, J. Zhou, Y. Q. Shen, W. H. Huang, C. Q. Zhang, M. N. Rahamanc and D. P. Wang, J. Mater. Chem. B, 2014, 2, 8547–8557.
47. R. Govindan, G. S. Kumar and E. K. Girij, RSC Adv., 2015, 5, 60188–60198.
48. M. P. Marlin, M. Sylvain, M. C. Jean, P. Elodie and C. Jérôme, J. Eur. Ceram. Soc., 2012, 32, 2765–2775.
49. S. Y. Choi, S. Coonrod, L. Estroff and C. Fischbach, Acta Biomater., 2015, 24, 333–342.

---