Ultra-porous titanium oxide scaffold with high compressive strength
- First Online:
- 1.3k Downloads
Highly porous and well interconnected titanium dioxide (TiO2) scaffolds with compressive strength above 2.5 MPa were fabricated without compromising the desired pore architectural characteristics, such as high porosity, appropriate pore size, surface-to-volume ratio, and interconnectivity. Processing parameters and pore architectural characteristics were investigated in order to identify the key processing steps and morphological properties that contributed to the enhanced strength of the scaffolds. Cleaning of the TiO2 raw powder removed phosphates but introduced sodium into the powder, which was suggested to decrease the slurry stability. Strong correlation was found between compressive strength and both replication times and solid content in the ceramic slurry. Increase in the solid content resulted in more favourable sponge loading, which was achieved due to the more suitable rheological properties of the ceramic slurry. Repeated replication process induced only negligible changes in the pore architectural parameters indicating a reduced flaw size in the scaffold struts. The fabricated TiO2 scaffolds show great promise as load-bearing bone scaffolds for applications where moderate mechanical support is required.
Bone tissue engineering offers an alternative strategy for restoring damaged, diseased, and resorbed bone tissue. In this approach, natural bone tissue is generated from osteogenic cells to replace the missing tissue with the aid of a porous three-dimensional scaffold [1, 2]. For the repair and regeneration of bone tissue, the 3D scaffold structure must provide bone forming cells a substrate to attach to and allow the newly formed tissue adequate space for growth . Hence, highly porous scaffolds with high surface-to-volume ratio are necessary for bone tissue engineering applications [4, 5, 6]. A large surface area-to-volume ratio assists cell attachment and growth and is therefore necessary in order to achieve a high cell density within the scaffold, whereas a large interconnected pore volume is required in order to allow bone cells to migrate into and subsequently proliferate within the scaffold interior [7, 8, 9, 10].
Together with the physical properties of the scaffold material, the pore architecture also determines the mechanical properties of the 3D scaffold structure. A bone scaffold construct should match as closely as possible the mechanical properties of natural bone and provide sufficient support for the repair site in a load-bearing environment [11, 12, 13]. Because increased porosity and pore size result in reduced mechanical properties and, consequently, diminish the structural integrity of the scaffold, there is a practical upper limit for the porosity and the pore size that can be tolerated [6, 14].
Titanium dioxide (TiO2) is a biocompatible material, which has also been reported to have bioactive properties [15, 16]. Ceramic TiO2 is therefore a promising material for scaffolds for tissue engineering purposes. Fabrication of highly porous TiO2 scaffolds with a well interconnected pore network and the ability to promote fibroblast and osteoblast cell adhesion onto the entire scaffold surface has previously been reported [17, 18]. However, these scaffolds have limited mechanical properties and may thus have inadequate strength to serve as scaffolds in applications where good mechanical support is required. Furthermore, differences have been found in the cytotoxicity of different TiO2 raw powders that can be used in the fabrication of the scaffolds . Based on the results presented in the aforementioned study, Kronos TiO2 powder was selected as the most appropriate raw powder for the scaffold production.
The aim of the present study was to fabricate highly biocompatible ceramic TiO2 scaffolds with improved mechanical properties suitable for bone tissue engineering applications. The objective was to produce scaffolds with compressive strength values above 2 MPa, which can be considered as the lower limit for the compressive strength of trabecular bone, and also to identify the key manufacturing steps that govern the strength of such scaffolds.
2 Materials and methods
2.1 TiO2 raw powder
Commercial TiO2 powder (Kronos 1171, Kronos Titan GmbH, Leverkusen, Germany) was used for the preparation of the ceramic slurry. The TiO2 powder was cleaned in 1 M NaOH solution prior to use. The cleaning was performed by soaking the TiO2 powder in 1 M NaOH and subsequently rinsing it in deionised H2O. The water was drained out and the powder was dried at 70°C for at least 24 h prior to sieving (Analysette 3, Fritsch GmbH, Idar-Oberstein, Germany). After sieving, powder with particle sizes between 50 and 100 μm was collected and used for the preparation of the ceramic slurry.
2.2 Scaffold fabrication
The ceramic slurry for scaffold fabrication was prepared by gradually adding 60 or 65 g of TiO2 powder to 25 ml of sterilised water and the pH of the slurry was kept at 1.7 for the entire duration of the stirring with small additions of 1 M HCl. For double-coated scaffolds, the slurry for the second coating contained 40 g of TiO2 powder in 25 ml sterilised water. The TiO2 powder was stirred in to the water in several steps at low rotation speed of 1,000 rpm (Dispermat Ca-40, VMA-Getzmann GmbH, Reichshof, Germany). The temperature of the slurry was adjusted to room temperature during the low rotation speed stirring. When the slurry mixture was homogenous, the rotation speed was increased to 5,000 rpm and the slurry was stirred for 5 h at this rotation speed. Temperature of the slurry was reduced to 15°C during high speed stirring.
The polyurethane (PU) foam templates (60 ppi, Bulbren S, Eurofoam GmbH, Wiesbaden, Germany) were washed and cut to appropriate size as described in . The cylindrical templates were then immersed in the ceramic slurry, and excess slurry was either squeezed between two polymer foam sheets or centrifuged (1 min @ 1,000 rpm; Biofuge 22R Heraeus Sepatech, Osterode, Germany) out of the foam templates to ensure that only a thin layer of slurry covered uniformly the entire surface area of the polymer template without blocking the pores. The samples were then placed on a porous ceramic plate and allowed to dry at room temperature for at least 16 h before sintering.
For the burnout of the polymer, the scaffolds were slowly heated to 450°C at a heating rate of 0.5 K/min. After 1 h holding time at 450°C, temperature was raised to 1,500°C at a rate of 3 K/min and the sintering time at this temperature was set to 40 h. The sintered scaffolds were then cooled back to room temperature at the cooling rate of 5 K/min (HTC-08/16, Nabertherm GmbH, Bremen, Germany).
2.3 Material characterisation
The chemical surface compositions of both as-received and cleaned Kronos TiO2 raw powders were investigated by the means of X-ray photoelectron spectroscopy (XPS in a KRATOS AXIS ULTRADLD, Kratos Analytical, Manchester, UK) using monochromatic Al Kα radiation (hν = 1,486.6 eV). The XPS data analysis was performed using the CasaXPS (computer-aided surface analysis for X-ray photoelectron spectroscopy) software package (Casa Software Ltd., UK).
The particle size distribution of the TiO2 slurry was determined by the means of laser diffraction (Mastersizer 2000, Malvern Instruments, UK). Samples were collected for particle size measurement at several time points during high speed stirring (5,000 rpm).
The crystal structure of the sintered TiO2 scaffolds was examined by X-ray diffraction (XRD), performed with a Siemens D5000 X-ray diffractometer equipped with a Göbel-mirror providing parallel Cu Kα radiation (Siemens, Munich, Germany). The acquisition was done using grazing incidence configuration.
2.4 Pore morphology
The initial visualisation and optical observation of the microstructure of the prepared scaffolds were performed using a scanning electron microscope (TM-1000, Hitachi High-Technologies, Tokyo, Japan). The scaffolds were mounted on aluminium stubs with conductive carbon tape and viewed with backscattered secondary ions at 15 kV accelerating voltage.
Micro-computed tomography was used to determine non-destructively the three-dimensional microstructure, porosity, and porous interconnectivity of the scaffolds. The samples were mounted vertically on a plastic sample holder and scanned with desktop 1172 micro-CT imaging system (Skyscan, Kontich, Belgium) at 6 μm voxel resolution using a source voltage of 100 kV and a current of 100 μA with 0.5 mm aluminium filter. Samples were rotated 180° around their long axis and three absorption images were recorded every 0.400° of rotation. These raw images of the samples were reconstructed with the standard SkyScan reconstruction software (NRecon) to serial coronal-oriented tomograms using 3D cone beam reconstruction algorithm. For the reconstruction, beam hardening was set to 20% and ring artefact reduction to 12.
The mean pore diameter distribution was found by measuring the material thickness on the inverse model, which was generated by setting the threshold value between 0 and 65. Additional noise was again removed using the ‘despeckling’ function, which removed all objects smaller than 500 voxels and not connected to the 3D body. 3D models of the scaffolds were generated through algorithm adaptive rendering.
2.5 Compressive strength
Compressive mechanical testing (Zwicki, Zwick/Roell, Ulm, Germany) was conducted to evaluate the mechanical strength of the prepared scaffolds. Compression tests were performed in accordance with DIN EN ISO 3386 at room temperature using a load cell of 1.0 kN. The preloading force was set to be 0.5 N. The scaffolds were compressed along their length axis at a compression speed of 100 mm/min until failure. The force and displacement were recorded throughout the compression and converted to stress and strain based on the initial scaffold dimensions.
2.6 Statistical analyses
Different data groups were compared through a one-way analysis of variance (one-way ANOVA) test followed by post hoc tests for pairwise comparisons (Holm–Sidak method). Normality and equal variance tests were performed prior to the testing. Statistical significance was considered at a probability P < 0.05. A correlation study was performed with a bivariate regression analysis, Spearman Rank Order correlation, using the computer software SigmaStat version 3.5 (Symantec, St. Louis, USA). The results were interpreted as follows: small correlation if 0.1 < |r| < 0.3; medium correlation if 0.3 < |r| < 0.5; and strong correlation if 0.5 < |r| < 1 . A negative r indicated a negative correlation, whereas a positive r indicated a positive correlation.
3.1 Chemical surface composition of the TiO2 powder
The chemical surface composition (at.%) of as-received and cleaned Kronos TiO2 powders
Cleaned with 1 M NaOH
3.2 Particle size and stability of the TiO2 slurry
Some slurries were also prepared by dispersing the powder at pH 9–11, but these slurries developed such high viscosities that the slurry became too thick and unmanageable even before all the powder had been added into the dispersion. Normal viscosity was only restored when pH was lowered to values between 1.50–1.75.
3.3 Mechanical and pore architectural properties of the prepared scaffolds
There was also finer-scale porosity present at the foam struts which can be observed in Fig. 2. Triangular voids were found in the interior of some scaffold struts, whereas in most struts one wall of the triangular void left by the polymer template had collapsed, or folded in, thus eliminating the hollow space from the strut interior. Finer-scale porosity was also observed as long cracks and voids, particularly close to the stems of the scaffold struts.
Selected structural parameters of the TiO2 scaffolds derived from the micro-CT data displaying the mean values ± standard deviation (n = 20) for each parameter
89.6 ± 2.1
87.1 ± 2.8
87.6 ± 2.8
7.5 ± 1.6
8.6 ± 2.1
7.6 ± 3.1
Surface area-to-volume ratio
57.9 ± 9.1
53.7 ± 5.4
54.9 ± 4.7
410 ± 24
396 ± 16
392 ± 14
73.5 ± 16.5
70.3 ± 9.8
70.0 ± 8.1
Structure linear density
1.45 ± 0.21
1.85 ± 0.17
1.76 ± 0.21
Degree of anisotropy/strut
1.41 ± 0.18
1.42 ± 0.09
1.40 ± 0.08
Degree of anisotropy/pore
1.28 ± 0.14
1.17 ± 0.04
1.22 ± 0.07
2.14 ± 0.02
2.19 ± 0.01
2.17 ± 0.01
2.13 ± 0.03
2.14 ± 0.02
2.14 ± 0.02
Correlation study between compressive strength and selected pore architectural characteristics and processing parameters of the prepared scaffolds (n = 55, ** P < 0.01, * P < 0.05)
Spearman’s correlation coefficient
Surface area-to-volume ratio
Degree of anisotropy
XRD data revealed that the sintering process used in the fabrication of the TiO2 scaffolds produced TiO2 with rutile crystalline structure (data not shown).
3.4 The effect of processing parameters on scaffold properties
Neither the change in the solid content of the ceramic slurry nor the second replication process was observed to cause considerable differences in the overall microscopic appearance among the scaffolds. Moreover, the micro-CT analysis revealed that the porosity of the scaffolds was not markedly altered by the change in the solid content or the repeated replication process. A statistically significant difference was nevertheless found between the samples having different solid content. However, the increase in the solid content of the slurry was observed to cause only a 2.8% decrease in the average porosity (Table 2).
In previous studies, polymer sponge method has been successfully utilised in the fabrication of highly porous TiO2 bone scaffolds. However, with a maximum compressive strength of 1.2 MPa, these scaffolds lack the necessary mechanical integrity required for scaffolds intended for load-bearing applications [17, 18]. In the present study, the optimal processing parameters described in  were adjusted for the production of TiO2 scaffolds from cleaned Kronos TiO2 raw powder with the aim of producing ceramic foams with improved mechanical strength. Particular attention was paid on identifying the key processing parameters that may contribute to further improvement in the compressive strength of the ceramic foam structure.
4.1 Slurry preparation
Commercial ceramic powders often contain traces of impurities that remain from the powder manufacturing process . Such impurities may interfere with the densification and thus hinder the grain growth and lead to poor consolidation of the green body [23, 24]. Adsorbed impurities on the particle surface may also interfere with the short-range forces acting between the particles, therefore altering their behaviour when dispersed in aqueous solutions . Some impurities, namely phosphorus and potassium, were detected on the surface of the raw powder. Cleaning the Kronos powder with 1 M NaOH solution removed the potassium totally and considerably reduced the amount of phosphorus originally deposited on the powder surface. Moreover, the biocompatibility of sintered TiO2 samples has been found to improve due to the cleaning of the raw powder in a recent study. The cleaned Kronos TiO2 powder also performed better in terms of biocompatibility compared to several TiO2 powders from other manufacturers .
Ceramic slurry with fine particle size and narrow particle size distribution is generally required in order to produce ceramics with small and uniform grain size, which is known to contribute to better strength values [26, 27]. In this study, a lab dissolver was used for the dispersion and milling of the powder, and the milling time was adjusted to 5 h at high rotation speed of 5,000 rpm as this procedure has previously been proven to result in the smallest particle size with a narrow size distribution as well as a uniform dispersion of the particles . However, agglomeration of the small particles may counter the desired effects achieved by the decreased particle size in the ceramic slurry, and therefore the ability to control the stability of the dispersion is crucial for the effective processing of optimal ceramic slurries. Zeta potential is a very important physical parameter that influences the stability of the aqueous ceramic slurry, and its relationship to slurry rheology is well established [28, 29, 30, 31]. Zeta potential may be understood as the electrical potential of the electric double layer of a surface-charged particle and it can be used as the measure of the electrostatic repulsive forces which stabilise the slurry. At low zeta potentials these forces experience a minimum, and thus the maximum in viscosity and shear stress has been found at the vicinity of isoelectric point (iep), the pH of zero zeta potential [29, 32]. Since the stability of the dispersion is the highest and the viscosity the lowest at pH at which the zeta potential deviates the most from the iep, the pH of the dispersion should be adjusted accordingly.
The cleaning procedure has been found to shift the iep of the Kronos TiO2 powder from 1.7 to more neutral pH, resulting in an iep at 4.8, which is closer to the theoretical values reported for TiO2 . It is well known that many water-soluble impurities, such as phosphates, may cause a substantial shift in the iep, particularly in concentrated suspensions [33, 34]. Therefore, it is likely that the observed shift in the iep of Kronos TiO2 powder was caused by the reduction of adsorbed phosphate ions on the particle surfaces due to the cleaning process. However, sodium was detected in the powder in relatively high quantities after the cleaning. Sodium, at relatively strong ionic strengths, has been found to affect the viscosity, particularly in alkaline pH range, and shift the iep toward higher pH values in TiO2 dispersions [35, 36, 37]. When dispersing the TiO2 powder at pH 9–11, the behaviour of the slurry changed abruptly after half of the powder was dispersed and the dispersion became coagulated. However, normal viscosity was restored when the pH was adjusted to 1.7 with additions of acid (HCl). According to Sabetrasekh et al.  the zeta potential of the TiO2 raw powder reaches values below −60 mV at pH above 8, while at pH 1.7 it is only +30 mV. Zeta potential of ±45 mV has been found sufficient to maintain stability in anatase dispersions at low ionic strengths, although using lower solid content . It may be that the amount of sodium ions, which remained from the cleaning procedure, increased above the critical concentration in the dispersion as more powder was added, thus reducing the stability and rheological properties of the slurry.
Solid content also influences the stability of the dispersion. At high solid content, the repulsive forces between the surface-charged particles decrease due to overlapping and compression of the electric double layer as the particle separation distance in the slurry approaches the double layer thickness [33, 38]. Since very high solid contents (~70 wt%) are used in the preparation of the TiO2 slurry, the relatively poor stability may be further reduced as the particles become smaller and the particle separation distance reduces during milling with the lab dissolver, thus resulting in the flocculation of the finer particles in the slurry. Therefore, despite the long stirring time, the reduction in particle size did not prove as effective as previously reported . Instead, the particle size was found to reach a minimum after only 1 h of high speed stirring and further stirring only increased the average particle size to a certain level due to the agglomeration of the smallest particles (Fig. 1). After 6.5 h milling time, the dispersion coagulated and the amount of finer particles in the slurry increased notably. However, these finer particles only became detectable after storing diluted slurry samples for 24 h prior to particle size measurement, indicating the flocculated state of these particles in undiluted slurry. Thus, the stability of the slurry needs to be improved in order to produce slurries with finer particle sizes. This may be achieved by optimising the cleaning procedure: total removal of sodium from the powder with more vigorous rinsing should lead to better stability at pH range 7–9, where zeta potentials below −50 mV may be reached with minimal addition of alkaline solution, such as NaOH. It was also evident that the impeller of the lab dissolver was rather inefficient in milling the powder particles. Replacing the impeller with a better designed rotor blade should, together with improved slurry stability, result in reduced particle size in the TiO2 slurry.
4.2 Scaffold fabrication
One of the key steps in the replication process is the preparation of ceramic slurry with suitable properties for coating the polymer sponge as uniformly as possible. The quality of the ceramic coating on the polymer template is strongly dependent on the rheological properties, such as the viscosity, of the ceramic slurry . As was previously mentioned, solid content and stability of the dispersion are the most important parameters that affect the rheological properties. In the present study, it was observed that even a small change in the solid content resulted in altered sponge loading. As the TiO2-to-H2O ratio was increased from 2.4 to 2.6, the viscosity of the slurry increased notably. From the SEM and micro-CT analyses, it was observed that the slurry with increased solid content resulted in a thicker and more uniform coating on the polymer sponge. The improved compressive strength values of the scaffolds fabricated from the denser slurry seem to verify the notion of the more favourable sponge loading achieved using the slurry with increased TiO2/H2O ratio. With better control of the dispersion stability, slurries with higher solid content may possibly be prepared while maintaining the optimal rheological properties. This may lead to further enhancement in the sponge loading.
Mechanical pressing was used instead of centrifuging to remove the excess TiO2 slurry from the polymer templates. This method was found more suitable due to the relatively high viscosity of the slurry and also resulted in a thicker coating on the foam strut. This explains the notable increase in the strut thickness of the fabricated TiO2 scaffolds in comparison to the average strut thickness reported in an earlier study , which contributed to the improved mechanical strength observed in the present study. Also, since the 60 ppi PU foam was used as the scaffold template instead of 45 ppi PU foam, the resulting scaffolds had a smaller pore size and a higher number of struts per cross-sectional area than those produced by Fostad et al. , which partly explains the considerably higher compressive strength of the fabricated TiO2 scaffolds.
As expected, multiple replications resulted in further increase in the compressive strength. For recoating, thinner slurry was prepared in order to ensure uniform coating on the scaffold struts, and since the mechanical pressing was not applicable in the case of recoating, the excess slurry was centrifuged out of the ceramic preform. The compressive strength of the scaffolds was notably increased due to the second replication process reaching to the values of 1.63–2.67 MPa at 82–90% porosities. In comparison, the compressive strength of trabecular bone is typically 2–12 MPa . Nevertheless, the strut thickness and porosity of the scaffolds remained virtually unchanged in spite of the recoating process. One possible explanation for the observed increase in the compressive strength with no apparent change in the strut thickness may be linked to the folded nature of the scaffold struts. During the recoating process, the thinner slurry may be deposited in the folds of the struts, thus strengthening the strut structure without causing an apparent change in the strut diameter. Wherefore, subsequent recoating processes are anticipated to cause a detectable increase in the strut thickness. The observed decrease in the fractal dimension of the scaffolds struts seems to verify this since the added coating layer in the folds reduces the complexity of the strut structure. The strong correlation between strength and fractal dimension remains positive as a more noticeable increase in fractal dimension was observed due to the increase in the solid content. In addition, the recoating process eliminates large flaws, such as cracks and voids, remaining in the ceramic foam structure after the first replication. As the cracks and voids in the ceramic body have a debilitating effect on the mechanical properties, the removal of these flaws, or even the reduction in the flaw size, is expected to contribute to significantly improved strength values . The recoating was found to have only a negligible effect on the interconnectivity of the scaffold. Nonetheless, repeated recoating processes are expected to cause further reduction in the interconnectivity of the TiO2 scaffolds, particularly at larger minimum connection sizes . This is due to decreased porosity and blocking of pore openings as a result of the increasing strut thickness .
4.3 Pore architecture
In bone tissue engineering, the scaffold must provide an optimal microenvironment for osteogenesis. Therefore, a non-degradable scaffold intended for this purpose should exhibit a highly interconnected pore structure with porosity higher than 70% and an average pore size exceeding 300 μm to ensure viable vascularisation and bone formation through the entire scaffold [42, 43]. The fabricated TiO2 scaffolds had average pore sizes close to 400 μm and porosities between 76 and 94%. Therefore, these scaffolds are expected to allow sufficient tissue ingrowth and vascularisation for bone regeneration. Moreover, the remarkably high interconnectivity (>92%) through connections as large as 200 μm is more than likely to allow for excellent permeability, and therefore, to guarantee sufficient transport of nutrients and metabolism products for viable osteogenesis within the scaffold interior. Interconnectivity values above 95% have previously been reported for various bone scaffolds at similar porosities with the fabricated TiO2 scaffolds [20, 44, 45]. However, as the minimum connection size was increased to 200 μm, the accessible pore volume generally reduced more drastically than was seen in the case of the TiO2 scaffolds. Shi et al.  reported interconnectivity of less than 80% through openings larger than 200 μm, while the accessible pore volume was found to be less than 60% at the same minimum connection size in a study conducted by Hacker et al. .
Highly porous and well interconnected TiO2 scaffolds were fabricated using polymer sponge method. The fabricated scaffolds displayed compressive strength values above 2.5 MPa at 80–90% porosities. Considerable improvement was therefore observed in the compressive strength of the fabricated scaffolds in comparison to the strength values previously reported for TiO2 bone scaffolds. The improved strength was achieved by the optimisation of the processing parameters and without compromising the high porosity or the desired pore architectural features. Properties of the ceramic slurry, such as solid content, viscosity, and particle size distribution, are some of the key processing parameters that govern the strength of the ceramic foams. Recoating of the ceramic foams is another important processing step in producing foams with improved mechanical properties. Due to the high porosity and extremely high interconnectivity of the fabricated TiO2 scaffolds, the compressive strength of the TiO2 scaffolds may be improved even further with multiple recoating processes. However, subsequent recoating processes are anticipated to cause a detectable increase in the strut thickness, wherefore the desired pore architectural features, such as pore size, porosity, and interconnectivity, may eventually be compromised. Improvement in the stability of the ceramic slurry may also allow further enhancement in mechanical strength due to improved microstructure of the ceramic. Combined with the excellent biocompatibility and bioactive properties of TiO2, the highly interconnected pore structure of the scaffolds is expected to provide an excellent microenvironment for bone regeneration. Therefore, the fabricated TiO2 foams show great promise as load-bearing scaffolds for bone tissue engineering applications where moderate mechanical support is required.
This study was supported by Eureka-Eurostars Project Application E!5069 NewBone. The authors acknowledge Mr. Martin F. Sunding (Department of Physics, University of Oslo) and Dr. Ola Nilsen (Department of Chemistry, University of Oslo) for their assistance with the XPS and XRD analysis, respectively. Also many thanks to Mr. Geir Hansen (Malvern Instruments Nordic AB) for borrowing his Mastersizer 2000 at no charge.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.