Sie sind auf Seite 1von 12

Available online at www.sciencedirect.

com

Journal of the European Ceramic Society 34 (2014) 107118

Robocasting of 45S5 bioactive glass scaffolds for bone tissue engineering


Siamak Eqtesadi a , Azadeh Motealleh a , Pedro Miranda a, , Antonia Pajares a , Alexandra Lemos b , Jos M.F. Ferreira b
a

Departamento de Ingeniera Mecnica, Energtica y de los Materiales, Universidad de Extremadura, Escuela de Ingenieras Industriales, Avda. de Elvas s/n, 06006 Badajoz, Spain b Department of Materials Engineering and Ceramics, University of Aveiro, CICECO, 3810-193 Aveiro, Portugal Received 14 May 2013; received in revised form 2 August 2013; accepted 4 August 2013 Available online 4 September 2013

Abstract Three dimensional scaffolds with controlled pore architecture were prepared from 45S5 Bioglass powders by robocasting (direct-writing) using carboxymethyl cellulose (CMC) as the single processing additive. A comprehensive sintering study of the resulting structures was performed within the temperature range 5001050 C. Robocast scaffolds with interconnected porosities ranging from 60 to 80% were obtained for a xed scaffold design. All scaffolds exhibited compressive strengths comparable to that of cancellous bone (213 MPa), including those sintered at temperatures below the crystallization temperature of 45S5 bioactive glass. These strength values are substantially higher than any previously reported data for 45S5 Bioglass scaffolds and imply that robocasting is the rst technique which can be considered suitable for producing vitreous 45S5 scaffolds with a sufcient mechanical integrity for any practical application. Moreover, this process will enable the development of 45S5 Bioglass scaffolds with customized external geometry, and optimized pore architecture. 2013 Elsevier Ltd. All rights reserved.
Keywords: 45S5 bioactive glass; Robocasting; Direct-writing; Scaffold; Sintering; Strength

1. Introduction The 45S5 Bioglass developed by Larry Hench and colleagues at the University of Florida,1 with a composition of 45% SiO2 , 24.5% CaO, 24.5% Na2 O and 6% P2 O5 (in weight %, wt.%), is a commercially available inorganic biomaterial with a long history of success as bone replacement.2,3 This glass is claimed to be highly bioactive and both osteoinductive and osteoconductive, enabling its application in bone regeneration and tissue engineering.47 However, despite the fact that after more than 40 years of research no other calcium phosphate or bioactive glass composition has surpassed 45S5 Bioglass biological performance, its commercial success is somewhat limited and calcium phosphates remain market leaders as bone substitute materials.3 The main reason for this limited success lays in the difculties found when processing 45S5 bioactive glass powders into

Corresponding author. Tel.: +34 924 289 300x86735; fax: +34 924 289 601. E-mail addresses: pmiranda@unex.es (P. Miranda), jmf@ua.pt (J.M.F. Ferreira). 0955-2219/$ see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jeurceramsoc.2013.08.003

porous structures suitable for use as scaffolds for bone regeneration: that is, with a porosity appropriate for inducing bone ingrowth and the sufcient mechanical integrity to be handled during implantation and, if possible, to replace bone structural function during the whole tissue regeneration process. The problem lies in the small sintering windowtemperature difference between glass transition temperature (550 C) and that of the onset of crystallization (610 C)associated to 45S5 glass particles, which prevents any signicant degree of densication to occur before crystallization.8 However, extensive densication is required to obtain a mechanically competent 45S5 Bioglass -based scaffold in a foam-like structure, which would otherwise be made of loosely bonded particles and too fragile to handle. Accordingly, fabrication of porous scaffolds from 45S5 bioactive glass powder through traditional processing methods, such as foam replication, is made by using high sintering temperatures (1000 C), where the material is almost fully crystalline.9,10 Unfortunately, the bioactivity of such crystallized structures is retardedthe onset of apatite formation in simulated body uid (SBF) increases from 10 h to 2225 hover that of vitreous 45S5 material.11

108

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

Solid freeform fabrication (SFF) methods12 provide a means to improve the mechanical performance of scaffolds13,14 by providing a much greater level of control over pore architecture and more regular strut morphologies than in conventional techniques. In fact the regularly interconnected pores with large interconnection sizes of scaffolds fabricated by these additive manufacturing techniques allows one to signicantly reduce the overall porosity of the structure required to allow colonization by bone tissue.1518 However, despite the aforementioned advantages there are still few studies reporting the successful application of SFF techniques to fabrication of 45S5 Bioglass derived scaffolds.19,20 We recently added to this exiguous literature by reporting the rst successful attempt to fabricate 45S5 Bioglass -derived scaffolds via a colloidal processing route such as robocasting.21 Robocasting is an extrusion-based, direct writing technique, that utilizes high solids loaded paste-like suspensions (inks) to build 3D structures through a layer-wise deposition of extruded cylinders.2225 Critical rheological parameters for the printing ink include its apparent viscosity, yield stress and dynamic elastic modulus. An ink with appropriate rheological properties must be able to ow through the nozzle (i.e. low viscosity under high stress) and possess excellent shape retention upon deposition (i.e. high yield stress and elastic modulus at rest).25,26 To achieve this goal, the starting powders are usually dispersed in water with the addition of an anionic dispersant to obtain a concentrated suspension. The obtained owable slurry is then coagulated by adding suitable concentrations of salts or cationic species, or by changing the pH of the suspension to turn it into an extrudable paste. In the aforementioned preceding paper we reported how the difculties concomitant to the wet processing of 45S5 Bioglass powders, which are associated to its relatively high sodium content (24.5 wt.%) and the consequent pH increase resulting from leaching of Na+ ions into the dispersing water, can be overcome by using carboxymethyl cellulose as a single but multifunctional (dispersant, binder) additive.21 The current work extends the information on the robocasting ink optimization and rheological properties and reports the results obtained on a comprehensive sintering study, including mechanical characterization, performed on 45S5 Bioglass robocast scaffolds with a xed three-dimensional design. 2. Experimental procedure 2.1. Materials and ink preparation

zirconia container and balls as milling media and ethanol as dispersing medium, to provide particles in the 110 m size range, which was found optimal for the preparation of the concentrated suspensions required for robocastinglarger sizes seriously hamper colloidal stabilization, while smaller particles tend to reduce the maximum achievable solid loading. The slurries were dried at 60 C and the powder was sieved through 106 m and 75 m stainless steel sieves to eliminate large particle agglomerates. The size distribution of the resulting particulates was measured using a laser-scattering particle size analyser (Mastersizer 2000MU, Malvern, UK). Based on the ink optimization process described in a previous work21 and some additional details provided in Appendix, the optimal recipe for the 45S5 Bioglass robocasting ink uses carboxymethyl cellulose (CMC) as a single multifunctional additive, acting both as dispersant and binder to provide the required rheological behaviour. According to these results, 1 wt.% (relative to solid content) of CMC was rst dissolved in deionized water and then 45 vol.% of 45S5 glass powder was added to the suspension in small batches between successive mixing procedures performed in a centrifugal planetary mixer (ARE-250, THINKY, JAPAN), for a total mixing time of around 20 min. 2.2. Robocasting of bioglass scaffolds A robocasting device (A3200, 3D inks, Stillwater, OK, USA) was employed to fabricate three-dimensional structures of 45S5 Bioglass with the optimized inks, containing 1 wt.% of CMC as single processing additive and a solids loading of 45 vol.%. The ink was housed in a syringe and extruded through a conical nozzle (inner diameter, d = 410 m) by the computer-controlled robotic system. The position of the nozzle moved following the CAD model designed previously in the control software (Robocad 3.0, 3D inks, Stillwater, OK, USA). The scaffold design consisted of tetragonal mesh with a centre-to-centre spacing between adjacent rods within a layer of s = 2d = 820 m and a layer height h = 287 m; with external dimensions of 12 mm 12 mm 12 mm. The ink owed through the nozzle at the volumetric ow rate required to maintain a constant deposition speed of 10 mm s1 . To prevent a non-uniform drying of the structure during assembly, the deposition process was carried out within a parafn oil reservoir at room temperature. After deposition, the samples were removed from the bath and dried in ambient conditions for at least one day before thermal processing. 2.3. Thermal processing of assembled structures

The bioactive glass with the 45S5 composition, containing 45% SiO2 , 24.5% CaO, 24.5% Na2 O and 6% P2 O5 (in weight %, wt.%), was supplied by Mo-SCI Corporation (USA). Carboxymethyl cellulose (CMC, Mw = 250,000, Lamberti Iberia s.a.u., Castelln, Spain) with a viscosity of a 2 wt.% aqueous solution in the range of = 13 Pa s, as indicated by the supplier, was used as the sole processing additive. The as-received powder was milled for 4 h in an attrition mill (Model 01-HD, Union Process, Akron, OH), using high-purity

Thermal debinding kinetics of 3D constructs was evaluated using thermo gravimetric analysis, TGA, and differential thermal analysis, DTA, (SETSYS Evolution-16 SETARAM). Samples of as-received 45S5 bioactive glass powder and of dried robocasting inks (for 1 day at 100 C), were heated in air at 5 C min1 up to 1200 C. Based on these kinetic data, the following thermal treatment was selected for the removal of the polymeric additive and sintering of the as-printed structures.

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

109

Thermal burning out of CMC was performed in air at 400 C for 1 h, with a heating rate of 1 C min1 . Following the binder removal, the constructs were sintered in air for 1 h at different temperatures up to 1200 C, with a heating rate of 2 C min1 . The sintered samples were cooled in furnace at a rate of approximately 10 C min1 .

2.4. Microstructural and mechanical characterization of 45S5 bioactive glass scaffolds The density of the scaffolds, s , was determined from the mass and external dimensions of the structures. The porosity, p, was then calculated as: p=1 s = 1 rel th = 2.7 g cm3 (1)
Fig. 1. Particle size distribution of the as-milled 45S5 Bioglass powder.

3. Results and discussion is the theoretical density of bulk 45S5 where th Bioglass .3 The linear shrinkage, L, on the samples was calculated at specic temperatures from the variation of their external dimensions as: L(%) = Li Lf 100 Li 3.1. Scaffold preparation The particle size distribution of the as-milled 45S5 Bioglass particles used in preparing the inks for robocasting is shown in Fig. 1. The particles had a wide distribution of sizes, with a median diameter of 4.3 m. Fig. 2 shows SEM micrographs of an as-cut scaffold created using the optimized ink recipe and geometric design described in Section 2, followed by sintering at 1000 C for 1 h. Note the regularity of the structure and the straight geometry on the rods, attesting the quality of the ink used for its fabrication. 3.2. Sintering and microstructural characterization Fig. 3 presents TGA and DTA results of 45S5 Bioglass as-received powder and of the as-dried ink. The as-received powder underwent a slight weight loss up to 900 C, with the more marked changes in the range of 200400 C (Fig. 3a). This weight loss is associated to the evaporation of free water and the release of surface OH groups, a phenomenon reected as an endothermic effect, especially observed between room temperature and 200 C in the DTA plot (Fig. 3b). After recovering the base line, a second endothermic dip centred at 560 C is observed, which is attributed to the rst transition temperature of the glass (Tg1 ). This is followed by an exothermic effect starting around 600 C and with a maximum at about 685 C (Tp1 ), which is attributed to crystallization, in agreement with previously reported values.8,27,28 A second exothermic peak centred at 850 C (Tp2 ), attributed to the formation of a secondary crystalline phase, is followed by a new endothermic valley associated to a second glass transition (Tg2 910 C) of the remaining glass matrix.21 Melting clearly occurred above 1100 C, as deduced from the associated endothermic valley. The weight changes (Fig. 3a) and the corresponding thermal effects (Fig. 3b) were greatly enhanced in the dried ink, probably as a result of a more extended surface hydration

(2)

where Li is the initial dimension of the specimen at room temperature, and Lf the dimension after sintering at the selected temperature. Both shrinkage and porosity were evaluated as the average of measurement performed on four different specimens. The microstructure of the sintered scaffold was examined using a scanning electron microscope, SEM (SEM, S-3600N, Hitachi, Japan). Internal dimensions (rod diameter, spacing, etc.) of scaffolds fabricated at different temperatures were determined as the average of a minimum of 100 direct measurements performed on SEM images. These dimensions were then used to estimate the macroscopic, pre-designed porosity from geometrical considerations assuming an ideal network of interpenetrating cylinders. X-ray diffractometry (XRD, PW-1800, Phillips Research, The Netherlands) using CuK radiation (step width 0.03 2 , angular interval 1065 2 , and count time 4 s per step) was performed on 45S5 Bioglass grounded samples after sintering at selected temperatures for identication of the phases present. The mechanical response of the scaffolds fabricated at each sintering temperature was evaluated by uniaxial compression in a universal testing machine (AG-IS 10KN, Shimadzu Corp., Kyoto, Japan). The compression force was applied in a direction perpendicular to the printing plane (i.e. orthogonal to the scaffold rods) by using a constant crosshead speed of 0.6 mm min1 . Nine parallelepipedic samples (3 mm 3 mm 6 mm) were tested for each condition. Prior to mechanical testing, the contact surfaces of each sample were ground to produce smooth parallel surfaces.

110

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

Fig. 3. Results of the simultaneous (a) thermal gravimetric analysis (TGA, weight loss) and (b) differential thermal analysis (DTA, heat ow) of 45S5 Bioglass as-received powder (light colour) and as-dried robocasting inks (dark colour).

Fig. 2. SEM micrographs of a 45S5 Bioglass tetragonal mesh scaffold fabricated by robocasting and sintered at 1000 C for 1 h: (a) corner view of an as-cut sample for use in uniaxial compressive tests and (b and c) side views of the same sample.

of the powder undergone while in contact with dispersion media (ball milling and ink preparation), and of the burnout of CMC additive. Indeed, the weight loss centred in the range of 700800 C, but extending beyond these limits, is attributed to

the burnout of carbonaceous remnants from CMC degradation. This degradation occurred in two successive steps, originating two exothermic peaks at 225 C and 350 C (Fig. 3b), as conrmed by a TG/DTA test performed on pure CMC (results not shown). This suggested the use of a debinding treatment at 400 C for 1 h prior to sintering. Interestingly, the thermal events Tg1 and Tp1 occurred in the ink, respectively, at about 60 C and 30 C below the corresponding temperatures in the as-received 45S5 Bioglass powder, implying a concomitant shift of the intermediate exothermic effect associated to the nucleation step. This shift is attributed to a change in the composition at/near the particles surface as a result of leaching of modier cations (Na+ , Ca2+ ) during wetmilling, and also to the reduction in the particle size.29 Besides, Tp2 and Tg2 cannot be distinguished for the dried ink, possibly because they are also shifted to lower values and became overlapped with the peak for rst crystallization. The apparent increase of melting temperature of the ink is also consistent with the leaching of modier cations into the milling media. These results are in good agreement with the XRD results shown in Fig. 4. The 45S5 Bioglass remained amorphous up to above 550 C but there is evidence of crystallization already at 600 C. The angular location and intensity of the new peaks

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

111

Fig. 4. X-Ray diffractograms of 45S5 as-received and as-milled powders, and ground scaffolds sintered at the indicated temperatures. Sodium carbonate (Na2 CO3 , JCPDF database 19-1130) peaks are clearly visible on the as-milled powders, and barely on the sample sintered at 550 C. Peaks at 600 C, correspond to Na2 CaSi2 O6 (JCPDF database 77-2189) crystallized from the glass. At 1000 C, Na2 CaSi2 O6 remains the main crystalline phase with some additional peaks, noted by and , corresponding to Na2 Ca4 (PO4 )2 Si2 O4 (JCPDF database 32-1053) and tetragonal zirconia (ZrO2 , JCPDF database 72-7115), respectively.

Bioglass

Fig. 5. Evolution of macroscopic shrinkage and scaffolds internal dimensions as function of sintering temperature. Data represent mean values with standard deviations as error bars.

appearing upon crystallization of 45S5 Bioglass closely match the standard JCPDF database 77-2189, conrming that the major crystalline phase is Na2 CaSi2 O6 . Minor peaks of a secondary phase, attributed to Na2 Ca4 (PO4 )2 Si2 O4 (JCPDF database 321053), were also identied on samples sintered at 1000 C, all in good accordance with previous reports.8,3032 There are, however, some peaks already observable at 550 C and below, which correspond to sodium carbonate (Na2 CO3 , JCPDF database 191130) formed during the milling process in ethanol, with some small contribution of tetragonal zirconia from the milling media and container. The sodium carbonate, deposited onto the surface of the particles during milling, partially dissolves in water during ink preparation, which produces a substantial decrease in the intensities of the peaks corresponding to this extraneous phase. Actually, the peaks more clearly observable in the dried ink or in samples sintered at 550 C, located at 30.2 and 35.3 2 , can also be appreciated even at 1000 C, and are therefore attributed to the overlapping tetragonal zirconia phase rather than the sodium carbonate. Nonetheless, the presence of even small amounts of sodium carbonate in the XRD results support the explanation for the systematic down shift of thermal events (except melting) in Fig. 3 based on a changed composition, rich in Na+ cations, at/near the particles surface. Fig. 5 shows the evolution of scaffolds macroscopic linear shrinkage and internal dimensions with the sintering temperature, with standard deviations as error bars. The data at room temperature correspond to as-dried green structures. The structure shrinkage during sintering was essentially isotropic, and the scaffolds kept the designed parallelepipedical shape upon sintering for 1 h up to 1050 C. Above 1050 C or for longer sintering times at this temperature, indications of melting/softening of

the material became evident to the naked eye: the scaffold specimen deformed under its own weight and the shape was distorted, with some material accumulation at the bottom. When treated at 1200 C for 1 h the samples completely melted and collapsed with almost no remaining macroporosity. Due to the severe distortions they presented, samples sintered above 1050 C were not characterized. In good agreement with previous reports, the scaffold contraction progressed steadily with increasing sintering temperature, although at a somewhat faster rate around the two glass transition temperatures (500550 C, Tg1 , and 800850 C, Tg2 )8,28 ), with the same trend being observed for all the internal dimensions of the scaffolds. As shown in Fig. 6, the reduction in the different dimensions did not translate into any appreciable change in the pre-designed volumetric macroporosity, which remained constant at around 50% for all sintering temperatures. Consequently, all the densication produced during sintering, from an as-dried green body porosity of 80 2% to a minimum of 60 2% after sintering

Fig. 6. Evolution of scaffold porosities as function of sintering temperature. Data represent mean values with standard deviations as error bars.

112

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

Fig. 7. SEM micrographs of the rod surfaces of 45S5 Bioglass -derived scaffold before and after sintering at the indicated temperatures.

at 1050 C, was associated to a reduction of the microporosity within the glass-ceramic rods. Meaning that the rods reduced their initial 60% porosity (which roughly agrees with the 45% solid loading of the inks) to around a nal 20% porosity at the highest sintering temperature. Failure to achieve a high densication of the rods even at temperatures so close to the melting point is not an unexpected result, given the poor sintering ability of 45S5 Bioglass even under pressure-assisted sintering.31 Again, the evolution of densication with sintering temperature found in this work (Fig. 6) is in agreement with previous results:8,33 there is a steady densication of the scaffolds with increasing sintering temperature, but densication is signicantly enhanced around the two glass transition temperatures, and sluggish in-between. The evolution of in-rod microporosity described in Fig. 6 can be correlated with the microstructural analysis of the scaffolds. Fig. 7 shows SEM micrographs of rod surfaces of scaffold sintered at the indicated temperatures. No sintering occurred up to the glass transition temperature at 550 C, when the glass softened enough to allow some viscous ow, and even at that temperature the microstructure does not differ all that much from that of green samples (compare top images in Fig. 7), except for the evidences of inter-particle neck formation (see

inset in top-right image of Fig. 7). At around 600 C, just after the glass transition, the glass started to crystallize (Fig. 4), which severely slows down the densication process by reducing the species diffusivity and increasing the system viscosity. At this point, it is worth mentioning that the maximum percentage of crystallinity achievable in 45S5 Bioglass is about 80%,9 so that even at the highest sintering temperatures the material remains a glassceramic. Densication regains pace at around 850 C, when the residual phosphate-rich glassy phase softens and progresses appreciably with further increasing sintering temperature. Indeed, porosity at the rod surfaces is almost gone at 1000 C (bottom-left image in Fig. 7), although there remains a signicant level of internal (closed) porosity (Fig. 2). Up to this sintering temperature, the rod surface microstructure evidences the presence of crystalline grains homogenously embedded in a glass matrix. However, the intergranular glassy phase seemingly disappeared from the surface of the rods upon increasing the sintering temperature to 1050 C, or the holding time at 1000 C to 2 h and the naked grains became apparent (bottom-right image in Fig. 7). These changes were accompanied by an apparent increase in surface porosity, while the materials microporosity continued diminishing steadily with the chronothermic increments, as shown in Fig. 6. This

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

113

Fig. 8. Representative compressive stressstrain curves of 45S5 Bioglass derived scaffolds sintered at the indicated temperatures for 1 h.

Fig. 9. Evolution of the compressive strength of 45S5 Bioglass -derived scaffolds with sintering temperature. Data represent mean values with standard deviations as error bars. Cancellous bone strength (shaded band)36,37 and results for scaffolds from other authors10,19 are included for comparison. The total porosity of each structure is indicated as a label beside each data point.

suggests that the glassy phasewith a very low viscosity so close to the melting temperaturemight be draining from the surface towards the interior of the rods and aiding in the reduction of the aforementioned closed porosity. 3.3. Mechanical response of sintered scaffolds Fig. 8 shows representative stressstrain curves of scaffolds sintered at the indicated temperatures for 1 h that were tested under uniaxial compression in the direction perpendicular to the deposition plane. The scaffolds show an elastic response almost until the compressive strength of the structure (i.e. the maximum stress) is reached. At this point, the rst longitudinal cracks pops-in, breaking the unsupported rod segments close to the joints with adjacent layers, where maximum tensile stresses are located,34,35 and the applied stress drops. As cracking multiplies and propagates all over the scaffold, the stress continues decreasing progressively down to zero with some eventual temporary increases associated to a certain densication of the fractured structure under the applied compression, which is typical on cellular structures. The evolution of the compressive strength of robocast 45S5 Bioglass -derived scaffolds with the sintering temperature is shown in Fig. 9. It is evident, that strength increases steadily and strongly with sintering temperature, but there is no clear correlation of strength data with rod density (porosity is indicated as a label beside each data point in Fig. 9) or rod diameter, and the different stages observed in the data for the latter (Figs. 5 and 6) are not so clearly observed in Fig. 9. Especially indicative of this lack of complete correlation is the fact that while densication was improved in scaffolds sintered at 1050 C, there is a statistically signicant (p < 0.05) reduction in strength compared to structures sintered at 1000 C (from 13 1 MPa to 11 1 MPa). This reduction is attributed to the deleterious change in the starting aw population at the rod surfaces, where cracking initiates,34,35 upon the aforementioned disappearance of the

glass matrix (bottom-right image in Fig. 7). This result emphasises that surface defects are to be minimized whenever possible since their negative inuence on scaffold strength can counteract and even supersede the positive effect of a reduction in the total porosity of the structure. Thus, surface aw population effects, and maybe some toughness variations associated to the different microstructural and phase-compositional changes occurring in the material at different sintering temperatures, might be responsible for blurring the correlation between strength data and porosity/shrinkage results. In any case, it is evident from these results that, at least in terms of mechanical performance, the optimal condition for sintering 45S5 Bioglass -derived scaffolds fabricated by robocasting is 1000 C for 1 h. On the other hand, as shown in Fig. 9, when comparing these strength results with existing literature reports10,19 and cancellous bone properties it becomes evident that scaffolds produced by robocasting exhibit strength values far superior to anything previously reported for 45S5 Bioglass -derived scaffoldsunder the same sintering conditions, the enhancement is well over one order of magnitude (4000%). The reason for the enhancement of mechanical resistance over conventional scaffolds fabricated by foam replication10 should be found on the more regular pore architecture and thicker struts provided by robocasting. More surprising is the fact that other solid freeform fabrication methods such as stereolithography (STL),19 which in principle provide the same advantages, have not been able to duplicate these strength results even in scaffolds with lower porosities. Nonetheless, as the authors of the latter report point out, the reason for this may lay in a less than optimal selection of the sintering process.19 45S5 Bioglass is difcult to sinter, especially into complex scaffold geometries, and great care needs to be taken in selecting the thermal treatment parameters to avoid deleterious cracking during sintering. This must be especially critical in the case of STL, where larger amounts of organic binders (compared to robocasting) have to be burnt out in the

114

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

process. In any case, all robocast 45S5 Bioglass -derived scaffolds sintered between 550 C and 1000 C fall within or even surpass the range 212 MPa frequently quoted for cancellous bone.36,37 Consequently, robocasting provides, for the rst time, a means to produce mechanically competent Bioglass (not just Bioglass -derived) scaffolds. Indeed, the robocast scaffolds sintered at 550 C remain completely amorphous and exhibit a compressive strength that, while barely improved over that of as-dried green structures, lies at the lower end of cancellous bone range and is enough for safe handling during implantation. This could help overcome the main hurdle for the successful application of 45S5 Bioglass as a broad-use bone substitute material: the difculty to produce scaffold with sufcient mechanical integrity without reducing the outstanding bioactivity of amorphous 45S5 particles.3 At this point it is worth discussing some issues regarding porosity. Although the total porosities reported here (60% and 80% for scaffolds sintered at 1000 C and 550 C, respectively) are lower than typical values of conventional 45S5 Bioglass -derived scaffolds (9095%),10 conventional (foam replication) scaffolds with similar porosities as those reported here will still fall well short of robocast scaffolds in term of strength (as can be easily estimated by extrapolating reported data10 ). Moreover, there is ample evidence that, while such extremely high porosities might be required to attain signicant bone ingrowth in the tortuous architectures of conventional scaffolds, porosities such as those reported here are enough to allow bone regeneration and blood vessel growth in the case of calcium-phosphate robocast structures.1517 In any case, robocast scaffold porosities could be easily increased to match reported values by modifying the initial design, for example by increasing rod-spacing within each deposition layer. More in vitro, in vivo and degradation studies are nonetheless warranted to assess the full potential and possible limitations of these regular pore architectures with thicker struts produced by robocasting as compared to more conventional trabecular-like geometries. Finally, the authors would like to point-out that the mechanical performances of the robocast 45S5 Bioglass -derived scaffolds both in terms of strength and toughness could be further enhanced by impregnating them with suitable biodegradable polymers,3841 although this might come at the expense of reducing bioactivity.49 This might prove especially useful in the case of the less resistant but more bioactive glassy scaffolds sintered at 550 C. 4. Conclusions Robocasting technique provides a means to produce 45S5 Bioglass -derived scaffolds with customized external geometry and internal pore architecture with compressive strengths that are far superior to any previously reported values. This is achieved through the use of carboxymethyl cellulose (CMC) as a single, multifunctional additive that allows one to overcome the inherent difculties associated to the colloidal processing of 45S5 Bioglass and prepare the concentrated suspensions required by robocasting.

The enhancement in mechanical performance associated to the novel pore architectures, with thicker struts and wider interconnections, produced by robocasting is enormous: more than one order of magnitude compared to existing scaffold strength data and within or even surpassing cancellous bone strength range. This enhancement allows one to produce, by sintering at sufciently low temperature (550 C), fully amorphous robocast 45S5 Bioglass scaffolds, which preserve intact all the outstanding bioactivity of 45S5 Bioglass , with enough mechanical integrity for practical use. Therefore it could be said that robocasting, and perhaps also other solid freeform fabrication techniques, can overcome the main obstacle for the successful application of 45S5 Bioglass as a broad-use bone substitute material. The simplicity and versatility of the proposed robocasting process will greatly facilitate the development of patient-specic 45S5 Bioglass scaffolds and signicantly extend the current range of biomedical applications of this excellent bioactive material. Acknowledgments The authors would like to thank Prof. Angel Luis Ortiz for his assistance in the analysis of XRD data. This study was supported by the Junta de Extremadura and FEDER Funds under grant IB10006. The support from Department of Materials Engineering and Ceramics, CICECO, University of Aveiro, is highly appreciated. Siamak Eqtesadi thanks the Universidad de Extremadura for his grant (Ref. 1105). Appendix. Optimization of robocasting ink composition and rheology In the optimization of the inks to be used in robocasting, different polyelectrolyte dispersing agents were tested: (i) a sub-group of anionic nature: ammonium polymethacrylate (Darvan C; mol. wt. = 2000; R.T. Vanderbilt Co., Norwalk, CT); poly (methylvinyl ether) (EasySperse; ISP, Wayne, NJ); Targon 1128 (BK Ladenburg GmbH, Ladenburg, Germany); and (ii) a sub-group of cationic ones: tetramethylammonium hydroxide (TMAH, Moses Lake Industries, Inc., USA); Polyethyleneimine (PEI, Low Mw, Aldrich, USA). Zeta potential ( ) measurements as a function of the pH (changed by adding drop by drop appropriate amounts of 0.1 M HCl to the suspension) were performed for the 45S5 Bioglass particles dispersed in water in the absence and in the presence of 1 wt.% (relative to solid content) of each different dispersant using a Zeta-potentiometer (Zetasizer Nano ZS, Malvern, UK). At each pH value, an excessive number (100) of measurements was set, and the instrument automatically stopped when a constant zeta-potential value was measured, typically after 2025 scans. The preparation of colloidal suspensions using different added amounts of each of the abovementioned dispersants ranging from 0.5 to 2 wt.% was attempted. Anionic dispersants were rstly attempted for ink preparation because, according to the traditional approach, they are expected to adsorb at the surface of

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118 Table A1 Maximum solid loadings achieved in aqueous 45S5 bioglass suspensions with different added amounts of the various polyelectrolyte dispersants tested. Dispersant Type Maximum achievable solid loadings (vol.%) 0.5 29.8 28.1 22.5 21.2 19.5 1.0 31.9 29.2 23.7 22.5 22.2 1.5 33.8 31.5 25.1 23.8 23.4 2.0 35.3 33.1 27.8 26.5 26.7

115

Dispersant concentration (wt.%) Darvan C EasySperse Targon 1128 TMAH PEI Anionic Anionic Anionic Cationic Cationic

the solid particles and stabilize the suspensions, which can in a later stage be coagulated by adding cationic additives.26,42 However, as shown in Table A1, a maximum of 35 vol.% solids loaded suspensions could be hardly achieved (only in the presence of 2 wt.% Darvan C), which is somewhat low for robocasting (optimal would be around 4045 vol.%). Even worse results were obtained using cationic polyelectrolytes as dispersants (Table A1). The explanation for these difculties can be found by looking at the -potential results plotted in Fig. A1. For all dispersion conditions other than in the presence of PEI, the -potential curves show a continuous decrease from a common isoelectric point at around pH 1.5close to the values reported for silica,18 suggesting that 45S5 glass particles exhibit a silica type surfaceup to a pH of about 56 where the trend inverted slightly to again resume its drop after reaching a relative maxima at about pH 8. The intermediate relative maxima imply that Ca2+ , CaOH+ ions leached from the 45S5 Bioglass, and not just H+ and OH ions, are playing a role in the determination of -potential.21 The fact that the electrophoretic curves are not signicantly shifted (except in the case of PEI) from that of the bare particles suggests that there is not a signicant adsorption of the selected polyanions on the surface of 45S5 Bioglass particles, which severely limits their effectiveness as dispersants. This is

Fig. A1. Zeta-potential of 45S5 Bioglass particles versus pH. Data for bare particles (hollow symbols) and after the addition of a xed amount (1 wt.%) of the indicated anionic (dark symbols: EasySperse; Targon 1128; Darvan C) or cationic (light symbols: TMAH and PEI) dispersants are included.

attributed to electrostatic repulsion between anionic polyelectrolytes and the negatively charged 45S5 bare particles, and to the fact that TMAH ions are not prone to be specically adsorbed onto silica-like surfaces.43 On the other hand, the strong upwards shift in the electrophoretic curve to positive potentials in the presence of the cationic PEI proves that PEI species do specically adsorb at the surface of the 45S5 Bioglass particles, as previously reported on similar silica-like surfaces.44 However, the resulting isoelectric point (pH 10) coincides with the pH of the dispersion environment created by the 45S5 glass powder in waterindeed, pH augments from 6 to 10 within the rst 30 s due to Na+ /H+ ionic exchange.45 This implies the absence electrostatic interactions among the particles under these conditions, and the system will tend to occulate under the action of Van der Walls forces. This cannot be avoided by lowering the pH of the dispersing medium since this has the counter effect of accelerating the 45S5 Bioglass dissolution, which makes pH to increase again.4648 Moreover, considering that anionic/cationic interactions of species added to the system in different successive steps are the base of the most commonly used mechanism to promote the rheological transition from a uid-like behaviour (starting suspension) to a paste-like character (usable in directwrite assembling techniques), it became clear that this traditional approach could not offer successful prospects for robocasting processing of 45S5 Bioglass . Based on all the above considerations, we have adopted a completely different and simplied approach: searching for a single additive that could act as both dispersant and binder for robocasting. In this pursue, carboxymethyl cellulose (CMC) was regarded with interest due to its aptitude to play different but complementary roles as processing additive.21 The less polar character of the CMC chains in comparison to the anionic polyelectrolyte molecules facilitates surface adsorption. Such adsorption is likely to occur, however, through small chain segments, with laces and tails extending to the solution, since it is not electrostatically driven (both CMC molecules and the surface of the particles have the same negative polarity). The predicted conguration of the CMC molecules adsorbed at the surface of the 45S5 Bioglass particles is likely to play a steric effect which enabled reaching signicantly higher solid loadings, up to 45 vol.%, compared to anionic polyelectrolytes (Table A1). In particular, high molecular weight CMC (Mw = 250,000, Lamberti Iberia s.a.u., Castelln, Spain, with a viscosity = 13 Pa s for a 2 wt.% aqueous solution) was found more appropriate for the preparation of inks for robocasting since longer chains

116

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

Fig. A3. Effect of the amounts of CMC on the yield stress, y , and the average elastic modulus at low stress (0.11 Pa), G , of 45 vol.% 45S5 Bioglass suspensions. G data were calculated from previous results.21 Error bars represent standard deviations (from tting in the case of y ).

The initial data plotted in Fig. A2a (up to 3 s1 ) were tted with the HerschelBulkley model,29 which is described as: = y + K( 0 )n (3)

Fig. A2. Plots of (a) shear stress and (b) apparent viscosity as a function of shear rate for 45 vol.% 45S5 Bioglass suspensions with indicated amounts of CMC as binder.

induced a stronger thickening effect, which was attributed to a bridging occulation mechanism.21 The ow properties of 45 vol.% suspensions prepared using different amounts of this CMC as the sole binding agent were evaluated using a rheometer (Bohlin CVO, Malvern, UK). Flow measurements were conducted at shear rates between 0.2 and 100 s1 using a cone and plate conguration (4 , 40 mm, and gap of 150 m). Fig. A2 shows the shear stress versus shear rate (Fig. A2a) and the apparent viscosity versus shear rate curves (Fig. A2b). All systems exhibit an evolution of shear stress versus shear rate that approaches the Bingham plastic behaviour, typical of suspensions containing polymeric binder/thickening agents. The polymer molecules help forming a network structure under rest or near rest conditions, which is then gradually disrupted under the applied shear stress eld as the polymeric molecules align along the ow direction.

where 0 is the shear rate, is the shear stress, y is the yield stress, n is the shear thinning exponent and K is the viscosity parameter. The yield stress is the shear stress in the absence of shear motion and represents the minimum stress leading to deformation of the static form of the material. The model parameters were estimated as n = 0.66 0.06 and K = 580 50 Pa sn for suspensions containing up to 1 wt.% of CMC and n = 0.36 0.03 and K = 2600 200 Pa sn for samples with 2 wt.%. Fig. A3 shows the evolution of the estimated yield stress and the average elastic (storage) modulus, G , at low shear stress of suspensions with added CMC. Modulus values were calculated as the average of the virtually constant G data within the 0.11 Pa stress range from previous work.21 These parameters reect, respectively, the strength and stiffness of the attractive network structure resulting from interactions between 45S5 Bioglass particles, CMC and water. In the absence of CMC, both the yield stress and the modulus of the suspension are very low. Upon the addition of CMC, the yield stress increases linearly with the amount of binder up to a concentration of 1 wt.%, decreasing at 2 wt.%. This thickening effect promoted by CMC might be attributed to chain entanglement of the molecules adsorbed at the surface of the particles and/or to an increased viscosity of the dispersing medium. On the other hand, the elastic modulus increases rst very slowly, then dramatically (by 2 orders of magnitude) at 1 wt.% CMC additions, and nally decreases slowly at higher CMC concentrations. The dramatic change in modulus observed for 1 wt.% CMC indicates a uid-to-gel transition attributed to the occurrence of bridging-occulation, in which the longer polymeric chains might adsorb at the surface of different neighbouring particles. Increasing the amount of CMC further leads to a higher degree of surface coverage and a denser adsorbed layer, which decreases the probability of one polymeric chain to adsorb in more than one particle, thus hindering

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118

117

bridging-occulation and reducing both the modulus and the yield stress. Since high values of this two parameters are desired features for robocasting inks, 1 wt.% was considered as the optimal CMC amount. This concentration yielded maximum network stiffness (>100 kPa), and yield stress (>1 kPa) at a relatively low viscosity (Fig. A2), which facilitates owability through the deposition nozzle. Indeed, such suspensions exhibited both very good owability (unlike suspension with higher CMC concentrations) and excellent shape retention capacity upon deposition, which made them ideal inks for direct writing techniques. References
1. Hench LL, Splinter RJ, Allen WC, Greenlee TK. Bonding mechanisms at the interface of ceramic prosthetic materials. J Biomed Mater Res 1971;5:11741. 2. Hench LL. The story of Bioglass . J Mater Sci Mater Med 2006;17: 96778. 3. Jones JR. Review of bioactive glass: from Hench to hybrids. Acta Biomater 2013;9:445786. 4. Xynos ID, Edgar AJ, Buttery LDK, Hench LL, Polak JM. Gene-expression proling of human osteoblasts following treatment with the ionic products of Bioglass (R) 45S5 dissolution. J Biomed Mater Res 2001;55:1517. 5. Hattar S, Asselin A, Greenspan D, Oboeuf M, Berdal A, Sautier JM. Potential of biomimetic surfaces to promote in vitro osteoblast-like cell differentiation. Biomaterials 2005;26:83948. 6. Reilly GC, Radin S, Chen AT, Ducheyne P. Differential alkaline phosphatase responses of rat and human bone marrow derived mesenchymal stem cells to 45S5 bloactive glass. Biomaterials 2007;28:40917. 7. Hench LL, Polak JM. Third-generation biomedical materials. Science 2002;295:10147. 8. Boccaccini AR, Chen Q, Lefebvre L, Gremillard L, Chevalier J. Sintering, crystallisation and biodegradation behaviour of Bioglass-derived glassceramics. Faraday Discuss 2007;136:2744. 9. Chen QZZ, Thompson ID, Boccaccini AR. 45S5 Bioglass -derived glassceramic scaffolds for bone tissue engineering. Biomaterials 2006;27:241425. 10. Chen Q, Mohn D, Stark WJ. Optimization of Bioglass scaffold fabrication process. J Am Ceram Soc 2011;94:418490. 11. Filho OP, La Torre GP, Hench LL. Effect of crystallization on apatite-layer formation of bioactive glass 45S5. J Biomed Mater Res 1996;30:50914. 12. Tay BY, Evans JRG, Edirisinghe MJ. Solid freeform fabrication of ceramics. Int Mater Rev 2003;48:34170. 13. Miranda P, Pajares A, Saiz E, Tomsia AP, Guiberteau F. Mechanical properties of calcium phosphate scaffolds fabricated by robocasting. J Biomed Mater Res A 2008;85A:21827. 14. Fu Q, Saiz E, Tomsia AP. Direct ink writing of highly porous and strong glass scaffolds for load-bearing bone defects repair and regeneration. Acta Biomater 2011;7:354754. 15. Abarrategi A, Moreno-Vicente C, Martnez-Vzquez FJ, Civantos A, Ramos V, Sanz-Casado JV, et al. Biological properties of solid free form designed ceramic scaffolds with BMP-2: in vitro and in vivo evaluation. PLoS ONE 2012;7:e34117. 16. Lan Levengood SK, Polak SJ, Wheeler MB, Maki AJ, Clark SG, Jamison RD, et al. Multiscale osteointegration as a new paradigm for the design of calcium phosphate scaffolds for bone regeneration. Biomaterials 2010;31:355263. 17. Abarrategi A, Fernandez-Valle ME, Desmet T, Castejn D, Civantos A, Moreno-Vicente C, et al. Label-free magnetic resonance imaging to locate live cells in three-dimensional porous scaffolds. J R Soc Interface 2012;9:232131. 18. Simon JL, Michna S, Lewis JA, Rekow ED, Thompson VP, Smay JE, et al. In vivo bone response to 3D periodic hydroxyapatite scaffolds assembled by direct ink writing. J Biomed Mater Res A 2007;83A:74758.

19. Tesavibul P, Felzmann R, Gruber S, Liska R, Thompson I, Boccaccini AR, et al. Processing of 45S5 Bioglass by lithography-based additive manufacturing. Mater Lett 2012;74:814. 20. Comesa na R, Lusqui nos F, del Val J, Lpez-lvarez M, Quintero F, Riveiro A. others Three-dimensional bioactive glass implants fabricated by rapid prototyping based on CO2 laser cladding. Acta Biomater 2011;7:347687. 21. Eqtesadi S, Motealleh A, Miranda P, Lemos A, Rebelo A, Ferreira JMF. A simple recipe for direct writing complex 45S5 Bioglass- 3D scaffolds. Mater Lett 2013;93:6871. 22. Lewis JA, Smay JE, Stuecker J, Cesarano III J. Direct ink writing of threedimensional ceramic structures. J Am Ceram Soc 2006;89:3599609. 23. Lewis JA. Direct-write assembly of ceramics from colloidal inks. Curr Opin Solid State Mater Sci 2002;6:24550. 24. Cesarano III J. Review of robocasting technology. Boston, MA, USA: Materials Research Society; 1999. p. 1339. 25. Cesarano III J, Segalman JR, Calvert P. Robocasting provides moldless fabrication from slurry deposition. Ceram Ind 1998;148:94102. 26. Michna S, Wu W, Lewis JA. Concentrated hydroxyapatite inks for directwrite assembly of 3-D periodic scaffolds. Biomaterials 2005;26:56329. 27. Chatzistavrou X, Zorba T, Kontonasaki E, Chrissas K, Koidis P, Paraskevopoulos KM. Following bioactive glass behavior beyond melting temperature by thermal and optical methods. Phys Status Solidi (A) 2004;201:94451. 28. Lefebvre L, Chevalier J, Gremillard L, Zenati R, Thollet G, BernacheAssolant D, et al. Structural transformations of bioactive glass 45S5 with thermal treatments. Acta Mater 2007;55:330513. 29. Chatzistavrou X, Zorba T, Chrissas K, Kaimakamis G, Kontonasaki E, Koidis P, et al. Inuence of particle size on the crystallization process and the bioactive behavior of a bioactive glass system. J Therm Anal Calorim 2006;85:2539. 30. Xin R, Zhang Q, Gao J. Identication of the wollastonite phase in sintered 45S5 bioglass and its effect on in vitro bioactivity. J Non-Cryst Solids 2010;356:11804. 31. Chen QZ, Xu JL, Yu LG, Fang XY, Khor KA. Spark plasma sintering of solgel derived 45S5 Bioglass -ceramics: mechanical properties and biocompatibility evaluation. Mater Sci Eng C 2012;32:494502. 32. Chen Q, Boccaccini AR. Coupling mechanical competence and bioresorbability in Bioglass -derived tissue engineering scaffolds. Adv Eng Mater 2006;8:2859. 33. Bretcanu O, Chatzistavrou X, Paraskevopoulos K, Conradt R, Thompson I, Boccaccini AR. Sintering and crystallisation of 45S5 Bioglass powder. J Eur Ceram Soc 2009;29:3299306. 34. Miranda P, Pajares A, Saiz E, Tomsia AP, Guiberteau F. Fracture modes under uniaxial compression in hydroxyapatite scaffolds fabricated by robocasting. J Biomed Mater Res A 2007;83A:64655. 35. Miranda P, Pajares A, Guiberteau F. Finite element modeling as a tool for predicting the fracture behavior of robocast scaffolds. Acta Biomater 2008;4:171524. 36. Jones JR, Ehrenfried LM, Hench LL. Optimising bioactive glass scaffolds for bone tissue engineering. Biomaterials 2006;27:96473. 37. Hench LL. Bioceramics. J Am Ceram Soc 1998;81:170528. 38. Martinez-Vazquez FJ, Perera FH, Miranda P, Pajares A, Guiberteau F. Improving the compressive strength of bioceramic robocast scaffolds by polymer inltration. Acta Biomater 2010;6:43618. 39. Martinez-Vazquez FJ, Perera FH, van der Meulen I, Heise A, Pajares A, Miranda P. Impregnation of -tricalcium phosphate robocast scaffolds by in-situ polymerization. J Biomed Mater Res Part A 2012, http://dx.doi.org/10.1002/jbm.a.34609. 40. Peroglio M, Gremillard L, Chevalier J, Chazeau L, Gauthier C, Hamaide T. Toughening of bio-ceramics scaffolds by polymer coating. J Eur Ceram Soc 2007;27:267985. 41. Martinez-Vazquez FJ, Miranda P, Guiberteau F, Pajares A. Reinforcing bioceramic scaffolds with in-situ synthesized -polycaprolactone coatings. J Biomed Mater Res A 2013, http://dx.doi.org/10.1002/jbm.a.34657. 42. Miranda P, Saiz E, Gryn K, Tomsia AP. Sintering and robocasting of betatricalcium phosphate scaffolds for orthopaedic applications. Acta Biomater 2006;2:45766. 43. Jiang D. Gelcasting of carbide ceramics. J Ceram Soc Jpn 2008;116:6949.

118

S. Eqtesadi et al. / Journal of the European Ceramic Society 34 (2014) 107118 48. Tilocca A, de Leeuw NH. Ab initio molecular dynamics study of 45s5 bioactive silicate glass. J Phys Chem B 2006;110: 258106. 49. Peroglio M, Gremillard L, Gauthier C, Chazeau L, Verrier S, Alini M, Chevalier J. Mechanical properties and cytocompatibility of poly(epsilon-caprolactone)-inltrated biphasic calcium phosphate scaffolds with bimodal pore distribution. Acta Biomater 2010;6: 436979.

44. Sun J, Gao L. Dispersing SiC powder and improving its rheological behaviour. J Eur Ceram Soc 2001;21:244751. 45. Cerruti M, Greenspan D, Powers K. Effect of pH and ionic strength on the reactivity of Bioglass- 45S5. Biomaterials 2005;26:166574. 46. Zeitler TR, Cormack AN. Interaction of water with bioactive glass surfaces. J Cryst Growth 2006;294:96102. 47. Tilocca A, Cormack AN. Modeling the waterbioglass interface by ab initio molecular dynamics simulations. ACS Appl Mater Interfaces 2009;1:132433.