Rapid manufacturing of bioceramic implants by direct writing.

Tallis, Andrea ; Jones, Jason ; Wimpenny, David 等

1. INTRODUCTION

The complexity of geometry found in the maxillofacial area means that off the shelf prosthetics are an impractical solution to meeting the needs of surgeons when tackling bone loss in the skull. A UK government funded research project seeks to create a viable process for manufacturing custom made ceramic based prostheses for the maxillofacial region using the rapid prototyping technique of Direct Writing, sometimes known as Solid Freeform Fabrication. In this process a ceramic slurry mixture is extruded to create a thin filament of material which can be built up layer by layer and 'freeze cast' by rapid cooling to create a complex 3D geometry specifically designed to match the needs of the patient.

This work seeks to carry out the initial optimisation of the slurry formulation in order to obtain a mix that is capable of both freeze casting and accurate extrusion.

2. LITERATURE

Rapid prototyping of ceramics has previously been confined to Selective Laser Sintering techniques, however the Direct Writing of ceramics has recently received considerable research interest, e.g. Wang and Shaw (2005) looked at using dental porcelains in RP applications to directly fabricate dental implants. Morrissette and Lewis (2000) carried out work on aqueous alumina and PVA gelcasting suspensions, this technique however requires large quantities of organic additives to promote cross linking in the green body making burn out more complex.

Laurie et al (1992) showed how a modified sol-gel processing technique could be used to make bulk castings of ceramic material when an aqueous solution of colloidal silica is used as the liquid phase of the slurry. Rapid freezing of parts causes an irreversible gelling reaction to take place and near net shape castings can be made with reduced levels of both cracking and shrinkage compared with the conventional sol-get processing route. Use of this gelation method would therefore reduce the need for organics, simplifying the slurry mixture.

However, when freeze casting aqueous alumina slurries in bulk the addition of small amounts of glycerol to the slurry mix serves as a cryoprotectant, effectively disrupting the crystalline behaviour of water contained in the slurry (Sofie and Dogan, 2001). As water crystals in the freeze cast body translate into microstructural defects in the dried sample this homogenisation of crystal growth leads to a much more even porosity without the large scale defects that can be associated with the process.

Substrates suitable for 3D scaffolds must exhibit both microporosity to aid cell attachment and an interconnected macroporosity to enable vascularisation to take place (Hench, 1991). The even microporosity generated by the processing conditions of the freeze cast slurry, coupled with the interconnected macroporosity that the rapid prototyping technique allows us to 'write' into the structure makes this an ideal substrate for cell growth and proliferation.

3. METHOD

In the trials conducted at DeMontfort University the ceramic slurry was made from a mixture of mono-modal alumina powder with a sub-micron particle size (CT3000SG, Almatis, Germany) and a sol made up of an aqueous suspension of colloidal silica, a dispersant and glycerol. The slurry was prepared by mixing the sol and powder in a ball mill (Retsch PM100, Germany) to ensure homogeneity before being transferred to either the rheometer or extrusion apparatus. Four different slurry formulations were prepared, A--D, these consisted of identical quantities of powder mixed with varying quantities of sol to give a range of material viscosities.

Trials to determine the viscosity of each of the slurry formulations were carried out using a Bohlin Instruments CVO 120 HR rheometer (Malvern, England) fitted with a 4[degrees] / 40 mm cone and plate geometry. Isothermal controlled shear ramps from 0.1/s to 100/s were conducted at 21[degrees]C with a solvent trap fitted.

A PVM syringe pump (Sapphire Engineering, UK) fitted to an XYZ robot was used as the extrusion apparatus. The 5 ml extrusion syringe is fitted with a syringe tip with an internal diameter of 900 [micro]m. This system was programmed to write a freestanding triangular structure with wall length of 22 mm composed of 10 layers. Extrusion started with a tip height of 1 mm above the glass substrate and was raised by 1 mm for each subsequent layer. The rate of travel of the extrusion head was set to 900 mm [min.sup.-1] and the syringe piston was set to dispense 5 [mm.sup.3] [s.sup.-1].

Immediately after extrusion, samples were sealed in a polypropylene box which was then placed in a bath filled with an isopropanol / dry ice slurry with a temperature -70[degrees]C. After one hour the samples were removed and allowed to dry for several days on the bench top before measuring.

[FIGURE 1 OMITTED]

Samples that did not have uniform extrusion due to trapped air within the syringe were discarded, these only occurred in sample set A. Wall height was measured to the nearest 10 um on each of the three walls of the samples apart from sample set D where excessive slumping made measurement impossible.

4. RESULTS

Figure 1 shows examples of extruded structures. The left hand image shows a discarded sample of material A where the extrusion has been affected by trapped air within the syringe body, one third of samples constructed from slurry formulation A had to be discarded due to this effect. The central image is from sample set B, formulation C produced similar parts to B. The right hand image is from sample set D where it is possible to see that excessive slumping of the material means that control of part geometry is impossible.

In order to determine the effect of viscosity on the slumping behaviour of the different slurries the height of the ten layer wall was measured with the exception of sample set D where slumping was extreme. The mean and standard deviation of these sample sets is set out in Table 1 above.

The slumping behaviour of slurries is determined by their rheological characteristics. The viscosity of each of the samples A--D was measured whilst the rate of shear they were subjected to was linearly increased. Figure 2 shows a graph plotting sample viscosities as a function of shear rate.

The degree of shear experienced by the slurry as it passes through the extrusion tip can be calculated using

Shear rate = 4Q/[pi][r.sup.2] (1)

where Q is the volume flow rate and r is the radius of the extrusion orifice. This gives us an applied shear rate of 70 [s.sup.-1], at this level of shear the viscosity range of the formulations is very narrow ranging from 6.8 Pa s for A down to 1.6 Pa s for D.

5. DISCUSSION

[FIGURE 2 OMITTED]

The rheological plots in Fig 2 show that all of the slurry formulations exhibit shear thinning (pseoudoplastic) behaviour. This means that as the level of shear that the material is exposed to is increased, its viscosity decreases. This characteristic makes these slurries suitable for extrusion as the action of passing through the syringe orifice subjects the slurry material to considerable shear stresses. Once the shear is removed the viscosity of the slurry increases once again, resisting further flow of the material and allowing the construction of self supporting 3D structures.

The results of this trial show us that there is a viscosity 'window of opportunity', the viscosity of sample set D is insufficient to prevent a degree of slumping that makes precise control of geometry impossible. Whereas the increased viscosity of sample set A means that air pockets can become trapped within the slurry in the syringe body of the extrusion head. As a consequence, when these voids reach the syringe tip no material is extruded until the trapped volume of air has been expelled. This leaves a gap in the wall which on the next layer of the program effectively means that the tip height above the substrate has been increased and is too far above the substrate, allowing the extrudate to distort before settling on the next previous layer. Thus irregular parts like those illustrated in the left hand image of Figure 1 become more likely and part reproducibility decreases. When formulation A extrudes successfully, uniform parts are produced, giving the lowest standard deviation value from each of the sample sets, however the high degree of wastage from this group makes this too an undesirable slurry mix.

Both formulations B and C yield parts with good reproducibility without the problem of trapped air volumes within the syringe. This would indicate that both of these formulations fall within the desirable viscosity 'window' for this process. Further, the slightly lower standard deviation of sample set B would indicate that this formulation is closer to the optimum and should be adopted for further trials.

6. FURTHER RESEARCH

Further work includes sintering these samples to determine whether firing shrinkage is influenced by changes in formulation. In addition, tensile testing of samples made with each of the slurry mixes should be carried out in order to determine whether the strength of the finished artefact is affected by changes in formulation.

7. REFERENCES

Hench, L. (1991) Bioceramics: From Concept to Clinic, Journal of the American Ceramic Society, 74, 4, (March 2005) 1487-510, ISSN 1551-2916

Laurie, J, Bagnall, C, Harris, B, Jones, R, Cooke, R, Russell-Floyd, R, Wang, T, Hammett, F. (1992) Colloidal suspensions for the preparation of ceramics by a freeze casting route, Journal of Non-Crystalline Solids, 147 & 148, (August 1992) 320-325, ISSN 0022-3093

Morissette, S. & Lewis, J. (2000) Solid Freeform Fabrication of Aqueous Alumina-Poly(vinyl alcohol) Gelcasting Suspensions, Journal of the American Ceramic Society, 83, 10, (October 2000) 2409-16, ISSN 1551-2916

Sofie, S. & Dogan, F. (2001) Freeze Casting of Aqueous Alumina Slurries with Glycerol, Journal of the American Ceramic Society, 84, 7, (July 2001) 1459-64, ISSN 1551-2916

Wang, J. & Shaw, L. (2005) Rheological and extrusion behaviour of dental porcelain slurries for rapid prototyping applications, Materials Science and Engineering A, 397, (April 2005) 314-321, ISSN 0921 5093

Tab. 1. Mean wall height of 10
layer structures from sample sets

Sample Mean Standard
Set ([micro]m) Deviation
 ([micro]m)

A 10 239 127
B 10 126 169
C 10 269 185