Preparation of 3D Printable HAp-based Composite Powder for Binder Jetting Process

The purpose of this study was to prepare a 3D printable powder composed of hydroxyapatite and biocompatible polymers such as chitosan, dextrin, and polyvinylpyrrolidone for the binder jetting process. The relationship between powder properties such as owability and particle size distribution, as well as printing quality were investigated in the binder jetting process. For this purpose, 3D printable powder and an appropriate solvent were designed and demonstrated by hydroxyapatite and water-soluble polymers such as carboxymethyl chitosan, dextrin, and PVP. Results showed that a combination of 60%wt. hydroxyapatite, 28%wt. carboxymethyl chitosan, 10% wt. dextrin, and 2% wt. PVP, with controlled particle size distribution according to the Dinger-Funk equation led to the best print quality. Finally, ash dipping of the 3D printed parts in chitosan solution resulted in increases of compressive and Young's modulus from 1.3 and 10 MPa to 7.4 and 125 MPa, respectively.


Introduction
Every year, millions of patients suffer from bone tissue loss due to pathological events such as trauma, disease, or surgery. Consequently, bone reconstruction procedures are widely implemented using autografts or allografts to improve bone healing in a critically sized defect. However, these bone scaffolds and grafts suffer from many limitations, originated from the traditional shaping method, which makes synthetic alternatives an attractive option. Calcium phosphate (CaP) ceramics are superior and promising candidates to repair the bone defect since they decrease the problems of the rejection reaction and donor scarcity. Furthermore, they have a long history of synthetic bone replacement due to their chemical similarity to bone minerals [1][2][3]. Hydroxyapatite (Ca10(PO4)3OH) is widely used due to the excellent physiological and environmental biocompatibility, as well as compatibility with synthetic and natural polymers such as polysaccharides and proteins (e.g. collagen). Consequently, a combination of HAp and biocompatible polymers creates a functional biomaterial for medical and veterinary applications [4].
The excellent application potential of additive manufacturing (AM) technology, more commonly known as 3D printing, in bone tissue engineering is widely noticed in previous works [5][6][7][8][9][10][11][12]. Among AM techniques, only direct extrusion 3D printing and binder jetting 3D printing can fabricate the scaffolds at low temperatures [13]. Thus, these techniques are highly bene cial since they provide the potential to create composites with synthetic or biological polymers [1]. Binder jetting (BJT) is an AM technology suitable for the production of high complexity 3D components using ceramic or metal precursor powders, with the potential for various applications, wherein a roller mechanism spreads the powder material, forming a bed, and then an inkjet print head moves in the XY plane of the machine selectively jetting the binder agent [14]. In the binder jetting processes, the binder can be applied in three different routes [15]. In the rst one, called binder jetting in place, binder material is diluted in an appropriate solvent, and then jetted onto the powder bed. The second route, called binder pre-mixing, is to mix binder uniformly and ceramic powder together, followed by milling and mixing of the mixture [16,17]. In the third way, called binder pre-coating, the ceramic powder and binder particles are dissolved in the binder solvent to obtain a mixture slurry. Then, the slurry can be dried or spray dried, milled, and sieved to obtain a binder-coated ceramic powder [18][19][20].
The major challenge in scaffold fabrication via BJT is to select the type and amount of binder, which leads to high green strength, and the method of binder added to the powder stock [33]. Unfortunately, despite the importance of the powder properties, which profoundly in uences printing accuracy, there has to date been a limited number of practical scaffold 3D printing works showing the relation between the powder properties and the printing outcome [3].
As a controversial subject, the powder size and shape is a crucial factor determining the quality of the printed components by in uencing the surface roughness, owability, and packing density of powder bed, which are essential parameters for the layer-based additive manufacturing techniques. The challenge is to select an optimum particle size; a small particle size may increase the packing density. However, it quickly causes the formation of large and stable agglomerates, which leads to a decrease in owability.
Although large particles have good owability, they cause lower packing density and a rough surface [3,13]. It was suggested in previous works that the multi-modal powders, by a mixture of ne and coarse powders in a particular ratio, lead to the optimization of the challenging parameters. This is because ne powders ll the gaps among coarse powders, resulting in increased overall packing density and improved mechanical properties of the printed part [34,35].
The principles mentioned above suggest that a complicated relationship must exist between powder properties and the nal additive manufactured scaffold properties. Due to the given theoretical and technical limitations, this study aims to prepare a 3D printable powder using HAp/biocompatible polymer composite for the BJT process. The relation between relevant powder properties, printability, and nal properties of 3D printed parts was investigated here.

Powder preparation
Hydroxyapatite (Merck, Germany, CAS No. 1306-06-5, 100% purity), chitosan (CMS) (Sigma-Aldrich, Germany, 100% purity), dextrin (Bonfozhan Kimia, Iran, 100% purity), and polyvinylpyrrolidone (PVP) (Rahavard Tamin, Iran, 100% purity) were used as the primary materials. Firstly, chitosan was transformed into Carboxymethyl Chitosan (CMCTS) through a synthesis process to obtain a watersoluble composition according to a previous report by Lu et al [36]. Considering that HAp 60wt.%: Chitosan40wt.% is the composition as an arti cial bone, different arrangements of HAp and polymer binders were mixed in aqueous solution with 1:4 powder to water ratio according to were dried at 60 °C for 24 hours, ground, and passed through different sieves to adjust particle size distribution. The standardized test methods were used to measure the apparent density and tap density of the prepared powders [37,38]. In previous reports [35,39,40], the Hausner rate was de ned as a quantitative number showing the owability of printable powders, which is de ned as the ratio of the tapped density to the apparent density.

3D Printing of samples
The printability of prepared powders was evaluated by 3D printing of hexagonal shape with a 20 mm diamter and a 5 mm height using a homemade BJT machine, working with a piezoelectric printing head.
3D printing parameters of the device were xed at 200 µm layer thickness, powder roller spreading speed of 50 cm/sec, layer printing intervals of 2 secs, printing speed of 1 cm 2 /sec, and binder saturation of 50%. Table 2 shows the composition of the water based binder. The best 3D printable powder was selected among the samples having the optimum values of spread and owability, binder adhesion, and solution to binder ratio. The nal printed parts had high dimension accuracy, appropriate density, and desired mechanical strength of a biocomposite. Infrared spectroscopy analysis (Shimadzu S 8400, Japan) was utilized to examine likely changes in polymers structure. Density measurement of printed samples was carried out according to the C373-88 ASTM standard [41,42].
The compressive strength of samples was measured by a universal testing machine (SANTAM SAF-50, Iran) with a uniaxial force rate of 0.5 mm/min. Mechanical test samples were 3Dprinted in the form of a cylinder, with 5 mm diameter and 10 mm height, according to the ISO5833 standard. Post-print processing operations included ash dipping of the 3D printed samples in a 0.5 wt.% chitosan solution, followed by drying at 40 °C for 24 hours, which was carried out to increase the strength of printed samples.

Results And Discussion
3.1. Characterization of powdered batches Figure 1 represents the SEM image of hydroxyapatite and composite powders. Given the surface morphology and particle size change, it can be concluded that the HAp particles were mixed by the CMCTS, dextrin, and PVP to form new composite powder with larger granules, in that smoother surface can be an evidence for the presence of polymers.
According to a report provided by Inzana et al. [1], two sieves 80 (177 µm) and 120 (125 µm) were used for powders P1-P3 to remove large and ne particles, respectively, to attain high dimensional accuracy in the printed parts and prevent the agglomeration of ne particles. Considering a report by Bai et al. [35], remained powders under 80 and 120 mesh sieves were then mixed by a ratio of 30:70.
The Dinger-Funk equation was applied as an alternative approach for particle size distribution optimization of P4 and P5 powders assuming that all possible particle sizes were present in the nal distribution.
In Eq. 1, CPFT is the cumulative percent of particles ner than a given particle size D, D l is the size of the largest particle, D s is the size of the smallest particle and n is Dinger-Funk particle size distribution modulus. According to this Equation, considering a constant printing layer height thickness of the print layer in this study (200 µm) and a report presented by Spierings et al. [43], the highest powder owability and layer packing density are obtained when the D90 of powder is smaller than layer thickness (200 µm); moreover, D10 should not be lower than 5 µm. The sizes of the smallest and largest particle were considered 5 and 105 µm, respectively. In order to obtain the highest packing density, a n power of 0.37 was considered in Eq. 1.  Table 3 describes the characterization of optimized powders via the Dinger-Funk equation. The particle size distribution width was de ned as (D90-D10)/D50. The higher the numerical value obtained for this parameter, the broader the particle size distribution, and the smaller the numerical value, the narrower the particle size distribution [44]. In a study performed by Karapatiz [39], the criteria for proper particle size distribution were reported as follows: 1) D90 < t Layer 2) D50 / D10 ≥ 10

3) D90 / D10 ≤ 19
However, the value for D50/D10, calculated according to the Dinge-Funk equation (Table 2), is smaller than 10 and did not pass the criteria expressed by Karapatiz; this condition was ignored in this study. However, it is within the range reported by Spiering et al. [43] Since the thickness of the printed layers is 200 µm in this process, the condition of Expression 1 was satis ed in both samples.  Table 4 represents the tap and apparent densities of P1-P5 powders, and the Hausner ratio. An increase in the PVP percentage in P2 powder led to decreased apparent and tap densities, followed by the rise of the Hausner ratio. This effect is attributed to declined HAp granule sizes due to the dispersing behavior of PVP [45,46], and consequently, an increase in the van der Waals force followed by diminishing the lling properties of the powder. The effect of particle size and density reduction by PVP was somewhat reduced in powders P3-P5 due to the use of dextrin along with CMCTS and PVP.
The lowest and the highest Hausner ratios were 1.22 and 1.3 for P2 and P3 powders, respectively (Table 3). A comparison of P1 and P2 reveals that adding PVP to P2 results in ner granules. PVP has both hydrophilic and hydrophobic groups and functions as a surfactant resulting in particle size reduction and an increase in the van der Waals force, leading to the formation of low-ow agglomerates [47]. Well-owing powders can move quickly and, by occupying less space and reducing the cavities between the powder particles, lead to an increase in density. Low-ow powders also reduce density due to their inability to move. From what deduced before, it can be concluded that the P1 powder with the lowest Hausner ratio has the best owability.
In order to evaluate printability, P1-P5 powders were printed as a 3D hexagonal shape model, named A1-A5, respectively. Layer shifting was observed during 3D printing of A1 sample with the lowest Hausner rate, due to a high ow rate in this sample, and consequently, the fast distribution of the layers by the roller. Moreover, this sample showed high shrinkage, and hence, the edges of the model have a protruding protuberance on the surface. The high dissolution rate of CMTC in water and its fast reaction with HAp powder caused rapid bonding of the particles to each other and, consequently, occurring different shrinkage of outer and inner layers.
Due to the presence of PVP in the P2 sample, bonding between powder particles is improved in one layer compared to powder P1. In fact, due to the property of surface lm formation, this polymer was used as the third component to eliminate layer shifting and edge protuberance in sample A1. The PVP powder particles swell as they moisten and form a surface lm following the polymer chains. This swelling is low at low molecular weights and very high at high molecular weights. The higher the molecular weight of the PVP used, the higher its ability and the more particles form the lm will be [48]. Adding the PVP eliminated layer shifting and edge protuberance, but it lowered the green strength.
Dextrin was added to P3 powder to increase the green strength of the printed sample. The green strength of the A3 sample increased due to the improvement of bonding between the particles. According to a patent by Liu. et al. [49], the presence of at least one carbohydrate as an active binding agent is required in the binder used in the BJT process.
By obtaining good green strength after reaching an ideal composition, the effect of particle size adjustment in P4 revealed good owability and packing density of the A4 sample. An increase in the slurry mixing time in the P5 sample improved the bonding of powder particles and resulted in the highest green strength in printed samples (Fig. 2).

Characterization of composite components
Cylindrical shape samples with 5 mm diameter and 10 mm height were 3d printed from P5 powder, as revealed in Fig. 3.
SEM images (Fig. 4a and 4b) show the printed sample before and after ash dipping in the chitosan solution, respectively. Immersion in the chitosan solution decreased the porosity in the sample due to partial closure of some surface [50].
Figure 4a clearly shows that large and small pores remained form incomplete binder penetration in some parts of the powder bed during the 3d printing [51]. In the powder-bed AM method, particles with different sizes become separated by moving powder; the coarser particles form a coarse particle front at the top of the powder layer due to the better ow, and the ne particles fall to the bottom of the layer. Consequently, the non-lling of the voids occurs between the larger particles by the ne particles and eventually causes the creation of the cavities. These results are consistent with a report by Jacob et al. [52] who found that ner particles moved away from larger particles by accumulating in the lower layers of the powder bed.
The porosity of the printed parts was measured by the Archimedes method. Printed parts without any processing operation showed average and maximum porosity levels of 42% and 45.31%, respectively, ranging between 40 and 45%. The ash dipping of the printed sample in the chitosan solution reduced the porosity. The porosity of these samples was in the range of 30-36%, with an average of 34% and a maximum of 36.42%. Mean apparent green density and bulk density of respectively 1.74 g/cm 3 and 1.01 g/cm 3 were calculated for A5, and 1.86 g/cm 3 and 1.18 g/cm 3 for A5-D samples for A5. The increase in both densities in the A5-D samples can be attributed to the decreased porosity percentage, and to the chitosan introduced arti cially into the printed sample structure. The signi cant differences in the densities of both groups can be attributed to the presence of closed porosities and a lack of a good correlation between the pores.
In the CMCTS spectra, the peak at 1612 cm − 1 belongs to the carboxyl group (-COOH) added to the polymer. Since the carboxyl group usually appears at 1730 cm − 1 , its displacement in the corresponding spectrum is due to the conversion of the -COOH group to the carboxylic acid salt (-COONa   Compressive strength and modulus of elasticity in the A5 specimen are 1.3 and 10 MPa, respectively. These values for the ash dipped sample, A5-D, increased to 7.4 and 125 MPa, respectively, which can be attributed to the amino groups of chitosan protonated in the acetic acid solution. Since nitrogen atom in the amine group (NH 3+ ) is more electronegative than hydrogen, nitrogen becomes negatively charged. On the other hand, nitrogen in the amine group has two free electrons in its valence layer. Due to the Ca + 2 ion in hydroxyapatite, nitrogen leads to a covalent bond by sharing two valence electrons with calcium. By arti cial incorporation of chitosan into the ash dipped sample structure, the covalent bond between the chitosan amine group and the hydroxyapatite metal ion results in increased compressive strength and modulus of elasticity in this sample [54]. On the other hand, as noted in the porosity measurement section, there is an 8% reduction in the porosity of the A5-D sample compared to the A5 sample, which improves the mechanical properties.
Chavana et al. [55] reported a maximum compressive strength of 16 MPa and 30% porosity for a composite including 40% hydroxyapatite, 60% chitosan, and lactic acid printed using the BJT process and immersed in 40% wt. Lactic acid solution. By comparing the results of this study with previous reports, the strength obtained for the A5-D specimen showed acceptable results regarding the amount of hydroxyapatite.

Conclusion
The best composition among the powders was P5 with 60% HAp, 28% CMTC, 10% dextrin, 2% PVP with a particle size width of 2.10, and a Hausner ratio of 1.26 compared with the other powders.
Both the apparent and tap densities can be increased by optimizing the particle size distribution using the Dinger-Funk equation; in addition, it results in suitable powder owability and optimum print quality.  The printed sample with height to diameter ratio of 1 to 0.5

Figure 4
Scanning electron microscope images of a) A5 before ash dipping in Chitosan solution and b) A5-D after ash dipping in Chitosan solution.