3Y-TZP DLP Additive Manufacturing: Solvent-free Slurry Development and Characterization

Accepted: January 21, 2021 Vat photopolymerization (VP) stands out among ceramic additive manufacturing processes for its ability to print sub-100 micrometer complex features. One of the main challenges of this process is the preparation of a homogeneous and stable ceramic slurry with a high solid load and low viscosity. In this work, different dispersants and resins were tested, aiming to provide a solvent-free slurry suitable for DLP additive manufacturing. Disperbyk-111 and PEGDA 250 stood out in the tests, providing a 40 vol% ceramic slurry with no noticeable sedimentation and viscosity of 2.3 Pa.s at 30 s -1 despite the relatively high specific surface area (15 m 2 /g) of the 3Y-TZP powder used compared to powders usually used for VP slurries. The adsorption of Disperbyk-111 on ceramic particles surface was investigated by FTIR. Finally, ceramic bodies were 3D printed, debound and sintered at 1500 ºC for 2 h, confirming the ability to manufacture


Introduction
A critical step in ceramic vat photopolymerization (VP) is the preparation of ceramic slurries that meet the necessary process requirements 1,2 . These suspensions must have a high ceramic loading to ensure low porosity and high flexural strength. A solid loading of at least 40 vol% is required to avoid post-processing defects 3 and high shrinkage that enhances the generation of distortions in the product geometry after sintering. On the other hand, the higher the solid load, the higher the viscosity of the slurry 4 . Some studies have used high viscosity (up to 18.8 Pa.s) ceramic slurries in equipment with special system for forming new layers such as Admaflex 130 (ADMATEC Europe BV, The Netherlands) 5 and Ceramaker 300 system (3D CERAM, France) 6 . However, suspensions whose viscosities do not exceed 3 Pa.s are preferred for common VP equipment 2,3,7 . Higher viscosities prevent the recoating of constant and homogeneous layers, and therefore, defects can be generated when a surface tension balance is not achieved 8 . Moreover, stability is a key factor in the performance of ceramic slurries for VP additive manufacturing 4, 9,10 .
Although some studies use aqueous acrylamide-based ceramic suspensions 7,[11][12][13][14][15][16] , which provide low-strength green parts 5,17 , most suspensions used for ceramic VP are based on hydrophobic non-aqueous systems. The monomer 1,6-hexanediol diacrylate (HDDA) has been used in the vast majority of such work due to its low viscosity and its ability to form crosslinked organic networks 18 . HDDA has been used in association with other monomers such as ethoxylated pentaerythritol tetraacrylate (EPTA) 19,20 , polyethylene glycol diacrylate PEGDA 17 , ethoxylated pentaerythritol tetraacrylate (PPTTA) [21][22][23] , and 1,1,1-Trimethylol propane triacrylate (TMPTA) 4,24,25 , or as a single monomer [26][27][28] . Moreover, an appropriate photoinitiator must be chosen to initiate the photopolymerization reactions, considering mainly its compatibility with the adopted monomers and the relationship between its light absorption spectrum and the wavelength emitted by the light source 29,30 . Lastly, the hydrophilic character of the ceramic particles requires the use of suitable dispersants to promote efficient dispersion in non-polar resins 31 . Figure 1 shows the viscosity of zirconia photosensitive slurries used in additive manufacturing by VP 1,32-36,38 according to volume fraction (at a shear rate of 30 s -1 ) and the dotted rectangle identifies the preferred characteristics for suspensions (at least 40 vol% solid loading and up to 3 Pa.s viscosity). It is important to note that the three studies that achieved these characteristics used solvent-based slurries. Another interesting point is that the zirconia powders used in the related literature have a specific surface area between 6 and 9 m 2 /g, as seen by the labels in Figure 1, consistent with the protocol for the development of photocurable ceramic suspensions created by Gonzalez et al. 30 which suggests that the specific surface area should be around 7 m 2 /g. Other related works that used alumina 26,31 or piezoelectric material 17 also follow this trend. Li et al. 5 found that powder's specific surface area is the characteristic that most effects the suspension viscosity.
*e-mail: italo.camargo@usp.br According to this work, the larger the specific surface area of the powder, the greater the amount of liquid needed to wet the surface and the smaller the amount of free liquid in the suspension, resulting in increased viscosity. Also, larger specific surface areas favor agglomeration 5,38 .
An attempt to use 3Y-TZP powders with a specific surface area of 15 m 2 /g in photopolymerizable slurries was recently made; however, a large amount of solvent was used to decrease the viscosity, which provided linear shrinkage higher than 34% 39 .
In this work, different dispersants and resins were tested, aiming to provide a solvent-free low viscosity slurry with relatively high specific surface area 3Y-TZP powder suitable for digital light processing (DLP) additive manufacturing. Rheological behavior and stability were investigated for different formulations and 3Y-TZP detailed parts were fabricated using 40 vol% ceramic slurries.

Materials and Methods
Solvent-free photosensitive 3Y-TZP suspensions made of ceramic powder, resin, photoinitiator and dispersant were prepared, characterized, and then a suitable ceramic slurry was used to manufacture ceramic parts by DLP additive manufacturing, as described next.
The slurry preparation started with the mixing of the resin with the dispersant, followed by the gradual addition of the ceramic powder and photoinitiator (2 wt% of the resin). Then, the suspension was ball-milled with zirconia balls for 24 hours to break up agglomerates and homogenize the suspension.

Slurry Characterization
3Y-TZP powder and 3 wt% of DISPERBYK-111 were mixed in ethanol, ball-milled with zirconia balls for 24 hours, rinsed with ethanol and oven-dried at 100 °C for 24 hours. The powder Fourier Transform Infrared Spectrum (FTIR) was obtained by a spectroscope (VERTEX-70, Bruker, Germany).
The rheological behavior of the ceramic slurries was characterized using a rotational viscometer (DV2T extra, Brookfield, Canada) between 23 and 25°C. In the flow curves, the rheological parameters were adjusted according to the Herschel-Bulkey model 41 (Equation 1).
where τ is the shear stress, τ e is the yield stress, K is a constant, γ is the shear rate, and n is the power index.  Each suspension's stability was investigated by sedimentation tests, as previously done by some related works 4, 10,25,26,31,42 . Thus, 10 ml of each suspension was poured into graded tubes. The measuring tubes were kept undisturbed for 30 days and the sedimentation volume fraction was recorded as a function of time.
Ceramic slurries with 15 vol% of 3Y-TZP and PEGDA 250 as monomer were chosen to analyze the effects of dispersants on rheological behavior and stability. Although a higher percentage of ceramics is desired for the slurries used for DLP additive manufacturing, preliminary analyzes need to be performed at lower concentrations in order to find suitable components, otherwise, most of the initial test suspensions would have very high viscosities, outside the measurement range of the viscometer, and could not be compared. In addition, high solid loading makes it difficult to visualize sedimentation and related works performed these tests in a concentration below 20 vol% 6,26,31,38,43 .
Once the best dispersant was defined, the solid loading could be increased for testing different resins. Ceramic slurries with 30 vol% of 3Y-TZP and 2 wt% of Disperbyk-111 were chosen to analyze the effects of using different resins on the rheological behavior of the suspensions. In this case, solutions with high concentration (e.g 40 vol%) would still provide slurries with viscosity outside the viscometer's measurement range.
Next, ceramic slurries with PEGDA 250 as the monomer and 2 wt% of Disperbyk-111 were chosen to analyze the effects of solid loading on the viscosity of the suspension. In these experiments, it was found that a suspension with 40 vol% solid loading and 2wt% of dispersant BYK 111 generates viscosity greater than 3 Pa.s. Hence, suspensions with the same ceramic content and with 1 and 3 wt% of BYK were analyzed, looking for a suspension with a viscosity suitable for the process. Although the viscosity of the suspension with 15 vol% did not change significantly at concentrations of BYK111 between 1 and 3 wt%, it is expected that this variation may have significant effects in highly concentrated suspensions.

DLP additive manufacturing of 3Y-TZP ceramic parts and characterization
The fabrication of the green parts was performed through a home-built top-down DLP 3D printer whose design and validation were presented by Camargo et al. 46 . The parts were then rinsed with isopropyl alcohol (IPA) and dried on the oven for 12 h at 100 °C. Finally, the samples were debound and sintered at 1500 °C for 2 hours in a box furnace (Blue M, Lindberg). The linear shrinkage was determined using a digital micrometer. The density of the sintered parts was measured based on Archimedes' Principle using an analytical balance (AUW220D, Shimadzu). in shear rate), as shown in Figure 2b. In addition, Figure 2c shows the flow curve.

Influence of different dispersants
The rheological parameters were adjusted according to the Herschel-Bulkey model. The coefficient of determination (R 2 ) related to all the suspensions was higher than 0.999, indicating a good fit to the model. High yield stress is generally considered to be an obstacle to the spreading of new layers 3,31 and the yield stress tends to rise with increasing solid loading 4,31 . Therefore, Disperbyk-111 and Disperbyk-2001 have the advantage of having a negligible yield stress in 15 vol% suspensions. All the other dispersants had yield stress higher than 1.0 Pa, with Triton X-100 reaching the highest value (11.1 Pa). For these reasons, Disperbyk-111 was chosen for the following rheological tests and also for the DLP additive manufacturing experiments. Figure 3a shows that the slurry using PEGDA 250 had the lowest viscosity (under 0.5 Pa.s at 30 s -1 ) even though, on its own, this monomer has a considerably higher viscosity than HDDA. These are two bifunctional monomers with similar molar mass; however, HDDA is hydrophobic and PEGDA 250 is hydrophilic with polar groups on the polymer chains, which makes it more compatible with ceramic particles with hydroxyl groups on their surfaces. Hydrophilic monomers lead to well-dispersed ceramic particles in suspensions 47 . Contrarily, nonpolar/hydrophobic monomers do not effectively disperse ceramic particles 31,47 . Hydrophobic monomers do not prevent agglomeration 38 and tend to provide higher viscosity dispersions 48 . Lastly, Figure 3b shows the flow curve with the Herschel-Bulkey model which again proved to be a good fit (R 2 >0.999) and indicating the absence of yield stress for all slurries analyzed. Figure 4 shows the viscosity at a shear rate of 30 s -1 for suspensions with different ceramic volume fractions. The Krieger-Dougherty model was used to estimate the maximum solid fraction. The maximum solid volume fraction found m φ = 52,2%, providing a satisfactory fit (R 2 ≅0.999). As φ approaches m φ , any increase in the solid volume fraction of the suspension represents a large increase in viscosity.

Influence of solid loading
The highest possible solid loading that provides adequate viscosity to the process (<3 Pa.s) is desired. However, 40 vol% exceeded this viscosity limit in the dispersant concentration analyzed (2 wt% of BYK) and the effect of the dispersant concentration is discussed again in the next section

Influence of dispersant concentration
Even though a concentration test has already been presented for Disperbyk-111 with suspensions of 15 vol%, in that case, the viscosities were very low and no significant differences could be observed. Therefore, a new test was performed for suspensions with higher solid loading. Ceramic slurries with 40 vol% of 3Y-TZP and PEGDA 250 as the monomer were chosen to analyze the effects of the concentration of Disperbyk-111 on viscosity, as shown in Figure 5a. The suspension with 1 wt% of Disperbyk-111 was very viscous and could not be measured because it was outside the measuring range of the rotational viscometer. An increase in the concentration of Disperbyk-111 caused a decrease in viscosity and 3 wt% of this dispersant provided a 40 vol% ceramic slurry viscosity suitable for the suitable for DLP  additive manufacturing (2.3 Pa.s at 30 s -1 ). The shear-thinning behavior presented by these suspensions is characteristic of VP ceramic slurries 21,42,43,47,[49][50][51][52][53] . Such behavior is desirable in VP 1,54,55 because it avoids sedimentation of the suspension in a stationary state, and adequate flow when a shear rate is applied to the slurry 32 . Figure 5b shows the shear stress-shear rate curves. According to the Herschel-Bulkley model, the suspension with Disperbyk-111 presented a yield stress of 5.9 Pa (R 2 >0.9999). Such a moderate value of this property has been considered desirable to avoid spontaneous flows 56 .

Stability
The effects of dispersants in the stability of the ceramic suspensions is shown in Figure 6. The suspension with Disperbyk-111 had the best performance with a retained volume fraction of 98% after 30 days. Moreover, all the sedimentation tests using this dispersant with a solid loading of 40 vol% presented no noticeable sedimentation, indicating well-dispersed slurries with stability suitable for DLP additive manufacturing. Figure 7a is the FTIR spectrum of Disperbyk-111. The peaks at 1,099 and 1,730 cm -1 , present in the FTIR spectrum of this dispersant, correspond to the stretching vibration peaks of COC and C=O 33 . Figure 7b shows that the spectrum of the zirconia powder modified with Disperbyk-111, after being rinsed and dried, presents similar stretching vibration (at

Printed Parts
Ceramic slurries with 40 vol% of 3Y-TZP, 3 wt% of Disperbyk-111, and PEGDA 250 were chosen to manufacture some parts by DLP additive manufacturing. Figure 8 shows a 3D printed part before and after sintering and Figure 9 shows sintered printed bars with a magnified view of the printed layers indicating parts with no visible warping or delamination. The relative density of the sintered parts was found to be greater than 95% and the shrinkage was 25.0% in the X-Y plane and 27.3% in the Z direction. Despite the significant shrinkage difference, such results are in agreement with what has been presented in the literature, in which related studies obtained even greater differences 38 . The use of a solvent-free formulation provided a significant decrease in shrinkage compared to related works. For example, Amaral et al. 2020 39 produced parts with 34 vol% shrinkage using the same powder in a solvent-based formulation.

Conclusions
In this work, a solvent-free low viscosity slurry with relatively high specific surface area 3Y-TZP powder suitable for DLP additive manufacturing was developed and characterized. Disperbyk-111 and PEGDA 250 proved to be good choices for the formulation of ceramic VP formulations. 3 wt% of this dispersant provided low viscosity slurries with no noticeable sedimentation due to the steric stabilization created on the ceramic particles. Although PEGDA 250 is a higher viscosity monomer than HDDA, it generated slurries with lower viscosities, possibly due to its hydrophilicity, which matches the character of the ceramic particles' surfaces. Photosensitive 3Y-TZP slurries with a solid loading of 40 vol% and viscosity less than 3 Pa.s were used to manufacture parts with sub-100 µm details in a DLP 3D printer.
The variations of VP processes (building direction, recoating system, light source, etc.) and possible applications in various areas (sensors, fuel cells, heat exchangers, dental, bioengineering, etc.) may require different requirements regarding the slurry (rheology, stability, curing parameters) and post-processing requirements (solid-state reactions, densification, grain growth). For a specific project, the specifications must be made respecting the prerequisites of the project and the process. In this work, 3Y-TZP with a relatively high surface area was used because it is a widely used universal ceramic and the characterization of the photopolymerizable suspension developed can be used in a wide range of applications. The influence of adding such components as defoamers, plasticizers and sintering additives may be the subject of future work.

Acknowledgement
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior -Brasil (CAPES) -finance code 001 6. References