Digital Light-Processing (DLP) 3D Printing of Magnesium Phosphate Minerals and Cements

Digital light processing (DLP) enables the fabrication of complex 3D structures based on a photopolymerizable resin usually containing a photo initiator and an UV or photo absorber. The resin and thus the final properties of the printed structures can be adjusted by adding fillers like bioceramic powders relevant for bone‐regeneration applications. Herein, a water‐based and biocompatible poly(ethylene glycol diacrylate) (PEGDA) resin containing the photo initiator lithium‐phenyl‐2,4,6‐trimethylbenzoylphosphinate (LAP) enables the production of 3D structures via DLP. The addition of calcium magnesium phosphate cement (CMPC) powder, acting as photo absorber, leads to higher accuracy of the final structures. After curing the printed construct in a diammonium–hydrogen phosphate (DAHP) bath for hardening, the resulting mechanical properties can be adjusted without post‐process sintering. Solid loading of up to 40 wt% CMPC powder is possible, and the resins are investigated regarding their rheological behavior and printability. The resulting constructs are analyzed in respect to their surface morphology using scanning electron microscope (SEM), porosity, phase composition using X‐ray diffraction (XRD) methods, as well as mechanical properties influenced by the hardening process using DAHP for different durations.


Introduction
Additive manufacturing (AM) is a promising field of research especially for the fabrication of patient-specific implants for regenerative medicine. Digital light processing (DLP) is one of these AM techniques capable of producing different materials into 3D scaffolds. DLP uses a layer-by-layer process to fabricate 3D structures by exposing a thin layer of resin to a light source to initiate a spatially controlled polymerization reaction within the resin. This step is repeated with fresh resin for each layer resulting in a 3D structure. Templates and print files can either be designed by computer-aided design DOI: 10.1002/adem.202201330 Digital light processing (DLP) enables the fabrication of complex 3D structures based on a photopolymerizable resin usually containing a photo initiator and an UV or photo absorber. The resin and thus the final properties of the printed structures can be adjusted by adding fillers like bioceramic powders relevant for boneregeneration applications. Herein, a water-based and biocompatible poly(ethylene glycol diacrylate) (PEGDA) resin containing the photo initiator lithium-phenyl-2,4,6trimethylbenzoylphosphinate (LAP) enables the production of 3D structures via DLP. The addition of calcium magnesium phosphate cement (CMPC) powder, acting as photo absorber, leads to higher accuracy of the final structures. After curing the printed construct in a diammonium-hydrogen phosphate (DAHP) bath for hardening, the resulting mechanical properties can be adjusted without postprocess sintering. Solid loading of up to 40 wt% CMPC powder is possible, and the resins are investigated regarding their rheological behavior and printability. The resulting constructs are analyzed in respect to their surface morphology using scanning electron microscope (SEM), porosity, phase composition using X-ray diffraction (XRD) methods, as well as mechanical properties influenced by the hardening process using DAHP for different durations.
biocements. To the best of our knowledge, processing of dual-setting cements was solely performed by hand-casting and such cements have not yet been used in AM by now. This is somewhere surprising as there are clear advantages of this approach, for example, it would open the way to process temperature-sensitive ceramic phases due to the absence of a temperature treatment post printing.
Applying the concept of dual-setting cements to DLP printing requires the development of processable resins with prerequisites such as a suitable viscosity as mentioned earlier. One potential material could be a water-based resin matrix, which can be photopolymerized as well as contains embedded cement particles. The latter must be nonreactive during printing, but with the possibility to initiate the setting reaction post printing for hardening of the structure. This was achieved in the current study by choosing poly(ethylene glycol diacrylate) (PEGDA) for the resin formulation as it is miscible with water and can be rapidly polymerized into a 3D network. Furthermore, PEGDA is also known for its biocompatibility, and has already been used for DLP printing of soft tissue implants. [21,22] As a cement system, a calcium-substituted magnesium phosphate with the general formula Ca 0.75 Mg 2.25 (PO 4 ) 2 was added to the PEGDA resin. This phase is in a first attempt not reactive in an aqueous environment, which allows processing by DLP without time constraint using lithium-phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) as photo initiator. A cement-setting reaction is then initiated post printing by immersion of the samples in diammonium phosphate solution to form struvite as setting product. Such struvite forming biocements are of high clinical interest for bone replacement as they combine excellent mechanical performance with a high-bone-regeneration capacity as recently demonstrated in large animal studies. [23]

Results and Discussion
To prepare the water-based PEGDA resin for DLP printing, the calcium magnesium phosphate cement (CMPC) powder was first wet milled for 16 h. The particle-size distribution of the milled CMPC powder (Figure 1a) shows a narrow distribution of particle diameters, ranging from 0.5 to 20 μm with the mean diameter D 50 = 3.45 AE 0.07 μm. D 10 and D 90 are 1.37 AE 0.06 and 8.25 AE 0.09 μm, respectively. The PEGDA resin was prepared at a PEGDA:water ratio (PWR) of 1:1, and a LAP content of 0.25 wt% and a CMPC fraction of 25-45 wt% were then added to the solution. Preliminary experiments showed that the viscosity of the cement-filled resin with 45 wt% loading was too high to enable processing by the DLP printer and leading to insufficient post flow when tilting the tank between each layer. Therefore, a maximum of 40 wt% cement powder within the resin was used for further printing experiments.
The results of the rheological study of the PEGDA resin containing 40 wt% CMPC powder are shown in Figure 1b indicating shear-thinning behavior, which is desirable for optimal spreading in the printer tray. [24] The upper viscosity limit for printable resins is %1000 Pa s for a shear rate of 0.01 1 s À1 and 3 Pa s for 10 1 s À1 , [25,26] whereas the PEGDA resin studied here resulted in approximately 250 Pa s (shear rate of 0.01 1 s À1 ) containing 40 wt% CMPC powder.
Printing experiments were initially performed on cuboid test specimen to optimize exposure time as a critical printing parameter to obtain precisely resolved structures ( Figure S1a, Supporting Information). Overall, the precision of the printing results of unloaded PEGDA resins improved significantly with increasing PEGDA content resulting in a PWR of 1:1 for further investigations, unless otherwise stated. Without the CMPC as photo absorber, the selection of the appropriate exposure times was crucial. Inaccurate exposure quickly leads to incomplete structures when using too short exposure times (Figure S1b, Supporting Information) or to samples with undefined dimensions due to too long exposure times ( Figure S1c, Supporting Information). Surprisingly, already in the case of the PEGDA resin containing 40 wt% CMPC powder, the cement particles acted as potent photo absorbers, which significantly reduced the risk of overcuring and resulted in an improved printing precision. However, too long exposure times lead to layer shifting www.advancedsciencenews.com www.aem-journal.com ( Figure S1d, Supporting Information) resulting in severe printing artifacts. This effect was further enhanced using a more viscous resin containing 50 wt% CMPC powder ( Figure S1e, Supporting Information). Since the addition of the cement powder increases viscosity of the resin, this high cement content of 50 wt% also led to unfinished layers due to the insufficient post flow when tilting the tank between each layer. Figure 2 displays structures printed with PEGDA resin with 40 wt% CMPC and optimized printing parameters. Figure 2a shows an osteotomy wedge with the corresponding CAD file in Figure S2a, Supporting Information. The dimensions of the printed structure precisely match those of the digital template ( Figure S2, Supporting Information). Figure 2b depicts a small cube with channels of various diameters used to determine the minimum printing resolution with the CAD file shown in Figure S2b, Supporting Information. Only the lower row with channel diameters of 1.3-1.6 mm is clearly recognizable within the cube. Smaller channels were found to be non-continuous through the sample and were only present as indentations on the sample surface. This was caused by slight overcuring of the structure, which can also be seen in the indentation in the upper right area of the cube. The hollow cylinder shown in Figure 2c and the corresponding CAD file in Figure S2c, Supporting Information, is also slightly overhardened, but this is particularly limited to the fine, inner struts. In contrast, the outer dimensions match more closely the 3D model. While the outer shape of the tooth shown in Figure 2d corresponds closely to the model ( Figure S2d, Supporting Information), the channels in the interior, however, are not present. Here, the fine structures could also not be resolved during printing due to overcuring. It can be concluded that the maximum resolution of the PEGDA resin with 40 wt% CMPC powder is about 1-2 mm for channels, while the overall dimensional accuracy was approximately 200-300 μm.
Here, a further improvement of the precision, especially for fine structures such as channels and struts, is necessary and might be achieved by both the use of an additional photo absorber to prevent overcuring as well as reducing resin viscosity and hence increasing its flowability. The latter is thought to be achieved by using a bi-or trimodal particle-size distribution enabling the use of thinner layers and leading to enhanced layer finish. [27] In addition, additives such as sodium citrate will increase the surface charge (Zeta-potential) of the particles to prevent particle agglomeration and settling. [28] This will also enable higher filler contents to increase the mechanical performance by reducing microporosity of the printed samples. [27] Figure 2e shows the Fourier transform infrared spectroscopy (FTIR) spectra of samples printed with PEGDA resin containing 40 wt% CMPC. Spectra of non-cross-linked PEGDA resin with 40 wt% CMPC, as well as PEGDA, served as reference. The absorption bands did match with the recorded spectra of distilled water and CMPC raw powder, and the absorption band at 1720 cm À1 is due to the C═O bond of PEGDA, which was present in all the samples. [29] In the wet samples, a strong O-H vibration of water at 1640 cm À1 is overlapping with the absorption band of the C═C double bond at 1620 cm À1 . [29] By freeze-drying, this water peak was eliminated, and it could be shown that also the absorption band of the C═C double bond present in PEGDA disappeared, demonstrating that cross-linking of PEGDA via a polymerization of double bonds was successful during printing.
The printed samples were stored in diammonium phosphate solution to initiate a setting reaction for up to 14 days. This had a strong effect on the mechanical performance of the samples as shown in Table 1, since the newly precipitated cement crystals  grow and interlock, resulting in greater consolidation over a longer period. While the addition of CMPC particles to the resin alone more than tripled the compressive strength and Young's modulus compared with the pure PEGDA resin, immersion in diammonium-hydrogen phosphate (DAHP) solution for 4 days increased the compressive strength by nearly one order of magnitude at the same PWR. The latter had also a strong influence on the mechanical performance. Here, the PEGDA content in the solution was increased stepwise from 20 to 50 wt%. The lower limit of 20 wt% PEGDA was chosen to ensure curing of the resin during DLP printing. Furthermore, slight inhomogeneities in the resin already occurred from 30 wt%, which led to phase separation at even lower proportions, since the CMPC particles were no longer evenly distributed in the resin, but sedimented during processing.
Compressive strength increased sharply up to a PEGDA content of 40 wt% and then dropped again slightly at a PWR of 1:1. The Young's modulus, in contrast, reached a maximum of 85.9 AE 3.1 MPa at 30 wt% PEGDA. The highest values were thus present for a PWR of 2:3, that is, 40 wt% PEGDA in solution, with the difference from the properties of the resin with a PWR of 1:1 being in the range of standard deviations.
The last section of Table 1 contains the specimens printed in two different directions and cured over time with a PWR of 1:1. As expected, the values of the specimens printed upright were significantly higher than those of the longitudinally printed specimens. The reason for this is that DLP printing is a layerby-layer fabrication process. Since the layers are oriented parallel to applied load for longitudinal printed samples, this facilitates fracture propagation along the layer boundaries during mechanical loading. In contrast, a perpendicular layer orientation (upright printed samples) increased compressive strength by a factor of %2 to a maximum of 17.3 AE 1.1 MPa. Although the Young's modulus increased slightly for the longitudinally printed specimens with a curing time of 14 days compared to the 4 days of curing, this change is not significant due to the high standard deviation. The large variation of the results is probably due to structural inhomogeneities caused by printing errors in these specimens. Overall, the values could thus be optimized by a high PEGDA content, the addition of cement particles, an extended curing time and the orientation of the layers perpendicular to the loading direction. Furthermore, as previously mentioned, a bi-or trimodal particle distribution and the addition of dispersant agents may further decrease the standard deviation and the inhomogeneities within the bioceramic loaded resin, and likely lead to improved mechanical properties of the printed samples by porosity reduction. [27] Thus, the mechanical properties of trabecular bone, which has a compressive strength of 4-12 N mm À2 and a Young's modulus of 100-500 MPa, were sometimes even exceeded. [30] The setting reaction in DAHP solution was monitored by X-ray diffraction (XRD) analysis ( Figure 3 and Table 2) by both analyzing the surface and bulk volume of the samples.
According to previous studies, setting should result in the formation of struvite and minor phases of brushite and newberyite. [31] Surprisingly, this could not be confirmed for the samples from the current study. While the diffraction patterns of the CMPC raw powder contained only stanfieldite (s) and farringtonite (f ) peaks, all three printed and post-hardened samples showed further peaks of a struvite (*) and whitlockite (!) phase. While the struvite peaks are clearly recognizable in the diffraction patterns, the correspondence with the whitlockite structure was only evident in the evaluation by Rietveld refinement analysis. In contrast, the expected by-products, newberyite and brushite, could not be detected.
As shown in Table 2, phase transformation during aging in DAHP occurred mainly at the surface and penetrated only slowly into the bulk of the specimens, as the total percentage of struvite and whitlockite phase in the ground powder samples increased from 2.8% and 1.7% after 4 days to only 5.7% and 2.1%, respectively, after 14 days.
However, XRD analysis of the surface indicated a struvite and whitlockite content of 14.9% and 2.5% even after 4 days immersion in DAHP. The reason for the formation of whitlockite as a minor phase during immersion in DAHP is unclear. A possible explanation is linked to the reaction conditions within the PEGDA matrix, whereas initially struvite is formed within the surface-near volume of the sample due to a high supersaturation with ammonium, phosphate, and magnesium ions. In contrast, the concentration of ammonium and phosphate in the inner volume is expected to be much lower, which is thought to reduce struvite precipitation and to favor the formation of less soluble whitlockite. Another explanation would be the conversion of struvite in the presence of calcium ions (derived from the CMPC powder), which was also observed after the heterotopic The porosity and relative pore volume of test specimens printed with PEGDA resin containing a 40 wt% CMPC before and after immersion in DAHP for 4 days is displayed in Figure 4a.
The overall porosity of the test specimens printed with PEGDA resin was 8.6% directly after printing, which decreased to 3.3% after 4 days of curing. The reason for this is the precipitation of struvite crystals during aging in DAHP, which have a lower density compared to the CMPC particles and hence occupy more space in the sample volume. Especially small pores with a diameter up to 1 μm seem to be closed, while new and larger pores with a size up to 100 μm are formed. Figure 4b-d shows scanning electron microscope (SEM) images taken directly after printing using PEGDA resin containing 40 wt% CMPC powder. Figure 4b shows the surface of the printed body at 2500x magnification, revealing the individual CMPC particles embedded in the surrounding PEGDA matrix and Figure 4c,d shows the side view of the printed structure indicating the individual layers of the DLP process. In contrast, the SEM image of the printed sample after hardening in DAHP appears more uniform due to the PEGDA mainly covering the crystals with only a few isolated crystals present at the surface (Figure 4e) potentially indicating sedimentation. Figure 4f-g shows the side few of the cured samples in DAHP indicating the single layers similar to the side view of the samples directly after printing (Figure 4c-d). However, the magnified side view in Figure 4g highlights the rougher surface of the sample side consisting of large struvite crystals growing during the hardening process. Overall, the SEM images support the findings of the previously discussed porosity and relative pore volume representing the formation of larger struvite crystals after hardening in DAHP solution.
Although the overall porosity decreases, the change in poresize distribution due to the aging in DAHP may have a positive effect on bone formation in vivo. Since the formation of mineralized tissue is only possible from a pore diameter of >1 μm, the closing of the smaller pores by cement crystals has no negative consequences. Furthermore, the formation of new, larger pores may enable the formation of lamellar bone by an ingrowth of blood vessels. [33] Overall, the porosity after aging was slightly below that of cortical bone, which is 5-13%. [34] However, the lower porosity of the cured structures may lead to observed higher mechanical stability and strength due to the higher density and homogeneity of the material. [35] 3. Conclusion CMPC powder dispersed within a PEGDA resin is a promising combination for DLP printing resulting in well-defined 3D structures. Due to the use of the biocompatible and water-based PEGDA in combination with the inorganic bioceramic powder, this resin enables the use for applications in bone regeneration. The addition of 40 wt% CMPC powder within the PEGDA matrix acted as photo absorber resulting in more defined structures compared to the structures fabricated using the PEGDA without CMPC powder. After printing, the CMPC can partially undergo a setting reaction in DAHP solution, which results in an adjustable increase of the mechanical performance. This is clearly an advantage compared with currently available resins including bioceramic powder, which usually require sintering steps post printing for densification including a large shrinkage of the structures. It even enables the adjustment of the resulting mechanical properties to those of trabecular bone.
Since the reaction product still contained pyrophosphate, two further sintering steps at 1050 and 1100°C for 5 h, respectively, were performed. Finally, the sintered cake was manually crushed, sieved >355 μm, and ball milled in ethanol for 16 h in a planetary ball mill (PM400, Retsch GmbH, Germany). The particle-size distribution was determined by laser diffraction in isopropanol (LA-300, Horiba Ltd.). A water-based resin was obtained by mixing 50 wt% PEGDA with water and 0.25 wt% LAP as photo initiator. This resin was mixed with the previously described cement powder at different solid contents between 25 and 45 wt%. Since the viscosity of the cement filled resin was too high at 45 wt% loading to enable a processing by the DLP printer, a maximum of 40 wt% cement powder within the resin was used for the printing experiments.
The flow behavior of the resins used was investigated using a rheometer (Physica MCR 301, Anton Paar GmbH). Shear viscosity was measured as a measure of flow resistance as a function of shear rate. Controlled shear rate tests were performed in plate-plate geometry, where the plate diameter was 25 mm, and the viscosity of the resins was measured with increasing shear rate in a range of 0.1-10 1 s À1 at 22°C.
DLP Printing: The structures were printed using a DLP printer (Original Prusa SL1, Prusa Research a.s.) and then post-processed in a curing and washing machine (CW1, Prusa Research a.s.). The structures for DLP printing were created using "Fusion 360" software (Autodesk Inc.) and output directly in the program as an image file. For the printing process, a layer height of 50 μm was set and the x-y resolution was 47 μm. Finally, the objects were post-cured in a polymerization device (Otoflash G171, NK-Optik GmbH) with light flashes in the wavelength range from 280 to 580 nm.
Except for the mechanical properties test specimens, the PWR of the water-based resin was always 1:1 and the LAP content was 0.25 wt%. A CMPC content of 40 wt% was then added to the solution. The printed structures were shaken out five times in a snap lid jar filled with distilled water. After drying, the objects were post-cured in the curing machine for 3 min under UV light and then post-cured in the polymerizer with 500 light flashes.
Hereafter, the structures were cured in 3.5 M DAHP solution for 4 and 14 days, respectively, in a water bath at 37°C. To prepare the solution, the DAHP was dissolved in distilled water for 1 h at 40°C with stirring.
Characterization Methods: The pore-size and volume distribution of the printed samples was determined using a mercury porosimeter (Pascal 140 and 440, Thermo Fisher Scientific Inc.). In this process, a vacuum was first created in the sample vessel to degas the sample. The vessel was then filled with mercury to a specified volume level. During analysis, the pressure was increased from the vacuum, forcing the mercury into the pores. Based on the changed volume level in the dilatometer, the pore size and pore volume could thus be determined. Of the test specimens printed with PEGDA resin, one sample after printing and one cured for 4 days in DAHP were examined.
A FTIR spectrometer (Nicolet iS10, ThermoFisher Scientific Inc.) was used for FTIR measurements on the polymerization of the PEGDA matrix. The aim of the measurement was to verify the cross-linking of the PEGDA resin obtained by DLP printing through polymerization of the C═C double bonds. For this purpose, the spectra of a mortared printed test specimen of PEGDA resin containing 40 wt% CMPC were recorded after printing and after freeze-drying in the range 4000-600 cm À1 with a resolution of 4 cm À1 . For drying, the body was first placed in liquid nitrogen for 15 min and then the water was removed from it for 72 h in a freeze-dryer (Epsilon 2-4 LSCplus, Martin Christ Gefriertrocknungsanlagen GmbH). For comparison, the spectra of CMPC crude powder, PEGDA resin containing 40 wt% CMPC, distilled water, and PEGDA were also measured.
To investigate the influence of posttreatment on the phase composition of the printed structures, they were cured in 3.5 M DAHP in a water bath at 37°C for 4 days. All test specimens were then ground into a fine powder, or a flat cylindrical test specimen was used as a solid for surface analysis. The CMPC raw powder was also analyzed as a reference measurement. The samples were analyzed using an X-ray diffractometer (D8 Advance DAVINCI Design X-ray Diffractometer, Bruker AXS GmbH) with CuKa radiation at an accelerating voltage of 40 kV and a current of 40 mA in an angular range of 2θ = 7-70°. Each measurement was performed in triplicate. The resulting diffractograms were evaluated for qualitative and quantitative compositional analysis using "Diffrac.EVA" and "Diffrac.TOPAS" software (Bruker AXS GmbH). Peaks were analyzed using the International Centre for Diffraction Data (ICDD) database, checking matches with reference samples of stanfieldite (PDF 11-0231), farringtonite (PDF 33-0876), struvite (PDF 15-0762), and whitlockite (PDF 42-0578).
The mechanical properties of the printed specimens were investigated in a compression test setup. The compressive strength and Young's modulus were measured using a testing machine (Z010, ZwickRoell GmbH & Co. KG) with a 10 kN load cell, where the preload was 1 N and the testing speed was 1 mm min À1 .
The surface of the printed samples was examined using a (SEM, Gemini Crossbeam 340, Zeiss GmbH). Before microscopy, the specimens were glued to a specimen holder with Leit-C and sputtered with a 4 nm thick platinum layer in the coating device (Leica EM ACE600, Leica Microsystems GmbH).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.