Tough, Resorbable Polycaprolactone‐Based Bimodal Networks for Vat Polymerization 3D Printing

Vat polymerization allows for the accurate and fast fabrication of personalized implants and devices. While the technology advances rapidly and more materials become available, the fabrication of flexible yet tough resorbable materials for biomedical applications remains a challenge. Here, a formulation that can be 3D printed with high accuracy using vat polymerization, yielding materials that are tough, degradable, and non‐toxic is presented. This unique combination of properties is obtained by combining a long‐chain polycaprolactone macromonomer with a small molecule cross‐linker. A wide range of properties is achieved by tuning the ratio of these components. The use of benzyl alcohol as a non‐volatile, benign solvent enables fabrication on a low‐cost desktop 3D printer, with an exposure time of 8 s per 50‐micron layer. The 3D‐printed networks are tough and elastic with a tensile strength of 11 MPa at 116% elongation at break. Cells attach and proliferate on the networks with a viability of >91%. The networks are fully degradable to soluble products. This new 3D printable material opens up a range of opportunities in biomedical engineering and personalized medicine.

have been developed, vat polymerization (VP), which includes stereolithography (SLA), digital light processing (DLP), and 2-photon polymerization, is acclaimed for its high resolution, reliability, and large freedom of design. [2,3] The main drawback of VP is the specific need for a photocurable liquid resin that solidifies at an acceptable rate into a solid that is strong enough to withstand the mechanical forces applied during the printing process. Although the palette of available resins is steadily growing, [4] obtaining high print fidelity while maintaining certain properties is still challenging. These properties include toughness, extensibility, biocompatibility, and biodegradability; and particularly the combination of them. [5,6] The most common strategy for formulating a biodegradable VP resin is to end-functionalize hydrolyzable oligomers of molecular weight typically <4 kg mol −1 with reactive groups, obtaining macromonomers that can be cross-linked into a network through photo-initiated polymerization. Oligomers typically consist of lactide, [7] caprolactone, [8,9] ethylene glycol, [10] polyoxazoline, [11] trimethylene carbonate [12] or poly(propylene fumarate). [13,14] To obtain a liquid resin out of these solid macromonomers, reactive or non-reactive diluents are often employed, or heat.
The formulation of resins for VP is usually a balance between reactivity and final network properties. Final networks tend to be brittle because of high cross-link densities and a lack of energy-dissipating mechanisms. One approach to improve toughness is to build bimodal networks from two distinct precursors; one of higher and one of lower molecular weight. This concept, pioneered by Mark with silicone rubbers [15] has been applied to VP by Van Bochove et al using poly(trimethylene carbonate) macromers with high (10300 g mol −1 ) and low (700 g mol −1 ) molecular weights. [12] In this work, we apply this concept for the first time to polycaprolactone (PCL) for DLP 3D printing. We hypothesized that the combination of a long chain macromonomer and a small multifunctional cross-linker (XL) would give adequate reactivity for DLP 3D printing, as well as improved mechanical properties. To investigate this hypothesis, PCL-dimethacrylate with molecular weight 10 kg mol −1 and pentaerythritol tetraacrylate (PETA) were combined in various ratios, and used with a low-toxicity, low vapor pressure solvent to print on budget DLP apparatus at room temperature. Besides printability, thermomechanical properties, degradation, and cytotoxicity were assessed.

Resin Formulation
PCL-dimethacrylate (PCL-2MA) macromonomers were obtained by quantitatively converting the terminal hydroxyl groups on 10 kg mol −1 PCL to methacrylate groups, as confirmed by 1 HNMR spectroscopy ( Figure S1, Supporting Information). A high degree of functionalization was obtained (99%+), and offered a yield of 80-90%. Photocurable resins were formulated from PCL-2MA macromer, PETA cross-linker, diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide (TPO) photoinitiator and a non-reactive diluent. For selecting a diluent, 9 candidates were shortlisted from a range of 50 potential solvents for PCL. [16] The 41 solvents excluded were deemed too harmful to health, too volatile for use in an open vat 3D printer, or incompatible with the resin tank. Of the 9 most promising solvents, benzyl alcohol outperformed all others for its ability to dissolve 10 kg mol −1 PCL oligomer, while also being one of the least volatile as indicated by a low vapor pressure of 0.09 mmHg at ambient temperature ( Figure S2, Supporting Information). Hence, benzyl alcohol was selected for further use.
In selecting relative amounts of PCL-based macromer and cross-linker (XL) for our resins, we considered the reactive double bond concentration in the formulated resins (important for reactivity), as well as the resulting cross-link density in the final network (important for mechanical properties). Table 1 shows the selected resin compositions, with their corresponding reactive double bond concentrations including methacrylates from the PCL-2MA macromer and acrylates from the cross-linker indicated by [CC] in mmol g −1 of resin. M c,n and M c,w represent the hypothetical number-average and weightaverage molecular weights between cross-links in the final network, assuming complete double bond conversion. Compositions were selected to obtain a logarithmic spread of values for M c,n with an approximate factor of 3 between subsequent groups. While the double bond concentration scales comparably with the M c,n , the values for M c,w show a very different distribution, with relatively high values even at moderate additions of XL. This forms the basis of our hypothesis that good mechanical properties may be obtained without compromising resin reactivity, by using a combination of low-and high-molecularweight (macro)monomers.

Gel Properties of the Networks
When photo-cured in moulds, all resins yielded sturdy rubbery gels apart from pureXL, which cured into a hard, brittle glassy polymer network. For all resins, inclusion rates of macro monomer into the networks was high as expressed in their gel content values of >89%. Upon free swelling in benzyl alcohol, loosely cross-linked noXL networks increased 19 times in mass, while the densely cross-linked pureXL networks showed swelling of only ≈1%. Upon extraction of unreacted components (predominantly non-reactive diluent) and drying, shrinkage occurred in an isotropic, reproducible, and hence predictable manner. All bimodal network discs shrunk by ≈25% in a both thickness and diameter. This corresponds to 58 vol.% and therefore is consistent with the 54.5 wt.% of benzyl alcohol that was used in these resins, as benzyl alcohol has a somewhat lower density than both polymer components. As the noXL networks contained additional sol fraction, their shrinkage was slightly higher. The pureXL networks showed negligible dimensional shrinkage.
Infrared spectroscopy confirmed the complete conversion of reactive double bonds in the noXL and lowXL networks ( Figure 1). Networks with higher cross-linker contents showed unreacted double bonds at 1634 cm −1 (CC stretching vibration from both the PCL-bound methacrylate and the acrylate in the PETA cross-linker), with double bond conversion in the range of ≈50-70% (Table 1). The spectra of PCL-2MA macromer, PETA cross-linker, and complete middleXL resin are given for reference in Figure S7 (Supporting Information). Resin compositions range from PCL-2MA only (no XL) to XL only (pure XL, i.e., no PCL-2MA) before addition of TPO photoinitiator (1 wt.% for casting and molding or 5 wt.% for VP printing). M c,n and for M c,w are the hypothetical number-and weightaverage molecular weight between cross-links in the final networks at 100% double bond conversion, respectively. Gel content is the weight fraction of reactive materials (excluding diluent) incorporated in the network (after extraction). Swelling ratios are mass ratios between the equilibrium swelling state in benzyl alcohol and the extracted and dried state. 1-Directional shrinkage is based on diameter and thickness of extracted and dried discs compared to as-cured or "green" discs (including benzyl alcohol and unreacted resin). Double bond conversion determined from the ratio of CC to CO peak intensity in FTIR spectra ( Figure  1 and Figure S7, Supporting Information) of the networks compared to their corresponding resin (middleXL and pureXL, n = 3) or to an interpolated value between these two (highXL, [n = 1]). For all measurements of gel content, swelling ration and shrinkage (n = 10).

Thermomechanical Properties
Dynamic mechanical analysis was performed to reveal the influence of XL content on thermal transitions and stiffness at relevant temperatures (Figure 2 Table 1. Figure 2. Thermomechanical properties. a) Temperature-dependent storage modulus (E') of photo-cured networks recorded using DMTA in tensile mode. Glass transitions are less pronounced for higher XL content, and absent in the pureXL network. A second decrease corresponding to melting is seen for noXL and lowXL only. Representative curves from n ≥ 3. b) Stress-strain curves measured in a static tensile test at 37 °C. Representative curves from n ≥ 6. c) Elasticity of middleXL networks at 37 °C. Permanent set/unrecovered strain was recorded 15 min after unloading at incremental strains. All groups are statistically different (One-way ANOVA with Games-Howell post hoc, p < 0.05) except for 1% and 5% strain (n = 3).
Only the noXL and lowXL networks showed a second drop in storage modulus (59.3 ± 7.8 °C for noXL and 46.6 ± 8.1 °C for lowXL, tan δ peak), indicating the presence of a crystalline phase in these materials. This was corroborated by their opaque appearance (while the networks with higher XL content were transparent) as well as by DSC measurements ( Figure S3, Supporting Information). With a melting temperature of 63.7 °C and melting enthalpy of 87.9 J g −1 , the noXL network showed almost as high crystallinity as non-cross-linked, high-molecularweight PCL (63.5 °C and 95.1 J g −1 , respectively). The addition of XL significantly reduced the ability of the PCL chains to crystallize. The lowXL network still showed a melting temperature of 54.9 °C and melting enthalpy of 64.6 J g −1 , but none of the networks with higher XL content showed any crystallinity at all.
The XL content appeared to strongly influence the properties at physiological temperature, including modulus ( Table 2) and ultimate properties (Figure 2b). The Young's modulus was ≈0.2 GPa for the semi-crystalline noXL and lowXL networks, while for amorphous middleXL this was approximately halved. By further increasing the XL content the modulus showed a sharp increase, with the highXL and pureXL networks both being glassy at 37 °C. Although the networks containing high levels of XL showed high strength, their low elongation at break revealed their brittleness, expressed in their low toughness values ( Table 2). The noXL and lowXL networks typically yielded at around 12% strain and 8 MPa stress, followed by necking and further elongation exceeding the extensibility limits of the DMTA without reaching a breaking point. Static tensile tests were also performed on a universal tensile tester, confirming the mechanical behavior of all networks seen in DMTA (Table S1, Supporting Information).
In contrast to all other compositions, the middleXL did not fail (neither yield nor break) at low strain, but stretched to more than double its original length before breaking. As a result, toughness values (area under the stress-strain curve) were 9x and 65x that of highXL and pureXL, respectively. We set out to investigate to which extent this high strain was reversible. To this end, tensile testing at incremental strain values was performed ( Figure 2c). The unrecovered strain (also termed permanent set) was insignificant after release of up to 5% strain, and small but significantly higher for higher strain values. Even after straining to 80% (corresponding to ≈8 MPa of tensile stress), 68.8% of the strain was recovered elastically 15 min after unloading, indicating tough elastic behavior.

Reactivity of the Resins
The reactivity of the resins was assessed through photorheometry, which measures the increase in shear moduli as the photo-cured network develops. The shear storage modulus of the solvent-swollen gel gives an indication for the stiffness of the as-built "green part" in VP 3D printing. The time taken to reach the gel point (expressed as the cross-over point where the storage modulus surpasses the loss modulus) after the light was switched on decreased rapidly with increasing XL content, indicating a higher reactivity (Table 3; Figure S4  Estimated using rubber elasticity theory [17] : M c = RTρ/E'; b) Values at yield point, sample did not break; c) Extensibility limits of the DMTA instrument reached, sample did not break; d) Samples were extremely brittle and broke during temperature sweep, before a rubbery plateau was reached a considerable effect on resin reactivity. The time taken to reach 95% of full cure was less affected, with the noXL, lowXL, and middleXL showing similar values. As expected, the viscosities of the uncured resins decreased upon addition of XL (a low-viscous liquid) and are all within the range of commonly reported viscosities for DLP resins. [18] While photo-rheometry gives a relative comparison of reactivity between different compositions, a more pragmatic assessment of resin reactivity in VP printing is obtained through composing a working curve. This shows the thickness of a cured layer as a function of exposure time in a VP 3D printer as can be seen for the middleXL resin in Figure 3. By extrapolation to layer thickness zero (i.e., the gel point) a critical time of 2.53 s was found. The slope of the working curve represents the penetration depth D p ; the value of 95.6 µm is deemed appropriate for DLP printing with layer thickness of 50-200 µm. [19]

Four Different Measures for Cross-Link Density
The cross-link density of polymer networks can be measured, and expressed, in various ways. Figure 4 shows measures obtained from photo-rheometry on solvent-swollen networks (maximum storage modulus), dynamic tensile testing on extracted networks (Young's modulus at rubber plateau), and solvent-induced equilibrium swelling (mass swelling ratio). All show a stark influence of XL content, which is consistent between the different experimental modalities. The equilibrium swelling ratio correlated with the hypothetical molecular weight between cross-links through a power law: Q = 0.077 . M c,n 0.6 with coefficient of determination R 2 equal to 0.997, highlighting the predictable swelling based on resin composition.

Vat Polymerization 3D Printing
Encouraged by the results of the mechanical and reactivity assessment, we set out to test printability on an unmodified lowcost desktop DLP 3D printer (Autodesk Ember). As the highXL networks showed stiff and brittle behavior similar to many commercially available VP resin materials, we focused our attention to the middleXL resin. Several models were attempted, all resulting in well-formed structures ( Figure 5). Small features can be discerned, such as the 0.3 mm curved strut size for the gyroid model (Figure 5a), sharp-edged 0.15 mm pores for the Voronoi tower (Figure 5c), and 0.15 mm size struts for the tracheal stent (Figure 5f). Individual layers can be seen at high magnifications (Figure 5b,f); these were 50 µm thick during fabrication, and ca. 33 µm after removal of benzyl alcohol along with unreacted components. The design files for the tracheal stents depicted in Figure 5d-f were collegially made available by researchers at ETH Zürich who previously printed these designs in poly(D,L-lactide-co-ε-caprolactone) resins using hot lithography and demonstrated their usefulness in a rabbit model. [20] The limits of printability of this degradable material on a low-cost desktop 3D printer were further investigated through the printing of test parts containing spans, through-holes, and embossed features ( Figure S5, Supporting Information).
Attempts at printing with the lowXL resin were less successful; parts started to form but either detached from the build head or broke from the shearing forces during the peel action in-between layers.

Accelerated Degradation
To assess whether the photo-cured networks are prone to hydrolytic degradation, studies were performed by boiling polymer   discs in 1 m NaOH, and assessing their mass loss over time (Figure 6). Similar to the mechanical behavior, three groups could be discerned showing distinct characteristics. First, pureXL discs showed only 39% mass loss at 165 h (1 week), with the solution turning cloudy indicating non-degraded material shed by the discs; hence mass loss was due to disintegration rather than degradation. On the contrary, all networks that included PCL macromer fully degraded within a much shorter time. The degradation occurred via surface erosion, with the discs retaining their structural integrity until near-complete mass loss, while their thickness gradually decreased. Importantly, a clear solution was observed for these networks shortly after the discs lost their integrity, indicating complete degradation and dissolution. The middleXL and highXL networks lost mass rapidly, with the discs being completely degraded in 7.5 and 6 h, respectively. The noXL and lowXL networks showed a similar linear loss profile, albeit considerably slower. Their solutions were observed to be clear within 55 and 50 h, respectively.

Cytotoxicity
The potential toxicity of the middleXL material was assessed using two cell lines. First, networks of middleXL were incubated in cell culture media for 1 week. Then, this media was added to a monolayer culture of MC3T3-E1 pre-osteoblast cells for 24 h, after which metabolic activity was assessed using resazurin (Figure 7a). The metabolic activity of cells exposed to incubated media was indistinguishable for mid-dleXL, high molecular weight PCL (HMW PCL) reference material, and the control that was media incubated without material indicating no toxic components had leached from either of the polymer materials. The aging of media during incubation caused a small but insignificant decrease in metabolic activity compared to when using fresh media, ascribed to a loss in potency of growth factors and other serum components.
MC3T3-E1 cells were also used in a direct contact test. Cells were seeded and cultured for up to 11 days on discs prepared from middleXL, HMW PCL, on glass coverslips or in a tissue culture polystyrene (TCPS) well plate as positive controls. The cells exhibited their normal fibroblastic phenotype; adhering, spreading and colonizing the surface of all materials by day 7 (Figure 7b top row). Live/dead staining on these cells, followed by measurement of green (live) and red (dead) area yielded a measurement of viability of adherent cells ( Figure S6   area was green on both day 1 and day 3. For the middleXL material, the relative area stained in green was 91% and 96% on days 1 and 3, respectively, indicating high viability. The metabolic activity of adhered cells was measured with resazurin at 3 different time points. After three days of culture, the metabolic activity of cells on middleXL was lower than on TCPS (p = 0.006), most likely due to a lower seeding efficiency or delayed adherence (Figure 7c). Over the 11 days of culture there was an increase in metabolic activity on all materials due to cell proliferation, though on middleXL this increase was small with low statistical significance (p = 0.048).
The direct contact test was also performed using a mouse macrophage cell line (J774A.1) to gain insights into the innate immune response to the material. These cells are semiadherent and showed the same normal, rounded morphology on all materials and glass control (Figure 7b, bottom row). The metabolic activity of the macrophages cultured on middleXL for 3 days was slightly but significantly lower than the TCPS control (p = 0.03). However, after 7 days the metabolic activity on middleXL was higher than on TCPS, albeit not statistically significant (p = 0.13). This indicates the high viability of these cells as well when exposed to the 3D printable PCL-based polymer network. Over the 7 days of culture, there was a slight increase in metabolic activity for this cell type as well on all of the materials, although the difference was of low statistical significance for middleXL (p = 0.0502).
Adv. Funct. Mater. 2023, 33, 2213797 Figure 7. Cytotoxicity assessment of the printable biomaterial. a) Viability of MC3T3-E1 pre-osteoblast cells (measured using resazurin) exposed to either fresh media, or to media incubated with middleXL, HMW PCL, or with no material (incubated media) (n = 4). Results are normalized to the fresh media control. The statistical significance of the differences observed was analyzed via one-way ANOVA with Tukey's test. b) Fluorescence microscopy images of MC3T3-E1 and J774A.1 macrophage cells cultured for 7 days on middleXL, HMW PCL, and glass coverslips. Blue = cell nuclei (DAPI), green = actin (phalloidin); scale bars are 100 µm. Metabolic activity of c) MC3T3-E1 (n = 5) and d) J774A.1 (n = 6) cells cultured on middleXL, HMW PCL or TCPS control for up to 11 days. Results are normalized to surface area (mm 2 ) to account for differences in disc size. Cells lysed with Triton were used as a death control in all assays. The statistical significance of the differences observed was analyzed by using an unpaired t-test.

Discussion
In this work, we present a new resorbable, photocurable biomaterial. We demonstrate its high print fidelity in DLP 3D printing, its outstanding mechanical properties, as well as its non-toxicity and ability to be hydrolyzed. To our knowledge, this material (middleXL) presents a unique combination of high strength and stretchability that is unprecedented in biomaterials for light-based 3D printing. Additionally, the identification of a suitable non-reactive diluent enabled 3D printing on an unmodified low-cost desktop SLA with minor environmental or health concerns.
Previous demonstrations of PCL-based resins required modifications such as a heated resin tray, [8,9] and/or the use of low molecular weight precursors [8,21] that compromised the mechanical properties of the final networks. Elomaa et al. demonstrated high print fidelity using a resin based on PCL macromers of 1.5 kg mol −1 . [8] However, the tensile strength of 2.55 MPa and 19% strain at break were respectively 4.4× and 6.0× lower than for the new material presented here, making the new middleXL material 27× tougher. Farzan et al. reported a much higher strain at break of 105% using a PCL-based polyurethane macromer in combination with a diluent printable at room temperature, but at 1.6 MPa the tensile strength, and hence toughness, was still nearly an order of magnitude below that of middleXL. [22] In the same study, the researchers managed to double the strength and triple the toughness of their networks by including poly(ethylene glycol) in their macromer, yet still achieving less than half of what is demonstrated in this work. In fact, a literature search on DLP printable flexible resorbable biomaterials showed almost all were less tough (measured by area under the curve equals energy absorbed in a tensile test) than middleXL; [14,[23][24][25] only networks containing high-molecular-weight macromers requiring customized heated platforms showed higher values. [12,20,26] These references exclude hydrogels, which are much more brittle by nature. Mechanical properties, particularly toughness, are typically quite poor for DLP printed materials compared to engineering plastics and thermosets processed otherwise, and are often under-reported in scientific literature. As hypothesized, the combination of high-molecular-weight macromers and a small cross-linker molecule yielded tough networks, with the cross-linker also contributing to high reactivity that aided 3D printing; 8 s curing time per 50 µm layer is relatively fast for resorbable resins consisting of functionalized polyester oligomers, and only marginally slower than most non-resorbable resins that have a higher double bond content.
Unsurprisingly, increasing the cross-linker content of the resins resulted in higher maximum shear storage moduli of the solvent-swollen gels, increased tensile storage moduli of the extracted solvent-free network at the rubbery plateau, and a decrease in equilibrium swelling ratio. These bulk analysis methods all reveal effects of increasing cross-link density, however they do not unravel local network topology and chain dynamics. These may be examined using methods such as double quantum NMR [27] or x-ray scattering. [28] Perhaps one of the most striking of our findings is that networks containing high levels of PETA cross-linker were readily hydrolyzed in an accelerated degradation study, even faster than networks consisting solely of PCL macromers. The susceptibility of ester bonds to hydrolysis varies greatly with the local chemical environment, dictating the accessibility of water molecules to access the cleavage site regardless of the local water concentration. Whereas, for example, glycolic acid esters in hydrophobic glassy poly(lactide-co-glycolide) are readily cleavable, hydrogels of methacrylated poloxamer show no decline in stiffness after 40 weeks in PBS at 37 °C despite being highly hydrated. [29] Furthermore, when boiling networks prepared from methacrylated poly(D,L-lactide) in 1.0 m NaOH even for 3 weeks, the resulting polymethacrylate chains were still decorated with one lactic acid unit on each monomer, connected through a stable ester bond. [30] As PETA itself, and most formulations it is used in, form dense glassy networks virtually impenetrable to water, hydrolytic degradation is not typically observed. However, the possibility for the ester bonds in PETA to be hydrolytically cleaved in a suitable local chemical environment has been irrefutably demonstrated in thiol-ene networks containing PETA and a non-degradable dithiol. [31] Interestingly, the networks containing relatively high amounts of cross-linker degraded about 8x faster than those with little or no cross-linker. This cannot be explained by the crystallinity of the latter networks, as the degradation experiment was performed above their melting temperatures. Rather, it is thought that the presence of PCL facilitates the uptake and access of water to the cleavable ester bonds in PETA, as well as those in PCL itself. Accelerated conditions strongly alter the progression of degradation; PCL degrades by surface erosion at pH > 11 whereas it proceeds through bulk degradation at intermediate pH, including in vivo. [32,33] Therefore, the main outcome here is not the observed rates of degradation, but the fact that degradation into fully water-soluble components was achieved for all networks apart from pureXL. Accelerated degradation under acidic conditions proceeds slower than under alkaline conditions and more throughout the bulk, thus may give a better representation of relative degradation rates. [32,34] Alternatively or in combination, degradation can be accelerated while preserving the degradation mechanism by increasing the temperature, [35] particularly if no thermal transitions (T g or T m ) exist between the test temperature and body temperature. [36] However, biodegradability should eventually be confirmed in an in vivo model relevant to the intended biomedical application. Interestingly, the addition of 1 part PCL-2MA to 2 parts PETA (by weight) in highXL rendered the resulting network degradable under accelerated conditions, as well as more deformable and tougher.
The cytotoxicity of the printable middleXL material was assessed using the pre-osteoblast cell line MC3T3-E1 and the macrophage cell line J774A.1. Cell lines are often used to assess the toxicity at early stages of material development. [37] The extract test performed following the methodology described in ISO10993 reliably showed that no toxic compounds had leached from the polymer. This was expected as the material had been thoroughly extracted using a Soxhlet apparatus. To complete the assessment, direct contact tests were conducted by seeding the cells on polymer discs made with the middleXL formulation. In corroboration with the extract test, no evidence of toxicity was detected, although there were some differences in the growth rate when compared with standard tissue culture plates, especially at day 11. This was presumably due to the optimized surface coating on TCPS compared to the PCL-based samples that require further treatment to improve cell attachment. [38] As the J774 cells of the macrophage cell line as are semi-adherent, the surface properties should have less of an influence on the cell growth, which was observed in Figure 7d. While cell lines were used in this study, ultimately for the fabrication of medical devices, the biocompatibility of the material should be investigated with primary cells in vitro and by implanting the devices in vivo to evaluate the physiological responses elicited in the body.

Conclusion
Tough, bimodal polycaprolactone-based networks were 3D printed at room temperature on an unmodified low-cost desktop DLP 3D printer. The linear 10 kg mol −1 PCL-2MA macromer was dissolved in a benign solvent, benzyl alcohol, and mixed with pentaerythritol tetraacrylate cross-linker to enhance reactivity. The resulting networks were shown to possess superior mechanical properties to previously DLP-printed PCL networks, and to many resorbable vat polymerization materials. The networks were further shown to fully degrade under accelerated conditions, and nontoxic to cells either through leachables or in direct contact.
Synthesis: The PCL diol oligomer was functionalized by reacting the terminal hydroxyl groups with methacrylic anhydride under a nitrogen atmosphere at 130 °C in the presence of K 2 CO 3 as a proton scavenger and small quantities of anhydrous toluene added to decrease viscosity and achieve smooth stirring. A molar excess of 50 mol% MAAh and K 2 CO 3 was used. Proton-nuclear magnetic resonance spectrophotometry ( 1 H-NMR, CDCl 3 , Bruker AVIII 300 MHz) was used to determine the degree of functionalization of the macromer (PCL-2MA). The reaction was continued until a high degree of functionalization (99%+) was reached (≈3 days), after which the macromer was precipitated from pre-chilled hexane (−80 °C), and filtered. The yield was 80-90%.
Sample Preparation: Photo-curable liquid resins were formulated by mixing PCL-2MA macromer and PETA cross-linker in various ratios, with a fixed amount of 55 wt.% benzyl alcohol (5 wt.% for pureXL) and 1 wt.% of TPO photo-initiator. Strips were formed by casting resin on a microscope slide with a casting knife (Elcometer 3580) set at 400 µm. The slide was positioned within an Autodesk Ember 3D printer. A print file with five strips of 5 × 50 mm was prepared within Autodesk Print Studio and loaded into the printer. The resin was illuminated for 60 s, after which the slide was removed, and the strips lifted off gently using a scalpel. Any superficial excess resin was removed from the strips with a paper towel, before post-curing for 90 min (45 min each side) within a UV cabinet (UVP CL-1000 L, 365 nm, 3 mW cm −2 ). Discs (6 mm Ø, 1.6 mm thick) were prepared from the same resins by injection into custom-made PTFE molds and photo-curing in the UV cabinet in the same manner as the strips. Both strips and discs were extracted in isopropanol for 3 days within Soxhlet apparatus and dried in an oven at 80 °C until achieving a constant mass.
Materials Characterization: Gel properties were obtained by weighing discs directly after photo-curing (m 0 ), at equilibrium swelling (at least 24 h in benzyl alcohol) after removal of excess solvent from the surface (m sw ), and after extraction in IPA followed by drying at 80 °C until a constant weight was achieved (m dry ). Gel content (%) was calculated as m dry /m 0 with correction for benzyl alcohol content of the non-extracted discs, and swelling ratios were calculated as m sw /m dry . Shrinkage reported is the average of the decrease in thickness and diameter of the discs from the as-cured state to the extracted and dried state. Infrared spectroscopy was performed in Attenuated Total Reflectance (ATR) mode on a Thermo Scientific Nicolet iS5 FTIR Spectrometer with DTGS KBr detector applying Happ-Genzel apodization (16 scans). A TA Instruments Q800 Dynamic Mechanical Thermal Analyser (DMTA) was employed for dynamic and static testing. Oscillatory tensile testing (frequency = 1 Hz, amplitude = 15 µm) was performed while the sample was heated from −100 to +100 °C (+160 °C for highXL and pureXL) at 10 °C min −1 . Static tensile testing was performed through a force ramp at 3 N min −1 at 37 °C. The elastic strain recovery of the middleXL strips was assessed at 37 °C through a series of incremental strains (1, 5, 10, 20, 40%) applied at 10% min −1 each followed by a 15 min recovery period in-between steps. Differential scanning calorimetry (DSC) was performed on extracted and dried network samples (ca. 10 mg) using a DSC 204 Phoenix (NETZSCH, Germany) with aluminum pans and pierced lids, under a nitrogen flow. Two consecutive heating ramps from 0 to 100 °C at 5 °C min −1 were applied, with 5 min hold time at the extremes and cooling at 10 °C min −1 in-between. Melting points were determined as the temperature corresponding to the maximum heat flow, while melting enthalpies were derived from integrating the melting peaks with respect to baseline using NETZSCH Proteus software.
Photo-rheometry was performed on an Anton Paar MCR302 rheometer with quartz lower plate allowing to illuminate the sample area with 365 nm UV using an Omnicure S2000 spot UV curing system (Excelitas Technologies Corp.) that resulted in an irradiance of 0.76 mW cm −2 at the sample contact surface of the quartz plate. Irradiance measurements were performed using a Radiometer R2000. Additionally, use of a Peltier hood provided fine temperature control, and protected the sample from ambient light. With this setup, the sample was irradiated uniformly while being subjected to constant oscillatory small amplitude strain. A parallel plate of 15 mm diameter was used as the upper geometry with a gap size of 0.05 mm. All measurements were performed at a controlled temperature of 25 °C.
Resin gelling behavior was monitored throughout a time sweep where constant oscillations were applied at a fixed frequency of 40 rad s −1 with a target strain of 0.3%, which was in the linear viscoelastic region of the materials. These shear conditions were maintained for between 30-60 s prior to turning on the UV light source to establish baseline moduli. The gel point was considered to reside at the time where the storage www.advancedsciencenews.com 2213797 (10 of 11) © 2023 The Authors. Advanced Functional Materials published by Wiley-VCH GmbH (G′) and loss moduli (G″) were equal providing G′ was increasing over several orders of magnitude.
Vat Polymerization 3D Printing: For use on the Autodesk EMBER, resins described above were further supplemented with 0.2 wt.% 2,5-bis(5-tert-butyl-benzoxazol-2-yl)thiophene (BBOT) as photoabsorber and 0.1 wt.% hydroquinone as an inhibitor. The concentration of TPO photo-initiator was 5 wt.% for printing; for discs and strips described above the larger thickness combined with the high extinction coefficient of TPO prohibited the use of 5 wt.%, hence 1 wt.% was used there. The working curve was compiled by curing drops of resin on a microscope slide for set times, removal of excess uncured resin, and measuring the thickness of the solidified layer using a Mitutoyo dial thickness gauge.
The gyroid design was generated as previously described, [4] the CAD files for the tracheal stents were kindly provided by researchers at ETH Zürich, [17] while that of the Voronoi tower was downloaded from Thingiverse (https://www.thingiverse.com/thing:24123; originally uploaded by user @Dizingof) under CC BY 4.0 license. Other test parts were designed using Autodesk Fusion CAD software.
Printing was performed on an Autodesk Ember DLP 3d printer with polypropylene resin tray, using 20 s for the first layer, 13 seconds for 10 burn-in layers, followed by 8 s for model layers, with a z-step height of 50 µm. The separation slide velocity was also reduced to 1 rpm for all layer types. Printed parts were washed first in benzyl alcohol to remove unreacted resin, then in IPA, and finally air-dried prior to visualization using a Leica Stereozoom S9D stereomicroscope and GXCAM-U3 Series 5MP USB-3 Superfast, C-Mount Microscope Camera + GXCapture-T Software.
Degradation: A solution of 1 M NaOH was prepared, of which ≈30 mL was added to a 50 mL wide neck round bottom flask (RBF) placed within a heating block on a hotplate with a condenser. The solution was brought to boil prior to addition of the discs to be degraded. Discs were arranged into pairs, their initial combined weight taken, and added to the boiling solution. The discs were removed at each timepoint, dabbed dry of superficial water, and weighed together, before being added back to the solution. This was repeated until no measurable mass could be obtained. This process was repeated for replicates across the different material groups and the NaOH solution replaced each time.
Cytotoxicity: All incubations were performed at 37 °C and 5% CO 2 . Prior to testing, polymer discs and glass coverslips were sterilized by immersion in 70% ethanol for 20 min, drying in a laminar flow cabinet and washing three times with PBS. For the extract test, discs (n = 4) were then incubated in 250 µL of complete media for six days. MC3T3-E1 cells were seeded on a 96-well plate at a density of 70 cells mm −2 . After two days of culture, the media was replaced by the supernatants containing the leachables from the polymer. The cells were incubated in this media for 24-48 h and a resazurin assay was performed to assess cell viability.
For the direct contact tests, discs were incubated in complete media at 37 °C overnight to allow adsorption of serum proteins. Cells were then seeded at 200 cells mm −2 for MC3T3-E1 (n = 5) and 2000 cells mm −2 for J774A.1 (n = 6). Metabolic activity was quantified using resazurin after 3, 6, and 11 days of culture for MC3T3-E1, and 3 and 7 days for J774A.1. The control groups were glass coverslips, TCPS (well plate) ± treatment with Triton-X100 (lysed cells as death control).
Resazurin assays using a 0.05 mm working solution prepared by 80× dilution of a stock solution (1 mg mL −1 resazurin sodium salt in PBS) with phenol red-free, serum-free cell culture media. Discs and coverslips were transferred to an unused well to exclude resorufin signal from cells growing on the well bottom. 250 µL of the working solution was added to each well and incubated for four hours. The viability was quantified by transferring 100 µL of supernatant to a black 96-well plate and measuring fluorescence in a plate reader (Cytation 3 Imaging Reader, Agilent BioTek, Santa Clara, CA) at λ Ex = 532 nm and λ Em = 590 nm at gain 50. The fluorescent intensity was normalized to the surface area of the different materials, i.e., to the number of cells initially seeded on each material.
Live/Dead Assay: The viability of MC3T3-E1 cells grown on middleXL was also evaluated by performing a live/dead staining. The polymer discs and glass coverslips (n = 3) were incubated in calcein AM (2 µm) and propidium iodide (3 µm) dissolved in phenol-red free, serumfree media for 30 min. Cells grown on glass coverslips were used as a control. Calcein AM stained living cells in green, while propidium iodide stained dead cells in red. After that, they were placed in a coverslip cell chamber and imaged in a Leica SP8 TCS 3× STED laser scanning confocal microscope at excitation wavelengths of 485 and 530 nm for calcein AM and propidium iodide, respectively, and 20× magnification. For each sample, a tile scan was captured with a total of 172 frames using two channels. The software used was Leica Application Suite X (Leica Biosystems, Wetzlar, Germany). The area covered by live and dead cells was analysed with ImageJ.
Statistical Analysis: Experimental data were all collected from the raw data produced by the instrumental testing results. Data were processed using Microsoft Excel, then exported to Originlabs Origin for figure plotting. Tables, bar charts, and scatter plots with error bars all display mean ± standard deviation (mean ± SD) of the number of replicates given in the respective captions (n = "x"). Box-and-whisker plots (all in Figure 7) present the individual values (filled diamonds) as well as their distribution of measured values across the second and third quartiles (box) and first and fourth quartile (whiskers), the median (horizontal line), and mean (open square). The statistical comparisons of the cell viability between the four experimental groups investigated in the extract test were evaluated using a one-way ANOVA with Tukey's post hoc test. In the direct contact tests, the statistical comparisons between two different groups were evaluated using an unpaired t-test, whereas a paired t-test was used to evaluate the differences between two different time points of the same material group. In both cases, the differences were deemed significant if p < 0.05. GraphPad Prism 9 was used for these analyses.
Statistical comparison of the elasticity of middleXL networks (Figure 2c) was evaluated using a one-way ANOVA with Games-Howell post hoc test and deemed significant if p ≤ 0.05. Minitab 2020 was used for the statistical analysis.
Statistical comparisons were all homoscedastic Student's t-tests, apart from tests between time points within one material that were paired Student's t-tests.

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