Soft Hydrogel Shapeability via Supportive Bath Matching in Embedded 3D Printing

Embedded 3D printing is widely adapted for fabricating architected soft and non‐self‐supporting hydrogels for applications ranging from tissue engineering to soft robotics. Although the matching between hydrogels and supportive baths sets the foundation in embedded 3D printing, the rule‐of‐thumb for supportive bath selection and creation is not yet established. Herein, the “shapeability” of distinct classes of hydrogel inks (i.e., pH‐responsive, photo‐crosslinkable, thermal‐sensitive, chemically crosslinkable monomeric, and cationic and anionic inks) in diverse representative support baths (i.e., gelatin slurry, agarose fluid gel, Carbopol and oil‐based baths) is evaluated. The results show that the dominate mechanisms for interfacial instabilities, including diffusion‐driven or charge‐driven, can be predicted by evaluating the composition pairing of the pre‐crosslinked hydrogel ink and supportive bath. Based on this, a general and simplistic guideline for supportive bath selection to attain hydrogel shapeability is proposed. The approach can widen the spectrum of hydrogel materials that can be structured on‐demand for a plethora of functionalities.


Introduction
The need to create soft and bioinspired functions with sustainable or biomedical materials has called for novel techniques to shape hydrogel materials in three dimensions. [1][2][3][4] With moduli in the pascal to hundreds of kilo-pascal range, soft hydrogels are generally not self-supporting as high aspect-ratio structures. [5,6] In this context, the concept of hydrogel 'printability' has been proposed and investigated extensively for in-air extrusion printing. [7] Hydrogel printability can now be enhanced by utilizing rheological modifiers or photo-crosslinkable moieties to increase crosslinking speed and controllability, thereby attaining better shape definition during the in-air deposition process. [8][9][10][11] However, excessive ink modification could sacrifice the inherent characteristics of hydrogel inks. To mitigate this problem, embedded 3D printing (EMB3D) has been proposed based on the widely-accessible extrusion 3D printing technology. [6,[12][13][14][15][16][17] The method involves the use of yield stress viscoelastic materials (also known as Bingham pseudoplastic materials) as support baths (Figure 1a), which permits smooth translation of nozzle in the bath and the subsequent trapping of the deposited materials in place for further crosslinking. [14,18] To date, various studies have proven the potency of this method to create hydrogel structures for soft robotic and tissue engineering applications; recent works have further established the critical rheological properties of inks and baths required for the printability of an ink in a matrix medium (i.e., embedded printability) [19][20][21] (Figure 1a,b). For example, an ideal support bath should possess thixotropic behavior for mitigating crevice formation upon nozzle departure, [19,22] a yield stress characteristic for smooth nozzle translation, and a sufficient storage modulus to mechanically support the printed ink. In addition, the ink-bath system should exhibit a sufficient but not too high Oldroyd number (Od) for minimizing the yielded region around the nozzle and reducing distortion of the printed structure. [19,20,23] While the ink-bath rheology is imperative to embedded printability, the end shape of the embedded hydrogels is determined by the interplay between the crosslinking kinetics of the inks and a good balance of the chemical and physical ink-bath interaction at their interfaces. Various forces (e.g., gravity, concentration gradient, ionic interaction, and the ink-bath interfacial tension) affect the shape evolution of the printed object over time until the object is fully crosslinked (Figure 1b,c). [14,20] While the gravity- Rheological and physical-chemical factors that determine the shape fidelity of the hydrogel constructs produced via embedded 3D printing. a) Schematic showing an embedded printing process assisted by the yield stress characteristics of the support bath. When the shear stress ( ) induced by the nozzle movement exceeds the yield stress of the bath ( y,bath ), the support bath becomes fluidized, allowing a smooth translation of needle in the bath. As the nozzle departs, the bath rapidly recovers its elastic behavior, trapping the extruded ink in place for further crosslinking. b) Common instabilities during embedded printing, and the corresponding properties served to alleviate them. Thixotropic time refers to the duration for the bath to recover its zero-shear viscosity after the applied shear stress is removed. c) "Embedded shapeability" after printing. Various factors can act on the printed object to change the intended shape after printing. A good embedded shapeability is resulted from good crosslinkability and shape-maintaining ability of the ink in the support bath. d) Shear elastic (G ′ ) and loss (G " ) moduli versus shear stress plots of different support baths, measured at a frequency of 1 Hz. The dotted line indicates the yield stress.
driven sagging evolution can be avoided through matching the densities of the ink and the bath, other factors that are subjected to multiple ink-bath properties add extra complexities in bath selection for different types of hydrogel inks. Indeed, despite different support bath systems have been reported for EMB3D (e.g., gelatin slurry, [6,15,24] Carbopol microgels, [14,25] and agarose [16,26] ), research on embedded printing has been mostly restricted to a few types of supramolecular hydrogel materials as inks, such as proteins and polysaccharides. However, hydrogel materials can be formed from a wide variety of species, from small monomers (a few tens Dalton) to supramolecular polymers (millions Dalton) that have different ionic charges and responsive behaviors. In addition, bath-specific constraints, such as bath stability during ink crosslinking, limited strategies in boosting crosslinking rate of the ink inside the bath and the challenge in preventing the spontaneous/premature diffusion of un-crosslinked hydrogel precursors, further restrict the generalizability of the method. Therefore, exploring the 'shapeability' of different types of hydrogel-forming species in support baths could provide strategies for their fabrication via embedded printing.
Here, we investigated the 'embedded shapeability' of different hydrogel inks using a variety of support baths with diverse characteristics, from oil-based support baths that eliminate inkbath diffusion to the mainstream aqueous support baths (i.e., gelatin slurry, [6,15,24] Carbopol microgel, [14,25] and agarose fluid gel [16,26] ) in literature. Their diverse rheological and chemical properties are denoted in Figure 1d, and Table S1, Supporting Information. The hydrogel inks tested here were pH-responsive, photo-polymerizable, thermal-sensitive, chemically crosslinkable monomeric, and cationic and anionic inks (see Table S2, Supporting Information, for their characteristics), covering most hydrogel materials that have been implemented in biomedical and soft robotic applications. We refer "embedded shapeability" to the capabilities of the ink-bath system to preserve the shape of the as-deposited ink in the bath throughout the crosslinking process (i.e., shape-maintaining ability and crosslinkability, Figure 1c). Embedded printing of chemically crosslinkable small monomeric ink. a-i) Crosslinkability phase diagram of the monomeric acrylamide ink in different support baths. n/a = not applicable. a-ii) Representative photo of the as-printed hydrogel in a Carbopol support bath. Scale bars = 5 mm. Acrylamide inks were not crosslinkable in aqueous baths regardless of whether the baths were supplemented with or without APS and TEMED. b) Proposed fabrication scheme. c) Linear correlated plot of the melting peak obtained from differential scanning calorimetry, indicating that the fatty acid support baths can be fluidized easily via heating. d) The extrusion of a continuous filament with pre-crosslinked alginate added to the acrylamide ink. Scale bar = 500 μm. e) Fabrication of a tri-leaflet aortic valve using our proposed scheme showed in Figure 2b. The trileaflet structure was able to stop liquid back flow, resembling the unidirectional flow characteristics of native valve. Scale bars = 10 mm.
Our results reveal two dominating instabilities affecting embedded shapeability in an aqueous bath system. They are molecular species transportation and charge-driven interaction between the ink and the bath. These instabilities can be predicted by the compositions of the hydrogel-forming ink and the bath, prior to experimentation. In other words, the bath choices can be informed based on the ink composition. We envision that the findings from this study will widen the utility of this technology for fabricating diverse hydrogel materials, and overcome the cost-intensive trialand-error challenges for bath selection.

Diffusion-Driven Interface Instability
Good embedded shapeability is more likely to achieve when the crosslinking kinetic of the ink species is faster than the diffusion kinetics of the major molecular constituents in both the ink and the bath. Since hydrogels can be formed from a range of molecules of varying molecular weight, from small molecules to macromolecular systems (a few tens to a few millions Dalton), these species will have different diffusion kinetics in aqueous support baths. Figures S1 and S2, Supporting Information, denote the characteristic diffusive timescale (t) of the ink and the bath species used in this study, estimated based on the Stokes-Einstein law (D = kT/6 R, where D is diffusion coefficient of the species, k is Boltzmann Constant, T = 298.15 K, R is solute hydrodynamic radius, and is solvent viscosity) and the approximation equation for diffusion time (t = L 2 /D, where L is diffusion length). Dependent on the size of the species, the diffusive timescale of the species can vary over two orders of magnitudes. In the following sections, we explore the embedded shapeability of different hydrogel inks according to the molecular sizes of the constituent species, from small to large molecular sizes.

Chemically Crosslinkable Monomeric Hydrogels
Hydrogels synthesized from monomers via wet chemical crosslinking, such as free radical polymerizations, have been used for tuning their properties (e.g., mechanical properties). [27] One representative example is polyacrylamide (PAAm), which is widely-used in biomedical applications. PAAm is usually prepared from acrylamide monomers via free radical vinyl polymerisation with the aid of N,N′-methylene-bis-acrylamide (bis-AAm) as crosslinkers, ammonium persulfate (APS) as oxidizing agents and tetramethylenediamine (TEMED) as catalysts. Using the above ink formulation as the representative ink, in which the smallest molecular weight constituent is ≈70 Da, we found that none of the mainstream aqueous support baths in literature (i.e., Carbopol and agarose fluid gel) enables satisfactory crosslinking of the acrylamide ink into shaped hydrogels (Figure 2a, and Figure S3, Supporting Information). We hypothesize this poor crosslinkability is associated with the rapid diffusion of the low molecular weight species of the ink in aqueous media compared with the crosslinking timescale (see Figure S1, Supporting Information). Accordingly, we simulated the concentration profiles of the species within the ink-bath system via finite element modelling ( Figure S4, Supporting Information), assuming pure liquid diffusion for simplification.
Further modelling work can be conducted with a multi-scale diffusion model that considers the hindrance effect caused by the mesh size of the hydrogel network. [28] As all the ink species have similar molecular weights (≈10 2 g mol −1 ) and diffusion coefficients (≈10 −9 m 2 s −1 ) ( Figure S1, Supporting Information), the concentrations of the ink species rapidly reduce to half of their initial values within ≈40 min, as illustrated in our simulated results ( Figure S4b-ii,c-ii, Supporting Information). As a consequence, the amount of reagents at the ink zone might become deficient for crosslinking into shaped hydrogels. Interestingly, our simulation results suggest that the diffusion of Carbopol species from the bath phase to the ink phase is negligible ( Figure  S5, Supporting Information), with the concentration remains almost constant within 3 h timescale of simulation, owing to the macromolecular size of the bath molecules.
To overcome the pronounced diffusion problem of the small molecular ink species, we suggest the use of an oil-based support bath, such as stearic acid and oleic acid (a mixture of fatty acid), for embedded printing. Figure 2b depicts our proposed fabrication scheme. The oil-based mixture prohibits diffusion, and stearic acid is chosen due to its ability to be fluidized at physiological temperature, which is advantageous for the subsequent bath removal step ( Figure 2c). In addition, to prevent the low viscosity acrylamide ink from breaking up into droplets in the oil-based bath due to large interfacial energy, a pre-crosslinked alginate-Ca 2+ slurry can be added to the ink as a sacrificial component to strengthen its viscosity. This pairing of support bath and ink enables embedded printability (Figure 2d), while the print can be subsequently chemically crosslinked in the bath via free radical polymerisation. After crosslinking, the printed construct of PAAm can be collected via warming the bath to 37°C, and the alginate component can be sacrificed by immersing the construct in a Ca 2+ chelating EDTA solution. Using this approach, a geometrically-complex construct with the shape of an aortic valve was created from the acrylamide ink ( Figure 2e). Such an approach can be applied to other chemically crosslinkable monomeric inks, such as acrylic acid, to widen the spectrum of materials that can be fabricated via embedded printing.

Photo-Crosslinkable Hydrogels
Photo-polymerisation is a highly attractive method for preparing hydrogels through UV or visible light illumination with photoinitiators. [29] Photoinitiators usually have a fast diffusion timescale due to its small molecular size ( Figure S1, Supporting Information). The amount of photoinitiators in the ink zone plays a key role in photopolymerization and hence embedded shapeability. In general, there are three ways to incorporate photoinitiators in an embedded printing system (Figure 3a): 1) both the support bath and the ink are supplemented with photoinitiators; 2) only the ink is supplemented with photoinitiators; and 3) only the support bath is supplemented with photoinitiators. While reduced usage of photoinitiators is always advantageous for reducing cytotoxicity and cost, there is no clear consensus on how polymerisation efficiency is influenced by different supplementation ways of photoinitiators.
To explore the embedded shapeability of photopolymerizable inks under different supplementation methods, we exploited an ink composed of PEGDA (MW ≈ 700 Da) with sodium hyaluronate (molecular weight, MW ≈ 1.9 MDa) as a rheological modifier and Irgacure 2959 as photoinitiators (MW = 224.25 Da).
All the aqueous support baths tested here are highly transparent to allow efficient light penetration (Figure 3c). Comparing the three supplementation scenarios (Figure 3a), greatest polymerisation efficiency was achieved when both the ink and the bath were supplemented with the same concentration of photoinitiators, C supp. (Figure 3a-i). This would result in an almost zero net flux of photoinitiators across the ink and the bath, and the equilibrium photoinitiator concentration (C ∞ ) is the highest among the three scenarios ( Figure S6, Supporting Information). A gelled hydrogel structure can be attained within a short photo-exposure time of <5 min (Figure 3a-i). In stark contrast, when only the ink was supplemented with photoinitiators, the polymerisation efficiency was significantly diminished due to the net outflow of the photoinitiator from the ink to the bath (Figure 3a-ii). A prolonged UV-exposure time (>30 min) was therefore necessary for attaining a sufficiently crosslinked structure. In comparison, the photo-polymerisation efficiency was substantially improved when only the bath was supplemented with photoinitiators ( Figure 3a-iii). This can be explained by the fact that the equilibrium photoinitiator concentration of scenario 3 (Figure 3a , as the bath volume is typically much larger than the ink volume (v bath > v ink ) (also denoted in our simulation results, Figure S6, Supporting Information). Furthermore, when only the bath is supplemented with photoinitiators, the inward transport of photoinitiators from bath to ink facilitates polymerisation at the construct boundary. The crosslinked polymeric network further hinders the diffusion of ink to the bath. Hence, more efficient photo-polymerisation was achieved in a bath-supplemented system (scenario 3) compared with the inksupplemented system (scenario 2).
Similar outcome was also observed when no sodium hyaluronate rheological modifier was supplemented in the ink (Figure 3b, and Figure S7, Supporting Information). The PEGDA ink can be crosslinked efficiently into shaped constructs within 10 min of UV exposure when both the ink and the bath were supplemented with photoinitiators ( Figure 3b-ii), while the ink was not able to be sufficiently crosslinked when photoinitiators were present in the ink only (Figure 3b-iii). This indicates that polymerisation is sufficiently faster than the diffusion of PEGDA even though the ink viscosity is relatively low (≈20 mPa s [30] ). Overall, the results suggest that a balanced amount of photoinitiators in the ink and the bath is required for minimizing light exposure time, and when minimized usage of photoinitiators is preferred, a system with the bath supplemented with photoinitiators has better polymerisation efficiency than an inksupplemented system.

Protein-Based Hydrogels
Protein-based bioinks, such as collagen, are commonly crosslinked via thermal gelation at physiological temperature. Literature has suggested that at least an hour of incubation in bath is needed to ensure sufficient gelation of protein-based inks prior to the bath removal. [6,15] Although proteins are large size molecules with an average molecular weight of over 50 kDa, [31] the long thermal gelation time can lead to poor shapeability owing to the prolonged diffusion. Hence, choosing a bath with thermal stability is the key for preserving the shape of proteinbased inks during the long gelation time. As demonstrated in Figure 4b-i, all support baths enable similar print accuracy. The deviations are about 10-20%, comparing the areas of the as-printed constructs with the CAD model. However, the printed collagen construct failed to crosslink into shaped constructs in a HEPES-supplemented gelatin slurry support bath (Figure 4a), where HEPES was used for neutralizing the acidic collagen. The crosslinking of collagen and the sol-gel transition of gelatin occur at a similar temperature range around physiological temperature. Thus, intermixing of the two species is likely if gelatin slurry is used as a supporting material for collagen. Nonetheless, we note that previous studies have demonstrated the fabrication of geometrically-intricate protein-based biological constructs using the gelatin slurry approaches. [6,15] Optimization of the ink conditions (i.e., pH, concentration, molecular weight, and sources) and the print setting (i.e., infill density) is therefore imperative to the crosslinkability and the final shape of inks that require a long gelation time above 37°C. On the other hand, both the Carbopol and agarose fluid gel baths enable successful gelation of collagen inks using a non-optimized experimental setting. A good embedded shapeability was achieved with the 0.2% Carbopol bath (Figure 4b-ii), indicating its suitability for fabricating thermal-sensitive protein-based inks due to its thermally-stable and sufficient storage modulus.

pH-Responsive Hydrogels
pH-responsive hydrogels, such as polyacrylic acid (PAA), can swell or deswell tremendously under pH changes. [32] Swelling is a water transport behavior governed by the ionization degree of the functional groups of hydrogels (Figure 5a). For example, when the environmental pH is higher than the pKa of the carboxyl groups of anionic hydrogels (pH > pKa), more carboxyl groups on the polymer backbones become deprotonated. The ionized hydrogels imbibe more water and swell to reduce the electrostatic repulsion between the ionized functional groups until an equilibrium is reached. [33] Although this behavior has attracted great interest in drug release and soft robotic applications, [34] we observe this swelling behavior significantly discourages the possibility to construct pH-responsive hydrogels with aqueous support baths. As depicted in Figure 5b-i, the embedded printed PAA structure swelled enormously in an aqueous bath of xanthan gum with pH ≈ 5.8, resulting in a poorly defined structure. Although shape definition can be remarkably improved through utilizing a pH condition unfavorable for swelling (i.e., a highly acidic xanthan gum bath in this case), an aqueous bath still cannot completely avoid the thermodynamically favorable water transportation from the bath to the ink (Figure 5b-ii). In addition, conventional support baths, such as Carbopol and gelatin slurry, are incompatible with harsh pH conditions as the viscosity of the baths are greatly weakened at these conditions. [25,35] To fully suppress the swelling problem, we propose oil-based support baths, such as fumed silica-mineral oil baths that impart shear-thinning behavior and appropriate rheology (Figure 5c), for constructing pH-responsive hydrogels. Oil-based media inhibit water transportation, hence enabling no swelling effect (Figure 5d). To reduce the interfacial tension between the aqueous ink and the hydrophobic bath that could lead to shape evolution of the printed construct, the bath can be supplemented with stearic acid as surfactant and rheological modifier (Figure 5e-ii, and Figure S8, Supporting Information). With this, the shape of the printed photopolymerizable PAA construct can be effectively preserved in the bath for more than 50 min (Figure 5e-iii), which is sufficient for various crosslinking strategies (see Table S2, Supporting Information). After photopolymerization, the construct can be collected from the oil bath and rinsed with soapy acidic water to remove any residual oil (Figure 5f). A higher concen-tration of stearic acid was not employed as this increases material usage and compromises the bath transparency ( Figure S9, Supporting Information), which makes observation difficult and might reduce photo-polymerisation efficacy.
With the above established bath formulation, pH-responsive morphing systems can be fabricated with extrusion printing. As shown in Figure 5g, and Video S1, Supporting Information, through designing the print path of the PAA pH-responsive construct to induce a structural stiffness inhomogeneity, the resulted constructs can be morphed into different shapes when there is a pH change in their environment, such as rolling, helix, twisting and saddle shapes. This fabrication method can be readily extended to create complex 3D pH-responsive hydrogel functional devices.

Ionic Crosslinkers in Anionic Hydrogels
Typically, when embedded printing anionic hydrogels, an ionic crosslinking agent is mixed with the support bath for initiating ionic crosslinking immediately after deposition. The construct is then further crosslinked with additional ionic crosslinking agent after printing, followed by releasing the construct via fluidizing the bath (Figure 6a). Extra caution should be taken when using ionic-strength sensitive support baths for fabricating such class of hydrogels. This is exemplified with an embedded printing system comprised of alginate ink and Carbopol support bath. First, supplementing ionic crosslinkers (e.g., CaCl 2 ) to low concentration Carbopol support baths greatly degrades the bath rheological properties (i.e., viscosity and storage moduli, Figure 6b-i, and Figure S10, Supporting Information), even though the bath is only at a low ionic strength of 11 mm. The hydrogel cannot be printed into the intended shape in the bath owing to the inferior storage modulus (Figure 6c-i). As a consequence, a higher concentration of Carbopol is required to attain suitable rheological properties for embedded printing. However, when a higher concentration Carbopol bath is used, the Carbopol polymers will precipitate upon the addition of extra ionic crosslinking agent due to salting out effect [36] (Video S2, Supporting Information). The aggregating Carbopol polymers will exert a drag force on the soft printed structure, reducing the printed shape fidelity. The drag problem is more severe when a high concentration of Carbopol is used. Although the drag problem can be reduced with the use of a lower concentration of crosslinking agents (e.g., 11 mm CaCl 2 ), this might lead to insufficient ink crosslinking. On the other hand, the drag problem is not observed in ionic strength-insensitive agarose fluid gel and gelatin slurry baths. As shown in Figure 6b-ii, the storage moduli of agarose fluid gel and gelatin slurry are insensitive to ionic strength. The printed structures can be successfully crosslinked and collected in the baths (Figure 6c-ii,iii), accompanied with only a small change in the areas of the constructs (Figure 6d). This suggests that the agarose fluid gel and the gelatin slurry bath are compatible with the ionic crosslinking mechanism of anionic hydrogels. c) The shape definition of ionically crosslinkable sodium alginate constructs fabricated using different support baths i) 0.2 w/v%, 0.5% and 1% Carbopol, ii) agarose fluid gel, and iii) gelatin slurry. All support baths were supplemented with 11 mm CaCl 2 . Scale bars = 5 mm. Photos of constructs printed in 0.5% and 1% Carbopol were taken after removing the Carbopol aggregate in the baths. d) Absolute area deviation (|Δ area |) of the letter P constructs and the inner circle of the letter P constructs after crosslinking and bath removal, in comparison with the as-printed constructs. The x-axis indicates the constructs fabricated using different support baths.

Cationic Hydrogels
While cationic hydrogels are a popular class of hydrogels in biomedical applications due to their affinity to the negatively charged cell membranes, [37] fabrication of cationic hydrogels via embedded printing has been rarely demonstrated in literature.
In view of this, we investigated the capabilities of the mainstream support baths for fabricating constructs made of chitosan, a classic natural antimicrobial cationic hydrogel derived from crustaceans, such as shrimps and crabs. The printability of two chitosan inks was evaluated. They were 3 w/v% chitosan from shrimp shells and a 13 w/v% 600-800 kDa chitosan ink. The chitosan from shrimp shells' ink has a lower viscosity and storage modulus ( Figure S12, Supporting Information). When cationic chitosan is printed in anionic Carbopol support baths (Figure 7a), the ink was dragged inside the bath, failed to form the intended shape. This behavior is due to the rapid increase in the ink storage modulus (G′ ink > G′ bath ) caused by the electrostatic crosslinking between the oppositely charged ink-bath system (Figure 7di). As G′ ink surpasses G′ bath , dragging results. [19] Although the dragging issue is less pronounced in a higher concentration Carbopol bath that has a larger G′ bath , the resulting shape fidelity is still severely diminished. Hence, anionic baths are poorly suited for fabricating cationic inks. On the other hand, dragging of ink does not take place in agarose fluid gels and gelatin slurry baths, which are non-ionic and positively charged at physiological pH, respectively. [38] The rheology of the chitosan ink is of similar magnitude after interacting with agarose fluid gel and gelatin slurry baths (Figure 7d), suggesting minor interaction between the ink and the bath. However, the agarose fluid gel bath might not be suitable for fabricating low viscosity chitosan ink (3% chitosan from shrimp shells) owing to its lower storage modulus and yield stress (Figure 7b). Remarkably, the gelatin slurry bath is well suited for fabricating both types of cationic chitosan inks (Figure 7c). The printed constructs are well-defined and have dimensions similar to the CAD models (Figure 7e). The example here reveals the importance in controlling the ink-bath electrostatic interaction for embedded shapeability.

Conclusions
EMB3D has enabled patterning of low viscosity hydrogel inks into complex freeform structures that are previously unattainable using conventional in-air printing. [39] With increasing demands for complex functional structures made from a wide range of hydrogels for various applications, a guide to aid the selection of support baths is of necessity. Figure 8 informs the decision for selecting suitable 3D printing approaches for fabricating different categories of hydrogel materials (i.e., photo-polymerizable, protein-based, pH-responsive, and cationic and anionic hydrogels). The work here emphasizes the importance of mass transport and electrostatic interaction at the ink-bath interface on final shape definition. It is envisioned that one can predict suitable types of support baths based on the ink composition. For example, the bath should be selected based on the rule of thumb to minimize electrostatic interaction during printing and net flux of constituent species that will affect the intended shape. Although oil support baths with non-ionic surfactants can intrinsically inhibit all transport events and reduce interfacial tension at the ink-bath interface, the residual oil within the construct can sacrifice its biocompatibility and mechanical integrity due to the poor fusion between filaments. Hence, the required degree of transport inhibition should be decided based on the molecular weights of the ink species, the ink crosslinking kinetics, as well as application needs. With the appropriate choice of embedded printing strategy, it is expected that a wider range of hydrogels can be effectively fabricated for inducing diverse functionalities.
Preparation of Inks: All concentration written in this article is expressed as w/v% unless otherwise specified. The photopolymerizable sodium hyaluronate ink was prepared by dissolving a 3 w/v% sodium hyaluronate in DI water, followed by mixing the solution with PEGDA and a 10 w/v% Irgacure 2959/ethanol at a volumetric ratio of 10:2:1. The photopolymerizable PEGDA ink was formed by mixing PEGDA (700 Mn), DI water, and 10 w/v% Irgacure 2959/ethanol at a 2:8:1 volumetric ratio. The thermo-gellable collagen ink was directly used without any modification. To prepare the chemically crosslinkable acrylamide monomeric ink, sodium alginate and calcium chloride powders were dissolved into a precursor solution of 12 w/v% acrylamide, 0.25 w/v% bis-acrylamide, 0.2 w/v% APS and 0.01 w/v% TEMED, making the resultant ink with 3.5 w/v% sodium alginate and 0.15 w/v% CaCl 2 . For the pH-responsive ink, poly(acrylic acid) (PAA) solutions were prepared by dissolving the PAA powder at 25 w/v% in 0.1 m NaOH under stirring for a week. The PAA The black and blue arrows indicate preferable routes, while the grey dotted arrows indicate feasible but not optimal routes. The blue arrows indicate ink-bath systems with equal photoinitiator concentration and optical clarity.
solution was mixed with PEGDA and 10 w/v% Irgacure 2959 in ethanol at a volumetric ratio of 10:1:1 to allow for photo-polymerisation. Two cationic chitosan inks were used for printing. They were a 3 w/v% high molecular chitosan derived from shrimp shells (C3646, Sigma) dissolved in 0.1 m acetic acid and a 13 w/v% 600-800 kDa chitosan (10636695, Thermo Scientific) dissolved in 1 m acetic acid. A photopolymerizable chitosan was prepared by mixing the 13 w/v% chitosan stock solution with PEGDA and 10 w/v% I2959/ethanol at a volumetric ratio of 10:1:1. A chitosan solution derived from dissolving 3 w/v% chitosan from shrimp shells (10636695, Aladdin) in 1 m acetic acid was used for rheological measurements showed in Figure 7d. Sodium alginate inks were prepared by mixing 10 w/v% sodium alginate mixed with 200 mm CaCl 2 at a volumetric ratio of 10:3. To avoid inhomogeneous gel formation, the CaCl 2 solution was added slowly to the solution of sodium alginate under vigorous stirring.
Preparation of Support Baths: The gelatin slurry, Carbopol, and agarose fluid gel support baths were prepared following the reported protocols in literature. [6,14,16] In brief, the 4.5 w/v% gelatin was dissolved in either 11 mm CaCl 2 at ≈50°C. The dissolved solution was then stored overnight in a 4°C fridge and was kept in a −20°C freezer for 1 h prior to the blending step. This step was to prevent the gelatin jelly from melting during the following blending and centrifugation steps. Before blending, the gelatin jelly was mixed with an 11 mm CaCl 2 solution or 10 mm HEPES in a volumetric ratio of 2:3. The mixture was blended for 90 s at 'speed 1' using a blender (VonShef 4-in-1 blender). The blended slurry was then transferred into 50 mL conical tubes, and centrifuged at 3600 g and 5°C for 4 min. The supernatant was removed and replaced with the dissolving medium. The mixture was then resuspended using a spatula and centrifuged again. The centrifugation step was repeated until no white foam was observed at the top of supernatant. The produced gelatin slurry was stored in a 4°C fridge when not in use and could last for ≈3 days. Three Carbopol ETD 2020 solutions (0.2, 0.5, and 1 w/v%) were prepared by stirring Carbopol in DI water for ≈5 h until completely dissolved. The Carbopol solutions were then neutralized to pH ≈ 7 by addition of 10 m NaOH. The resulting microgel was then centrifuged to remove bubbles. Agarose fluid gel was prepared by dissolving the agarose power in hot water at 0.5 w/v% concentration, followed by cooling the solution to room temperature under constant shear using a magnetic stirrer. Fumed silica in mineral oil bath was prepared by first dissolving 0-2 w/v% stearic acid in mineral oil at 50°C. When it was fully dissolved, fumed silica was added to mineral oil at 6 w/v%, respectively. The dispersion was then stirred for ≈30 min and was then degassed with a vacuum for at least ≈30 min until most bubbles were removed. Xanthan gum support baths were prepared by dissolving 1.3 w/v% xanthan gum in either DI water or 0.1 m citric acid under stirring using a magnetic stirrer for at least 2 h until complete dissolution. For the support baths used for embedded printing photo-polymerizable inks, the baths were either supplemented with the same concentration of photoinitiators or without any photoinitiators. To prepare the oil-based fatty acid support bath for printing chemically crosslinkable acrylamide inks, 7.5 w/v% of stearic acid was dissolved in oleic acid at 60°C. The resultant homogeneous mixture was allowed to cool down to room temperature prior to printing.
Embedded Printing of Hydrogels: Prior to printing, the ink was centrifuged at 1000 g for 3 min to remove bubbles. A self-designed custom-made 3D extrusion-based printer was used for all the printing experiments. [40,41] The 3D STL files of the printed structures were either custom-designed or downloaded from Thingiverse (www.thingiverse. com) ("All Alphabet Letters" by 6brueder). The ink was drawn into a 1 mL syringe, and the syringe was loaded to the syringe holder of the setup. 27G needles from Adhesive Dispensing Ltd. (inner needle diameter = 0.2 mm) were used in all printing experiments. A petri dish with dimension 35 × 10 mm (Ø × H) filled with support baths was loaded to the stage. Syringe heating, stage heating and UV light were applied when required. To further crosslink the alginate ink, after printing, 1 mL of 500 mm CaCl 2 was added per 7 mL of the bath for further crosslinking and the baths were released after 5 min. This setting enables sufficient Ca 2+ in the bath for ink crosslinking and bath fluidization ( Figure S11, Supporting Information). To release the printed structures after printing, the printed constructs in gelatin slurry were released by warming the support bath at 37°C. Carbopol support baths were liquified by adding a 1 w/v% NaCl solution to the bath, unless otherwise specified. Agarose support baths do not exhibit a sol-gel transition, hence gentle agitation was used to release the printed objects. The printed collagen constructs were kept at an incubator at 37°C for 5 h before bath releasing. The printed PAA structures were UV crosslinked and were directly collected from fumed silica-mineral oil support baths after crosslinking, and rinsed with soapy acidic water (1 m citric acid). For the chemically crosslinkable acrylamide monomeric ink, the printed construct was kept in the fatty acid support bath for 24 h prior to the removal to ensure complete polymerisation of the PAAm. To retrieve the printed construct, the fatty acid bath was fluidized at 37°C, and rinsed with 50 v/v% ethanol solution. The retrieved construct was further immersed in a 0.1 m EDTA solution for removing the alginate constituent.
Characterization of Rheology: The rheological properties of the inks and the baths were measured using a rheometer (HR-2, Discovery Hybrid rheometer system, TA instruments) with a parallel plate (20 mm in diameter, 1000 μm gap size). Oscillatory amplitude sweep measurements were carried out at 1 Hz with a shear strain range of 0.0002% to 9999% at 25°C. Shear viscosity measurements were carried out with a shear rate range of 0.01 to 500 s −1 . All rheological measurements were performed at 25°C unless otherwise stated. The rheological measurements of gelatin slurry were performed at 4°C. The measurements of chitosan inks were conducted with a pristine chitosan solution and the same solutions after interacting with 1% Carbopol, agarose fluid gels, and gelatin slurry for 1 min. The chitosan inks were collected from the baths by a syringe.
Characterization of the Melting Peak: Thermal properties were measured using a PerkinElmer 4000 DSC equipment (Waltham, USA) under nitrogen atmosphere. Approximately 4 mg of the sample material in a closed aluminum tray was referenced against an empty closed aluminum tray. A multi-stage heating and cooling program between −30 and 100°C with a scanning rate of 20°C min −1 was used. The melting temperatures were examined except the first heating stage to remove thermal history.
Finite Element Simulation: Finite element modelling was performed in COMSOL Multiphysics using the Transport of Diluted Species interface under the Chemical Species Transport module. A 2D geometry model was built ( Figure S4a, Supporting Information), composed of an ink phase (an inner circle with a diameter of 5 mm) and a bath phase (an outer circle with a diameter of 35 mm). The initial concentration of the species was computed based on the ink formulation used in the study. Time-dependent concentration of the species within the ink and the bath phases were simulated based on Fick's laws of diffusion and the diffusion coefficients of the species were computed from the Stokes-Einstein law.

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