3D Printing Low-Stiffness Silicone Within a Curable Support Matrix

Abstract

Embedded 3D printing processes involve extruding ink within a support matrix that supports the ink throughout printing and curing. In once class of embedded 3D printing, which we refer to as “removable embedded 3D printing,” curable inks are printed, cured, then removed from the uncured support matrix. Removable embedded 3D printing is advantageous because low-viscosity inks can be patterned in freeform geometries which may not be feasible to create via casting and other printing processes. When printing solid-infill geometries, however, uncured support matrix becomes trapped within the prints, which may be undesirable. This study builds on previous work by formulating a support matrix for removable embedded 3D printing that cures when mixed with the printed silicone ink to solve the problem of trapped, uncured support matrix within solid-infill prints. Printed specimens are shown to have a nearly isotropic elastic modulus in directions perpendicular and parallel to the printed layers, and a decreased modulus and increased elongation at break compared to specimens cast from the ink. The rheological properties of the support matrix are reported. The capabilities of the printer and support matrix are demonstrated by printing a variety of geometries from four UV and addition-cure silicone inks. Shapes printed with these inks range by nearly two orders of magnitude in stiffness and have failure strains between approximately 50 and 250%, suggesting a wide range of potential applications for this printing process.

Graphical Abstract

1. Introduction

Silicone elastomers are used in a wide variety of applications due to their favorable material properties such as low elastic modulus, high elongation at break, and biocompatibility. Examples of such applications include soft robots [1–2], prostheses [3], MR phantoms [4–7], surgical training models [6, 8–12], and biomimetic models [13–16]. Our lab uses silicone to create biomimetic, life-size (centimeter-scale) vocal fold models to study the biomechanics and flow-structure-acoustic interactions of voice production [14]. These vocal fold models are patterned after the structure of human vocal folds by being composed of multiple layers of silicone, the most flexible of which has a modulus of elasticity on the order of hundreds of Pascals.

The presently-described work has been motivated by challenges and limitations associated with fabrication processes often used for vocal fold models and other models that incorporate ultra-soft silicones. Such models are often fabricated via traditional casting processes in which positive molds are machined or 3D printed from rigid materials, and negative molds are created by casting around the positive molds. Casting is performed by pouring liquid silicone into negative molds, curing the silicone, and demolding. Simple, monolithic geometries can be created using a single mold, while more complex models may be cast using multi-part molds. Methods that incorporate dissolvable geometries in casting have also been developed [17–20]. Casting can be used to create high-resolution models with high fabrication yield and is often the method of choice when fabricating silicone models in large quantities. However, creating a low quantity of custom models (e.g., patient-specific geometries) via casting can be relatively inefficient and time-consuming because it requires new positive and negative molds. Further, fabrication yield can be low when casting exceptionally soft silicone models, particularly with a modulus below approximately 1 kPa, due to challenges with demolding the ultra-soft cured material.

3D printing offers several potential improvements over traditional casting methods for fabricating vocal fold (and other similarly-composed) models, including elimination of demolding steps, ability to create geometries which would be difficult to cast, and faster turnaround times for parametric studies. However, existing 3D printing processes, such as those discussed in the following paragraphs, appear to be unable to fabricate solid-infill models with adequate material adhesion and sufficiently-low elastic modulus values for our application.

Some material properties of silicone, such as low uncured viscosity, long cure time, and low elastic modulus, pose challenges for 3D printing. Despite these inherent challenges, many processes spanning several additive manufacturing technologies have been developed for printing silicone. These include vat polymerization [21–22], material jetting [23–25], binder jetting [26], and hybrid methods [27–28]. Several reviews provide more insight into soft 3D printing and silicone 3D printing [29–31]. We examined material extrusion processes as candidate solutions for our application due to their versatility, ease-of-use, and ability to pattern multiple materials. Silicone material extrusion processes are here grouped into three categories which are described in the following paragraphs: direct ink writing, complete-matrix-cure embedded 3D printing, and removable embedded 3D printing (comparisons between these processes are found in Table S1, supporting information).

In direct ink writing (DIW), a print material (“ink”) is extruded through a translating nozzle in a predefined path to create complex geometries [32]. Due to the low viscosity of uncured silicone, the inks used for DIW are often made to be self-supporting by adding filler particles and other rheological modifiers. Formulating inks to be self-supporting, however, complicates ink preparation and alters the material properties of the cured material (e.g., increases elastic modulus), which may be undesirable. While many studies have successfully printed silicone using DIW [33–42], we determined the process was unsuited for our application because modifying ink rheology also increased the stiffness of the cured silicone beyond the desirable range.

Embedded 3D printing is a term that has been used to describe a general process in which prints are created by extruding ink into a gel-like “support matrix” [43]. In one class of embedded 3D printing, the entire support matrix is cured while the printed inks may be curable or non-curable. We here refer to this specific class as “complete-matrix-cure embedded 3D printing.” In complete-matrix-cure embedded 3D printing, a reservoir of the desired shape is first filled with support matrix, then ink is extruded into the matrix through a needle. The support matrix supports the ink during the printing process [44–45] and then the entire matrix cures. Complete-matrix-cure embedded 3D printing is a hybrid of casting and 3D printing because model outer geometry is determined by the reservoir cavity while inner features are determined by the printed ink. Because the matrix in complete-matrix-cure embedded 3D printing is inherently curable, both curable and non-curable inks may be incorporated into prints. Curable inks may be used to create composite structures while non-curable inks may remain inside the reservoir or be evacuated to create well-formed cavities. Silicone complete-matrix-cure embedded 3D printing processes utilize a support matrix composed of silicone, regardless of ink composition [43,46–48].

In another class of embedded 3D printing, the entire support matrix is not cured, but rather the printed ink is cured and then removed from the uncured support matrix. We here refer to this as “removable embedded 3D printing” to denote removal of the embedded print. In removable embedded 3D printing, a reservoir is filled with support matrix and curable ink is then extruded into the support matrix through a needle (see Figure 1). After printing, the ink structure is cured and removed from the uncured support matrix. Several removable embedded 3D printing processes utilizing silicone inks have been described [49–57], though a variety of other materials have also been printed, including collagen, hydrogels, and cells [58–64]. Here we distinguish “complete-matrix-cure embedded 3D printing” from “removable embedded 3D printing” because in the former the entire support matrix is cured and the printed inks may be composed of curable or non-curable inks, while in the latter the entire support matrix is not cured and the inks are curable to allow for the embedded print to be removed from the matrix. Removable embedded 3D printing processes are known by a variety of names including freeform reversible embedding [58,64], writing in the granular gel medium [49], printing in liquid-like solids [59], printing-then-gelation fabrication [60], gel-in-gel printing [61], rapid liquid printing [52], and fluid extrusion bioprinting [62]. Many bioprinting processes can also be classified as removable embedded 3D printing processes [59,62,65–66].

Figure 1.

Printer hardware setup and printing process summary. A) Isometric view of CNC mill with custom extruder used for printing. B) Illustration of ink, syringe, needle, support matrix, print, and reservoir during printing. C) Side-view illustration and D) images of several instances during the printing of a cube (C) and a 2.5 cm-tall ‘Y’ shape (D). E,F) Illustration of cube (E) and image of ‘Y’ shape (F) after removal curing and removal from support matrix. In (F), the Y is deformed under its own weight due to its exceeding flexibility.

In previous work [56–57], our lab determined that removable embedded 3D printing could be utilized to fabricate ultra-soft prints from low-viscosity UV-curable silicone materials. Solid-infill prints were fabricated by printing the silicone within a micro-organogel support matrix developed by O’Bryan et al. [53], then subjected to tensile testing to determine their elastic moduli in directions perpendicular and parallel to the printed layers. Synthetic vocal fold models were also printed and subjected to vibratory testing. The printing method, experiments, and testing results are discussed in Romero [56] and Romero et al. [57]. One disadvantage of this method was that after being cured and removed from the support matrix, the solid-infill prints contained uncured support matrix material trapped within the printed structure. Romero [56] and Romero et al. [57] found that this trapped uncured material decreased the stiffness of the printed material and caused poor material adhesion and a non-homogeneous material distribution. Print defects, poor material adhesion, and altered material properties have similarly been noted in other works [51, 54, 58]. Many studies appear to have avoided these defects by printing thin-walled and hollow shapes with overlapping layers [49, 53–54, 58]. Silicone vocal fold models, however, are solid-infill shapes that cannot be printed by overlapping layers in this manner. Observational experience by Romero [56] and the authors of this study showed that adjusting print settings to reduce the amount of trapped uncured material in solid-infill geometries resulted in shapes with unsatisfactory geometry and resolution, as well as a non-homogeneous material distribution. Additionally, removing the trapped uncured material after model fabrication was challenging or impossible. Consequently, the focus of this work was to solve the problem of trapped, uncured support matrix material to enable fabrication of solid-infill centimeter-scale silicone prints with low elastic modulus values.

It is here shown that the two seemingly competing requirements of the support matrix - its presence around and within the print during printing and its absence within the print after curing - can be solved by formulating a support matrix that both supports the ink during printing and cures within the print during the curing process. Importantly, the support matrix only cures within the printed ink structure, but not at other locations within the reservoir, allowing the print to be removed from the reservoir after the curing process. In this work we report the development of this support matrix, here termed a “curable support matrix” due to its ability to cure when locally mixed with the printed ink. As will be discussed, this curable support matrix enables solid-infill prints to be fabricated via removable embedded 3D printing by supporting the ink during printing, then curing within-but not around-the printed ink structure.

In the following sections we describe the processes and setup used for printing as well as methods for preparing the curable support matrix and silicone ink. The curing property of the support matrix is investigated by examining mid-sections of prints. Modulus and elongation at break values are reported. The rheological properties of the support matrix are compared to published results of similar printing processes. The capabilities of the printer and support matrix are then demonstrated by printing a variety of geometries from four different silicone inks. This printing process is then compared to other material extrusion processes for printing silicone. Lastly, conclusions are presented, limitations of the print process are discussed, and directions for further research are outlined.]

2. Material and Methods

2.1. Support Matrix Preparation

The support matrix was prepared by adding fumed silica (Ure-Fil 9, Smooth-On) at 3.0 wt.% to silicone oil (Silicone Thinner, Smooth-On), mixing for 4 minutes at 2000 rpm until homogeneous using a planetary centrifugal mixer (DAC 150.1 FVZ-K SpeedMixer, FlackTec), and then vacuum degassing for approximately 2 minutes using a vacuum pump (Pittsburgh) and vacuum chamber (ThermoScientific).

The two components of the support matrix, fumed silica and Silicone Thinner, are commonly used to modify the uncured and cured properties of silicones. Fumed silica causes shear-thinning behavior and yield stress [45] in uncured silicone and increases the stiffness of cured silicone. Fumed silica was thus chosen to be a rheological modifier so the support matrix would have desirable properties for printing (see discussion in Section 3.4). Silicone Thinner, on the other hand, decreases the uncured viscosity and cured stiffness of silicone. Silicone Thinner was chosen as the main component of the support matrix because it cures when mixed with other silicone components (i.e., the silicone ink). By formulating the support matrix from these two components, the support matrix cured within but not around the silicone print.

2.2. Ink Preparation

The silicone ink used to print the test specimens is described below, while three other inks used for demonstration prints are described in Section 4. The test specimen ink was prepared by combining the base and catalyst of a UV-curable silicone (UV Electro 225–1, Momentive) at a 15:1 (base:catalyst) ratio by weight, then adding Silicone Thinner at a 1:3 (base+catalyst:Thinner) ratio by weight to decrease the mixed viscosity and cured material stiffness. The ink was optionally colored by adding approximately 0.01 wt.% pigment (SilcPig, Smooth-On) before mixing. These components were mixed at 2000 rpm for 2 minutes until homogeneous, then vacuum degassed for approximately 2 minutes and subsequently loaded into a plastic 3 ml syringe (CareTouch).

The base:catalyst ratio of the silicone was different than recommended in the datasheet [67] because preliminary testing showed that adding Thinner at a 1:3 ratio required more catalyst to adequately cure the ink. The large amount of Thinner added to the silicone is consistent with previous studies in which the authors have added Thinner at high ratios to substantially lower the stiffness of silicones for fabricating vocal fold models [14].

2.3. Printing Process

The printer was made from a 3-axis linear CNC mill (Zen Toolworks, https://www.zencnc.com/) retrofitted with a custom extruder as described in Romero [56] (see Figure 1A,B). Print geometry in STL format was converted into Mach3-flavored G-code using a 3D slicing software (Cura). Next, the G-code was modified by adding instructions to reset all axis locations at the beginning of the print and move the needle vertically out of the support matrix reservoir once printing was complete (approximately three lines of G-code were added). The G-code was then loaded into the printer control software (Mach3). Further details regarding the printing process, analysis, and results can be found in Greenwood [68]. Before printing began, an ink-filled syringe was affixed with a 25G stainless-steel blunt-tip dispensing needle (0.26 mm ID, 0.52 mm OD, 38 mm-long, Jensen Global) and loaded into the extruder. A clear plastic reservoir was filled with support matrix and secured to the print bed. Next, the needle tip was positioned at the desired start location approximately 5 mm above the inside bottom of the reservoir. During printing, ink was extruded through the needle as it translated through the support matrix following instructions contained in the G-code (see Figure 1C,D and Video S1, supporting information).

2.4. Print Settings & Performance Evaluation

Select print settings modified in the slicing software (Cura) for creating the tensile test specimens are included in Table 1. The infill density and infill pattern were determined by preliminary testing [69] designed to improve print geometry and resolution. A discussion of these settings is included in Section 3.1 and visualized in Figure 2A,B. Print performance was evaluated by qualitatively examining the top, front, and side faces of 12 printed cube specimens (one cube shown in Figure 2C–E and all 12 cubes shown in Figure S1, supporting information) under a microscope (Leica Microsystems, M125C), comparing the cubes to the desired geometry, and subjecting the cubes to tensile tests (discussed in Section 2.7).

Table 1.

Select print settings which were modified in the slicing software (Cura) for printing the tensile-test specimens.

Print Setting	Value	Note
Layer Height	0.26 mm	Needle inner diameter
Nozzle Diameter	0.26 mm	Needle inner diameter
Infill Density	60%	Determined in preliminary testing to yield satisfactory geometric fidelity
Infill Pattern	45° Rectilinear	Other infill angles were tested, yielding similar results
Perimeters	None	Perimeters not used for tensile test specimens, but used for demonstration prints
Filament Diameter	8.66 mm	Syringe inner diameter
Print Speed	24 mm/s	Various speeds were tested, yielding similar results
Travel Speed	24 mm/s	Various speeds were tested, yielding similar results

Figure 2.

A) Illustration side view showing two cubes with infill densities of approximately 100% (left) and 60% (right). The top half of each cube is shown as a cross-section to enable visual comparison between the infills. B) Cut-away detail view of ink structure printed at 60% infill density, illustrating where support matrix material becomes trapped within the ink structure during printing. When printing at 60% infill density, the remaining volume of solid-infill prints (approximately 40%) is thus composed of support matrix. C–E) Images of a printed and cured cube specimen used for print performance evaluation and tensile testing. All 12 cubes are shown in Figure S1, supporting information. C) Isometric view of the printed cube. Arrow indicates the ‘tail’ of silicone which is deposited as the needle is pulled vertically out of the support matrix after printing is complete. D) Illustration overlay showing the edges of the cube (solid lines) and the approximate locations where prints were cut (dashed lines) to create the mid-sections for imaging. E) Top, front, and side faces of the printed cube. Lines are spaced 10 mm apart for visual inspection of the print’s dimensional accuracy.

2.5. Print Curing and Removal

After printing, the reservoir was placed inside a UV curing bed (Dreve Polylux 2000) for 10 minutes to cure the printed UV silicone ink. It was observed that the support matrix and reservoir were sufficiently translucent for UV light to penetrate to and cure the printed geometries. The UV light intensity and penetration depth are likely affected by the reservoir material and thickness, concentration of fumed silica within the matrix, and pigment in the inks (only a small amount of pigment was added to colored prints to ensure complete curing). Further investigation will be helpful to understand the spatial and time limitations of this UV curing process.

During the curing process, the curable support matrix solidified within the print. As documented in Sections 3.2 and 3.3, the curing property of the support matrix was verified by observing the inner mid-sections of cut-apart cured prints and tensile testing results. Once cured, the print was removed from the support matrix and gently cleaned with paper towels and, optionally, acetone. We observed that careful removal of well-structured prints left no cured ink residue within the reservoir, allowing the remaining support matrix to be reused with little risk of print defects from remaining ink.

2.6. Support Matrix Curing Tests

Support matrix curing tests were performed using two cubes printed from clear silicone ink (prepared as described in Section 2.2, without pigment) into two different support matrices (see Figure 3). The first cube was printed into a matrix prepared as described in Section 2.1 and containing a dispersion of copper particles (nanopowder, <100 nm BET, CAS #7440–50–8), while the second cube was printed with the same ink into an unmodified support matrix. Mid-sections of these cubes were obtained by cutting specimens into three roughly-equal sections (parallel with the cube’s front face, see Figure 2D) with scissors and using the middle section.

Figure 3.

Evaluation of cured prints by examining the mid-sections of cubes printed into support matrices with a dispersion of copper particles (row A) and unmodified (row B). Columns: i) Support matrices used for printing. Dashed lines in (Bi) show edges of reservoir. ii) Cured cubes which were printed from the same unpigmented (i.e., clear) silicone into their respective support matrix (A or B). iii) Microscope image of the cube mid-section on a darkened background. Apparent yellow tint on edges is a result of lighting setup. Coloring of mid-section in Aiii is distinctly copper-colored. iv) Magnified, multifocus image with light source behind the mid-sections. Speckles are observed in image Aiv, showing the presence of trapped copper particles, which indicates that the support matrix remained trapped within the cured print. The few speckles observed in image Biv are likely due to defects in the material. v) Semi-transparent image with edges marked with dashed lines showing evidence of print layers. Red lines in Av and Bv have been added to approximately denote regions of apparent separation between layers on the edge.

Three observations were conducted to investigate several aspects of the printing and curing processes. First, to determine if support matrix material remained inside the print during the printing and curing processes (which was expected due to the 60% infill density, see Figure 2A,B and Section 3.1), the two cubes were visually compared. Second, to determine if the trapped support matrix cured within the print, the mid-section of both cubes were observed under microscope. Third, to investigate the extent of mixing during printing and to help identify any potential variations in material homogeneity that could have been a result of printing, the dispersion of copper particles within the first cube’s mid-section was observed.

2.7. Cured Print Tensile Tests

Tensile testing was performed on 12 printed and six cast specimens fabricated from the same batch of silicone ink using a procedure and analysis similar to that which was developed by Romero [56] and Romero et al. [57] for testing similar 3D-printed specimens. The geometry for tensile test specimens was a 1×1×1 cm cube (see Figure 2C–E). The process for fabricating and testing the specimens (from 3D model to tensile testing) is summarized in Figure S2, supporting information. For the tensile tests, half of the printed and cured cubes were oriented as printed with their layers perpendicular to the force direction (“perpendicular-oriented cubes”) and half were rotated after printing and curing so their layers were parallel with the force (“parallel-oriented cubes”) (see Figure 4A). The printed specimens were tested in these two orientations to measure the effect of layer orientation on the cured elastic modulus and to enable comparison with prints fabricated by Romero [56] and Romero et al. [57]. Before testing, each cube was glued between two acrylic plates using a silicone adhesive (SilPoxy, Smooth-On) which was cured overnight.

Figure 4.

Tensile testing setup and results. A) Illustration showing the setup and orientation of the cast and printed cubes during tensile testing. B) Several instances during tensile testing of a perpendicular-oriented printed cube, labeled with percent strain. C) Engineering stress plotted from −10 to +20% strain for all 18 cube specimens (six samples for each specimen type: cast, perpendicular-oriented, parallel-oriented). Markers are placed every five data points. D) Detail view of (C) from 18 to 20% strain showing the spread between the stress-strain curves for perpendicular-oriented and parallel-oriented specimens. Markers are placed at each data point. E) Elastic modulus for each specimen type. Columns show average modulus. Markers show individual specimens and are spread so each is visible; standard deviations are given in the text .F) Engineering stress vs. engineering strain curves for all 18 cube specimens, with markers placed at the failure strain of each specimen. G) Failure strain for each specimen type. Columns show average failure strain. Markers show individual specimens and are spread so each is visible; standard deviations are given in the text.

Tensile testing was performed on a uniaxial tensile tester (Instron 3342) fitted with custom mounts to hold the acrylic plates (see Figure 4B). A precycle procedure was performed on each specimen for 10 cycles at a rate of 100 mm/min between −10% and +25% tensile strain. Next, a tensile load was applied at a rate of 30 mm/min until failure. Engineering stress-strain data were calculated for each specimen, and a second-order polynomial fit was applied to the data between ±10% strain. Elastic modulus was calculated as the tangent modulus at 0% strain. Failure strain was calculated as the strain at which stress was maximum.

2.8. Support Matrix Rheological Tests

To characterize support matrix rheology, measurements were performed on a rotational rheometer (AR2000-ex, TA Instruments) using a stainless steel, sandblasted 40 mm parallel upper plate (519400.901, TA Instruments) and a roughened lower Peltier plate with a gap of 1000 μm. The lower plate was roughened by applying adhesive-backed sandpaper to the surface (extra fine 320 grit, Gator Power). Roughened plates were used to decrease wall slip (see Figures S3–S4, supporting information). All rheological tests began by pre-shearing the support matrix sample at a rate of 10 s⁻¹ for 30 s then equilibrating at zero shear for 60 s to remove loading history and allow the structure to rebuild to its steady state (see Figure S5, supporting information). For all tests, the Peltier plate was temperature-controlled at 20°C. The dynamic yield stress, or the minimum stress required for maintaining flow, and shear thinning viscosity of the support matrix were measured by decreasing shear rate from 500 to 0.01 s⁻¹ in 60 s and fitting the data to a Herschel-Bulkley model. The static yield stress, or the stress required to initiate flow, was calculated by ramping shear stress from 1 to 100 Pa for 120 s while measuring viscosity and shear strain. The thixotropic response was measured by running a 3-Interval Thixotropy Test at shear rates of 0.1 and 46 s⁻¹ (see Equations S1 and S2, supporting information). Thixotropic recovery time was measured as the time required for the sample viscosity to recover to 95% of its rest viscosity.

3. Results and Discussion

3.1. Print Performance

Through iteration a set of print settings were identified (discussed in Section 2.4 and shown in Table 1) that resulted in high-quality prints (see Figure 2 and Figure S1, supporting information). This iteration showed that some print settings (e.g., changing infill angle) could be changed without causing a noticeable change in print quality, whereas changing other settings (e.g., infill density) did affect print quality. In previous preliminary work [69], an infill density of 60% was identified as a value suitable for achieving high geometric fidelity for solid-infill prints (see Figure 2), while printing with an infill density of 80% or 100% caused defects of extra material on the top of prints. Further investigation is needed to understand the optimality of an infill density of 60% and to fully explore the effects of print settings on print quality and material properties. We suspect that the support matrix inside the printed ink structure provided critical support to the low-viscosity ink during printing, though further investigation is also needed to verify this. As discussed in Sections 3.2–3.3, it is evident that the cured support matrix within prints enabled strong material adhesion and other favorable material properties.

Printed cube specimens were qualitatively observed under microscope to generally be approximately between 0.25 and 1 mm larger in width, depth, and height from the dimensions of the STL model (10×10×10 mm). This deviation in print geometry may be due to a combination of factors including the support matrix rheology, printer setup and design, possible extrusion imprecision, and the movement resolution of the CNC motors. It was observed that the printer hardware caused vibration of the needle during printing, which likely affected the precision of ink placement and resulting print quality. The specimens printed in this study and those printed by Romero and Romero et al. [56–57] were all fabricated using the same printer hardware, thus a comparison between these studies enabled visualization of the relative advantages of the support matrix on print quality. Qualitative comparison of images of the cube specimens in these studies showed that cubes printed within the present curable support matrix (Figure 2C–E) had improved geometry, resolution, and apparent material adhesion compared to those printed by Romero and Romero et al. [56–57].

Print time for each cube specimen was 8 minutes 26 seconds, which was comparable to the time required to print the same geometry on an FDM 3D printer (print times were also comparable for the other geometries printed in Section 4). We anticipate that creating new geometries via this process could, in some instances, be faster than casting because the most time-intensive fabrication step (i.e., mold fabrication) is not needed (for a discussion on fabrication times of this 3D printing process vs. casting, see Greenwood [68]).

3.2. Evaluation of Support Matrix Curing

Three support matrix curing tests were performed using two cubes printed from the same clear silicone ink. The first cube was printed within a copper-infiltrated support matrix (Figure 3Ai), while the second cube was printed into an unmodified support matrix (Figure 3Bi). As evident in Figure 3Aii, the distinct copper color of the printed cube indicates that a noticeable amount of support material remained dispersed inside the cube after printing and curing. When the mid-sections of these cubes were observed under microscope (Figure 3iii, iv), no observable voids or uncured support matrix material could be identified, which indicates that the support matrix cured within the prints. On the perimeters of sectioned prints, however, the edges of layers could be identified (see Figure 3iv,v). Although small (approximately 0.5 mm depth), these edges are undesirable. Further optimization of print settings (such as by adding perimeters, as was done in Section 4) or support matrix rheology may improve the quality of the print’s exterior.

The images shown in Figure 3iii,iv also show that the copper particles were uniformly distributed instead of being grouped together as if between printed layers and paths. The uniform distribution of particles suggests that some amount of mixing was present during printing. We expect that mechanical mixing may have been caused by motion of the support matrix around the translating needle during printing. Mixing via diffusion between the ink and support matrix could also have occurred. We believe this mixing to be the underlying cause of the support matrix curing within the prints but not elsewhere in the reservoir. The extent of mechanical mixing during printing should be examined using flow visualization in future work.

3.3. Tensile Testing Results

The elastic moduli of cast and printed specimens were tested to determine the effects of printing on material stiffness and isotropy. Tensile test results showed that the cast, perpendicular-oriented printed, and parallel-oriented printed cube specimens had average elastic moduli of 48.4, 19.6, and 17.8 kPa, respectively (see Figure 4C–E). Standard deviations for each set of specimens was 4.0 (cast), 1.1 (perpendicular-oriented), and 1.4 kPa (parallel-oriented). The printing process thus decreased the modulus to 40% of the cast modulus in the direction perpendicular to the layers (perpendicular-oriented cubes) and 37% of the cast modulus in the direction parallel with the layers (parallel-oriented cubes).

Due to the similarity in moduli of the printed specimens, we can conclude that the printed material is nearly isotropic in directions perpendicular and parallel to the printed layers. This similarity means that the decrease in printed modulus is likely not due to their layer-upon-layer fabrication as has been commonly observed in 3D printed parts [70–71]. Instead, the decrease in printed modulus is likely due to a higher concentration of Silicone Thinner within the print due to the cured support matrix material within the printed structure. Some decrease in elastic modulus was not unexpected for the printed cubes since the test specimens were printed at 60% infill density (thus the remaining volume consisted of support matrix, see Section 3.1 and Figure 2A,B). Further, because the support matrix cured within prints, the decrease in printed modulus was not as large as when printing into support matrices that did not cure (comparison in Section 5.3).

The failure strain of the printed and cast specimens was measured to evaluate the material adhesion within the cast and printed materials. The failure strain, found at the location of maximum stress, averaged 148% for cast specimens, 200% for perpendicular-oriented printed specimens, and 217% for parallel-oriented printed specimens (see Figure 4F,G). The standard deviations were 17.2% (cast), 21.8% (perpendicular-oriented), and 5.2% (parallel-oriented). These results indicate printing increased the failure strain of the specimens, which was possibly due to the increased amount of Silicone Thinner cured within the printed models and which suggests strong material adhesion within prints. A comparison of the failure strains in this study and those when printing within a non-curable support matrix is provided in Section 5.3.

After testing it was noted that no evidence of layers or paths could be observed within the fractured faces of printed specimens, providing further evidence of mixing of the ink and support matrix (discussed in Section 3.2). Separation between the edges of the layers, however, could be observed in the fractured parallel-oriented printed specimen (see Figure S6, supporting information) similar to those observed in sectioned prints (discussed in Section 3.2, Figure 3v).

3.4. Support Matrix Rheology

We characterized the shear-thinning viscosity and dynamic yield stress of the support matrix by fitting a classic Herschel-Bulkley model to the test data for the decreasing shear rate test (shown in Figure 5A). The Herschel-Bulkley fit gave a dynamic yield stress of 21.4 Pa, a consistency index of 0.16 Pa∙s, a flow or shear thinning index of 1.01, and a standard error of 8.42. The calculated yield stress for this test was a dynamic yield stress because the test was performed as shear rate decreased to zero. By way of comparison, O’Bryan et al. performed a similar test on a micro-organogel support matrix used for printing silicone and reported a shear stress of 21.4 Pa [53].

Figure 5.

Support matrix rheology. A) Viscosity (circular markers, left y-axis) and shear stress (diamond markers, right y-axis) are shown for a decreasing shear rate test. A Herschel-Bulkley curve was fit to the data to quantify both the shear-thinning viscosity and dynamic yield stress. The shear-thinning viscosity is measured as the slope of the linear portion of the viscosity, and the dynamic yield stress is measured as the asymptote of the shear stress as shear rate decreases. The dashed line in bottom right corner shows where shear stress asymptote would intersect the right y-axis. For visual clarity, only every other data point is shown. B,C) Static yield stress was found in two ways using data from the shear rate ramp test. B) Viscosity (circular markers, left y-axis) and shear stress (square markers, right y-axis) are shown during a shear rate ramp test. The static yield stress, found at the time of maximum viscosity, was 6.7 Pa. The x-axis begins at 40 s because the rheometer was incapable of gathering data at low shear stresses below approximately 5 Pa. The dashed lines indicates the maximum viscosity and static yield stress. C) The static yield stress, found as the y-intercept to the linear fit to slope of the data, was 21.1 Pa. Linear fit shown as dashed line.

The static yield stress was measured by ramping shear rate from 1 to 100 Pa. When found at the location of maximum viscosity, the static yield stress was 6.7 Pa (Figure 5B).

Alternatively, when viewing the same data in a shear stress versus shear rate curve, the static yield stress was 21.1 Pa (found as the y-intercept for the linear portion of the curve, Figure 5C). By way of comparison, Abdollahi et al. [54], who calculated yield stress using the same shear stress versus shear rate curve fit, reported a yield stress of 70 Pa for a support matrix made from Carbopol and optimized for printing silicone. Here we note that calculations of dynamic and static yield stresses are highly dependent on data analysis and measurement conditions [72]. Thus, the same data can be used to calculate different values and comparison between studies likely includes differences due to test parameters. Despite the differences between studies, the yield stress values for the present support matrix appear to be within the range of those reported previously.

The Herschel-Bulkley model parameters reported for the decreasing shear rate test also characterize the shear-thinning viscosity, showing that the viscosity decreases approximately one decade for each decade increase in shear rate (shear thinning index: 1.01). Reports of other removable embedded 3D printing processes for silicone do not include descriptions of shear-thinning behavior for their respective support matrices [49,51,53–54]. Reports of complete-matrix-cure embedded 3D printing processes do, however, indicate that their support matrices had shear-thinning viscosities [43,45–46,48]. This similarity was not surprising, as the matrices also contain fumed silica.

Results from the 3-Interval Thixotropy Test show that the support matrix fluidized to a low viscosity within 1 second on loading due to the shear-thinning viscosity (see Figure S7, supplementary information, transition between intervals 1 and 2). On unloading (transition between intervals 2 and 3), the support matrix took approximately 25 s to recover 95% of the rest viscosity. This thixotropic recovery time is slower than reported by O’Bryan et al. (< 1 s) [53] and longer than reported by Wehner et al. (approximately 200 s) [43], though the measurement method varied in each case. Further investigation will be necessary to determine if a faster recovery time could improve print performance.

Lastly, because the rheology of the support matrix was meant to support the low-viscosity ink, the support matrix could be evaluated by qualitatively examining the geometric fidelity of printed specimens. As discussed in Sections 3.1 and 4.2, the geometry of printed specimens was close to the desired geometry, which indicated that the support matrix was capable of supporting the inks until cured. While some of the printed geometries had overhangs (which are difficult to print without adequate support), the first layer of all prints was in effect a complete overhang because the first layer was printed above the bottom of the reservoir (and thus suspended within the matrix) instead of being printed along the bottom surface. As discussed in Sections 3.1–3.2, the support matrix may also have provided support within the printed ink structure.

4. Demonstration

To explore the capability and versatility of the curable support matrix and printing setup, we produced a variety of demonstration prints from UV and addition-cure silicone elastomers.

The demonstration prints were compared to shapes printed using PLA on an FDM 3D printer (Original Prusa I3 MK3S). The elastic modulus and elongation at break of printed cube specimens were tested using the process described in Section 2.7. Images of the prints are shown in Figures 6 and S8, in supporting information, in which the geometry and apparent stiffness of each is evident.

Figure 6.

Demonstration prints. The grid of images on the left shows prints fabricated from four silicone inks (columns 2–5) compared to shapes printed from PLA using an FDM 3D printer (column 1). The material properties of each ink are described in Table 2. Images A–C on the right show select shapes printed from the same UV-curable ink used for tensile testing. Grid: Cubes were printed, placed on a U.S. penny (row 1), and compressed by the weight of a go lf ball (approx. 45 g) to demonstrate the stiffness of each material (row 2). Other geometries including a ‘Y’ (row 3) and a gecko (row 4) further exemplify material stiffness. The base of each ‘Y’ and bunny was glued to a magnet before imaging. Geckos were placed near the tip of a mechanical pencil for imaging. All demonstration prints were fabricated using the same set of print settings and printed into the same support matrix formulation. The size of the geometries are as follows: cube (1×1×1 cm), ‘Y’ (2.5 cm height, 0.95 cm depth), gecko (4 cm length), bunny (2.5 cm height; also shown in Figure S8). A-C: Images of the ‘Y’ (A), bunny (B), and gecko (C) demonstration prints taken 2 months after fabrication. Inset images in A and B show the soft shapes being deformed by a gloved finger. Shapes in (A) and (B) are also shown in Videos S2 and S3, supporting information. Length of each scale bar is 1 cm.

4.1. Demonstration Materials

Four silicone elastomer inks were chosen that differed in curing method and ranged in uncured viscosity and cured stiffness. The first two inks were mixtures of Momentive UV Electro 225–1, the same UV-curable silicone used in other parts of this work. The first was mixed at the same ratio as described in Section 2.2 (1:3, base+catalyst:Thinner) and the second at a higher ratio (1:6) for lower-stiffness prints. The third ink was made from the addition-cure silicone Ecoflex 00–30 (Smooth-On), chosen for demonstration because it has a low stiffness, is commercially available at a relatively low cost, and is commonly used in soft robots [41,43,47–48,73], soft electronics [48,74–80], and biomechanics studies [4,6–7,14–16]. Sylgard 184 (Dow Corning), a high-stiffness addition-cure silicone, was selected as the fourth ink because it is also widely used in many applications. Ecoflex 00–30 and Sylgard 184 were prepared as described in their respective datasheets, and prints fabricated with these addition-cure materials were left to cure overnight instead of using the UV curing bed. As shown in Table 2, the demonstration print inks ranged in uncured viscosity between approximately 0.4 and 3.5 Pa∙s.

Table 2.

Material properties of mixed and printed silicone inks used for demonstration prints.

Material	Color	Curing method	Mixing ratio (by weight)	Uncured viscosity [Pa-s]	Print stiffness [kPa]	Failure strain [%]
UV Electro 225–1 (1:3)	Light Bluea)	UV	1:3 (base+catalyst:Thinner)	0.373	20.0	180
UV Electro 225–1 (1:6)	Dark Bluea)	UV	1:6 (base+catalyst:Thinner)	Not testedd)	5.3	198
EcoFlex 00–30	Greena)	Addition	1:1 (part A:part B)	3.0c)	15.2	260
Sylgard 184	Translucentb)	Addition	10:1 (base:catalyst)	3.5c)	430.3	47

^a)Pigmented;

^b)Unpigmented;

^c)Information from respective datasheets;

^d)Qualitatively appeared less viscous than UV Electro 1:3.

4.2. Demonstration Geometries

The demonstration print geometries were selected as proof-of-concept geometries to explore the capabilities of the printer and demonstrate the curing and supporting properties of the curable support matrix. Selected geometries included thick sections to highlight the curing property of the support matrix, overhangs to demonstrate the support provided by the support matrix and the apparent stiffness of each material, and features near the spatial resolution of the setup to show the ability of the matrix to preserve intricate features during printing and curing. The geometries were obtained from the following sources: the cube and ‘Y’ were modeled in SolidWorks and the gecko and bunny were obtained from Thingiverse (gecko: thing 2363148, bunny: thing 3731).

4.3. Demonstration Print Settings

The printing process, support matrix, and print settings were kept constant for all demonstration prints to enable comparison between the geometry and stiffness of shapes printed from the different silicone inks. The support matrix for each demonstration print was formulated as described in Section 2.1. Print settings used to create the G-code for each shape were the same as described in Section 2.4 and Table 1, with the exception that a single perimeter wall was also printed around each layer’s infill. The infill overlap parameter was set to −10% and the perimeter was printed after the infill for each layer.

4.4. Demonstration Tensile Testing

One cube of each material was printed and subjected to tensile testing as described in Section 2.7. Tensile test results, shown in Table 2, indicate that the printing process was capable of fabricating shapes with a wide range of cured material properties. For example, the stiffness of the cube specimens ranged between 5.3 and 430 kPa and the elongation at break ranged between 47 and 260%.

4.5. Demonstration Print Performance

In their undeformed states, the geometries of the demonstration prints were close to the desired geometries which, for this initial demonstration of concept, was deemed satisfactory. These geometries included shapes that could be difficult to cast using single-part molds but feasibly created using multi-part molds. Using multi-part molds, however, is generally undesirable because they complicate fabrication. Further, using multi-part molds does not seem to be feasible for our primary intended application (multi-layer vocal fold models). Future work, including improvements to printer hardware, will be necessary to fully investigate the range of geometries that can be printed.

Some apparent defects in the demonstration prints, visible in Figure 6, were a result of the low material stiffness rather than from printing. Other defects were due to printer hardware limitations. Because the extruder setup was not yet capable of sufficient material retraction (due to material compliance of the plastic syringes and play between components), material between the arms of the ‘Y’ and ears of the bunny were trimmed (images before trimming are shown in Figure S8, supporting information).

Larger prints were not fabricated because this study focused on solving the problem of trapped, uncured matrix to print solid-infill centimeter-scale geometries. Further work will be necessary to determine the ability of the printing process and support matrix to fabricate both larger and smaller models. Curing models while printing or utilizing addition-cure silicones may facilitate the fabrication of larger geometries.

5. Comparison with other Removable Embedded 3D Printing Processes

Because of the supporting and curing properties of the curable support matrix, the present removable embedded 3D printing process is distinct from, and includes some advantages over, other similar processes. Here we compare the present removable embedded 3D printing process with other removable embedded 3D printing processes. Most of the comparison is made with processes for printing silicone [49–57], with a few exceptions as noted. This comparison is by no means exhaustive, but instead intended to elucidate primary differences between the present and other similar studies.

5.1. Materials & Printing Process

The support matrices in previous studies for printing silicone have been composed of various materials including a hydrophilic Carbomer gel [49,51–52,54] and micro-organogels [53,56–57] (see Table 3, column 2). This study, on the other hand, utilized a support matrix composed of silicone oil and fumed silica which was formulated to be curable within the prints to solve the problem of trapped, uncured matrix within solid-infill prints.

Table 3:

Comparison of several silicone removable embedded 3D printing processes with the present study.

Reference	Support matrix materials	Silicone inks	Hollow or thin-shell geometries	Solid geometries	Needle IDa) [μm]
Bhattacharjee et al. [49]	Carbomer, water	Sylgard 184b)	Continuous knot, hemisphere, model octopus and jellyfish, nested Russian dolls, branched tubular networks	None reported	50
Fripp Design Ltd. [50,55]	Silicone base, silicone cross-linker	Silicone catalystc)	None reported	Dog bones, gears, earplugs	200
Hinton et al. [51]	Carbomer, water	Sylgard 184b)	Helix, tubes, helical tube, perfusable tube, perfusable bifurcation	None reported	400
Hajash et al. [52]	Carbomer, water	Not specified	Large-scale lattice geometries	None reported	150–3000
O’Bryan et al. [53]	Micro-organogel	Sylgard 184b) Mold Maxd) UV Electro (50:1):0.25e)	Model trachea, sinusoidal scaffold, perfusable network, pump, thin vertical and horizontal sheets	None reported	150–1000
Abdollahi et al. [54]	Carbomer, water	Sylgard 184b)	Cylinder, hollow cube with 1 open face, twisted vase, waterdrop vase, life-size toe, ear	None reported	965
Romero, Romero et al. [56–57]	Micro organogel (from [53])	UV Electro 225–1 (10:1):3e)	None printed	1 cm cube, vocal fold models	260
Present study	Fumed silica, silicone oil	UV Electro (15:1):3e) UV Electro (15:1):6e) Ecoflex 00–30f) Sylgard 184b)	None printed	1 cm cube, Y, gecko, bunny	260

^a)ID: inner diameter.

^b)Addition-cure silicone: Sylgard 184, Dow Corning.

^c)2-part RTV silicones.

^d)Condensation-cure silicone: Mold Max, Smooth-On.

^e)UV-cure silicone: Momentive UV Electro 225–1, ratio: (base:catalyst):Thinner.

^f)Addition-cure silicone: EcoFlex 00–30, Smooth-On.

The inks in most removable embedded 3D printing studies, including this study, did not depend on the composition of the support matrix but were instead independently curable [49,51–54,56–57]. For printing silicone, this meant that the inks consisted of silicone base polymers, cross-linkers, and catalyst materials (see Table 3, column 3). A variant of this general process exists (discussed in the following paragraph) which uses a catalyst ink that causes the support matrix to cure.

In a commercial silicone printing process (“PICSIMA”) [50,55], the ink is composed of a catalyst material while the support matrix is made from silicone base polymers, cross-linkers, and rheological modifiers. Because the support matrix cures due to the injected catalyst ink, this process was capable of fabricating solid-infill prints from silicone. The present study differs from the PICSIMA process because the ink in the present study contained silicone base polymers, cross-linkers, and catalyst materials and the support matrix was made from silicone base polymers and a rheological modifier. One implication of this difference is that the present process may be suitable for a wider range of inks, including both UV and addition-cure silicones, which may not be possible with the PICSIMA process.

5.2. Print Geometry & Print Settings

The geometries and needle inner diameter for the silicone removable embedded 3D printing processes are included in Table 3, columns 4–6. Most shapes fabricated in these processes were centimeter-scale, with the exception of Hajash et al. [52], who printed large-scale lattice shapes including small pieces of furniture.

As evident in the table, few of the processes fabricated solid-infill shapes. The PICSIMA process [50,55] was capable of fabricating solid shapes, though as mentioned in Section 5.1, the ink and support matrix material compositions were fundamentally different than this study and the other reported studies. Romero and Romero et al. [56–57] also printed solid-infill silicone models, though as noted in the Introduction, the printing process resulted in uncured support matrix trapped within prints. They found this trapped uncured material decreased the stiffness of the printed material and caused poor material adhesion and a non-homogeneous material distribution. In a related study of printing non-silicone materials, Hinton et al. [58] reported internal voids within solid-infill dog bones printed from alginate into a gelatin slurry. These voids were likely caused by the trapped, uncured slurry within the alginate prints because the prints were fabricated at a 50% infill density.

Instead of printing solid-infill shapes, most silicone removable embedded 3D printing studies reported fabrication of thin-walled and hollow shapes. Hollow shapes may be less susceptible to the undesirable effects of uncured support matrix within prints. To increase material adhesion within vertical sections, many of the studies increased layer overlap by decreasing layer height [49,51,53–54]. Even with overlapped layers, however, it appears that some defects remained within prints. For example, Abdollahi et al. [54] reported that horizontal and vertical faces of a hollow cube could not be printed with the same set of parameters. Similarly, Hinton et al. [51] observed poor lateral fusion between extruded PDMS filaments and noted that support material remained trapped within void spaces in prints. As was discussed, Romero and Romero et al. [56–57] also observed uncured support material trapped within prints.

In the solid-infill prints fabricated in the present study, on the other hand, internal voids and trapped, uncured support matrix did not appear to be evident. Instead, the prints were shown to have high material adhesion, a nearly isotropic elastic modulus, and a relatively high elongation at break. Instead of having overlapping layers to achieve material fusion, these prints were made using a 60% infill density and minimal layer overlap. As discussed in Section 3.2, fabricating solid-infill prints in this manner was instead possible because the support matrix cured when mixed within the printed ink. By comparing solid-infill prints in this study to similar prints fabricated by Romero and Romero et al. [56–57], it is further evident that the curable support matrix improved material adhesion within prints and thus resulted prints with better geometric fidelity and spatial resolution.

5.3. Print Stiffness & Failure Strain

We compared the tensile test results of this study to those presented by Romero and Romero et al. [56–57] to determine how printing in our support matrix affected the tensile properties of silicone. This comparison was possible because both studies used the same silicone mixing ratio, specimen dimensions, and tensile testing procedure. The major difference between the tests was the support matrix: Romero and Romero et al. [56–57] printed cube specimens within a micro-organogel support matrix [53] that was formulated for printing silicone, whereas we printed specimens within the curable support matrix. Another difference between the tests was the amount of UV catalyst in the ink: Romero and Romero et al. used a base:catalyst ratio of 10:1 whereas the present study used a ratio of 15:1 (the base+catalyst:Thinner ratio of both studies was 1:3). Because of this material difference and the inherent variability of silicone between mixtures, direct comparison of material values between studies was not feasible. However, comparisons can be made of the relative differences between the specimens cast and printed in each study.

For 1 cm cube specimens printed within the micro-organogel support matrix, Romero reported average elastic moduli (calculated using engineering stress and fitting a line to the stress-strain data from 0 to 10% strain) of 30.4 kPa, 5.2 kPa, and 11.0 kPa for cast, perpendicular-oriented printed, and parallel-oriented printed cube specimens, respectively [56]. Thus, printing within the micro-organogel support matrix decreased the modulus in the perpendicular and parallel directions to 17% and 36%, respectively, of the cast modulus. Romero and Romero et al. concluded that the decrease in modulus was a result of the printing pattern and trapped uncured support matrix material within the printed cubes [56–57]. A comparison between our work and that of Romero and Romero et al. indicates that printing silicone within the curable support matrix results in prints with more material isotropy and a higher relative stiffness in the perpendicular direction. This suggests that the material properties of the curable support matrix help mitigate some problems previously observed when 3D printing silicone, including the poor adhesion between print paths and layers. Furthermore, it is evident that the decrease in modulus when printing within the curable support matrix is a result of a higher concentration of Silicone Thinner rather than because of uncured material within the prints, as observed by Romero and Romero et al. Comparison of failure strain results with Romero and Romero et al. [56–57] was not directly possible, as those studies did not include specimen failure tests. They did, however, note that observable layer separation occurred during testing that required specimens to be discarded after testing. Because their specimens were only tested to 50% strain, we predict that printing within the curable support matrix allows prints to undergo significantly greater amounts of strain before failure.

Hinton et al. [58] printed solid-infill alginate dog bones and compared their stiffness and strain-to-failure properties with those of cast specimens. The printed material had a strain-to-failure that was approximately half of the cast material, and an elastic modulus about nine times lower than the cast material. They noted that part of the difference was because the dog bones were printed with 50% infill, which effectively reduced the true cross-sectional area and introduced voids within the prints. By comparing these results to those reported in this work, it is again evident that the curable support matrix improved material adhesion within prints and thus resulted in printed material properties which were closer to their cast counterparts.

5.4. Matrix Rheology

As in previous studies, the rheology of the support matrix in this study was formulated to support the printed ink during the printing and curing process. In Section 3.4, the rheology of the present support matrix was reported and compared with other studies. In general, the yield stresses of the present support matrix were close to those in previous studies, though measurement methods varied between studies. Similarly, the thixotropic recovery time in the present study fell between values reported in other studies, though variations in measurement methods likely affected this comparison as well.

Significant previous work has identified desirable parameters for support matrices in complete-matrix-cure embedded and removable embedded 3D printing processes [44–45]. The principles and parameters in these studies were used as guideposts during the formulation of the support matrix in this study. Because the purpose of this study was to demonstrate the curable nature of the support matrix and report the material properties of printed specimens, support matrix rheology was not significantly tuned. As such, further development of the support matrix will likely improve print results. This investigation could include tests to determine if different yield stress values or a faster recovery time would improve print performance.

6. Conclusions

In this work we have introduced a curable support matrix for removable embedded 3D printing of silicone elastomers which solves the problem of uncured, trapped support matrix within solid-infill prints. The curing property of the support matrix was tested by observing the mid-sections of printed cubes. The stiffness and elongation at break for cast and printed cubes were determined by tensile testing and these results were compared to previously-reported data from cubes printed within a comparable non-curable support matrix. The rheological properties of the support matrix were tested and compared to similar support matrices. Additionally, the capabilities of this printing process to create solid-infill prints was demonstrated by printing various solid-infill shapes from four UV and addition-cure silicone inks.

This study built on previous work by formulating a support matrix to cure when mixed within the printed structure but not elsewhere within the reservoir. As was discussed, the support matrix became trapped within prints due to the printing process and 60% infill density. The support matrix then cured within the print to enable strong material adhesion within the printed material. Evidence of mixing was observed and was likely critical to the support matrix curing. Further investigation, including flow visualization, will be necessary to understand the effects of material mixing on print performance and support matrix curing.

Tensile testing was performed on printed and cast cube specimens. Results showed that the printed cubes had a very low, nearly isotropic elastic modulus of approximately 20 kPa in directions perpendicular and parallel to the printed layers. As was discussed, the lower stiffness of printed specimens compared to the cast specimens was primarily due to cured support matrix within prints. If an increased final stiffness were desired, prints could be made with inks with a higher cured stiffness, as shown in the demonstration prints. Printed cubes exhibited a higher failure strain than cubes cast from the same material, indicating strong material adhesion within prints. These material properties suggest the potential application of this printing process in many fields of research such as for fabricating soft robots [1,41,43,48,73–74] stretchable electronics [48,75–81], and biomimetic models [13–16].

The support matrix provided support to the ink during the printing and curing processes. The shear-thinning viscosity of the matrix was close to that of complete-matrix-cure embedded 3D printing processes. The yield stresses and thixotropic recovery time of the matrix were within the range of other published values, but these values could likely be improved in future work. The successful fabrication of tensile test specimens and demonstration prints highlighted the overall success of the printing process and curable support matrix. The printed geometries ranged from simple cubes to more complex shapes. The solid nature of the prints demonstrated the ability of the support matrix to cure within prints made from different silicone inks. The materials used for printing included four different UV and addition-cure silicones which ranged in uncured viscosity and resulted in a very wide range of cured material properties. The stiffness of prints ranged by nearly two orders of magnitude between 5 and 430 kPa, and their failure strain ranged between approximately 50 and 250%. Further work will be necessary to determine the full range of silicone materials and viscosities that can be printed within the support matrix. We anticipate that this printing process could be used to fabricate novel silicone devices and expanded to other ink-matrix material formulations.

Many limitations were encountered with the printing process. For example, because the support matrix was translucent but not clear, details about the print process were difficult or impossible to directly observe, which slowed development and iteration. Similarly, initial tests indicated that the support matrix would not cure within condensation-cure silicones, perhaps due to the lack of water within the matrix. The relatively low spatial resolution of the printer motors and qualitatively-observed vibration of the extruder likely affected print quality and resolution. Similarly, the extruder setup limited print geometry to shapes requiring limited material retraction.

Potential improvements to the printing process and printer design include formulating the support matrix to be clear instead of translucent and adding retraction capability to the extruder to improve print resolution and performance. Other silicone inks [81] could be used to achieve low stiffness without problems we occasionally encounter with the current method (e.g., sticky surface, material leaching). We suspect adding fumed silica to the ink could improve resolution and decrease defects in the ink structure by giving the ink a yield stress and a shear-thinning viscosity, as has been done in select direct ink writing [39] and complete-matrix-cure embedded 3D printing methods [43,46,48]. We also suspect that this printing process could be capable of multi-material printing by printing with an array of needles as has been used to print silicone and other inks via direct ink writing [41]. Furthermore, improvements in resolution and geometric fidelity can likely be achieved with improved print settings and higher-precision equipment.

Highlights:

• Centimeter-scale, solid infill shapes are printed from soft silicone elastomers.

• A liquid silicone ink is printed via extrusion into a curable support matrix.

• The removable embedded 3D printing process is demonstrated using four silicone inks.

• UV and addition-cure silicone prints exhibit failure strains from approx. 50 to 250%.

• Prints exhibit a nearly isotropic elastic modulus ranging from five to 430 kPa.