UV-curable silicone materials with tuneable mechanical properties for 3D Printing

a file of such of a page and metadata, and formatting for readability, but the definitive version of This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Abstract In this paper, we present the development of a family of novel, UV-curable, highly flexible, 3D printable silicone-based materials, the mechanical properties of which can be tuned simply by varying the ratio of the polymerisable reagents within the formulation. This tuneability is achieved by exploiting the balance between in-cure phase separation and differential reactivity within the formulation to successfully produce composites structures via both casting and valve-based jetting processes. The structure and properties of both cast and 3D printed materials were examined in a range of compositions of silicone to acrylate between 30:70 and 70:30. The phase segregated structure, evidenced from distinct glass transitions, and the thermal stability of these materials were both shown to be insensitive to the composition ratio, whereas the elastic properties were strongly dependent on the composition. The stiffness could be made to vary from ~50 kPa to ~180 kPa by increasing the silicone content. This study will guide the formation of a new generation of Additive Manufacturing (AM) silicone elastomeric functional structures for various applications ranging from flexible electronics to regenerative medicine, which will benefit from local changes in mechanical properties within the same material family.


Introduction
Silicone elastomers are members of the organosilicone family of compounds and can be obtained via crosslinking functionalised polydimethylsiloxane (PDMS) at either elevated temperatures or at room temperature. The most common silicones used in AM are the room temperature cured silicones which utilize one of three curing reactions: (a) hydrosilation reaction in the presence of catalyst (addition cure system), (b) moisture curing reaction (condensation cure system), or (c) photopolimerisation in the presence of photoinitiator (UV cure system). The chemical structure and material properties of silicone elastomers make them useful in a wide variety of applications and, depending on which mechanical properties are of interests, different silicone formulations are used. In applications such as automotive, electronics, energy absorbers or thermal insulation [1][2][3], where stiff materials are required, silicones with high molecular weight which can be crosslinked in the presence of a platinum catalyst are used, and very often high surface area silica are added as a reinforcement agent [4]. For applications such as soft robotics [5], biomedical devices [6] and stretchable electronics [7], softer platinum-free formulations, void of silica, have been developed [8]. A range of soft silicones, thermoplastics [9] and thermosets [10], have been reported, including self-healing dielectric elastomers [11].
Recent developments in Additive Manufacturing (AM) technology have delivered the potential to fabricate components with complex, detailed and precise structures, thereby broadening the range of applications for various materials [12][13][14][15]. Thus, AM is becoming a viable alternative to traditional moulding techniques commonly used for silicone materials [16][17][18][19]. In practice, the principal AM technique capable of handling highly viscous fluids, such as silicones, is material extrusion [20][21][22][23][24][25][26][27]. However, the relatively limited complexity of the parts that can be produced with this technique and its low production efficiency present significant limitations.
Alternative methods such as Freeform Reversible Embedding (FRE) offer contact-based support-free solutions for the 3D printing of silicone [28]. Although FRE produces uniform, high-quality silicone features, its low throughput limits industrial use and its contact-based nature requires extensive and complex planning strategies.
In addition to material extrusion, a variant of the Material Jetting method, known as "valvebased" jetting, also enables the processing of high viscosity materials and feedstocks. Valvebased jetting can offer both high resolution and a relatively high throughput of material.
Compared to the contemporary layer-by-layer AM methodologies such as powder bed fusion that enable complex geometries associated with AM, material jetting offers the ability to codeposit materials within a pre-defined voxel pattern [19,20] thanks to the potential of using multiple heads. Valve-based jetting differs from traditional material jetting that uses thermal [29] or piezo-electric [30] inkjet heads by using micro-dispensing valves [31], that work on a combination of pneumatic and mechanical actuation. This allows discrete volumes (nl) of viscous fluids to be dispensed on demand. In our previous work, we have shown that high viscosity silicones can be processed with this technique and that latticed silicone structures can be successfully fabricated [31,32]. The technique has seen some commercialisation recently through Wacker with a similar technique for printing high viscosity materials [33].
Despite progress high viscosity valve-based jetting there has been limited activities on the development of silicone materials specifically for jetting methods. Studies focused on the viscoelastic properties of printable silicones [34,35], their mechanical properties [36] or biocompatibility [37] are available. However, these studies are on modified commercial formulations and do not investigate compatibility between the material and processing method and the consequent impact on component properties.
Silicones compatible with various AM methods have been reported to be two-part, temperature cured systems [22,24,38] and single component UV curables [39][40][41]. The advantage of using UV light to cure materials during printing surrounds the reduced printing time due to faster droplet pinning [42]. There are various formulations of UV-curable silicones with potential for AM [28,[34][35][36]43]. However, it is difficult to either adjust the chemical properties to tailor the final mechanical performance of the constructed device for a given application, and/or achieve the rheological properties appropriate for inkjet or valve-based printing. Therefore, developing a printable silicone formulation with easily tuneable mechanical properties would broaden their adoption within the AM community and beyond, accelerating adoption of AM technology and realising new innovative products based on material and design complexity.
Here we present novel UV curable silicone elastomers which successfully exhibit both the rheological properties for valve-based jetting and tuneable mechanical properties when part of the final desired structure. Furthermore, we demonstrate that this material property control is predictable by simply changing the ratio between the acrylate and silicone components within the structure. The influence of the silicone to acrylate ratio on curing behaviour, structural characteristics, and thermal properties are also reported, and linked to the macro material behaviour. Additionally, a hyperelastic rubber material model [44] was fitted to the mechanical data to understand the influence of composition on the crosslinked network achieved during cure.

Materials
Di-functional methacryloxypropyl terminated polydimethylsiloxane (PDMS-DMA), monofunctional 2-ethyl hexyl acrylate (EHA) and bis(2,4,6-trimethylbenzoyl)-phenylphosphine oxide (Irgacure 819) were utilised in this work. They were purchased from Gelest, Sigma Aldrich and BASF, respectively. All chemicals were used as received. Typically, silica fillers are incorporated into silicone elastomers as reinforcing agents. In this study silica was not added because of the complexity that would arise, especially in the context of 3D printing.

2.2.
General procedure for the synthesis of the paste formulations  intensity and a wavelength of 395 nm. The FE400 unit is identical to that used in the valvebased jetting system. The distance between the mould and the UV light was ~ 10 mm. The prepared samples were further characterised and mechanically tested.

Molecular characterisation: FTIR
To characterise the chemical properties of formulated and cast silicones samples before and after curing to determine the reduction in the quantity of vinyl groups, FTIR-ATR spectroscopy analysis has been performed. ATR spectra were measured on a Frontier Fourier Transform spectrometer (Perkin Elmer) based on the attenuated total reflection method [45]. Samples were placed directly on the 3mm in diameter diamond ATR crystal (Pike GladiATR module).
Absorption spectra of the samples were recorded at a constant temperature (25 °C) in the range 500-3500 cm -1 , at a resolution of 4 cm -1 . A background measurement was run for each sample before its FTIR analysis. The FTIR-ATR spectra were recorded using Spectrum software. A minimum of 3 samples were tested for each uncured and cured composition.

Cured Samples
TGA analyses were also conducted on cured and cast samples using the process highlighted above for the uncured components. A single sample was analysed in each instance.
DSC analysis was used to define the glass transition temperatures of the various cured g compositions. They were determined using a Linkam DSC600 in conjunction with the LNP96 liquid nitrogen cooling system. This setup was necessary to achieve a starting temperature of -150 °C. As this system is a single cell calorimeter, DSC count (proportial to heat flow) is measured. Samples of ~5 mg were loaded onto aluminium pans and placed into the crucible.
Following a N 2 purge, the samples were cooled to -150 °C and subsequently reheated at a rate of 5 °C min -1 through to a temperature of 20 °C. A minimum of 3 samples were tested for each composition.

Swelling Study
The sol content and swelling fraction were measured by the mass differential after incubation of the polymer network in toluene. First, polymer cubes (2 mm × 2 mm × 2 mm) were cut from crosslinked cast samples. The cubes were weighed to find the initial mass ( ), and i W immersed in toluene for 48 hours. The samples were removed from toluene and dried. Having removed any unreacted polymer, the samples were weighted again to determine the dry mass ( In practice, the process of obtaining the thickness measurement was complicated by the viscoelasticity and high compliance exhibited by the elastomers as the rubber thickness gauge probe applied a small force, causing the material to deform. Therefore, to calculate the thickness in the absence of the probe force, the elastic solution developed by Lebedev and Ufliand was used to correct for the deformation [48].
For tensile testing, rectangular specimens were attached to the drums of the SER 3-P apparatus using staple-like clips, strain was applied to the specimen by the counter-rotation of the drums as illustrated by Fig. 2. A small torque of 10 μN•m was applied to ensure that the specimen was straight prior to the test. The tests were carried out at a rotational speed corresponding to a constant true strain rate of 0.1 s -1 at room temperature (20 ± 1 °C). The raw data obtained from the tests can be utilised to determine the nominal strain  and the nominal stress . The former is defined as tru e e x p ( where t is the time, and the latter is defined as where T is the torque, a is the thickness of the specimen, w is the width of the specimen, R is the radius of the drums and  = (1 + ) is elongation.

General procedure for material jetting
A custom-built apparatus was utilised for the valve-based jetting process. It consisted of a moveable stage which was synchronised with stationary Nordson EFD jetting valves, as presented in [31]. Opening and closing of the nozzle is controlled by a driving pulse as highlighted on  The PDMS-DMA forms branching within the 3D cross-linked polymer network, the EHA acts as a diluent and lowers the formulation's viscosity, as well as reducing the branching density; Irgacure 819 is responsible for radical formation by absorbing UV energy.
FTIR analysis was performed both before and after UV polymerisation, to observe the changes in molecular structure that resulted from altering the composition and curing regime.
Representative FTIR spectra are highlighted in

Thermal Analysis
Thermogravimetric analysis (TGA) was conducted to study the thermal stability of the silicone samples prepared with different ratios of reagents. The TGA results are shown in Fig. 6 and the degradation temperature, which is defined as the temperature at the peak of the derivative of weight loss vs temperature, is listed for all the samples in Table 1 As can be observed in Fig. 6, the degradation of EHA monomer occurs below 200 °C while the dominant degradation step for unreacted PDMS-DMA starts at around 390 °C which corresponds to the split of Si-CH 3 bonds, followed by complete loss of pure PDMS-DMA above 540 °C [42]. The weight loss observed in the initial stage before 390°C is most likely related to the small percentage of low oligomer molecular weight present in the PDMS-DMA.
Once PDMS-DMA is mixed with EHA, and the mixture is cured, the degradation temperature is seen to increase when compared to pure EHA, suggesting the formation of cross-links between the components. The cross-linked polymers decompose in two stages. The mass loss of the cross-linked samples differs depending on the ratio between silicone and acrylate ( Table   1). A higher proportion of the mass was lost at the second peak temperature when more silicone was present in the composition.  [51] that the thermal degradation behavior of silicone elastomers obtained via a hydrosilylation reaction between the vinyl-terminated polydimethylsiloxane and the hydride-functional crosslinker is affected strongly by the stoichiometry of the network reactant and the degree of crosslinking. In the photopolymerised materials used in the current work, the characteristic peak temperatures are observed to be similar for all compositions. The first peak can be seen around ~ 420 °C and the second peak around ~ 485 °C. At present the reason for this is unknown.

Curing Behaviour
The curing behaviour of the UV curable polymer formulations was also found to influence the printing process and the final material properties. Fast curing of the material was predicted to contribute to improved print quality, resolution, and ease of fabrication of self-supported structures [42]. Furthermore, the degree of cure achieved during the reaction was also expected to influence the properties of the cross-linked polymer network, for example excess unreacted monomer may act as a plasticiser, or low cure may result in a lower branching density.
To characterise the curing behaviour, DSC analysis was conducted. Fig.6 presents the UV-DSC heat flow curves as a function of time for five different uncured silicone samples.
Negative values were observed for the heat flow, indicating that the chain polymerization of C=C bonds in the (meth)acrylate groups was exothermic, as would be expected. This is a consequence of the conversion of π-bonds in the monomer into σ-bonds in the polymer after curing [49].
where is equal to 86 kJ mol -1 , is 54.4 kJ mol -1 and r is the mass ratio of EHA 0 PDMS -DMA 0 the PDMS-DMA in the mixture. Additionally, since the PDMS has two methacrylate groups per molecule a factor of 2 is present outside the second square bracket [54,55]. determined is a similar fashion to α c and differs only by the fact that a factor of 2 is not present outside the second square bracket in equation (7). The data shows that the greater the silicone content in the sample, the smaller the exothermic peak. This was attributed to: (a), the limitation in radial diffusion caused by an increase in viscosity of samples with higher silicone content and; (b), the overall reduction in methacrylate functional groups present due to the larger molecular weight of the silicone reagent. Furthermore, the data in Table 2 shows that the conversion of acrylate bonds is similar in all of the samples, once the difference in molecular weight is taken into account, and reaches around 60% at completion. Interestingly, most of the reaction was completed after ~30 s. Thus, these results suggest that any observed changes in mechanical properties for different compositions will arise from differences in crosslink density within the cured material. Table 2 The mass ratio of PDMS-DMA in the mixture r, reaction enthalpy , total enthalpy Δ , the experimentally measured (α c,expt ) and molar corrected (α c,corr ) degree of cure for 0 various silicone:acrylate compositions.

Post-cure thermal properties
Additional DSC tests were conducted to attempt to determine the glass transition temperatures of the cured compositions. The transition temperatures were obtained from the peaks in the derivatives of the DSC count vs temperature curves, which are provided in the supplementary section (Fig. S6). The transition temperatures identified from these curves are shown on Fig.   8. respectively [56,57]. The observed discrepancy between the transition temperatures of the distinct phases and the pure polymers could be attributed to factors including, but not limited to, the instrument of choice, the heating/cooling rates and the molecular weight, but is unlikely to be associated with the degree of phase separation as no effect of component ratio was observed.
Further support of the existence of a phase separated structure is presented in the supplementary section ( Fig.S7-S11), where similarities between the TGA decomposition profile of the compositions and the profiles produced from the individual components scaled by the corresponding mass ratios are shown [58]. The final product structure will be a result of a combination of both, tendency for these formulations to rapidly phase separate during curing and fast reaction kinetics of EHA. The reaction kinetics of the EHA monomer is expected to be faster than PDMS-DMA, due to the steric bulk of both the full monomer structure and that around the reacting radical centre. Thus, the final structure will take the form of very lightly cross linked or thermoplastic EHA regions existing in conjunction with heavily branched /crosslinked PDMS-DMA domains. The observations from the subsequent swelling experiments (Fig. 5) conducted on cured specimens indicated that there are no thermoplastic EHA regions as no part of the material structure was observed to dissolve out. It is concluded that an extended three-dimensional network has been created by crosslinks between the surface of the domains and the matrix polymer.

Mechanical properties and modelling
The average uniaxial tensile stress-strain responses obtained from a minimum of three specimens per sample for all the silicone:acrylate systems explored in this study are illustrated in Fig. 9, and the error bars represent ± 2 standard deviations. A table of tensile properties has been presented in the supplementary section, see Table S1.  to the data. Although there are numerous applicable models in the literature, an Edwards-Vilgis (EV) model was selected due to the physical origin of its parameters [44]. Like many of the rubber models in literature, the EV model is intended for homogenous elastomers. Nonetheless, such models are frequently employed to successfully model the behaviour of heterogenous and filled elastomer systems, although interpretation of the parameters becomes more challenging.
The strain energy function of the EV model has the form: is Boltzmann's constant, T is the absolute temperature, is the cross-link density, Assuming isochoric deformation during the uniaxial tensile tests, the principal stretches are given by , where is the tensile stretch imposed on the specimen 1 / 2 1 2 3 a n d during the test [59]. The stress is obtained by applying these to equation (8), and differentiating with respect to . Model parameters were obtained by minimising the root-mean-square  (RMS) error between the experimental and model stress using the MATLAB lsqcurvefit function. In all cases, the EV function provided an excellent fit to the experimental data with RMS errors not exceeding 1.8kPa. The evolution of the EV model parameters is shown in Fig.   10 as a function of the composition.
As the EV function is relatively insensitive to the split between cross-links and slip-links, Fig.   10a reports the total N S + N C , which can be observed to increase systematically with increasing silicone content. This is attributed to the ability of the two methacrylate end groups of the PDMS to form cross-links with the mono-acrylate chains. An increase in cross-link density can also lead to an increase in the entanglement density, as there is a greater likelihood of trapped entanglements being present between cross-links [59]. The network densities of the different compositions were also determined on the basis of the swelling study, resulting in log 10 (N S + N C ) values ranging between 25.4 and 25.7 as the silicone content is increased from 30 to 70 wt%. The corresponding calculations have been provided in the supplementary material.
Although the change in network density with composition measured through swelling is more limited than that observed in the EV model, the trends are similar.  constraints increases, the chain length between these constraints decreases, leading to a reduction in the limiting elongation. Lastly, Fig. 10c shows that the slip-link mobility increases with increasing silicone content, most likely attributed to the greater flexibility of the silicone chains relative to the acrylate chains.
By expressing the evolution of EV parameters as simple functions of composition, a predictive model can be deployed as a tool to enable the fine tuning of properties for various applications.
As an illustration, Fig. 11 shows the measured (± 3 standard deviations) and model derived secant modulus E s,100% , a value typically quoted in material data sheets. E s,100% has also been included in the table of tensile properties provided in the supplementary section, see Table S1.
An increase in silicone content from 30 to 70 wt% increases E s,100% by a factor of ~3. 6 Fig. 11 The experimentally measured (symbols) (± 3 standard errors) and the secant modulus E s,100% predicted from the model (line) as a function of composition.

Validation of 3D Printing Capability
The custom material jetting apparatus described in section 2.5 was utilised to showcase how the formulation could be used to 3D Print initial arbitrary shapes. An example of a successfully printed silicone sample of the 70:30 composition is shown in Fig. 12. On closer inspection it can be observed that the surface is not completely smooth and air bubbles are present. Whilst material and process optimisation is still required, Fig. 11 shows the ability to use these formulations for 3D Printing.
These initial trials have shown that this family of flexible, tuneable materials can be used in 3D printing. The formulation strategy employed, in one step, creates a phase separated, lightly crosslinked structure. Considerations for future work will focus on the further printing process optimisation, and demonstrating the capability to produce 3D structures with locally tuned mechanical properties varying across all three dimensions by using multiple nozzles.

Conclusions
In this work, we have reported the creation of a family of novel UV-curable, highly flexible silicone compositions which are suitable for use in a valve-based jetting process. Furthermore, the design of these formulation successfully exploits the balance between in-cure phase separation and differential reactivity to produce composites structures with tuneable materials properties. Property tuning is achieved by simply varying the ratio of the polymerisable regents within the formulation. Thus, taking advantage of multiple jetting heads and correlation between the material composition and mechanical properties, it might be possible to fabricate functional devices with locally tuned, highly flexible mechanical properties via 3D printing.
However, more tests are required to make sure that different silicone:acrylate ratios are interfaceable. The formulation strategy is proposed to create a highly phase separated, lightly crosslinked structures that delivers the beneficial material properties by jointly inspiring and limiting the crosslink levels achieved and variability is created by the relative domain: matrix ratio.
The FTIR spectra of the samples before and after curing confirmed successful crosslinking reactions within all the compositions. The DSC study showed that the reaction rate varied from composition to composition as expected due to the differentiated reactivity of the two polymerisable components. However, this had no discernible impact on the overall time of the reaction, which remained within 30 s to reach the maximum cure level. The conversion of the C=C double bond reached ~ 60% for all of the samples, indicating that the observed changes in mechanical properties results from difference in crosslinking density. The presence of two distinct glass transitions was observed during DSC tests indicating that the sample is phaseseparated into silicone and acrylate phases. The TGA study demonstrated the first decomposition peak was observed to be similar for all composition and it was ~ 420 °C.
The results from the mechanical tests suggest that varying the silicone content results in a systematic variation in the stress-strain response. PDMS network. An EV model was fitted to the experimental data to study this variation. The network density is seen to grow with increasing silicone content due to the higher number of available reaction sites. This shortens the chain length, and hence the limiting extensibility is seen to drop. The increase in slip-link mobility with silicone content is attributed to the greater flexibility of the silicone molecule. It was shown that the model could be helpful in tuning the mechanical properties of this particular group of materials. A more detailed investigation into the mechanical properties will be required prior to application in the aforementioned applications, in particular the reversibility of the mechanical response, and the resistance to crack propagation. Lastly, an initial assessment was performed to validate the 3D printing capability of these materials, and results indicate that the material is printable. The work opens a possibility of employing developed silicone:acrylate formulations for producing 3D structures with locally tuned mechanical properties within the same material. The novel UV-cured silicone materials are softer than the classical platinum-catalyzed addition cured PDMS materials, and will be of particular interest in applications such as soft robotics, where soft elastomers with complex geometries are required.