Dielectric elastomer actuators (DEAs) are one of the soft actuator technologies commonly used in the soft robotics fields (Brochu and Pei, 2010; Anderson et al., 2012; Romasanta et al., 2015; Rosset and Shea, 2016; Gu et al., 2017; Hines et al., 2017; Rich et al., 2018; Shintake et al., 2018; Gupta et al., 2019) due to their excellent characteristics: large actuation strains [e.g., areal strain more than 1,000% (Li et al., 2013)], fast responses [e.g., actuation at 600 Hz (Ji et al., 2019)], and theoretically high electromechanical efficiency of up to 90% (Brochu and Pei, 2010). Simple structure and electrically driven nature of DEAs enable the fabrication of devices of various sizes, ranging from millimeter to meter scale (Rosset and Shea, 2016; Gu et al., 2017).

A typical DEA comprises an elastomeric membrane sandwiched between two stretchable electrodes. When subjected to a high voltage potential, electrical charges induced in the electrodes attract each other and generate an electrostatic force called Maxwell stress. Subsequently, the membrane shrinks in the thickness direction and expands in the planar directions. Moreover, elastomeric membranes used in DEAs are usually thin, ranging from 1 µm [a silicone elastomer, reported in ref (Ji et al., 2018)] to 1,000 µm [acrylic elastomer, VHB4910 (M Coporation., 2018)]. Therefore, multiple sets of the membranes and electrodes are stacked to yield useful actuation strokes in the thickness direction, like a muscle. Previous studies have presented various methods for fabricating stacked DEAs where the dielectric and electrode layers are alternately staked (Kovacs et al., 2009; Matysek et al., 2011; Ji et al., 2019), blade-casted (Li et al., 2018), spin coated (Lotz et al., 2011; Duduta et al., 2019), printed (Araromi et al., 2011; Reitelshöfer et al., 2016; McCoul et al., 2017; Chortos et al., 2020), or folded together (Carpi et al., 2007; Nguyen et al., 2014). However, these fabrication methods often require a dedicated setup with complicated processes or sometimes require laborious manual stacking of the layers, which may increase the failure rate of the actuators.

Herein, we propose a method to monolithically fabricate stacked DEA without alternately stacking the dielectric and electrode layers. In this method, stacked DEAs are mainly fabricated through two steps: 1) molding of an elastomeric matrix containing free-form microfluidic channels and 2) injection of liquid conductive material. The microfluidic channels possess the shape of stacked electrodes, which are formed by dissolving an inner part suspended in an elastomeric matrix. Owing to the use of molding process, the shapes of both the electrodes and elastomer structure can potentially be tailored by the geometry of the molded parts. Therefore, the proposed method is expected to provide high design flexibility for the creation of stacked DEA in various forms. Additionally, the relatively simple process may contribute to minimizing the failure rate of the actuators and the required fabrication setup. In this study, we investigate the feasibility of our method by fabricating and characterizing monolithic DEAs with multiple electrodes (2, 4, and 10). The fabricated actuators are characterized in terms of actuation stroke, output force, and frequency response.

The method proposed herein for fabricating monolithic DEAs combines two technologies employed in the microfluidics field: scaffold removal and vacuum filling. The former is a method used to create empty microfluidic channels by dissolving three-dimensional (3D) printed components suspended in an elastomeric matrix (Saggiomo and Velders, 2015). The latter is a hands-free injection technique where a liquid metal [eutectic gallium–indium, EGaIn (Dickey et al., 2008)] is pushed into microfluidic channels with dead-ends using a suction force because of the pressure difference inside and outside the channels (Lin et al., 2017). Liquid metals are compatible with DEAs when applied to DEAs as electrodes (Liu et al., 2013; Wissman et al., 2014; Pan et al., 2019; Piskarev et al., 2020; Tahidul Haque et al., 2020). Figures 1A–C depict images taken at different stages of the fabrication process of a monolithic DEA containing four electrodes, and Figure 1D represents the process flow. As shown in Figure 1E, the microfluidic channels are aligned such that opposing polarities face each other across an elastomer domain that acts as a dielectric layer. Subsequently, the entire structure functions similar to stacked DEAs. Figure 1F displays a monolithic DEA with 10 electrodes fabricated through the same process. Extension of both the inner part and mold allows the realization of an actuator with a high number of electrodes.

The experimental results demonstrate the successful implementation of the proposed method for fabricating monolithic stacked DEAs, from which the objective of this study—validation of the proposed method—was achieved.

Overall, the theoretical model, which considers the change in mechanical and material properties, well captures the behavior of the actuators, suggesting that the model is effective at designing actuators made using the proposed method. There are several possible explanations for the discrepancies between the theoretical and experimental values. The first is manufacturing error. If the thickness of the dielectric part is small, the electrostatic stress will increase and the actuation will be larger. The second explanation is that the changes in the mechanical and material properties due to the solvent immersion may be nonuniform throughout the structure. In this case, the dielectric part may be softer than that modeled, and thus, the observed actuation strokes will be larger. This effect may be significant, especially for the actuators with 10 electrodes because they have a large volume. The third is a slight reduction in the actuator volume due to the extraction of uncross-liked material during the immersion in the solvent (Glover et al., 2020). This may make the dielectric layers slightly thinner and, in turn, the actuation strokes larger due to an increase in the electrostatic stress. Lastly, while these three scenarios would increase the actuation strokes, the presence of passive elastomeric domains in the actuators would reduce the strokes. Therefore, the measured data likely represents a combination of these factors. Regarding the experimental results, the resonance observed in the frequency response characterization suggests that the actuation stroke can be magnified and controlled by matching the driving frequency. This characteristic may be particularly useful for devices that perform periodic motions, such as mobile robots and vibrating devices.

The actuation strain values observed for the actuators fabricated in this study range from 0.4 to 1.0% in an electric field of ∼20 V/μm. These numbers are noticeably small compared to those of stacked DEAs fabricated by other methods: 46% at ∼50 V/μm (Kovacs et al., 2009), 24% at 150 V/μm (Duduta et al., 2019), 15.5% at ∼12 V/μm (Carpi et al., 2007), 9% at 25 V/μm (Chortos et al., 2020), and 4.5% at ∼13 V/μm (Li et al., 2018). There are two main reasons behind this. The first is the fact that the electric field applied to our actuators is low (∼20 V/μm). The electric field is limited by the experimental equipment, which allows applied voltages of up to 10 kV. Applying higher electric fields below the breakdown strength of the elastomer leads to larger actuation strains. To produce this effect, one could simply use higher voltage or reduce the thickness of the dielectric layers. In the latter case, the thickness of the dielectric layers depends on the minimum allowable 3D-printed layer thickness for the scaffold. This minimum thickness depends on the performance and type of 3D printer, but a value around 100 μm can be reasonably expected given that the printer used in this study has a minimum layer height of 50 μm (Stano et al., 2020) and a dimensional accuracy of 50 μm (Msallem et al., 2020). However, as the performance of 3D printers continues to improve, the achievable minimum dielectric layer thickness in our method is expected to become smaller. The second reason for the low actuation strains of our actuators is that the Young’s modulus of dielectric layers (1.1–2.6 MPa) is significantly higher than that used in other studies. For example, Ecoflex 00-20 (Smooth-On), which was used by Li et al. (2018), has a Young’s modulus is ∼0.05 MPa (Jin et al., 2021). Within the same magnitude of electrostatic stress, the strain is greater for softer elastomers. Therefore, by identifying and incorporating softer elastomers that are compatible with our method, the actuation strains of our actuators can be expected to become larger. Candidates for such elastomers include the Ecoflex series described above and Sylgard 184 with a modified mixing ratio (Glover et al., 2020). It should be noted that this work will also require the exploration of suitable solvents and the investigation of their influence on the mechanical, material, and electrical properties of chosen elastomers.

Since this study aims to validate the feasibility of the proposed method comprising 1) molding of an elastomeric matrix containing free-form microfluidic channels and 2) the injection of a liquid conductive material, the shape of the tested actuators is limited to the simplest form. However, owing to the simple molding process and the use of 3D printed parts, our method provides high design flexibility; thus, it can be potentially applied to stacked DEAs of diverse forms. These actuator shapes include cylinder, triangle pole, ellipsoid, and other 3D free forms. In future studies, we will investigate the design flexibility of our novel method and the applicable materials as well as the extension of the model to predict the characteristics of actuators for a given specification. To assure prediction accuracy, the model will also consider the presence of passive elastomeric domains in the actuators as well as the volume reduction by immersion in the solvent. Related to this, the fabrication parameters should be optimized in order to reduce the non-uniformity of the mechanical and material properties within the actuator structure. In parallel, we will clarify how far the actuators’ performance can be improved, while considering the resolution of the mold and suitable materials. Lifetime and reliability are particularly important aspects for evaluating actuation performance. With the aid of future work mentioned above, our method is expected to derive DEA-based soft robots and devices of various forms in diverse applications.