MEMS-compatible structuring of liquid crystal network actuators using maskless photolithography

Figure 1(b) illustrates the chemical structures of compounds used in this work for fabricating the LCN actuator. Actuators are made by photopolymerization of mixtures consisting liquid crystal mesogens, crosslinkers, photosensitive elements, and photoinitiators. The measured absorption spectra of the photosensitive molecules for light-driven actuation are shown in figure 1(c).

For light-driven actuation of LC network, both photothermal effect and photochemical mechanism are used in this work. The difference in the actuation mechanism stems from different lifetimes of cis-isomers of used molecules M5 and M8, which are in the order of hours and seconds, respectively [33].

Molecules M2, M4, M7 and M8 are used to form photothermally actuated LCN actuator, in which a commercially available azobenzene-derivative Disperse Red 1 acrylate is used. Upon illumination, due to its short cis-lifetime, the molecules undergo fast trans-cis isomerization, leading to heat generation, and subsequent shape change of actuator [34, 35]. Molecules M2, M3, M4, M5 and M6 are used to form the photochemically actuated LCN actuator, in which, upon illumination, the bent shape of the cis-isomer disrupts the order of LC network, resulting in shape change of the actuator. Due to long lifetime of this cis-isomer, this shape change is preserved even after turning off the light and the original shape can be recovered by illumination with another wavelength or via thermal relaxation [33]. Detailed information on the chemical structures and composition used for each actuator can be found in the supplementary information, S1.

The LCN films are generated using the glass-cell filling technique. For this, glass substrates (24 × 24 mm) are cleaned by sonication in acetone, isopropyl alcohol, and de-ionized water for 5 min each and subsequently dried with nitrogen. This is followed by $\text{O}_\textbf{2}$ plasma treatment with a gas flow rate of 50 sccm at 0.2 mbar pressure for 5 min.

The alignment of liquid crystal mesogens is done with photoalignment technique, allowing alignment of the mesogens at arbitrary angles in one plane, thereby enabling the multi-directional motion of LCNs in more than one plane [32, 36–38]. The custom-built DMD-based optomechanical setup used in this work allows us to align the LC mesogens in a pixel-by-pixel fashion, with a resolution down to 20 µm [32].

For photoaligning LC meogens, 1 wt% solution of Brilliant Yellow (BY) in DMF (N,N-Dimethylformamide) is prepared and filtered through 0.2 µm PTFE syringe filter. 50 µl of BY solution is then spincoated at 3000 rpm for 20 s and is dried on a hotplate at 90 ^∘C for 5 min. This 5–20 nm thick BY layer acts as the photoalignment layer for alignment of the reactive mesogens. Substrates spin-coated with BY are then exposed to linearly polarized 450 nm light [32]. To obtain bending motion of the actuator, the top and bottom substrates of a glass-cell are aligned at 0^∘ and 90^∘ with respect to the polarizer, resulting in twist-bend alignment configuration [39]. The obtained 90^∘ twist-bend alignment between two parallel polarizers can be found in the supplementary information, S4.

The LC mixture is prepared by dissolving the monomers in dicholoromethane in a glass vial and is filtered through 0.2 µm PTFE syringe filter. The mixture is magnetically stirred at 80 ^∘C for 4 h to remove the solvent. To ensure that the solvent has completely evaporated, the mixture is then put in vacuum oven at room temperature for another 2 h. Glass-cells with different spacing thicknesses are fabricated using double-sided tape (Nitto Deutschland GmbH). The cells are then filled with above-prepared LC mixture at 90 ^∘C via capillary action. The cells are brought to their nematic phase 50 ^∘C at 2 ^∘C min⁻¹. The aligned LC films in the cells are then ready for structuring using photopolymerization as described in the next section.

The subsequent structuring of the LCN actuator is done by selectively polymerizing the network using a maskless photolithography technique. By spatially varying the illumination at the appropriate wavelengths, it is possible to structure the network in any arbitrary shape. Following exposure, the non-polymerized areas are developed in a solvent and thus removed. A schematic description of the process is given in figure 1(a).

The process for defining the pattern for polymerization is same as that for defining the pattern for aligning the BY molecules [32]. The desired image pattern is projected via a DMD projector onto the substrate. By changing the illumination pattern we may structure the network in any arbitrary shape in a pixel-by-pixel fashion. We employ the same projection setup for aligning the reactive mesogens and for structuring the network; the only major difference is that for polymerization the angle of polarizer is kept at 0^∘. We do not expect any effect of polarized light on the azobenzenes during polymerization.

After developing, the structured actuator may then be glued to a 3D-printed substrate via a stamp-and-stick technique [40].

As a proof of concept, different patterns, as illustrated in figures 2(a1)–(a4) were projected onto the substrate, wherein areas corresponding to the black regions will be polymerized. The resulting patterns are shown in figures 2(b5)–(b8) wherein the non-polymerized areas have undergone spherulite-like structure formation, confirming the successful selective polymerization of LC network.

Zoom In Zoom Out Reset image size

The projected image size with the optomechanical setup on the image plane is 4.5 mm. In figures 2(a2)–(a4), the projected image with area of $\mathrm{18\,\times 18\,mm^{2}}$ is split into a matrix of $\mathrm{4\,\times 4}$, yielding 16 squares. Then, each sub-square is individually illuminated and, with stitching, an area of $\mathrm{18\,\times 18\,mm^{2}}$ is thus selectively polymerized.

To quantify the spatial resolution for polymerization, the pattern shown in figure 2(a4) having different square side lengths and spacing between the squares was projected onto the substrate. Again, the black squares correspond to the areas which are polymerized. Figure 2(b8) illustrates the obtained polymerized structure with 100 µm thickness. The polymerized areas that were projected with squares with side lengths less than 1 mm do not have sharp edge (resembling almost a step function), hence observing polymerization in the unwanted areas. This is also evident from the figure 2(b5), wherein the polymerized areas are not of 1:1 size of their projected areas. From this observation, we can conclude that minimum achievable size for structuring the 100 µm thick actuator with the current optomechanical setup is in the order of 1–2 mm. We believe this difference between projected image size and obtained polymerized area, and thereby the limitation on minimum structure size, is not from the optical setup, but rather from how the chain-initiation reactions along the thickness of actuator take place. In comparison, using the process for the much thinner spin-coated LCEs (2–5 µm), the resolution can be on the order of few hundred microns [41]. Hence the structuring resolution is dependent on the thickness of material, as the bulk chain-initiation reactions probably do not allow for the same resolution for selectively polymerizing the material as can be obtained using standard photolithography and etching [29] or with two-photon polymerization [25, 28, 42]. It might be possible to increase the spatial resolution by using stronger demagnifying projection optics with appropriate numerical apertures or by using a DMD with smaller pixels, which would allow patterning with smaller pixel sizes.

For demonstrating the maskless photolithography as summarized in figure 2, LC network containing molecules M2, M4, and M7 with 100 µm thickness was used. To further quantify the selective polymerization procedure for LC mixtures for both photoactuation mechanisms, photothermal and photochemical, the respective LC mixtures (supplementary information, S1) with different thicknesses were also selectively polymerized. The resulting exposure doses are summarized in figure 3.

Zoom In Zoom Out Reset image size

For the photothermally-driven actuator, the photoinitiator (figure 1(M7)) has an absorption range 350–410 nm and for the photochemically-driven alternative, the photoinitiator (figure 1(M6)) has a broad absorption range 250–550 nm (see supplementary information, S2). An LED light source with an emission wavelength of 365 nm is thus used for polymerizing the network for photothermal actuation and an LED with an emission wavelength 554 nm is used for polymerizing the network with photochemical actuation. To avoid activation of azo-crosslinker, a bandpass filter is used to block wavelengths shorter than 520 nm.

As summarized in figure 3, the time required for polymerizing both LC mixtures used in this work is variable. This variation is primarily due to three main factors: photoconversion efficiency of photoinitiators; light intensity reaching at the image plane; and the ratio of acrylates to diacrylates in the mixture. The optical power reaching the sample/image plane for 365 nm LED was 90 µW, leading to higher exposure times. The power of the 554 nm LED was 25 µW at image plane, reduced to avoid overexposing the entire surface.

After selective polymerization, the glass-cells are submerged overnight in DI water at room temperature to easily release the films. After opening the cell, the substrate is submerged in diethyl ether to develop the unpolymerized areas. Diethyl ether is chosen to avoid any potential swelling of the polymerized network, which is a common issue with other organic solvents such as toluene and chloroform. The developing time for a 3 µm thick film is about 5 s and for 100 µm thick LCN actuator, the development time varies from 2–3 min. Finally, the obtained films are rinsed with DI water and subsequently dried with $\text{N}_\textbf{2}$ gas.

To anchor the LCN films, a 3D-printed substrate (Form 3+, FormLabs Inc.) is used as a prototype. To attach the actuator to the printed substrate, Araldite 2020 adhesive is used. This adhesive has low viscosity, making it easier to spin-coat and to achieve a variable thicknesses by changing the spin speed. In this work, 10 ml of the adhesive mixture is spin-coated at 5000 rpm, resulting in thickness of 4 µm ± 0.5 µm. The 3D printed substrate which also acts as a stamp, is dipped into the spin-coated glue and then is attached to the LCN actuator. The curing of adhesive is done at room temperature for 12 h. With this method, it is possible to achieve a uniform layer of adhesive between actuator and substrate.

The above results that we have mentioned demonstrate that by using DMD-based projection setup to structure the LCEs/LCNs based actuators does eliminates the need for using static masks which are cumbersome to create or the need of specialized photoresists that are not easily available. Using our approach of maskless structuring also allows to change the projected pattern onto the DMD within a few seconds, thereby decreasing the required time for processing.

Moreover, additional waiting times in-between the processing steps do not pose any limitation to the above mentioned fabrication capability. For instance, by curing the adhesive at 100 ^∘C, the curing time can be shortened from 12 h to a mere 15 min. However, curing at higher temperatures was not done in our work to avoid any potential thermal actuation. Furthermore, to speed up the release of structured films from the glass-cells, it is possible to use water-soluble sacrificial layers such as PVA or Dextran before spin-coating the photoalignment layer, so that upon submerging the glass-cells in DI water, the dissolution of the sacrificial layer will shorten the waiting time [29, 43].

Additionally, the fabrication capability was demonstrated on small substrates ($24\,\times 24\,\mathrm{mm}^{2}$). However, the optomechanical setup allows stitching the patterns over a full 100 mm wafer. Hence, both techniques, photoalignment and selective polymerization, can be easily scaled up to fabricate the actuators on a full 100 mm wafer. In addition to this, it is also possible to combine standard microsystem fabrication techniques with these processes to further increase the compatibility of LCE/LCN based actuators with the MEMS technology.

Using the processes and materials described above, an array of actuator structures was manufactured and characterized to demonstrate the utility and practicality of these new LCN fabrication processes. By taking inspiration from previously done work [38, 44, 45] for actuating LCEs/LCNs, in this section we describe the actuation characteristics of structured actuators in more detail.

3.2.1. Photothermal actuation.

To demonstrate the feasibility of the methods for aligning, fabricating, and structuring the LCN actuators described above, the LC mesogens are aligned and selectively polymerized in the patterns shown in figure 4(a1). The structure is aligned in 90^∘ twist-bend fashion, meaning that the top and bottom substrates at aligned orthogonally with respect to each other. Due to the 90^∘ variation in the orientation of LCs, the top and bottom surface will experience non-uniform strain rates upon actuation, resulting in an enhanced bending motion [39]. Videos demonstrating this actuation may be found under supplementary material.

Zoom In Zoom Out Reset image size

After assembling the glass-cell with 60 µm spacing and filling in the LC mixture for photothermally driven actuation, the precursor is selectively polymerized and developed to yield the cross as shown in figure 4(a1). Figures 4(a2)–(a4) shows the photothermal response of the actuator when illuminated with 450 nm at 150 mW cm⁻².

In this actuator, as briefly noted above, upon illumination at room temperature, the azobenzene derivative, Disperse Red 1 acrylate, absorbs light and due to its short cis-lifetime, on the order of a few hundred milliseconds [33], the light absorption leads to fast isomerization and subsequently heat generation. This heat is transferred to the LC network, resulting in disordering of the network and subsequent shape change of the LCN. When the LED is switched on, within less than 5 s, the four flaps of the actuator bend and as soon as light illumination is turned off (figure 4(a4)), the actuator retains its original shape, demonstrating reversible actuation.

To more fully characterize the mechanical deformation of LCN actuators, a set of photographs as shown in figure 5(a) is used to analyze the changes in the bending angle. This characterization also helps to analyze how the material will behave when it is constricted at the center, for example to function as a micro-gripper. A 90^∘ twist-bend LCN film is thus aligned via the photoalignment technique, structured by selective polymerization and attached to the substrate via the stamp-and-stick technique, and the overall angle made by both the flaps is tracked when the illumination is turned on and off.

Zoom In Zoom Out Reset image size

Figure 5(b) shows the obtained bending response by the flaps with an irradiation of 81 mW cm⁻², leading to rise and fall times of 7 s and 19 s, respectively. The rise and fall times are measured from the maximum bending angle for values from 10% and 90%. Due to the 60 µm thickness of actuator, which is higher in comparison to other published works [33, 44, 46], any heat generated during actuation will take extra time to dissipate in its surrounding environment, resulting in higher fall times of 19 s.

To further analyze the response behavior as a function of input intensities, three different input optical powers are applied in a step response. The obtained results are summarized in figure 6(a) where for different input optical powers, different targeted angles within t_rise ≈ 6 s can be reached.

Zoom In Zoom Out Reset image size

Furthermore, figure 6(b) demonstrates that the actuator can be held at different bending states by irradiating with constant input intensity. For measurements in figure 6(b), input power is applied as a step function and the maximum bending angle is recorded at t = 40 s. This recorded value is then plotted here as a function of input power.

3.2.2. Photochemical actuation.

Figure 4(b1) illustrates the pattern for alignment and as well as for selective polymerization for photochemically induced actuation motion. This actuator is also 60 µm thick and is aligned in a 90^∘ twist-bend fashion. As previously mentioned, upon illumination with 365 nm at 15 mW cm⁻² (figure 4(b3)), the globular shape of cis-azobenzene crosslinker disrupts the order of the network, leading to actuation of the LCN. Since the half-lifetime of cis-azobenzenes used in this work is in the order of hours [33], reverse actuation motion takes place only after illuminating the actuator with 450 nm at 55 mW cm⁻² (figure 4(b4)), and is achieved in about 5 s.

Similar to the bending characterization for photothermal induced actuation, figure 7 summarizes the bending characteristics for photochemical induced actuation. A 60 µm thick film was attached to the substrate and the movement of the actuator was tracked in the similar fashion as illustrated in figure 5(a). When the actuator is illuminated with 1.2 mW cm⁻² at 365 nm, it regains its maximum bending position within ≈ 45 s. As soon as 450 nm LED with 9.8 mW cm⁻² is turned on at t = 140 s, 10$\%$ of maximum value is achieved within 7 s.

Zoom In Zoom Out Reset image size

Figure 7(b) summarizes the transient step response when different power intensities are applied as input to the actuator. Interestingly, the transient step response behavior of the actuator is dependent on the input intensity, unlike the response of photothermally driven actuators. At lower 365 nm intensities, the rise time behavior follows a steady linear increase, followed by constant actuated state when 365 nm LED is switched off. Increasing the intensity leads to shorter response times and higher bending angles. The fall response is also dependent on the intensity of 450 nm LED. At lower intensities, the original state of the actuator is not reached and the actuator position saturates after reaching a certain value.

Similar to figures 6(b) and 7(c) shows that the actuator can be held at different bending states depending on the intensity of 365 nm wavelength. For these measurements as well, input intensity is applied as a step response and the bending angle is recorded at t = 120 s.

For both actuator types just considered, the actuation response can be easily tuned by changing the length/width ratio and thickness [44, 46]; by changing the chemical system [21, 47]; or by simply increasing the input optical intensity [45].

3.2.3. Reproducibility.