1. Introduction

Devices based on the electrowetting-on-dielectric (EWOD) effect have long been of interest as a non-mechanical alternative for tuning the incident and return light in many versatile applications such as beam steering [1–5], aberration control [6–11], and biomedical imaging [12–17]. Modern EWOD devices commonly use an applied voltage to tune the shape of the interface between two immiscible liquids contained in a cavity. Recently, there has been particular interest in tunable, arrayed optical devices at the micron scale, ideal for integration in compact 3D imaging and sensing systems [18–22]. Focus-tunable microlens arrays have been previously demonstrated using liquid crystals [23–25]. However, these devices tend to suffer from slow actuation speed, limited tunability, and polarization sensitivity. Microlens arrays based on the EWOD effect offer an attractive solution due to their transmissive nature, low power consumption, and large range of tunability [26–29]. While EWOD devices have been fabricated from the millimeter [5,26,27,30–33] to the micron scale [34,35], demonstrations of micro-scale arrays of electrowetting lenses have been limited [18,22,36–38]. The majority of these demonstrations do not allow for individual actuation of each lens in the array, limiting their potential applications. Additionally, electrowetting array designs that are capable of individual addressability are commonly fabricated with large overall dimensions or utilize an electrowetting cavity geometry that jeopardizes the imaging quality of the overall device. For example, the electrowetting arrays fabricated by Smith et al. [38] and Kim et al. [37] are capable of greater than 80% fill factor but are not individually addressable. The individually addressable electrowetting array on a flexible dielectric sheet introduced by Grinsven et al. [36] demonstrates excellent imaging quality for applications such as integral and 3D imaging. However, the 10 mm pitch between each individual lens results in overall device dimensions too large for use in compact optical imaging systems such as head mounted fluorescence microscopy [12]. Similarly, the arrays designed by Lee et al. [39] demonstrate an accessible method of fabricating an array of electrowetting prisms capable of up to $\pm$ 18.5$^{\circ }$ beam-steering but their large scale limits their applications for compact optical systems. Alternatively, the fabrication process detailed by Hou et al. [40] allows for a high fill factor, individually addressable electrowetting array, but the use of square electrowetting cavities introduces unwanted optical aberrations, reducing imaging quality [34,41]. In this work, we present an individually addressable, micro-scale array of electrowetting-based lenses fabricated using standard microfabrication techniques. Our design leverages previous work on electrowetting arrays [36–38], but incorporates several key improvements to prior demonstrations. The use of standard microfabrication techniques allows the number of elements in our device to be easily scaled up while still maintaining overall device dimensions useful for compact imaging systems. By performing all of our fabrication on glass substrates, we can easily increase the number of lenses while maintaining an assembled device thickness of less than 1.5 mm. Patterning the electrical connections onto these glass substrates also allows us to significantly improve the fill factor of future designs to greater than 60% [7] while still allowing each lens to be individually addressed. Additionally, using cylindrical cavities with uniformly patterned electrodes allows our device to actuate over the full tunability range without introducing unwanted optical aberrations. Integration of micro-scale electrowetting devices in an array opens the door to potential applications for novel imaging systems [18,22,37], transmissive wavefront aberration correction [7,8,42], tunable phase arrays for optical encryption [43] and even energy harvesting utilizing the reverse EWOD effect [44].

The EWOD effect describes the use of an applied electric field to tune the surface tension of a solid-liquid interface. When a voltage is applied across a conductive droplet and an electrode separated by a thin dielectric film, the effective contact angle of the liquid droplet changes to balance the induced surface tension imbalance [45–49]. The change in contact angle of the liquid droplet as a function of applied voltage is described by the Lippmann-Young equation (Eq. (1)):

2. Optical design and fabrication

For our micro-scale array of electrowetting lenses to be used in a compact optical imaging system, we require that the design fulfills a few key constraints. The entire device must have sufficient transmission across the visible spectrum, each lens must be isolated from its neighbor to allow for individual actuation, and the surface quality of each lens must be comparable to a fused silica microlens array. Using these constraints, we developed a design incorporating three separately fabricated chips: a bottom glass chip containing electrical contacts needed for individual actuation, a photo-patternable glass chip (Schott Foturan II) used as both the vertical cylindrical cavities and sidewall electrodes at equipotential, and a top glass chip for sealing the liquids in our device. A cross section of our array design is shown in Fig. 1(a) and a conceptual design of a fabricated device in Fig. 1(b). Individual actuation is achieved by applying a potential difference between the electrodes on the bottom glass chip and the electrodes on the sidewalls of the etched cavities in the Foturan glass chip. Crucially, this design is fabricated using standard microfabrication techniques and enables individual actuation for each micro-scale lens in the array.

 

Fig. 1. (a) A 2D cross-section view of our microscale electrowetting lens array design including the bottom and side electrode chip used to individually address each lens as well as the top glass chip used to seal the liquids in each cavity. The sidewall electrode chip is covered in a dielectric layer, Parylene, and a hydrophobic layer, Cytop, used to increase the initial contact angle of the liquid in each cavity to 173$^{\circ }$ [27]. Each chip is separately fabricated then bonded together using an intermediate layer of patterned SU-8. (b) A 3D schematic of a 4 ${\times }$ 4 electrowetting array design concept of a fabricated bottom and sidewall electrode glass chip without the top glass chip. Each liquid cavity is uniformly coated with our side electrode which serves to provide the potential difference between the polar liquid and the walls of each cavity. The polar liquid in each cavity is individually addressed via contact pads on the outer edges of our bottom electrode chip. In this design, the bottom electrode chip has dimensions of 7.5 ${\times }$ 7.5 mm.

Download Full Size | PDF

For the bottom glass chip, a layer of positive photoresist is spun onto a glass wafer and patterned corresponding to our electrode configuration needed for individual actuation. Approximately 200 nm of gold and 10 nm chrome for adhesion are deposited, through sputter deposition, onto our patterned wafer. The excess in undesired regions is removed using a lift-off process. A microscope image of two neighboring patterned electrodes in a 3 $\times$ 3 array is shown in Fig. 2(a). Finally, a 50 $\mathrm {\mu }$m thick layer of SU-8 2050 is spun and patterned onto the bottom electrodes serving as a separation layer between the bottom actuation electrodes and the sidewall electrodes as well as a bonding layer between the bottom and sidewall chips.

 

Fig. 2. (a) A microscope image of two neighboring patterned individual electrodes on the bottom electrode chip for a 3 ${\times }$ 3 array of 500 $\mathrm {\mu }$m diameter electrowetting lenses. Each patterned electrode makes contact with the polar liquid in each cavity which can then be addressed using contact pads patterned on the outer edges of the bottom electrode chip. (b) An SEM image of an etched 3 ${\times }$ 3 array of 500 $\mathrm {\mu }$m diameter electrowetting lens cavities in Foturan photopatternable glass, constituting our sidewall electrode chip. After the Foturan glass is etched and the surface roughness is decreased using a reflow process, we uniformly coat the chip in our gold sidewall electrode, our dielectric layer, Parylene, and a hydrophobic layer, Cytop.

Download Full Size | PDF

To achieve at least 1:1 aspect ratio vertical cavities with smooth sidewalls at the micron scale, we used a 500 $\mathrm {\mu }$m thick photo-patternable glass chip (Schott Foturan II). Foturan glass contains dopants which initiate crystallization when exposed to UV light around 320 nm in wavelength and then tempered at 500-600$^{\circ }$C [55]. When submerged in a 10${\% }$ weight hydrofluoric acid bath, the crystalline regions preferentially etch at up to 20:1 selectivity compared to unexposed regions [56,57]. These etched regions form the cylindrical cavities ideal for filling with liquids for electrowetting actuation. Fig. 2(b) shows a scanning electron microscope (SEM) image of a 3 ${\times }$ 3 array of 500 $\mathrm {\mu }$m diameter etched Foturan cavities coated in a uniform layer of gold. Uniform meniscus actuation requires a smooth sidewall surface roughness, typically under 1 $\mathrm {\mu }$m average profile height deviation from the mean (Ra) [58–60]. To improve the surface roughness of the etched Foturan cavities, we perform a glass reflow process by uniformly heating each chip to 560$^{\circ }$C for one hour. Using a white light vertical scan interferometer (Veeco Wyko NT3300), we measured the reduced surface roughness on the top face of the etched chip to be below 0.75 $\mathrm {\mu }$m Ra. The surface quality on the vertical wall of each etched cavity coated in gold was qualitatively assessed using an SEM oriented at 45$^{\circ }$ with respect to the top face of the chip. It was observed that the surface was free of any defects greater than 1 $\mathrm {\mu }$m and was of a similar surface quality to the top face of the chip. Once the Foturan glass has been etched, the gold sidewall electrodes are uniformly deposited onto the entire chip. The chip is then coated with 3.6 $\mathrm {\mu }$m of a dielectric layer, Parylene, and dip-coated with a 10${\% }$ hydrophobic Cytop solution to increase the initial contact angle of the liquid droplet in each cavity. To bond the bottom electrode chip and the etched Foturan chip, we use an intermediate layer of patterned SU-8 2050 [61–64]. The etched Foturan chip is aligned onto the bottom electrode chip containing a layer of SU-8 and both chips are heated to the transition temperature of cross-linked SU-8 (175$^{\circ }$ C) while maintaining pressure for one hour, then allowed to cool to room temperature.

Once each chip has been fabricated, we assemble and characterize our devices. We fill each cylindrical cavity with a polar liquid, DI water, and a nonpolar liquid, 1-phenyl-1-cyclohexene (PCH) using a micro-drop dispenser head (Microdrop Tech) capable of depositing pico-liters of liquid using a piezo actuator. PCH is chosen as our nonpolar liquid due to its high refractive index contrast with DI water (${\Delta}n = 0.24$) and low density mismatch ($\Delta_\rho = 0.004\;g/L$) [27]. Each etched cavity is first aligned underneath the dispenser head and approximately half of the volume is filled with DI water. We then overfill each lens cavity with PCH and contain both liquids in our device by placing our top glass chip directly onto our sidewall electrode chip. To avoid rapid evaporation of the small volumes of dispensed liquid, we operate the micro-drop dispenser remotely in a humidity-controlled box. This liquid combination results in a 173° initial contact angle inside each lens cavity [27]. To address each individual lens, we wire bond our patterned bottom electrode chip to a custom-designed printed circuit board (PCB). This PCB is then connected to an 8-channel amplifier (OKOTech) using a pin connector.

3. Device characterization

An image of a fully assembled electrowetting array bonded to its corresponding PCB is shown in Fig. 3(a). The final device (attached to a custom PCB) is 12 ${\times }$ 13 mm in scale, though this can vary depending on the number of individual lenses in the array. To demonstrate individual actuation of a single lens in our array, we image our device through a 10x objective (0.22 NA) and tube lens (TTL200, f = 200 mm) onto a 4k DSLR camera (Canon EOS RP). Individual actuation of two neighboring 500 µm diameter lenses in a 3 ${\times }$ 3 array is shown in Fig. 3(b). The effective focal length of each lens changes with an applied voltage, shown by the underlying "Array" text no longer being in focus when the lens is actuated at 60 V. To ensure each lens receives the correct actuation voltage, we characterize the small variations in amplification for each channel of our multi-channel amplifier and normalize our input signal by these values. A critical step in the fabrication of our electrowetting arrays is the isolation of our conductive polar liquid in each lens cavity. If the conductive liquid is not isolated, the actuation of a single lens will change the contact angle of neighboring lenses. Our method of SU-8 bonding the sidewall and bottom electrode chip minimizes the possibility of the polar conductive liquid from one lens bridging the connection to a neighboring lens. Additionally, we do not expect to observe any electronic crosstalk between neighboring lens electrodes throughout our actuation range as their relatively large separation (65 µm at their closest point) and small width (15 µm) leads to a negligible parasitic capacitance. As can be seen in Fig. 3(b), each lens’s focal length can be individually tuned with no effect on neighboring lenses. This result is crucial if these devices are to be used in wavefront shaping systems as it allows individual control of discrete portions of an incident wavefront.

 

Fig. 3. (a) A fully assembled 3 ${\times }$ 3 electrowetting lens array before filling with DI water and PCH and sealing with the top glass chip. The fabricated and bonded bottom and sidewall electrode chip is attached to a custom PCB which is used to individually address each lens in the array via wire bonds made between the contact pads on the PCB and the bottom electrode chip. A sinusoidal input voltage signal is amplified through a multichannel amplifier and sent to our device using a pin connector attached to the PCB. (b) A microscope image of a 3 ${\times }$ 3 array of 500 µm diameter electrowetting lenses (left) and the individual actuation of two neighboring lenses in the array (right). As a 60 V input signal is applied to an individual lens in the array, the contact angle of the liquid-liquid meniscus changes and the underlying "Array" text appears out of focus.

Download Full Size | PDF

3.1 Focal length tunability

To characterize the focal length tunability of our micro-lens array, we set up a 4f imaging system consisting of a 25.4 mm focal length singlet and a 30 mm focal length doublet, in line with an electrowetting lens array actuated at various voltages. A 635 nm collimated laser source is directed through our array and imaged onto a beam profiler by the 4f system. As increasing voltage is uniformly applied to each lens in our array, the change in contact angle results in a change of effective focal length. We then translate our 4f imaging system along the optical axis until the smallest focal spot is recorded onto the beam profiler. This location corresponds to the effective focal position from our microlens array lying in the back focal plane of the first lens in our 4f system. By recording the change in distance between the first lens in our 4f system and the top of our array, we arrive at a relationship between the back focal length of a single lens in our system as a function of applied voltage, shown by the red plotted points in Fig. 4(a). The relatively short focal length of the lenses used in our 4f system increases the divergence angle of the focusing beam, allowing more accuracy when determining the focal length while still providing a large enough aperture to collect the highly diverging light exiting the electrowetting lens at higher contact angles. We show that a device coated in 3.6 µm of Parylene C is capable of focal length tuning from −1.7 mm to −∞ with an applied voltage of 0 and 75 V, respectively. At approximately 75 V, the liquid meniscus reaches 90° contact angle and the optical power of the electrowetting lens becomes 0. Further actuation would result in a liquid meniscus contact angle below 90°, causing the device to behave as a converging lens for our liquid combination. Our group has previously demonstrated the large range of contact angle tunability of an electrowetting device using the same liquid combination of DI water and PCH [27]. Accounting for the difference in dielectric layer (Parylene C rather than Parylene HT), the tunability of our electrowetting array matches well with the interfacial surface tension of 24.83 mN/m and initial contact angle of 173° found previously. Additionally, in both cases we find that the tunability of this liquid combination experiences contact angle saturation around 90° and the response of our device begins to deviate from that expected by the Lippman-Young equation [Eq. (1)].

 

Fig. 4. (a) A plot of the back focal length of an individual lens coated in Parylene C $(\varepsilon_{D}=3.10\varepsilon_{0})$ in an electrowetting array as a function of applied voltage. This is measured by directing a collimated laser source through an individual lens and imaging the beam onto a beam profiler using a 4f system. As the applied voltage increases, the position of the 4f system which optimizes the focal spot on the beam profiler changes an equal amount to the change in back focal length. (b) The liquid switching time of an individual lens in our electrowetting array. This is found by directing a collimated laser source through an individual lens in our array which is focused onto a photodetector through a 50 µm pinhole. The photodetector signal is recorded in time as the device is actuated from 173° to approximately 90° contact angle, shown by the blue curve. We find that the liquid-liquid interface rises from 2 to 99 % of its final steady-state value within 1.5 ms. The orange plot shows the height change of the center point of the liquid meniscus as we simulate meniscus actuation from 173° to approximately 90° contact angle using the fluid dynamics model in COMSOL multiphysics 6.1. The simulated liquid meniscus height reaches within 2.5% of its steady-state value within 1.9 ms, matching well with experimental results. The oscillations observed in the liquid meniscus height change resulted in intensity variations too small to measure experimentally on our photodetector.

Download Full Size | PDF

To verify these results, we model this system in Zemax OpticStudio ray-tracing software using the same model of lenses used for our 4f system. We then calculate the radius of curvature of the spherical liquid meniscus in a 500 µm electrowetting lens for a range of contact angles. For each radius of curvature calculated, we optimize the location of our 4f system according to the minimum focal spot size at the imaging plane. By recording the optimized distance between our simulated 4f system and the electrowetting lens, we arrive at a predicted relationship between liquid-liquid meniscus contact angle and back focal length. We then use the Lippman-Young equation [Eq. (1)] to relate the meniscus contact angle to the applied voltage for our given liquid-liquid system. From this, we plot the predicted back focal length as a function of applied voltage, shown by the blue line in Fig. 4(a).

3.2 Liquid response time

We also characterize the liquid switching time of a single lens in our device. To do this, we direct a collimated 635 nm laser source through a single lens in our array and focus the light through an objective and onto a photodetector with a 50 µm pinhole on the front. We record the intensity measured at the photodetector over time as we actuate our device to approximately 90° contact angle using a sinusoidal input voltage first generated with an analog data acquisition card (DAQ, NI PCIe-6738) and amplified from a multi-channel amplifier (OKOTech) [27,29,65]. These results are shown in the blue plot in Fig. 4(b). The lens is actuated at time t = 0 in this figure with an input sinusoidal voltage signal with an RMS value of 70 V. The intensity measured on the photodetector rises from 2 to 99 % of its steady state value within 1.5 ms.

We can predict the liquid meniscus response of an individual lens in our array by solving the time-dependent Navier Stokes equations using the fluid dynamics model in COMSOL Multiphysics. This simulation uses the Lippman-Young equation [Eq. (1)] to modulate the meniscus contact angle assuming an initial contact angle of 173°. The displacement of the center point of the liquid-liquid interface is monitored in time as we actuate to approximately 90° contact angle. Our simulation shows good agreement with our experimental results, reaching 2.5% of its final steady-state value within 1.9 ms. A plot of the simulated meniscus height in time as we actuate our meniscus to 90° contact angle is shown by the orange plot in Fig. 4(b). Because our two liquid system is well density matched (${\Delta}\rho$ = 0.004 g/L) [27,31,66], the response time of our device is governed by the viscosity and dimensions of our individual lenses. Our group has previously measured the dynamic viscosity of PCH as 3.24 mPa${\cdot }\textrm{s}$ [27]. As we actuate our electrowetting lens, a standing wave is generated along the liquid-liquid meniscus with resonance characteristics corresponding to the roots of the zeroth order Bessel function [67,68]. In millimeter scale electrowetting lenses without input voltage shaping, these oscillations are observed as under-damped intensity oscillations recorded on the photodetector [27,29,65,69]. However, the micro-scale nature of the devices presented here lead to oscillations on the liquid-liquid interface too small to be distinguished on the photodetector in this system.

3.3 Liquid meniscus profile

The surface quality of an individual 500 $\mathrm {\mu }$m diameter lens in our electrowetting array was investigated to support the use of our device in optical phase arrays and compact imaging applications. To do this, we telescope our incident collimated 635 nm laser beam down to approximately 375 $\mathrm {\mu }$m diameter so that the incident light fills approximately 75${\% }$ of the aperture of an individual lens in our array. We directed this beam onto a Shack-Hartmann wavefront sensor (ImagineOptic HASO4). The wavefront and corresponding Zernike coefficients were recorded over various device contact angles from approximately 130$^{\circ }$ to 90$^{\circ }$. A reference wavefront taken through the same optical system but without an electrowetting device was then subtracted from the wavefront recorded at each contact angle to minimize aberration contributions from optical components prior to our electrowetting lens. The reconstructed wavefront from a device actuated to 130$^{\circ }$ contact angle and the same device actuated to 100$^{\circ }$ is shown in Fig. 5(a) and 5(b).

 

Fig. 5. (a) The reconstructed wavefront of a collimated laser source passing through a single lens in an electrowetting array actuated to approximately 130$^{\circ }$ contact angle. The RMS of the higher order Zernike contributions to the wavefront is measured to be less than $\lambda$/30. (b) The reconstructed wavefront when the same lens is actuated to approximately 100$^{\circ }$. In both cases we find that the higher order aberration contributions have no significant change when actuating our lens to various contact angles, matching well with predictions from literature [26,34,41]. (c) A plot of the contribution of defocus to our reconstructed wavefront as a function of decreasing meniscus contact angle. We observe that, as our device contact angle decreases, the contribution of defocus also decreases.

Download Full Size | PDF

The derived Zernike coefficients show that the aberration contributions from our device is dominated by defocus $(Z^{0}_{2})$. As a verification for our wavefront reconstruction, we also see that the contribution of defocus decreases as we actuate our device closer to 90$^{\circ }$ contact angle. We track the defocus contribution as a function of applied voltage in Fig. 5(c). Ignoring tilt and defocus, the higher order Zernike terms had a negligible contribution to the resultant wavefront. The RMS contributions from these coefficients correspond to a wavefront variation of less than $\lambda$/30, which is comparable to a similarly sized static fused silica micro lens [70,71] and small enough to support their use in phase-sensitive applications. For a more detailed view of the relative contributions of higher order aberrations, see Supplement 1, Fig. S1. Higher order aberration contributions do not significantly increase as lens actuation voltage increases, matching well with prior work and theoretical predictions [26,34,41,72].

4. Conclusions

We have designed and demonstrated a two-dimensional, individually addressable electrowetting micro lens array based on cylindrical EWOD lens design considerations from our previous work [26,27,41]. This design is fabricated entirely through standard 2D microfabrication techniques and allows each lens to be individually tuned without affecting neighboring lenses, justifying their potential use in optical phase arrays and aberration correction [7]. We have shown that devices coated in Parylene C display a large focal length tunability, matching theoretical predictions of −1.7 mm to $-\infty$ with an applied voltage of 0 and 75 V. Because of their small scale, each lens can be actuated to within 1${\% }$ of its steady-state value in less than 1.5 ms, facilitating their use in applications that place importance on fast switching speeds. Finally, we showed that the contribution of higher order aberration terms is comparable with a static fused-silica micro lens [70,71] and does not significantly increase throughout the device actuation range, supporting their use in phase-sensitive applications. From these results, we have shown the potential of these devices to be used in a variety of applications from compact novel imaging systems, to optical phase arrays.

This work represents a strong proof-of-concept for a scalable, individually addressable array of micro-scale electrowetting lenses. This design is distinct from previous work as it allows for the individual actuation of each lens in a high fill factor electrowetting array while maintaining imaging quality and a small overall device footprint. Future work involves expanding our current array design to a larger number of lenses, allowing for their use in applications such as integral imaging and microscopy. For these arrays to be used in settings outside of a lab environment, a more robust method of sealing the liquids in our device while minimizing the effects of evaporation and leakage is currently being investigated. Additionally, we are currently working on adapting this design to make an array of micro scale electrowetting prisms by incorporating multiple sidewall electrodes per liquid cavity. This would allow us to individually steer multiple beams to various regions of interest simultaneously, benefiting applications such as aberration correction and microscopy.