**Aperture Ratio Improvement by Optimizing the Voltage Slope and Reverse Pulse in the Driving Waveform for Electrowetting Displays**

**Zichuan Yi 1, Wenyong Feng 2,3,\*, Li Wang 3,\*, Liming Liu 1, Yue Lin 1,2, Wenyao He 2, Lingling Shui 1,2, Chongfu Zhang 1, Zhi Zhang <sup>1</sup> and Guofu Zhou 2,3**


Received: 3 November 2019; Accepted: 5 December 2019; Published: 7 December 2019

**Abstract:** Electrowetting display (EWD) performance is severely affected by ink distribution and charge trapping in pixel cells. Therefore, a multi structural driving waveform is proposed for improving the aperture ratio of EWDs. In this paper, the hysteresis characteristic (capacitance–voltage, C-V) curve of the EWD pixel is tested and analyzed for obtaining the driving voltage value at the inflection point of the driving waveform. In the composition of driving waveform, a voltage slope is designed for preventing ink dispersion and a reverse pulse is designed for releasing the trapped charge which is caused by hysteresis characteristic. Finally, the frequency and the duty cycle of the driving waveform are optimized for the max aperture ratio by a series of testing. The experimental results show that the proposed driving waveform can improve the ink dispersion behavior, and the aperture ratio of the EWD is about 8% higher than the conventional driving waveform. At the same time, the response speed of the driving waveform can satisfy the dynamic display in EWDs, which provides a new idea for the design of the EWD driving scheme.

**Keywords:** electrowetting display; aperture ratio; driving waveform; hysteresis characteristic; ink distribution; response speed

#### **1. Introduction**

In recent years, low-power and reflective displays which are readable in the sun are favored by many scientific researchers [1]. As a kind of reflective display technology, the electrowetting display (EWD) is receiving more and more attention [2], and its initial products have been successfully applied to various fields [3]. However, a stable multi-grayscale video display of EWDs has still not been implemented.

In 1981, researchers proposed the EWD technology [4]. It can display grayscale by controlling the movement of the colored ink. In the past few decades, various driving systems and image processing methods based on multi-level grayscale dynamic EWDs have greatly improved display quality [5,6]. We have also proposed a driving system for multi-grayscale display in an EWD based on a field programmable gate array (FPGA) chip, which provides an important reference for optimizing the dynamic display of EWDs [7]. In the field of EWD driving waveform, the driving method based on

amplitude modulation (AM) or pulse width modulation (PWM) has been proposed [8], and the driving voltage is used to control the shrinkage of the ink droplets. However, the ink in the EWD pixel shows a phenomenon of hysteresis, which has a negative effect on optical performance. The PWM compensates for the hysteresis to some extent, but the PWM means a reduction in response speed and it can reduce the effective frame for the video display, and the power consumption of the EWD grayscale display is also increased by using the PWM at the same time. In addition, a dynamic contact angle model is established based on the molecular dynamics theory; the shape evolution of droplets has been studied under different direct current (DC) driving conditions. In addition, the influence of liquid interface resonance on contact angle has also been studied under alternating current (AC) driving conditions [9]. However, there are only two types of driving waveforms for driving droplets: stepped DC driving signal and sinusoidal AC driving signal. The problem of ink splitting has been also proposed during coupling AC-common driving process [10]. A driving modulation scheme was used to improve the ink dispersion performance to a certain extent, but the ink showed instability during the driving process. In other driving modes, sinusoidal, bipolar, and single-pulse are used to drive electrowetting liquid lenses respectively, and it was found that the positive and negative polarities of the driving voltage have a significant effect on the charge trapping of liquid lenses [11]. This provides a possible direction for the optimization of the driving waveform for EWDs. Based on the EWD micro-space pixel units, it has been found that the ink motion shape can be controlled by applying driving voltage with different rising speed [12]. The EWD reflectivity was improved using this method, but the response time of the driving waveform became longer.

In order to improve the aperture ratio of EWD pixels and the shrinkage form of the ink, the relationship among driving voltage, ink shape, pixel aperture ratio, and response time are studied in this paper. Then, a driving waveform is proposed for EWDs to improve the display effect according to the motion behavior of the ink and the structure characteristics of the EWD. Compared with the conventional driving waveform, the proposed driving waveform can improve the performance of charge trapping, ink dispersion, aperture ratio, and response time.

#### **2. Principle of EWD Driving Waveform System**

#### *2.1. Electrowetting Equivalent Circuit Model*

Grayscale display is realized in EWDs by applying voltage to control the movement of colored ink droplets. Its essence is an optical switch, which has excellent grayscale display characteristics. A typical EWD pixel structure is shown in Figure 1a. The orange part is the pixel wall, which can control the range of ink movement. From the top down, the ink is flat, and the state is also the lowest energy state.

**Figure 1.** The electrowetting display (EWD) pixel structure and its equivalent circuit. (**a**) The EWD pixel structure without applied voltage; (**b**) simplified equivalent circuit diagram of an EWD pixel unit.

The voltage is applied between the upper and lower plates which are made up of indium tin oxide (ITO). According to the Lippmann–Young equation, with the increase of the applied electric field, the surface tension among the insulating layer, ink film, and water can be increased, and the contact angle between the interfaces will also increase. The original balance is broken by the electric field force which is generated by the voltage difference; the ink is squeezed away by the water, and the water contacts the surface of the hydrophobic insulation layer, then, the white base plate is exposed. The aperture ratio (white area ratio) WA of a pixel is defined as Equation (1).

$$\mathcal{W}\_{\rm A} = 1 - \left(\frac{\mathcal{S}\_{\rm ink}(\rm V)}{\mathcal{S}\_{\rm pix}}\right) \times 100\% \tag{1}$$

In Equation (1), Sink and Spix are defined as the surface area of the ink in a single pixel and the surface area of the entire pixel, respectively, and V represents the voltage applied on the pixel in the EWD. In addition, the area of the pixel wall is ignored in the calculation of the aperture ratio. The displacement of the ink shrinkage is determined by the applied voltage [13,14]. Theoretically, the radius of ink shrinkage is directly related to the contact angle, and the contact angle θ follows the Lippmann–Young equation.

$$\cos\theta = 1 - \frac{\text{CV}^2}{2\gamma\_{\text{CW}}} \tag{2}$$

In Equation (2), C is the capacitance of a single pixel unit area. V is the voltage applied on the pixel unit, and γOW is the ink–water interfacial tension. The capacitance–voltage (C-V) curve provides an important parameter for driving EWDs and is an important basis for the design of driving waveform voltage.

For the equivalent circuit model of the micro-space pixel unit, the ink droplets are driven under an alternating electric field, and the ink and pixel wall in the same dielectric layer are treated as a combined loop. The photoresist material of the pixel wall has low conductivity, and the simplified equivalent circuit diagram is shown in Figure 1b [15]. C0 is the ink capacitance, R0 is the ink resistance, CFP is the dielectric layer capacitance, RFP is the dielectric layer resistance, and U is the effective voltage applied to the ink in the pixel. Based on the equivalent circuit model, the resistance, capacitance of the insulating layer, and ink follow Ohm's law; then we can get their expressions of a single pixel.

Dielectric layer resistance is shown in Equation (3).

$$R\_{\rm FP} = \frac{\rm d}{\rm S\delta\_{\rm FP}}\tag{3}$$

In Equation (3), d represents the thickness of the dielectric layer; S represents the area of a single pixel unit; δFP represents the dielectric constant of the hydrophobic layer. The ink resistance is shown in Equation (4).

$$\mathcal{R}\_0 = \frac{\mathcal{H}}{\mathcal{S}\delta\_\mathcal{I}} \tag{4}$$

In Equation (4), H represents the thickness of the ink; δ<sup>I</sup> represents the dielectric constant of the ink. The dielectric layer capacitance is shown in Equation (5).

$$\mathbf{C\_{FP}} = \varepsilon\_0 \delta\_{\rm FP} \mathbf{S/d} \tag{5}$$

In Equation (5), ε<sup>0</sup> is the vacuum dielectric constant. The ink capacitance CI is shown in Equation (6).

$$\mathbf{C}\_{\mathrm{I}} = \varepsilon\_{0} \delta\_{\mathrm{I}} \mathrm{S}/\mathrm{H} \tag{6}$$

In Equation (6), δ<sup>I</sup> represents the dielectric constant of the ink; ε<sup>0</sup> is the vacuum dielectric constant. According to Equations (3) and (4), the effective resistance R is shown in Equation (7).

$$\mathbf{R} = \frac{1}{\mathbf{S}} (\frac{\mathbf{d}}{\delta\_{\rm FP}} + \frac{\mathbf{H}}{\delta\_{\rm I}}) \tag{7}$$

According to Equations (5) and (6), the single pixel cell effective capacitance C(H) is shown in Equation (8).

$$\mathbf{C}(\mathbf{H}) = \varepsilon\_0 \mathbf{S}(\frac{\delta\_{\rm FP}}{\mathbf{H}} + \frac{\delta\_{\rm I}}{\mathbf{d}}) \tag{8}$$

The critical starting voltage Vswith of a single EWD pixel unit can be estimated by Equation (9).

$$\mathcal{V}\_{\text{switch}} = \sqrt{\frac{2\pi^2 \mathcal{V}\_{\text{ow}}}{\mathcal{C}(\mathcal{H})\mathcal{S}}} \tag{9}$$

The insulating film Teflon used in the EWD pixel cell structure has a correspondence relationship with the threshold voltage [16].

$$
\cos\theta\_{\text{av}} - \cos\theta\_0 = \frac{\delta\varepsilon\_0 \mathbf{U}\_{\text{EW}}^2}{2\text{d}\chi\_{\text{ow}}} \tag{10}
$$

In Equation (10), θ<sup>α</sup> is the Lippmann contact angle; θ<sup>0</sup> is the static contact angle; δ is the dielectric constant of a single pixel. The relationship between the applied voltage UEW and the film thickness d of the insulating layer can be derived in Equation (11).

$$\mathbf{U}\_{\rm EW} = \sqrt{\frac{2\mathbf{d}\chi\_{\rm cov}}{\delta\varepsilon\_0} (\cos\theta\_\alpha - \cos\theta\_0)} \sim \sqrt{\mathbf{d}} \tag{11}$$

The voltage of the insulating layer is proportional to the film thickness. The insulating layer is easily electrically broken when the film thickness of the insulating layer is too thin, so the parameters of the film thickness of the insulating layer should be considered in the design of the driving waveform.

#### *2.2. Analysis of Ink Distribution in a Pixel*

In the process of driving an EWD, it is ideal to drive the ink to one corner of a pixel, as shown in Figure 2a, so we can get a max value of the aperture ratio at this time. However, once a conventional square driving waveform is applied, the ink may shrink to two or more corners, as shown in Figure 2b. The shape of the ink dispersion affects the EWD aperture ratio directly. As shown in Figure 3, one ink droplet A is divided into two droplets: B and C. Obviously, the covered area of the A droplet on the substrate SA is smaller than the sum area of SB and SC and the aperture ratio becomes smaller when the ink is divided into two parts.

**Figure 2.** Ink film distribution state when the pixel is driven by the driving waveform. (**a**) Ink shrinks to one corner; (**b**) ink shrinks to four corners.

**Figure 3.** Ink splitting diagram.

In an EWD pixel, the ink shape mainly has four shapes, as shown in Figure 4. Figure 4a is an ideal ink shape, the ink is all shrunk to one corner in a pixel. The maximum aperture ratio can be obtained, and the aperture ratio can reach 73.9%. However, if a conventional driving waveform is applied to the EWD panel, the ink may be dispersed into two parts as shown in Figure 4b, or dispersed into three parts as shown in Figure 4c, or dispersed into four parts as shown in Figure 4d. Obviously, the pixel can reach a max aperture ratio value when the ink distribution is stable. The aperture ratio values of different ink film distributions are shown as follows: 73.9%, 62.1%, 61.6%, and 60.7%, respectively. So, the greater the number of ink dispersions, the smaller the aperture ratio in the EWD pixel.

**Figure 4.** The ink distribution state and the corresponding aperture ratio with the action of the driving waveform. (**a**) The ink is all shrunk to one corner and its aperture ratio value is 73.9%. (**b**) The ink is dispersed into two parts and its aperture ratio value is 62.1%. (**c**) The ink is dispersed into three parts and its aperture ratio value is 61.6%. (**d**) The ink is dispersed into four parts and its aperture ratio value is 60.7%.

#### **3. Driving Waveform Design**

#### *3.1. Testing System*

In order to test the effects of the driving waveform, the experimental setup is designed as shown in Figure 5. In the tested EWD panel (2.7 inches diagonally), the size of a single pixel grid is 150 μm × 150 μm, and an entire panel contains 78,408 independent pixels. The width and height of the pixel wall are 15 μm and 5.6 μm, respectively, and the thickness of the insulating layer is 1 μm. The height between the ITO substrate and the upper cover in the pixel grid is 75 μm. The solvent C10H22 is used as a colored ink whose molecular concentration is 10 wt%, and the electrolyte is a sodium chloride solution whose concentration is 1.4 mol/L. The thickness of the spin-coated ITO glass substrate and the surface panel are 1.1 mm and 1.7 mm, respectively, and the impedance is 100 Ω/Sq. Teflon AF1600 is spin-coated on the surface of the glass substrate as a hydrophobic insulating layer, and the colored ink is injected into the pixel grid at a low speed (1 mm/s) by the grating filling method. Finally, the pixel unit is edge-sealed using a pressure sensitive element.

**Figure 5.** Optical testing system for EWDs. (**a**) AFG3052 arbitrary function generator; (**b**) ATA2022H high voltage amplifier; (**c**) testing board; (**d**) microscope; (**e**) pixels in an EWD; (**f**) computer.

An Agitek AFG-3052C function generator and an Agitek ATA-2022H high voltage amplifier were used as driving devices. The driving waveform was edited by the PC and transmitted to an arbitrary function generator. The high voltage amplifier can amplify the driving waveform and output the driving voltage to the EWD. The process of ink breakage is recorded by a camera and the video format is saved, then the EWD aperture ratio data is calculated by an image processing software. The system is shown in Figure 5.

#### *3.2. The Driving Waveform Structure*

In the design of the driving waveform, the C-V curve provides important parameters for the driving waveform design. The C-V curve of a typical EWD panel is shown in Figure 6, which can be tested by the optical testing system in Figure 5. The driving voltage steps from 0 V to 30 V, and its speed is 4 V/s. At each step voltage, a high-speed camera is used to record the real-time image of the EWD pixel and calculate the aperture ratio. In Figure 6, the threshold voltage of the ink rupture whose value is 15 V can be observed clearly. The capacitance of the pixel increases sharply when the driving voltage is higher than the threshold voltage, and the amount of the charge at the EWD three-phase contact line increases rapidly. However, the optical response is not linear with the change of the driving voltage, which is a significant hysteresis between the rising and falling of the C-V curve.

**Figure 6.** The relationship between the capacitance and the driving voltage in an EWD.

In addition, the threshold voltage for ink film reformation (red line) is lower than the threshold voltage for ink film dispersion (black line). According to the C-V curve, a large charge is rapidly accumulated in the hydrophobic insulator, which results in a sudden change of the electric field force which is likely to lead to ink dispersion when the capacitance value of the EWD pixel increases rapidly.

In order to avoid ink dispersion, we designed a new driving waveform with three stages. In the first stage the driving voltage starts from 0 V, and a reverse electrode pulse voltage of several milliseconds is applied to remove the electric charge which is trapped in the hydrophobic insulating layer. Hence, the polarization and the hysteresis phenomenon are avoided. The step voltage in this stage is raised to the threshold voltage (15 V) which is shown in the C-V curve. At the second stage a driving voltage with a slope of 0.4 V/s is applied, and the ink film is prevented from being dispersed during the rupture of the ink film, so the aperture ratio of the EWD can be kept at a stable value. During the third stage the duty cycle of the driving waveform is adjusted to keep the driving voltage at a high level to maintain the shrinkage state of the ink. A driving waveform with a rising slope for improving the ink dispersion phenomenon has been proposed [12], as shown in Figure 7a. The voltage rising time of the driving waveform is t2–t1 slower than the proposed driving waveform in this paper, as shown in Figure 7b.

**Figure 7.** The structure of driving waveforms. (**a**) The driving waveform with a rising slope. (**b**) The proposed driving waveform in this paper.

The wetting mechanism of EWDs is related to the properties of ions in liquid solution. Because of defects in dielectric layer, ions in liquid can be bound by the dielectric layer easily, and the binding strength of the negative ion is greater than that of positive ion. Therefore, a large number of negative ions are bound by the dielectric layer when the positive voltage is applied. However, the negative ion cannot be released immediately when the applied voltage drops to 0 V, which results in hysteretic response of EWDs. According to the behavior of the ion which is trapped in dielectric layer, charge trapping in wetting dielectric layer can be controlled by the frequency of the driving waveform, reverse electrode driving mode and duty cycle coefficient (K) of the driving waveform.

#### **4. Experimental Results and Discussion**

#### *4.1. The Frequency of the Driving Waveform*

The frequency of the driving waveform depends on two factors: the viscosity of the liquid and the charged ions in the hydrophobic insulating layer. The charged ion leads to the hysteresis effect of ink shrinkage and spread. In addition, the higher the ink viscosity, the higher the driving voltage or lower driving frequency. The conversion time of the charged ion can be shortened when the driving waveform has a high frequency, but there is not enough time to remove the charged ion trapped in the insulating layer, which leads to the accumulation of ions in the hydrophobic insulating layer; this is the main factor which leads to the hysteresis phenomenon of EWDs. Thus, the setting of the best frequency directly affects the performance of the driving waveform.

In the testing process, the frequency of the driving waveform must higher than 25 Hz, because the flicker can be discerned by human eyes when the frequency is lower than 25 Hz [7]. Therefore, the conventional driving waveform and the proposed driving waveform in this paper are tested with different driving frequencies (30–90 Hz), which are higher than 25 Hz. The experimental results are shown in Figure 8. The aperture ratio is increased at the beginning, and then, it gradually decreases in the process of driving waveform frequency change from 30 Hz to 90 Hz; the highest aperture ratio is about 60 Hz. Hence, the change of ink shrinkage is related to the frequency of the driving waveform, which determines the release rate of charged ions. The aperture ratio value is smaller when the frequency is lower.

**(b) Figure 8.** The relationship between the frequency and pixel aperture ratio. (**a**) Conventional driving waveform. (**b**) The proposed driving waveform in this paper.

In addition, the ink motion is mainly affected by charged ions which bind in dielectric layer, ink viscosity, and friction. However, the ink aperture ratio fluctuates when the ink motion lags behind the conversion of the driving waveform, which can lead to the decrease of the average aperture ratio, as shown in Figure 9. The aperture ratio of the pixel and the number of charges bound in the dielectric layer are decreased when the frequency is increased gradually. At the same time, the charged ion accumulated at the three-phase contact line is decreased, but the change speed of the charged ion cannot keep up with the voltage polarity conversion. In addition, the oscillation amplitude of the ink can be decreased when the frequency of the driving waveform is increased. So, T ≈ 1/60 s is the key cycle for the stable shrinkage of the ink in the EWD, as shown in Figure 9.

**Figure 9.** The relationship between aperture ratio (ink distribution) and the frequency of the driving waveform.

#### *4.2. The Duty Cycle of the Driving Waveform*

The aperture ratio of EWDs is determined by the voltage conversion frequency of the driving waveform. Nevertheless, the duty cycle of the driving waveform is another important factor for the performance of EWDs. In this paper, the experimental results show that the proposed driving waveform with a modulated duty cycle has better performance. Compared with the conventional driving waveform, the charged ion's behavior can be better controlled and the pixel aperture ratio can be improved by optimizing the duty cycle. In Figure 10, with the same driving voltage and key period (T ≈ 1/60 s), the performance of the conventional driving waveform and the proposed driving waveform in this paper are compared in respect to the different duty cycle coefficients. Obviously, the two kinds of driving waveform have a maximum aperture ratio when the duty cycle coefficient is K ≈ 0.65. So, the duty cycle of the proposed driving waveform is set as 0.65.

**Figure 10.** The relationship between the aperture ratio and the duty cycle in the driving waveform.

#### *4.3. The Performance of Driving Waveforms*

The aperture ratio and the response time are important factors in the driving waveform of EWDs. However, a slower slope of voltage rising speed can achieve a maximum aperture ratio in EWDs. In other words, the slower the voltage rising slope, the longer the response time required, which affects the display frame rate seriously.

As shown in Figure 11, the aperture ratio change trend of three driving waveforms are presented. Obviously, the shortest response time (8 ms) is achieved when the conventional driving waveform is applied, as shown in Figure 11a, but the ink is dispersed and the EWD aperture ratio is lowest. The response time becomes longer (65 ms) when a voltage slope is inserted into a driving waveform, as shown in Figure 11b; the ink is not dispersed and the aperture ratio of EWD increases significantly. In Figure 11c, the proposed driving waveform of this paper can achieve a shorter response time (15 ms). At the same time, the aperture ratio can also reach the maximum value (74%), which is about 8% higher than that of the conventional driving waveform. In Figure 11d, the proposed driving waveform can effectively control the shape of the ink, and a maximized aperture ratio is achieved. In order to obtain a clear comparison, parameter values among three kinds of driving waveforms are shown in Table 1.

**Figure 11.** Change process of the EWD aperture ratio under driving waveforms. (**a**) Conventional driving waveform. (**b**) Driving waveform with a slow slope. (**c**) The proposed driving waveform in this paper. (**d**) Comparison of aperture ratio and response time among three driving waveforms.



#### **5. Conclusions**

In order to solve the problems of ink dispersion, hysteretic response, and low aperture ratio caused by the defect of the EWD dielectric layer, a driving waveform with a reverse electrode pulse and an optimized voltage slope was designed in this paper. A series of experiments were executed to test the driving waveform, specifically to include the pixel aperture ratio, frequency, and the duty cycle of the driving waveform. The results show that the aperture ratio can reach its largest value when the frequency of the driving waveform is about 60 Hz and duty cycle coefficient is 0.65. Compared with the conventional driving waveform, the aperture ratio increased by about 8%, and the ink steadily shrunk without dispersion. Hence, the display quality of EWDs was improved by optimizing the driving waveform.

**Author Contributions:** Z.Y. and L.W. designed this project. W.F. and Z.Y. carried out most of the experiments and data analysis. Y.L. and W.H. performed part of the experiments and helped with discussions during manuscript preparation. L.L. and Z.Z. contributed to the data analysis and correction. L.S., C.Z., and G.Z. gave suggestions on project management and conducted helpful discussions on the experimental results.

**Funding:** This research was funded by the Key Research Platforms and Research Projects in Universities and Colleges of Guangdong Provincial Department of Education (No. 2018KQNCX334), Zhongshan Innovative Research Team Program (No. 180809162197886), Guangdong government funding (No. 2014A010103024), Zhongshan Institute high-level talent scientific research startup fund project (No. 416YKQ04), Project for Innovation Team of Guangdong University (No. 2018KCXTD033), and the National Key R&D Program of China (No. 2018YFB0407100-02, No. 2016YFB0401502).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Di**ff**erent Regimes of Opto-fluidics for Biological Manipulation**

**John T. Winskas 1, Hao Wang 1, Arsenii Zhdanov 1, Surya Cheemalapati 1, Andrew Deonarine 2, Sandy Westerheide <sup>2</sup> and Anna Pyayt 1,\***


Received: 23 October 2019; Accepted: 19 November 2019; Published: 21 November 2019

**Abstract:** Metallic structures can be used for the localized heating of fluid and the controlled generation of microfluidic currents. Carefully designed currents can move and trap small particles and cells. Here we demonstrate a new bi-metallic substrate that allows much more powerful micro-scale manipulation. We show that there are multiple regimes of opto-fluidic manipulation that can be controlled by an external laser power. While the lowest power does not affect even small objects, medium power can be used for efficiently capturing and trapping particles and cells. Finally, the high-power regime can be used for 3D levitation that, for the first time, has been demonstrated in this paper. Additionally, we demonstrate opto-fluidic manipulation for an extraordinarily dynamic range of masses extending eight orders of magnitude: from 80 fg nano-wires to 5.4 μg live worms.

**Keywords:** opto-fluidics; micro-manipulation; cells; microparticles

#### **1. Introduction**

Until now, to the best of our knowledge, the most common and reproducible methods of microscale manipulation have included mechanical and magnetic micro-manipulators [1,2] and various types of optical tweezers [3] including: holographic [4], plasmonic [5], antenna-based [6], and photonic crystal-based tweezers [7]. While these methods work well on smaller particles and nano-scale objects [8,9], they are limited in their application to biological specimens. They might induce severe heating of biological material [10] and phototoxicity [7]. More recently, magnetic levitation of cells has allowed for 3D manipulation of biological specimen and sensing capabilities that were unachievable using traditional 2D manipulation of cells, however it required the use of special magnetic fluids that are not compatible with many biological applications [2].

An alternative approach to cell manipulation is to use optofluidic tweezers (OFT) in capillaries [6] or thermally-induced current generation by absorption in amorphous silicon [11] using a light source focused through a microscope objective. These approaches are fundamentally different from the traditional optical and optoelectronic tweezers, and magnetic levitation, as they can attain much stronger forces [12,13]. Most recently, a thermo-plasmonic approach to OFT has appeared as the next generation of micro-manipulation techniques. In plasmon-assisted microfluidics, plasmonic heating is used to generate convective flow patterns [14–20]. Initially, simulations included the modeling of convection currents induced by photo-heating a single gold disk [14], which predicted very low fluid velocities in nm/s range. Later, it was demonstrated that localized heating of plasmonic structures could produce theoretically predicted toroid-shaped convection patterns [14–16]. Consequently, arrays of different plasmonic patterns were demonstrated [15], and this helped to somewhat increase the velocity. The currents could be generated by heating with different wavelengths of light in the near IR spectrum shining on a variety of plasmonic structures and rough metal films [14,17–20]. However, only small particles could be manipulated without causing significant heating. The main reason for that was a prior focus on optical properties of the whole system while ignoring its thermal properties. Later on, it was demonstrated that the thermal conductivity of the substrate plays an important role in the efficiency of the current generation. By adding continuous indium tin oxide (ITO) film to the substrate, the fluid velocity increased, and small particles were able to move with the speeds of up to 2 μm/s [21]. In addition to that, much faster flow can be demonstrated using thermo-plasmonic heating with the generation of a water vapor microbubble, but this requires a pretty high temperature of operation [22,23]. However, in all of these approaches to OFT the manipulated objects were exposed to significant amount of light radiation that could potentially damage cells or interfere with fluorescent imaging. Also, they were done using light coupled through a microscope objective, which made the whole setup inflexible in terms of independent observation and manipulation. Finally, they supported manipulation of small cells or particles on 2D surfaces, while we, for the first time, propose 3D manipulation and movement of much larger objects, truly pushing the limit of micromanipulation while using a very low controlling light power. This was accomplished because we were able to simultaneously optimize optical and thermal properties of our substrate to maximize the performance.

Our substrate contains a continuous bi-metal layer (Figure 1a), and this configuration has multiple advantages over traditional thermo-plasmonic structures. In comparison to small plasmonic patterns on transparent substrates, the continuous metal layer allows us to avoid the heating and phototoxicity that occurs due to direct light exposure of cells, while also providing the added advantage of, previously difficult, effective fluorescent imaging. The substrate is composed of a microscope glass slide with sputtered continuous bi-layer of chromium and gold. Light is absorbed by a thin layer of chromium, and the gold layer provides an interface that is biocompatible with cells and blocks the sample from light. Furthermore, the metallic bi-layer is optimized for maximum absorption by supporting a low-quality resonator, providing multiple passes of light through the chromium layer. This metallic bi-layer design can be optimized for various wavelengths by varying the thickness of both metals to optimize compound optical absorption and total thermal conductivity, allowing for the most effective micro-current generation. In this paper we demonstrate the safe and effective biological manipulation at two different wavelengths (532 and 808 nm) and show that the convective current generation can be efficiently supported with the help of the bi-metallic structure. Additionally, we demonstrate that opto-fluidic convection can be applied to several new biological functionalities including cell and particle capturing and levitation, fluid mixing, precise removal of biological material from surfaces, and cell/particle sorting. We also demonstrate that these phenomena can be optimized for ultra-efficient 3D manipulation of micro-scale biological objects, with seven orders of magnitude difference in mass, ranging from objects as tiny as silver nanowires to live organisms as large as an entire *Caenorhabditis elegans* worm.

**Figure 1.** Explanation of opto-fluidic manipulation. (**a**) A 2-D model showing the trapping and levitation of particles using the convective vortices generated by the laser excitation. The arrows show fluid velocity vectors, and the thickness of arrows corresponds to the speed of movement. (**b**) Optical microscope image showing stream of 5 μm polystyrene beads levitated by opto-fluidics. (**c–h**) Optical microscope images and corresponding three-dimensional models of the three regimes of opto-fluidic particle manipulation; (**c**,**f**) low power—no controllable particle movement; (**d**,**g**) medium power—trapping regime; (**e**,**h**) high power—levitation regime. In (**e**) the spot in the center looks out of focus because in that area particles are projected upwards.

#### **2. Methods**

A schematic of the experimental setup is shown in Figure 1a. To manipulate micro-scale particles, we coupled a laser light into an optical fiber that had been cleaved on one end. The fiber was inserted into a fiber holder attached to a micromanipulator allowing XYZ fiber movement. After aligning the cleaved fiber tip underneath the microscope objective, we placed the substrate between the fiber and the objective. We then pipetted a drop of a media containing particles or cells on top of the gold surface. Next, the laser was turned on and light from the optical fiber was shining under the substrate, through the glass layer and chromium. By the time it reached the gold, it was partially attenuated by the chromium layer, it was then reflected by the gold layer and again propagated through the chromium experiencing additional attenuation. The gold layer protects the sample from the excitation light and enhances the attenuation and heat generated by the metallic bi-layer. Since the chromium thickness (5–20 nm) is significantly smaller than thickness of gold (200 nm), the thermal conductivity properties of the metallic bi-layer are dominated by gold. This results in a highly efficient local fluid heating. Continuous laser heating of the substrate locally warms the fluid and it continuously rises to the top of the droplet, this results in the formation of a consistent vortex shaped as a toroid (Figure 1a). We used the horizontal part of the current to push objects towards the center of the heated spot and the vertical

part of the current to levitate particles with the stream of warm fluid upwards. This allows trapping relatively large objects, such as live nematodes and levitating a variety of smaller objects, such as red blood cells and micro-beads.

#### **3. Results**

Here we demonstrate that the intensity of the excitation light can be used to precisely control the fluid velocity and the forces applied to the microscopic objects. Our COMSOL simulations showed that the fluid velocity of the convective vortices is linearly proportional to the temperature of the heat source, while our experiments demonstrated that the temperature of the fluid is linearly proportional to the optical power. Thus, the fluid velocity is also linearly proportional to the optical power. This gave us the ability to easily and effectively adjust the fluid velocity and precisely control the size of the objects that we manipulated.

To study fluid velocity under different heating conditions, we used COMSOL multi-physics software that allows the simultaneous study of heat transfer and fluid dynamics. We simultaneously solve the Navier–Stokes equations and the conservation of energy equation to determine the velocity and the temperature fields. The micro-flow patterns formed by Rayleigh–Benard convection are analyzed.

Equations: Computation fluid dynamics is used to analyze the flow field. We are solving the continuity equation (Equation (1)), Navier–Stokes equations (Equations (2)–(4)), and conservation of energy equation (Equation (5)) for the corresponding initial and boundary conditions. Note that *u*, *v*, and *w* are components of the fluid velocity, ρ is density, *T* is the temperature, and *k* is thermal diffusivity. Conditions for Rayleigh–Benard convection are considered.

$$\frac{\partial \mathbf{u}}{\partial \mathbf{x}} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 \tag{1}$$

$$\frac{\partial \mathbf{u}}{\partial t} + \boldsymbol{u} \frac{\partial \mathbf{u}}{\partial \mathbf{x}} + \boldsymbol{v} \frac{\partial \mathbf{u}}{\partial y} + \boldsymbol{w} \frac{\partial \mathbf{u}}{\partial z} = -\frac{1}{\rho} \frac{\partial \boldsymbol{\delta}}{\partial \mathbf{x}} + f\_{\mathbf{x}} + \boldsymbol{v} \left( \frac{\partial^2 \mathbf{u}}{\partial \mathbf{x}^2} + \frac{\partial^2 \mathbf{u}}{\partial y^2} + \frac{\partial^2 \mathbf{u}}{\partial z^2} \right) \tag{2}$$

$$\frac{\partial v}{\partial t} + u \frac{\partial v}{\partial x} + v \frac{\partial v}{\partial y} + w \frac{\partial v}{\partial z} = -\frac{1}{\rho} \frac{\partial \delta}{\partial y} + f\_y + v \left( \frac{\partial^2 v}{\partial x^2} + \frac{\partial^2 v}{\partial y^2} + \frac{\partial^2 v}{\partial z^2} \right) \tag{3}$$

$$\frac{\partial w}{\partial t} + u \frac{\partial w}{\partial x} + v \frac{\partial w}{\partial y} + w \frac{\partial w}{\partial z} = -\frac{1}{\rho} \frac{\partial \delta}{\partial z} + f\_z + v \left( \frac{\partial^2 w}{\partial x^2} + \frac{\partial^2 w}{\partial y^2} + \frac{\partial^2 w}{\partial z^2} \right) \tag{4}$$

$$
\rho C\_P \left( \frac{\partial T}{\partial t} + u \frac{\partial T}{\partial x} + v \frac{\partial T}{\partial y} + w \frac{\partial T}{\partial z} \right) = -k \left( \frac{\partial^2 T}{\partial x^2} + \frac{\partial^2 T}{\partial y^2} + \frac{\partial^2 T}{\partial z^2} \right) \tag{5}
$$

The onset of natural convection is determined by the Rayleigh number Ra. For thermal convection due to heating from below, Ra <sup>=</sup> <sup>ρ</sup>0β*g*Δ*TL*<sup>3</sup> αμ , where ρ<sup>0</sup> is the reference density, typically picked to be the average density of the medium, *g* is the local gravitational acceleration, β is the coefficient of thermal expansion, Δ*T* is the temperature difference across the medium, *L* is the characteristic length-scale of convection, α is a thermal diffusivity, and μ is the dynamic viscosity. Preliminary data based on COMSOL simulations show the velocity of the observed convection exhibits an approximately linear relationship with the laser power. Arrows representing velocity vectors simulated using COMSOL are shown in Figure 1a. We extracted the fluid pattern from the simulations and combined it with the illustration of the setup to simplify the understanding of the current structure. The simulations of the laser light-controlled currents were conducted for substrate temperatures ranging from 30 to 50 ◦C. In simulations at all temperatures, the current structure was very similar to the one shown in Figure 1a. The main difference was that all velocities were proportionally scaling up. While fluid velocity measured in the center of the droplet was ~1.2 μm/s at 30 ◦C, it reached 6.5 μm/s at 50 ◦C.

#### *3.1. The Substrate Optimization, Fabrication, and Characterization*

In contrast with previous studies, which used different types of patterned plasmonic structures and nanoparticles, we chose continuous metallic bi-layer. We found out from our COMSOL simulations that the thermal conductivity of the substrate greatly affects the efficiency of the current generation and its velocity. Figure 2a demonstrates that by heating three differently structured substrates (continuous gold layer, continuous chromium layer, and gold islands) to the same temperature, we observe a noticeable difference in the fluid velocity. Since gold has the highest thermal conductivity, the resulting convective current induced in the fluid is the most powerful. A continuous layer of chromium is less efficient due to its lower thermal conductivity and gold islands that do not form a continuous thermally-conductive layer were observed to be the least efficient. Therefore, the optimal structure should be continuous and have high thermal conductivity, preferably closer to gold.

**Figure 2.** The substrate optimization. (**a**) COMSOL simulations were conducted to evaluate the most efficient substrate structure. (**b**) Finite difference time domain (FDTD) simulation results for substrate absorption optimization for 532 and 800 nm wavelengths varying chromium thickness from 0 to 100 nm while holding gold thickness constant at 200 nm. (**c**) FLIR IR camera imaging was used to measure the maximum temperature in a 1 μL drop of fluid. (**d**) Experimentally measured temperature increase while testing two substrates with 532 nm light with power ranging from 0 to 80 mW.

After determining the requirements for the thermal conductivity of the substrate, the next step was to take into consideration its optical and biocompatibility properties. A metallic bi-layer was chosen as a substrate because no single metal could simultaneously satisfy all our requirements: biocompatibility, high thermal conductivity, and high optical absorption for visible/IR wavelengths. Chromium was chosen for one of the layers for its high absorption for chosen wavelengths (532 and 800 nm) and gold was selected for its biocompatibility, thermal conductivity, and reflective properties. The individual metal thicknesses of the bi-layer were then studied to optimize overall performance. To optimize the optical absorption of the bi-layer we conducted finite difference time domain OptiFDTD simulations. The goal was to create a substrate that has opaque metallic bi-layer efficiently absorbing controlling light and isolating the sample from any light exposure. The gold layer thickness was chosen to be 200 nm for optimal reflection, while the thickness of the chromium layer was varied between 5 and 200 nm in the simulations (Figure 2b). Chromium and gold were both modeled as Lorentz–Drude dispersive materials. Optimization was conducted for wavelengths 532 and 800 nm because of wide availability of low-cost lasers with those wavelengths. In simulations the beams were linearly polarized, had a Gaussian profile, were emitting continuously, and were perpendicular to the substrate.

The absorption by bi-metallic substrate at both wavelengths with respect to the chromium thickness is plotted in Figure 2b. Since gold is highly reflective at both wavelengths, adding a layer of 200 nm of gold on top of chromium effectively doubles the propagation length of the light in the chromium layer. The absorption of the bi-layer is the highest for 5 nm of chromium at 532 nm and 20 nm of chromium at 800 nm. Interestingly, the optimized absorption by the layered structure was found to be 25%–30% higher than absorption by a single thick layer of Cr. This is the result of multiple reflections in the low-quality factor plasmonic resonator consisting of a metal bi-layer on a glass substrate.

Based on the results of the simulations, we fabricated substrates optimized for 532 and 808 nm lasers (5nm-Cr/200nm-Au and 20nm-Cr/200nm-Au, respectively), with the bi-layers deposited on glass slides by sputtering. After fabrication, the performance of the substrates was experimentally evaluated. All temperature measurements were conducted using a FLIR thermal camera. It was demonstrated that under local light exposure from an optical fiber, both substrates were able to efficiently absorb light and heat a 1 μL drop of water placed on top of the substrate (Figure 2c), reaching a maximum equilibrium temperature within 1–2 minutes after beginning of the experiment. As the power of the laser beam was increased from 0 to 80 mW, the temperature increased linearly from 20 to 40 ◦C (Figure 2d). This result, when combined with the linear relationship between the temperature and the fluid velocity vectors observed in COMSOL simulations, allowed us to infer a linear relationship between the fluid velocity in the generated convective vortex and the laser power.

Additionally, this result further demonstrated that an increase of 4 mW of the laser power results in a corresponding water temperature increase of 1 ◦C. Finally, it was experimentally confirmed that, similarly to the simulations shown in Figure 2b, the absorption at 532 nm is more efficient for chromium thickness of 5 nm comparing to 20 nm (Figure 2d), though both of them perform quite well.

#### *3.2. Sample Manipulation Using Optimized Substrates*

After optimizing the substrate, we demonstrated several new applications of 3D opto-fluidic manipulation to biological specimens including cell accumulation/trapping, projection, separation/filtering, followed by a viability study proving the safety of this approach to living cells. In the following experiments, we used human fibroblast cells suspended in a cell growth medium. The cells were treated with a 0.25% (w/v) Trypsin—0.53 mM ethylenediaminetetraacetic acid (EDTA) solution helping to temporarily keep the cells from bonding to the gold surface during experiments. We demonstrated the operation of the trapping regime at a controlling power of 15 mW using the 20 nm-Cr/200 nm-Au substrate optimized for 808 nm wavelength (Figure 3a–d). For this experiment a 5 μL drop of cell medium containing live fibroblast cells was placed on the substrate, and the optical fiber was aligned under an area of the substrate that initially contained no cells (Figure 3a). After 480 seconds, ten fibroblast cells were captured, and some of them from a distance greater than 250 μm (the field of view of the microscope).

To demonstrate the projection regime, we used a laser power of 47 mW applied to 5 nm-Cr/200 nm-Au substrate optimized for 532 nm wavelength. The sample was pipetted from a mixture of 1 μL of human blood and 1 mL of the medium containing fibroblast cells. The induced convection vortices were strong enough to levitate red blood cells vertically from the surface. After 260 seconds, we observed that both red blood cells and fibroblast cells were trapped in the center of the excitation area and there was a continuous stream of red blood cells projected upward (Figure 3e–h).

In the next set of experiments, it was observed that if we scan the optical fiber under the substrate (within the field of view), while operating in a high-power projection regime, we were able to successfully levitate and remove red blood cells from the surface without moving the fibroblast cells or trapping red blood cells (Figure 3i–l). The laser power used in these experiments was the same as in the previous experiment (47 mW). When optical fiber moves, the toroid current pattern does not have enough time to form, and the vertical water movement simply projects the red blood cells off the surface, like a miniature "pressure washing" system. These tree regimes—trapping, levitation, and pressure washing can be in parallel and differently applied to different types of particles with applications in trapping, mass-based filtrations/separations, measurements of masses of individual cells, and sensing applications.

**Figure 3.** Experimental demonstration of several opto-fluidic manipulation regimes. (**a**–**d**) Trapping of fibroblast cells during 480 second period. (**e**–**h**) Trapping of heavy fibroblast cells simultaneously with the projection of lighter red blood cells. (**e**,**f**) Microscope imaging plane of the substrate, (**g**,**h**) imaging above the substrate demonstrates stream of red blood cells being levitated. (**i**–**l**) Targeted "pressure washing" removing light red blood cells from the surface leaving in place heavy fibroblast cells. Optical fiber scans under the whole portion of the substrate visible in the field of view selectively removing one type of the cells from the surface.

Finally, the following experiments were conducted to demonstrate the extraordinary dynamic range of masses that could be manipulated using this technology, along with showing that this approach can also be used to build complex, multi-layer structures (Figure 4). First, we trapped live *C. elegans* worms using a 35 mW laser power (Figure 4a–d). The worms were still active after the manipulation, showing that the temperature increase did not damage them. The live worms are the largest reported living organisms manipulated using light. They have an average length of 1 mm and width of 50 μm. In contrast to the macroscopic worms, we show that the same technique can be used to manipulate objects as small as silver nanowires which have masses seven orders of magnitude smaller than that of worms (Figure 4m,n). Using our opto-fluidic manipulation we were not only able to capture silver nanowires from the solution but demonstrated the ability to levitate them and build multi-layer structures. In future this can be used for building more complex 3D nano-scale structures without the use mechanical manipulators. Additionally, we also demonstrated that our approach can be used for selective size-based trapping and assembly (Figure 4e–l). We were able to separate 20 μm polystyrene micro-beads from their mixture with 5 μm beads by trapping larger particles and projecting away smaller ones. The 20 μm beads assemble in multi-layer structure that can be potentially used for building photonic crystals and other complex ordered assemblies.

**Figure 4.** Demonstration of the extraordinary dynamic range of opt-fluidic manipulation and application to the assembly of complex structures. (**a**–**d**) Capturing live worms using 40 mW power. Worm length varied from 50 to 200 um. (**e**–**l**) Selective capturing of large particles and their multi-layer assembly. (**m**,**n**) A multi-layer silver nanowire structure built using trapping and levitation.

#### *3.3. Cell Viability*

To test the biocompatibility of the technology, a series of cell viability tests were conducted for different power regimes used in cell manipulation experiments (Figure 5). To ensure all cells used were studied under the same conditions, we integrated the substrate with a PDMS chip containing multiple identical wells. Each well was used to test cell viability after manipulation using one of the regimes, in addition to wells used for control experiments without any manipulations. The fibroblast cells were manipulated in low-power regime (3.7 mW), medium-power trapping regime (38 mW), and the high-power projection regime (78 mW) (Figure 5a1,a2,b1,b2,c1,c2). The cells in each well were manipulated for 150 seconds, after which, the whole chip was returned into an incubator, where the cells were cultured for 24 hours. During that time, the cells bonded to the gold surface and started growing. Bright-field microscopy demonstrated that the cells looked healthy and had an expected morphology (Figure 5a3,b3,c3). We then used a live/dead cell imaging kit to stain live cells with green-fluorescence-emitting dye, and dead ones with red-emitting dye. It was confirmed using fluorescence imaging that the cells appeared healthy and emitted a green fluorescence signal

(Figure 5a4,b4,c4). After a detailed examination, on average, less than one dead cell was observed per well, matching the results from the control well that was not exposed to any manipulations. This result confirmed that all regimes of opto-fluidic manipulation were biocompatible and did not decrease cell viability.

**Figure 5.** Demonstration of fibroblast cell viability after opto-fluidic manipulation. (**a1,a2**) Before and after low-power manipulation at 3.7 mW. (**b1,b2**) Before and after medium-power trapping at 38 mW. (**c1,c2**) Before and after high-power projection regime at 78 mW. (**a3,b3,c3**) Bright-field and (**a4,b4,c4**) fluorescent images of cells after 24 hours of incubation.

#### **4. Conclusions**

To summarize, we have demonstrated for the first time several new regimes of opto-fluidic manipulation. We showed the trapping of objects with the extraordinary seven orders of magnitude dynamic range of masses, from microscopic nano-wires to macroscopic live *C. elegans* worms. Furthermore, our studies have shown that we not only captured, but levitated, filtrated, and "pressure washed" a variety of objects. On top of this, we have observed, through cell viability testing, that this technology can be safely used, not only for capturing but also the levitation and separation of cells and other biological organisms. We also show that this method can be customized for different wavelengths of light. This opens the door for consistent, reproducible, and easily fabricated devices that can be implemented across a wide range of fields that require gentle and powerful biomanipulation of objects and cells, and small live organisms. We suggest that some of the potential applications might include filtering different types of cells, separation of bacteria or circulating tumor cells from biological samples. Additionally, capturing worms using our approach is very attractive, since all traditional mechanical ways to move the worms around require use of sharp metal tips and other objects that damage worms, while optical tweezers would require extremely high power that would kill them. At the same time, our approach allows gentle movement of live worms to a needed location for further imaging and analysis.

**Author Contributions:** A.P. lead the project, conceived and designed the experiments and simulations; J.T.W. and H.W. performed the experiments; S.C. conducted simulations, A.D. and S.W. helped with C. elegans experiments, A.Z. created visualizations, all authors contributed to manuscript writing.

**Funding:** Dr. Anna Pyayt's Lab was supported by NSF Grant # 1701081.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Micromachines* Editorial Office E-mail: micromachines@mdpi.com www.mdpi.com/journal/micromachines

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18