Efficient separation of bioparticles is often the critical, first step in biotechnology, cellular biology and biomedical analysis. Labelbased methods such as fluorescenceactivated cell sorting and magneticactivated cell sorting have been successfully used for this purpose. But they have limitations such as expensive equipment, timeconsuming process, and potential effects of the labels on downstream analysis. Alternatively, labelfree methods like density gradient centrifugation and membrane separation are commonly used to isolate target bioparticles from heterogeneous biofluid samples. However, they suffer from inherent problems of low purity, recovery rate, and viability.
Capillarity Enabled Large-Array Liquid Metal Electrodes for Compact and High-Throughput Dielectrophoretic Microfluidics Huichao Chai Junwen Zhu Yongxiang Feng Fei Liang Qiyan Wu Zhongjian Ju Liang Huang* Wenhui Wang* H C Chai, J W Zhu, Y X Feng, F Liang and W H Wang State Key Laboratory of Precision Measurement Technology and Instrument Department of Precision Instrument Tsinghua University Beijing 100084, P R China Email Address: wwh@tsinghua.edu.cn Liang Huang School of Instrument Science and Opto-Electronics Engineering Hefei University of Technology Hefei 230009, P R China Q Y Wu, Z J Ju The First Medical Center of PLA General Hospital Beijing 100853, P R China Accepted Article Keywords: liquid metal, capillarity, large-array electrodes, dielectrophoresis, microfluidics Dielectrophoresis (DEP) particle separation has label-free, well-controllable, and low-damage merits Sidewall microelectrodes made of liquid metal alloy (LMA) inherits the additional advantage of thick electrodes to generate impactful DEP force However, existing LMA electrode-based devices lack the ability to integrate large-array electrodes in a compact footprint, severely limiting flow rate and thus throughput Herein, a facile and versatile method is proposed to integrate high-density thick LMA electrodes in microfluidic devices, taking advantage of the passive control ability of capillary burst valves (CBVs) CBVs with carefully designed burst pressures are co-designed in microfluidic channels, allowing self-assembly of LMA electrode array through simple hand-push injection The arrayed electrode configuration brings the accumulative DEP deflection effect Specifically, we demonstrate to fabricate 5000 pairs of sidewall electrodes in a compact chip to achieve 10 times higher throughput in DEP deflection We applied the 5000electrode-pair device to successfully separate the mixed sample of human peripheral blood mononuclear cells (PBMCs) and A549 cells with the flow rate of 70 µL min−1 It is envisioned that this work can greatly facilitate LMA electrode array fabrication and offer a robust and versatile platform for DEP separation applications Introduction Efficient separation of bioparticles is often the critical, first step in biotechnology, cellular biology and biomedical analysis.[1, 2, 3] Label-based methods such as fluorescence-activated cell sorting and magneticactivated cell sorting have been successfully used for this purpose.[4, 5] But they have limitations such as expensive equipment, time-consuming process, and potential effects of the labels on downstream analysis Alternatively, label-free methods like density gradient centrifugation and membrane separation are commonly used to isolate target bioparticles from heterogeneous biofluid samples.[6, 7] However, they suffer from inherent problems of low purity, recovery rate, and viability Comparing to the above separation methods, emerging microfluidic techniques offer a promising solution with unique advantages.[8, 9, 10, 11] Among these, dielectrophoresis (DEP), as a typical electrodynamic method, is an attractive technique for bioparticle manipulation, and can achieve label-free, wellcontrollable, low-damage, and low-cost separation of bioparticles based on size and dielectric properties.[12, 13] With these capabilities, DEP has demonstrated a broad range of applications over the last few decades in bioparticle manipulation such as cancer cells,[14, 15] blood cells,[16, 17] bacteria,[18, 19] proteins,[20, 21] etc However, DEP techniques have long been limited by low throughput, which remains a key barrier locking its full potential.[22] This is primarily due to the limited range of DEP force, as the non-uniform electric field that generates DEP force rapidly decays with increasing distance from the electrodes In literature, DEP separation devices can be classified into two main types: insulator-based (iDEP) and electrodebased devices, based on the source of the non-uniform electric field.[23, 24] iDEP devices employ insulating This article has been accepted for publication and undergone full peer review but has1not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record Please cite this article as doi: 10.1002/adma.202310212 This article is protected by copyright All rights reserved Accepted Article structures like obstacles[25, 26] to produce a non-uniform electric field between remote electrodes However, the relatively large distance between electrodes in iDEP devices cause weak field coupling, inherently leading to low separation throughput In contrast, electrode-based devices with integrated microelectrodes in microfluidic chips can generate a strong non-uniform electric field for particle separation Various designs of planar electrodes, such as interdigitated,[27] castellated,[28] oblique,[29] arrayed,[30] and curved[31, 32] ones, have been proposed However, the electric field generated by planar electrodes decays exponentially with the distance from the electrodes along the vertical direction,[33, 34] which limits the channel height and thus the throughput Compared with planar electrodes, thick microelectrodes can maintain a more uniform electric field distribution in the vertical direction Correspondingly, DEP force can act on particles throughout the entire microchannel space, thereby increasing the throughput of particle separation While thick carbon[35] and metal[36] microelectrode designs have shown enormous advantages, their complex fabrication process and high costs restrict their applications Alternate materials, such as using ionic liquid[37, 38] and conductive polymers like Ag-polydimethylsiloxane (Ag-PDMS)[39, 40] and C-PDMS[41, 42], are easier to manufacture, but suffer from lower conductivity, which affects electric field coupling and limits their suitability for high-throughput separation Liquid metal alloys (LMAs) excel in electrical conductivity, flowability, and ease of patterning, standing as an ideal material for constructing thick electrodes in high-throughput microfluidic devices Injecting liquid LMA into microfluidic channel is a simple approach to co-fabricate and self-align thick electrodes and sample microchannel, and generates strong electric fields in the microchannel.[43] In this kind of method, barriers like micropillar rows are set up in the microchannel to separate the electrode channel from the main sample channel LMA melting at certain temperature (e.g 40◦ C) is injected into the designed electrode channels Due to surface tension, the inter-pillar regions prevent LMA from flowing into the main channel Then LMA solidifies at room temperature to form thick electrodes between the micropillars on both sides of the main channel Up to now, the fabricated electrodes have demonstrated great potentials [44] and undergone development in more electrode units to extend the DEP force action region [45, 46, 47, 48] However, existing co-fabrication methods for adjoining LMA electrodes and sample channels lack the ability to accurately control the LMA flow, which generally allows electrodes extending forward other than winding back [49, 50] This would result in a very large footprint for the microfluidic chip and yield poor efficiency in space utilization Due to the limitation in electrode fabrication, the separation throughput has been limited low (10 µL min−1 ,[44]) in consideration of milliliter clinical sample volumes.[51] In this study, we propose a method that can fabricate large-array thick LMA electrodes in a compact DEP microfluidic device, taking advantage of the passive control ability of capillary burst valves (CBVs) This method provides a long residence time in DEP manipulation by generating a long-range electric field gradient in the microchannel, enabling continuous, progressive, and high-throughput DEP particle deflection CBVs with different burst pressures are appropriately arranged in the microfluidic device (at the intersections) to achieve passive and accurate control of LMA flow In this manner, interdigital thick LMA electrodes can be self-assembled by simple hand-push syringe injection These dense electrodes allow progressive DEP deflection to achieve particle separations at high flow rates and high particle concentrations in the sample channel We design the microfluidic channels with repetitive electrode pairs, which can be extended in both row and column on a plane Through experiment, we validate progressive DEP deflection as particles (i.e., polystyrene=PS beads) pass by more and more electrode pairs Using PS beads and HeLa cells, we achieve 99.31% separation efficiency and 99.19% purity More importantly, we demonstrate to fabricate 5000 electrode pairs within a 3.3 × 1.4 cm2 area, and achieve DEP deflection of ∼40 µm under a flow rate of 100 µL min−1 (1.5 × 106 particles per min), which is 10 times the state-of-the-art throughput We further confirmed the high-throughput separation capability of the large-array LMA electrodes using a mixed sample of human peripheral blood mononuclear cells (PBMCs) and A549 cells with the flow rate of 70 µL min−1 In summary, this method significantly facilitates thick LMA electrode fabrication, and enhances the throughput in DEP particle sorting scenarios We believe that our work offers a new perspective for thick electrode fabrication and unlocks the full potential of DEP separation techniques in many applications This article is protected by copyright All rights reserved Accepted Article Figure 1: Conceptual diagram of the microfluidic device with large-array thick liquid metal electrodes for particle separation (A) Schematic illustration of the device and its basic structures for progressive DEP deflection (B) A close-up view of the repetitive DEP active regions (red) along the serpentine sample channel (C) Basic working flow of the proposed capillaritybased self-assembly of LMA electrodes array When injecting LMA, Path is filled first because w1 is greater than w2 Once the strong valve at the end of Path is reached, the weak valve is opened, allowing for the filling of Path 2 2.1 Results Design and fabrication of large-array liquid metal electrodes The microfluidic chip is designed with a single layer, as shown in Figure 1AB, consisting of a sample channel (blue) flanked by two electrode channels (golden) acting as positive and negative electrodes, respectively This article is protected by copyright All rights reserved Accepted Article 2.1 Design and fabrication of large-array liquid metal electrodes As a demonstration, the sample channel has a depth of 60 µm, two inlets for the sample and focusing sheath flows, and two outlets for collections of separated particles The electrode channels have two separate inlets for LMA filling in positive and negative electrode fabrication Along the sample channel, repetitive DEP active regions (red) are placed one by one to achieve progressive DEP deflection To fully use space, the microfluidic channel is wound back and forth Under this configuration, the key thing is to design the electrode channels such that LMA can autonomously select the flow paths during filling, thus self-assembles into a condensed electrode pattern We leverage the robust, passive liquid control ability provided by capillary burst valves (CBVs) to design the electrode channels Like our previous work[52], CBVs with different burst pressures are arranged appropriately in the microfluidic device to achieve passive and accurate control of LMA flow The electrode channels are connected to the sample channel via small orifices, which possess high burst pressure (Pstrong ) and serve as stopper valves to prevent LMA from entering the sample channel Additionally, as shown in Fig 1B, weak CBVs with appropriate burst pressure (Pweak ) are placed at each intersection between each electrode row and column to act as passive switching valves that enable path changeover when LMA flows Figure 1C shows the process of path switching during LMA filling, explaining the basic principle and workflow of the proposed capillary-enabled LMA electrode array self-assembly method Initially, when LMA is injected into the electrode channel and flows to the branch, it has the option to fill the channel along Path or Path Due to the higher burst pressure of the weak valve in comparison to the straight channel (Pstraight ), LMA firstly flows straight through Path till reaching the strong valve Then, because the burst pressure of the strong valve is greater than that of the weak valve, LMA breaks through the weak valve and fills Path This process endows the injected LMA with the ability to bifurcate flow By repeatedly performing this process, the electrode channels can be self-filled, resulting in the formation of a compact and large thick electrode array Note that each electrode pair has same structure and working condition in LMA filling, thus the electrode array can be extended indefinitely The reliability and versatility of the capillary-enabled LMA electrode array self-assembly method has been verified from three aspects, including the channel size, injecting pressure source, and applicable LMA materials As designed, successful LMA filling is highly dependent on the correct opening order of CBVs In particular, the burst pressures of these CBVs should be in the order of Pstraight < Pweak < Pstrong , (1) The burst pressures for straight channels can be calculated as ∆Pstr = −2γ cos θc ( 1 + ), w h (2) where γ is the liquid-gas surface tension, θc is the advancing contact angle, and w and h are the width and height of the microchannel Note the channels can be also designed as expansion channels (Burst pressures are modeled in Section S1, Supporting Information) Therefore, the key parameter is the channel width For h = 60 µm, the relationship between the burst pressure (∆Pstr ) and the inverse of the width ( w1 ) is positively linear (Fig S1C) According to the design criteria, the width of the straight channel must be greater than that of the switching valve, which is greater than that of the small orifices connecting sample channel, i.e., w1 > w2 > w3 , (3) where w1 is the width of the electrode channel, w2 is the width of the switching valve, and w3 is the width of the stopper valve Following the logic presented in Discussion Section, the microfluidic dimensions are determined as Table S1 We have successfully demonstrated self-assembly of the electrode pattern when the ratio of w1 and w2 is 1:2.5 Secondly, the self-assembly process is not affected by slight fluctuations in the injecting pressure source due to the high surface tension of LMA We demonstrate that even manual injection via syringe suffices for successful LMA electrode array self-assembly (Video S1) The LMA used in this work has a melting point of ∼40 ◦ C So, the LMA was totally liquefied by heating above 40 ◦ C and loaded to the pre-heated syringe to keep its liquid phase Throughout the injection process, the chip was placed on a hot plate set at 50 ◦ C (visible at the beginning of Video S1), while the tube connecting This article is protected by copyright All rights reserved Accepted Article 2.2 Finite element simulation of electric field and particle trajectory Figure 2: Simulation results for electric field distribution and particle trajectory (A) One pair of LMA electrodes in the microchannel for simulation (B-C) The electric field distribution at (B) side-view for section A-A and (C) top view when Vpp =20 V (D) Simulation of 10-µm particle trajectories through the (i) 1st, (ii) 15th, (iii) 30th, and (iv) 45th pair of electrodes under nDEP (20 Vpp and 100 kHz) the syringe to the chip was intermittently heated with a heat gun to prevent the LMA from solidifying After the injection was completed, the chip was taken off the hot plate, and the electrodes were allowed to solidify at room temperature Note there are no unfilled portions in the electrode channel or leaks into the sample channel Thirdly, three LMAs, including gallium, EGaIn alloys, and indium-tin-bismuth-leadcadmium alloys have been tested and validated for this approach Commonly used in the field of flexible electronics and microfluidics, these LMAs are successfully fabricated into workable electrodes (Section S2, Supporting Information) In this work, we adopt the indium-tin-bismuth-lead-cadmium alloy because of its availability for final device fabrication As the high surface tension properties required by this method are prevalent in liquid metal materials, we believe that this method can be applicable to other LMAs Using this method, we fabricated compact devices with large-array thick electrodes, i.e., 5000 pairs of electrodes on a 3.3 × 1.4 cm2 chip (Figure S3A) The self-assembled electrode arrays and an individual electrode pair are shown in Figure S3BC, where each electrode pair takes a classical, asymmetric electrode setup, just to create the non-uniform electric field required by DEP Note, the number of electrodes can be easily scaled-up upon demand 2.2 Finite element simulation of electric field and particle trajectory To investigate the non-uniformity of the electric field across the depth and width of the sample channel and its impact on DEP deflection of particles, we conducted simulations of electric field and particle trajectory using the electrode structure as depicted in Figure 2A The electric field distribution for one single electrode pair in the vertical and horizontal directions is illustrated in Figure 2B and 2C, respectively The simulation results confirm that the sidewall electrodes formed by LMA induce a strong non-uniform electric field across the two sidewalls, and this field is consistent throughout the depth of the sample channel This design offers three distinct advantages over planar electrode devices Firstly, the configuration of sidewall electrodes allows direct imaging of particles in the sample channel Secondly, there is no need to worry about the channel height-related electric field decay so that a higher separation throughput can be achieved with a greater channel height Thirdly, the high electrical conductivity of LMA minimizes circuit loss, allowing for DEP deflection with lower voltages The width of the stopper valve (i.e., w3 ) is effectively the width of each electrode,and has an important This article is protected by copyright All rights reserved Accepted Article 2.2 Finite element simulation of electric field and particle trajectory influence on the electric field distribution and thus DEP pattern in the sample channel To evaluate its effects, analytical modelling and numerical simulations have been conducted (Section S4, Supporting Information) It can be inferred that w3 affects the voltage drop across the sample channel and the nonuniformity or the gradient of the electrical field in the vicinity of the orifice To be specific, a smaller w3 yields higher values of electric field in the vicinity of the small orifice, but its workspace(i.e., a region with high gradient of electric field) is smaller To investigate the accumulative DEP deflection (ADD) effect generated by repetitive electrodes along the channel, we performed particle tracking simulations in a channel configured with 50 pairs of electrodes In the settings, the sample channel width is 100 µm The electrode pairs are spaced 250 µm apart the channel length direction and the orifice width in each pair is 10 µm The particles have a diameter of 10 µm and are squeezed by the sheath flow towards the positive electrode side The flow rate for the particles and sheath are µL min−1 and µL min−1 , respectively Figure 2D presents the simulation results for DEP deflection of particles passing through the 1st, 15th, 30th and 45th pair of electrodes under 20 Vpp and 100 kHz, respectively Initially, the particles span a band of approximate 25 µm in width along the positive electrode side As passing through the first pair of electrodes, they begin to experience deflection and focusing Specifically, the particles are progressively deflected from the positive electrode, with an ADD of 22.8, 27.6, 33.8 µm, after 15, 30, 45 pairs of electrodes, respectively The simulation results validate that DEP deflection can be increased by deploying more electrodes We qualitatively analyze how the forces counteracting the DEP force may affect the ADD effect As for the force that counter acts the DEP force in particle separation is primarily the drag force[8, 24] Here for separation, we are only interested in the forces along the channel width direction As the drag force is a passive force induced by the relative motion of particles and the fluid, the balance of the applied DEP force and drag force affects the motion and equilibrium position of the particles According to Stokes’ law, the drag force applied on a moving spherical particle is Fdrag = 3πµavt , (4) where vt is the relative velocity of the fluid in relation to the particle, µ is the dynamic viscosity of the fluid, and a is the particle diameter The time-averaged DEP force (FDEP ) acting on a homogeneous spherical particle suspended in a medium can be expressed by[53] FDEP = 2πa3 εm Re[KCM ]∇E , (5) where εm is the medium permittivity, E is the electric field strength, and Re[·] stands for the real part of a complex variable KCM is the Clausius–Mossotti (CM) coefficient Eqs and show that Fdrag ∝ a3 and FDEP ∝ a, i.e., the DEP force drops faster than the drag force with decreasing particle size At the equilibrium position, vt is nearly 0, Fdrag = FDEP , and FDEP is nearly This implies that for smaller particle sizes, the equilibrium position is closer to the positive electrode, as to be depicted later in Figure 4B and D Note that for a specific particle size, provided that other conditions are certain, the equilibrium position in the ADD effect is independent of the number of electrodes In practice, to fully harness the benefit of ADD effect, we can deploy enough number of electrodes such that the particle can be maximally deflected to its equilibrium position The DEP force exerted on a particle also depends on the medium To be specific, the medium conductivity σm and permittivity εm may affect the real part of Re[KCM ] (in Eq S6), which, in turn, determines the direction and magnitude of the DEP force If the DEP force pushes way a particle, it is called nDEP, otherwise it is pDEP Their turning point is determined by so-called crossover frequency, which can be simulated (Fig S6), or experimentally determined Notably, at frequencies below 10 MHz, σm plays a dominant role, whereby a smaller σm leads to a decreased crossover frequency Conversely, at frequencies exceeding 10 MHz, εm becomes the primary factor, wherein a smaller εm results in an increased crossover frequency In this study, we can perform both size-based and dielectric-property-based separation with the device by configuring the DEP settings to show its flexibility For size-based separation, we rely on nDEP to This article is protected by copyright All rights reserved Accepted Article 2.3 Progressive DEP deflection verification Figure 3: Progressive DEP deflection of 10-µm PS beads with sample flow rate of µL min−1 (A) Superimposed images showing the particle trajectory passing through the (i) 1st, (ii) 11th, (iii) 21st, (iv) 31st, and (v) 41st pair of electrodes under 30 Vpp and 100 kHz Note that (i) 1st and (iv) 31st electrode pairs are located at the turning point of the sample channel, so the number of negative electrode orifices is to accommodate the space requirements Scale bars: 50 µm (B) Progressive deflection vs the index of electrode pair under different voltages Sample size is about 300 deflect particles away from the positive electrode, and the deflection is greater for a bigger particle size For dielectric-property-based separation, we rely on both nDEP and pDEP, e.g., pDEP for one particles, and nDEP for other particles, or vice versa If only size-based separation is needed for mixed samples, inertial microfluidics is more simple and readily compatible with high-throughput analysis 2.3 Progressive DEP deflection verification For proof of concept, a device consisting of 50 pairs of thick LMA electrodes is first tested to verify the progressive DEP deflection of particles Sample of 10 µm diameter PS beads suspended in isotonic sucrose medium with a concentration of 4.85 × 106 mL−1 is used The trajectory of PS beads at a certain pair of electrodes is obtained by superimposing the images taken by the high-speed camera Figure 3A shows the particle trajectories through the 1st, 11th, 21st, 31st and 41st pair of electrodes with flow rate of µL min−1 and the DEP signal of 30 VP P and 10 kHz Initially, PS beads are focused by the sheath flow in a band of ∼30 µm on the positive electrode side While flowing through the first pair of electrodes, particles deflect ∼ µm, which may not be enough to allow sample collection from an alternate outlet from the original undeflected flow After passing more electrodes, particles have increasing deflections, i.e., 20 µm at 10th, 29 µm at 20th, 32 µm at 30th, 33 µm at 40th pair of electrodes These data confirm that the ADD is paramount (∼4 times increase after 40 pairs of electrodes) and would be effective in sample separation Next, the progressive DEP deflection is also obtained for the signal voltage - 50 V, revealing two salient features (Figure 3B) Firstly, the ADD effect is more pronounced at low voltages At low voltages (5-15 V), particle deflection after 40 pairs of electrodes is about 10 times greater than that of after a single pair The ratio decreases to about and times for medium (15 - 30 V) and high (30 - 50 V) voltages, respectively Secondly, the ADD effect shows a gradual saturation trend These two features suggest that this progressive DEP deflection method is suitable for low voltage and high-throughput particle separation applications This article is protected by copyright All rights reserved Accepted Article 2.4 Performance characterization with LMA electrode array Figure 4: Superimposed images of particle trajectories at the outlet for different (A) DEP voltage, (B) particle size, and (C) sample flow rate Scale bars: 50 µm Boxplots illustrating the particle deflection with varying (D) sample size, and (E) flow rate, under 10-50 Vpp and 100 kHz Sample size is about 300 2.4 Performance characterization with LMA electrode array We used the device with 50 pairs of LMA electrodes to study the relationship between deflection and three key parameters in working conditions of DEP, including particle size, flow rate, and signal voltage In order to visualize the deflection results for different experimental parameters, two outlets are placed in a way such that some or all particles would flow out of O2 rather than O1 when DEP is switched from off to on In experiment, the default settings are 10 µm for PS bead diameter, µL min−1 for sample flow rate, µL min−1 for sheath flow rate, 40 V (100 kHz) for DEP signal When characterizing the performance, only one parameter is changed while the default values are used for other parameters Firstly, particle size and DEP signal voltage have a direct impact on the DEP force exerted on the sample particle Theoretically, DEP force is directly influenced by the square of the electric field gradient and the cubic of the radius of the sample particle (Section S5, Supporting Information) Experimental results further verify the effect of these two parameters Figure 4A displays the eventual deflection of the PS beads at the outlet for different voltages As the voltage increases gradually from 10 V to 30 V, the deflection increases from 26.5 µm to 54.2 µm Correspondingly, the sample streamline is redirected from O1 to O2 Figure 4B illustrates the trajectory of PS beads at the outlet for different particle sizes For beads of 10 µm and µm, the deflections are 66.4 µm and 55.1 µm, respectively, with a relatively small difference Particles of both sizes could be pushed to O2 However, when the diameter is reduced to µm, the deflection at the outlet drops significantly to 31.8 µm, and thus PS beads are no longer separated to O2 Secondly, the sample flow rate, while not directly affecting the DEP force applied to the particle, influences the eventual deflection by affecting the duration of the DEP force Figure 4C shows the eventual deflection of the PS beads at different sample flow rates With sample flow rates of µL min−1 and µL min−1 , there is small difference for the eventual deflection (74.7 µm and 72.8 µm), and the sample is separated to O2 However, when the sample flow rate increases to µL min−1 , the eventual deflection This article is protected by copyright All rights reserved Accepted Article 2.4 Performance characterization with LMA electrode array Figure 5: Separation of PS beads and HeLa cells using the 50-pair-electrode device The superimposed micrograph (A) and probability density function (B) of the particle trajectories at the (i) 1st, (ii) 30th, and (iii) 50th (outlet) pair of electrodes when DEP is off The superimposed micrograph (C) and probability density function (D) of the particle trajectories when DEP is on Scale bars: 50 µm (E) Distributions of HeLa cells and PS beads at inlets and outlets when the DEP signal is 40 Vpp and MHz Sample size is about 300 (F) Purity of HeLa cells and PS beads in the inlet and outlet Sample size is about 300 significantly declines to 55.8 µm, and thus only partial PS beads are separated to O2 Note that the sample flow rate and sheath flow rate keeps a constant ratio (1:2) in experiment To further demonstrate size-dependent DEP deflection, we compare DEP deflection for PS beads of 10 µm, µm, and µm under different voltages The results in Figure 4D show that the eventual deflection is consistently and positively related to the particle size It is this size-dependent deflection that makes it possible to use the device for size-based separations Similarly, the deflection at the outlet for different sample flow rates under different voltages is shown in Figure 4E At flow rates less than µL min−1 , the deflections all exhibit a saturation-like phenomenon, i.e., the deflection does not decrease as the flow rate increases This indicates that the device with 50 pairs of electrodes is able to provide sufficient DEP action time to push the sample to the equilibrium position, where the drag force equals to the DEP force As the sample flow rate increases further, the deflection starts to decrease with increasing flow rate, meaning that the device is no longer able to provide enough DEP action time This phenomenon inspires the tactic of integrating more electrodes into the chip to allow higher flow rate and thus higher throughput in particle separation This article is protected by copyright All rights reserved 2.5 Accepted Article 2.5 Validation of dielectric-property-based separation with LMA electrode array Validation of dielectric-property-based separation with LMA electrode array We applied the 50-pair-electrode device to validate the DEP separation capability of the LMA electrode array using heterogeneous samples of HeLa cells and PS beads They are mixed at a concentration ratio of 2:1 Here, apart from the particle size, we also deliberately selected a proper DEP signal frequency to leverage the different dielectric properties of these two particles for separation The real part of the Clausius-Mossotti factor (Re[KCM ]) of PS beads and HeLa cells is calculated over a frequency range of 104 Hz to 108 Hz, using the homogeneous particle model and spherical single-shell model, respectively (Figure S5) Based on the data of Re[KCM ], a frequency of MHz is chosen such that HeLa cells are expected to experience pDEP force and PS beads to experience nDEP force According to Figure 4E, 40 Vpp voltage and sample flow rate of µL min−1 are applied in the experiment for effective deflection and reasonably high throughput The superimposed micrograph in Figure 5A shows the particle trajectories of the mixed sample passing through the 1st, 30th, and the last (i.e., outlet) pair of electrodes when the voltage is off Correspondingly, the positions of HeLa cells and PS beads in the mixed sample are obtained and fitted as normal probability density functions, as shown in Figure 5B There is almost no difference in the positions of HeLa cells and PS beads in the mixed sample, which flow in parallel with the sheath flow and exit the device from both outlets O1 and O2 Similarly, the results of the particle trajectories when the voltage is on are shown in Figure 5C and D With DEP on, PS beads start to deflect towards the negative electrode region under nDEP, while the HeLa cells are pulled by pDEP to the positive electrode region Through ADD of the electrode array, HeLa cells and PS beads exhibit a relatively clear demarcation when the mixed sample pass through the 30th pair of electrodes (Figure 5C-ii) Eventually at the outlet, the majority of HeLa cells enter O1 and most PS beads enter O2 , as indicated by the clearer demarcation (Figure 5C-iii) The particle trajectories at the outlet under DEP off or on can be seen in Video S2 Both separation efficiency and purity of HeLa cells and PS beads are quantified to characterize the separation performance (See Experimental Section for details) The microscopic images of the heterogeneous sample of HeLa cells and PS beads before separation, cell-dominated collection from O1 , and bead-dominated collection from O2 are shown in Figure S7 The counting results (Figure 5E) show that ∼99.3% of PS beads are separated to O2 , while 89.3% of HeLa cells move out from O1 Meanwhile, the purity of HeLa cells collected from O2 is only about 14.5%, whereas it is as high as 99.2% at O1 The purity of PS beads at O1 and O2 is 0.8% and 85.5%, respectively Compared to the initial concentration ratio of 2:1 at the inlet, HeLa cells and PS beads are well separated at the outlets 2.6 Demonstration of high-throughput dielectric-property-based separation with large-array LMA electrodes We fabricated and tested a device with 5000 pairs of thick LMA electrodes to demonstrate the full potential in high-throughput particle separation In first experiment, 10 µm PS beads are used with both sample and sheath flow rates of 100 µL min−1 (sample throughput = 4.5 × 106 particles per with the PS bead concentration of 4.5 × 107 mL−1 ), and the DEP signal is 60 Vpp and 100 kHz The superimposed micrograph and probability density function of the particle trajectories passing through the 1st, 500th, and 5000th pairs of electrodes are shown in Figure 6A-D, for DEP off and on Initially, the PS beads are hydrodynamically focused by the sheath flow from the negative electrode side When DEP is off, as the particles sequentially flow through the 500th and 5000th electrode pairs, their trajectory gradually and slightly deflects towards the center of the channel (Figure 6AB), which can be attributed to inertial forces When DEP is on, due to the high sample flow rate, particles pass through the effective electric region rapidly, resulting in almost no DEP-induced deflection observed at the 1st pair of electrodes (Figure 6C-i) When particles pass through the 500th electrode pair, which is hardly manufacturable in conventional DEP devices, the deflection is still small (6.2 µm, Figure 6C-ii) and not enough to direct the sample to O2 This means that a DEP device with 500 pairs of electrodes is unable to achieve effective DEP deflection required for high-throughput particle separation (e.g., 100 µL min−1 ) This limit is broken at the 5000th pair of electrodes, where the deflection (Figure 6C-iii) reaches 30 µm This would be sufficient enough to 10 This article is protected by copyright All rights reserved Accepted Article 2.6 Demonstration of high-throughput dielectric-property-based separation with large-array LMA electrodes Figure 6: High-throughput DEP separation with the 5000-pair-electrode device The superimposed micrograph (A) and probability density function (B) of the particle trajectories of 10 µm PS beads under flow rate of 100 µL min−1 without DEP passing through the (i) 1st, (ii) 500th, and (iii) 5000th pair of electrodes The corresponding superimposed micrograph (C) and probability density function (D) of the particle trajectories under 60 Vpp and 100 kHz at the same pairs of electrodes Scale bars: 20 µm Cell streamlines at the outlet showing continuous-flow separation of A549 from PBMCs with (E) DEP off and (F) DEP on under 45 Vpp at 100 kHz, at a sample flow rate of 70 µL min−1 Scale bars: 50 µm The corresponding probability density function of the cell trajectories at the outlet with (G) DEP off and (H) DEP on Sample size is about 300 Purity of A549 cells and PBMCs in the inlet (I) and outlet (J) Sample size is about 300 push the particles from their presumable outlet O1 to O2 The trajectories of PS beads passing through the 1st, 500th, and 5000th pairs of electrodes with DEP off or on can be observed in Video S3 It is worth noting that the flow rate can be set greater than 100 µL min−1 used in this work We limited the flow rate simply because higher a value would induce a driving pressure the PDMS-based chip could not withstand and thus break With stronger bonding, we believe that large-array thick liquid metal electrode chips would allow higher flow rates, and thus achieve higher throughput We further pilot tested the 5000-electrode-pair device for separation of a mixed sample comprising human PBMCs and A549 cells (3.6 × 107 PBMCs and 2.4 × 107 A549 cells per mL) Here, the Re[KCM ] values are calculated for PBMCs and A549 cells, and a frequency of 100 kHz is selected such that PBMCs exhibit nDEP and A549 cells exhibit pDEP During the experiment, the sample flow rate is maintained at 11 This article is protected by copyright All rights reserved Accepted Article 70 µL min−1 , and the DEP signal is 45 Vpp at 100 kHz Figure 6E-H illustrates the cell streamlines and probability density distribution at the outlet, demonstrating the successful continuous-flow separation of A549 cells from PBMCs The cell streamlines at the outlet with DEP off or on can be seen in Video S4 The microscopic images of the mixed sample of PBMCs and A549 cells before separation, cell-enriched collection from O1 , and PBMCs-dominated collection from O2 are shown in Figure S8 The purity of A549 was increased from 37.12% to 87.40% (Figure 6I-J) We also performed the separation experiments of MCF-7 cells from both mouse red blood cells (RBCs) (Fig S9 and Video S5) and horse WBCs (Fig S11 and Video S6), and the separation experiments of A549 and HeLa cells (Fig S13 and Video S7), demonstrating its wide applicability.To probe cell viability, after DEP separation, the collected samples of the exepriment with MCF-7 cells and horse WBCs were enriched by centrifugation, resuspended and stained utilizing common viability fluorescent dyes Calcein-AM and Propidium iodide (Fig S14) The MCF-7 cells and WBCs going through this device had 91.83% and 83.33% viability respectively, while the control group had 97.96% and 88.46% viability respectively These confirm that the cell viability can be almost maintained In literature, cancer cells are often spiked into blood cells in a very low ratio (e.g., 1/104 -106 [54] and 1/103 [55]) and higher enrichment factors (∼70 and 162 respectively) are achieved Here, our goal is to focus on demonstrating the improvement in flow rate, so we did not spike A549 cells in such low ratio and the enrichment factor was not that high But the purity was higher, which could be beneficial in many downstream analyses such as gene sequencing In terms of flow rate, 18-25 µL min−1 was used in DEP-only separation [54], and 63 µL min−1 was used for inertial-concatenating-DEP separation [55] By comparison, our device achieved 70 µL min−1 flow rate in the application experiment with PBMCs and A549 cells, superior to both works But in the latter work, for particles going through the DEP region, the flow velocity was significantly decreased (×1/200) to be suitable for DEP separation Our flow velocity was estimated to be orders of magnitude higher, which is an advantage Thus, with this high flow velocity, if we expand the height of our device, which is the comparable factor in thick electrodes as in planar electrodes in DEP separation application, our flow rate could be largely increased Note, if we take the same strategy in fabricating and assembling as the work [54] (which used PMMA for the main device and gaskets/bolts to withstand strong driving pressure), we can further improve the flow rate Discussion The dimensions of the microfluidic chip were determined in the following way (Section S13, Supporting Information) The sample channel width is presumably to be selected as great as possible in terms of throughput, however, it has to be compromised by considering the equilibrium position of DEP force counteracting on the drag force The equilibrium position can be experimentally determined (e.g., Fig 3B) For the stopper width, switching valve width, and electrode channel width, they have to satisfy w3