Microfluidic and lab-on-a-chip devices are transformative tools for carrying out chemical and biological reactions across a wide range of disciplines, from drug discovery to biomarker monitoring (Nahavandi et al. 2014;Regnault et al. 2018). These applications are driven by the small sample volume requirements, high device portability, and inexpensive setup and reaction costs compared to conventional approaches. Three key current challenges in the microfluidics/lab-on-a-chip community are (1) generating fluid flow cheaply and easily in point-of-care environments, (2) the ability to transport non-aqueous solvents, and (3) the fast and complete mixing of reagents (Cai et al. 2017;Chin et al. 2012;Renckens et al. 2011).
The most common approach to generate fluid flow in microfluidic devices is through syringe pumps, but these are unsuitable for point-of-care flow generation due to the bulky size and requirement of electricity. Portable plug-and-play pumps or valves have been designed to overcome this challenge, though they are often application-specific (e.g., fixed flow rates) (Feeny et al. 2016;Park and Park, 2019;Zhang et al. 2020). Capillary action-based flow is also used with hydrophilic materials such as paper or glass, although the flow rates can be slow and integration of complex designs (e.g., mixing and sensing modules) can be challenging (Carrell et al. 2019;Iliescu et al. 2012).
Many microfluidic devices are made of materials (like polydimethylsiloxane and paper) that are incompatible with non-aqueous solvents. This limits the scope of microfluidic devices (Channon et al. 2016). Cyclic olefin copolymers and hybrid materials such as fluorinated polymers coated onto PDMS can accommodate non-aqueous solvents, but the preparation of these materials is often more complex and expensive (Illa et al. 2010;Renckens et al. 2011). Paperbased devices are ideally suited for point-of-care environments, however, the flow rates are generally slow and the inks used to define channels are not stable in non-aqueous solvents (Cardoso et al. 2017;Park et al. 2016).
Micromixers are critical components to conduct chemical or biological reactions and/or to sense analytes in microfluidic or lab-on-a-chip devices (Jeong et al. 2010). Micromixers can be divided into active and passive mixers. Active micromixers employ an external stimulus to mix fluids, including magnetic beads, temperature, pressure, acoustics, digital droplets, or an electric field (Bayareh et al. 2020;Lynn et al. 2008;Shanko et al. 2019;Srinivasan et al. 2004;Ward and Fan, 2015). Active mixers typically provide a high degree of mixing and are especially useful for systems with low Reynolds numbers (Re < 1) to overcome low flow rates and high viscosity, but these mixers are unsuitable for pointof-care devices due to the added cost and increased size that comes with the integration of an external power source to a microfluidic device (Capretto et al. 2011). Passive micromixers rely on shrewd device designs to enhance advective and diffusive mixing without external forces, for example, serpentines, splitting then recombination of streams, multiphase flow, and integration of obstacles, grooves, or ridges (Bayareh et al. 2020;Chen et al. 2018;Feng et al. 2019;Seong and Crooks 2002;Shanko et al. 2019;Wang et al. 2019a;Ward and Fan 2015). Compared to active mixers, passive micromixers typically require longer channel lengths and longer mixing times to achieve complete mixing. Furthermore, there are very few effective passive micromixers for capillary-based flow systems (Gervais et al. 2011;Glavan et al. 2013;Lillehoj et al. 2010;Taher et al. 2017).
Among passive mixers, herringbone mixers are particularly popular due to the efficient mixing over short channel lengths. Herringbone devices feature an array of grooves in the wall of a microfluidic channel (Stroock et al. 2002). Fluid enters these grooves and recirculates into the main channel, generating complex circulating flows that result in fluid mixing. The speed and extent of mixing are controlled by the length of the herringbone, the channel dimensions, and the flow rate (Williams et al. 2008). Patterning herringbones into both the top and bottom channel walls has been shown to improve the mixing efficiency, although this is difficult to implement depending on the fabrication procedure and device materials (Choudhary et al. 2011;Feng et al. 2019;Hossain et al. 2011). Herringbone micromixers have been employed for micromixing in a range of applications (Liu et al. 2014;Wang et al. 2016), but rarely in point-of-care environments.
In this work, we demonstrate for the first time an asymmetric two-faced staggered herringbone passive micromixer combined with a passive/capillary action flow device. Our design is based on a self-pumping device made from glass and tape which is simple, inexpensive, and fast to manufacture. This novel fabrication approach provides fast flow rates and rapid mixing that have not been achieved previously in other passive flow herringbone devices. To demonstrate the scope of this point-of-care device, we investigate the application of the microfluidic device with two kinetically fast chemical and biological reactions.
Microfluidic devices with herringbone grooves were fabricated via a simple cutting and stacking method. Three sets of herringbone grooves were rastered onto each glass slide (Glass D, Nexterion) with a laser cutter (Epilog Zing), operated at a resolution of 500 dpi, 100% laser power, and 25% scanning speed. Each of the 3 sets of herringbones consisted of 15 herringbone grooves with a width of ~ 285-300 µm, depth of ~ 120 µm, and inter-groove spacing of ~ 530 µm (i.e., valley to valley), which was achieved through 8 laser engraving scans (Fig. S1). The angle between the long and short grooves is 105.7°. The channel height, herringbone groove depth were optimized to provide the most efficient mixing (Hossain et al. 2011;Williams et al. 2008). After fabrication, the glass slides were cleaned by ultrasonication in acetone for 10 min, rinsed with DI water, and dried with nitrogen.
To join the glass slides and define the channel height and width, double-sided tape (3M Scotch) was carefully cut with a laser ablator and aligned on the glass slide to form a ~ 3 mm wide Y-shaped channel (~ 50 µm channel height) over the herringbone structures (Fig. S2). A template was used to ensure proper alignment. Another glass slide with herringbone structures was placed on the tape layer and firmly pressed to form the microfluidic device. An asymmetric design was employed where the herringbone grooves in the top face of the channel were oriented in an offset pattern to the bottom face to enhance the mixing (Fig. 1a) (Choudhary et al. 2011;Hossain et al. 2011). A coating of 3M instant adhesive (SF 100) was applied to the sidewall with the channel inlets to prevent the undesired spreading of liquid droplets upon contacting the device entrance. Prior to experiments, the device was cleaned by flushing with ethanol, then water, and dried thoroughly with nitrogen.
To demonstrate the rapid mixing of our self-pumping microfluidic devices, Liquid droplets containing reagents were placed on a piece of superomniphobic paper, which displays extreme repellence to liquids with a wide range of surface tension (Movafaghi et al. 2019;Vahabi et al. 2017Vahabi et al. , 2018;;Wang et al. 2019c). The droplets were spaced accordingly to line up with the entrance of the microfluidic device. The device was moved toward the droplets until contact with the glass initiated fluid flow into the channel. The experiments were recorded with a digital camera (Canon) and then analyzed using NIH ImageJ software as described in the ESI.
Superomniphobic paper was fabricated through liquid phase silanization of Whatman filter paper 8, as previously described (Channon et al. 2019;Wang et al. 2019b). The filter paper was exposed to oxygen plasma for 10 min and then immersed in a mixture of 20 mL hexane and 600 µL of heptadecafluoro-1,1,2,2-tetrahydrodecyltrichlorosilane (FDTS) for 3 days at room temperature. The paper was then washed with hexane and dried with nitrogen.
The colorimetric reaction of nickel with dimethylglyoxime (DMG) was used as a representative inorganic reaction (Fig. S3). A solution of 2.25 mM nickel (nickel (II) sulfate hexahydrate, Sigma Aldrich) was prepared in 1:1 mix of ethanol and pH 5 acetate buffer. A solution of 4.5 mM DMG (Fluka) was prepared in 50% v/v ethanol: DI water. 50 µL droplets of each reagent were mixed in the device.
The enzymatic turnover of H 2 O 2 by horseradish peroxidase was used as a representative biological reaction to evaluate enzyme kinetics using the herringbone device (Fig. S4). Ten solutions of H 2 O 2 (Sigma Aldrich) were prepared in DI water between 0.001 mM and 0.06 mM (0.005-0.2% v/v). A 0.018 M solution of o-dianisidine dihydrochloride (Sigma Aldrich) in DI water was used for all experiments. A 32.5 U mL -1 solution of horseradish peroxidase (HRP, Sigma Aldrich, peroxidase from horseradish Type VI, > 250 U mg -1 ) was prepared in DI water immediately prior to use. All solutions were prepared in DI water instead of buffer due to the higher o-dianisidine (OD) solubility in DI water. In the device, a 30 µL droplet of a 50:50 mixture of OD and HRP was mixed with a 30 µL droplet of H 2 O 2 . Videos were recorded for each reaction and analyzed in ImageJ (ESI).
Given the challenges associated with mixing in passive flow devices, we sought to determine if we could use herringbone-based mixing devices to achieve mixing in a capillary flow device. Our device design (Fig. 1a) was adapted that used in multilayered paper-based microfluidic devices without mixers as well as the literature on staggered oneand two-faced herringbone microfluidic mixers (Channon et al. 2019;Choudhary et al. 2011;Lin et al. 2013;Stroock et al. 2002). The device itself is fabricated using a laser cutter to etch the herringbone grooves into glass slides and sealing with double-sided adhesive to form the microfluidic channel. The device design, dimensions, and characterization by optical profilometry are provided in the ESI (Fig. S1 andS2). The number and depth of the herringbone sets, as well as the height of the tape, were previously optimized and chosen for both efficient mixing and consistency of the device (Hossain et al. 2011;Williams et al. 2008). The staggered herringbone grooves in the top and bottom faces of the channel were offset yielding an asymmetric pattern. This provides enhanced mixing over herringbones in a single face or symmetric herringbones in two faces, and likely results in four circular flows across the channel cross-section as previously described (Choudhary et al. 2011). To establish fluid flow, reactants were placed as droplets on superomniphobic paper (contact angle ≈ 165° for ~ 10 µL, Fig. 1c), which displays extreme repellence to liquids with a wide range of surface tension (Movafaghi et al. 2019;Vahabi et al. 2017Vahabi et al. , 2018;;Wang et al. 2019c). Moving the device toward the droplets until contact with the glass initiates fast fluid flow (0.14 mL min -1 , 1 cm s -1 ) of the reactants into the channel due to capillarity.
To illustrate the enhanced mixing in the herringbone device, aqueous droplets of blue and yellow food dyes were mixed forming a green color. The pixel count for each color was measured in ImageJ before and after the herringbones (Fig. S3). Three device configurations were evaluated: a control device with no herringbone grooves, a device with one slide with ablated herringbone grooves, and one with two slides with herringbone grooves in an asymmetric pattern (Fig. 2, ESI videos 'BY dye control', 'BY dye 1 side herringbone', and 'BY dye 2 sides herringbone'). In the absence of herringbone grooves, mixing only occurred at the interface of the two colors (20% green pixels). The Reynolds (Re) and Peclet (Pe) numbers for our system are 0.98 and 490, respectively. This suggests flow in the devices is laminar, and mass transport is likely mixed diffusive and convective. In fact, in the control device without herringbones, an 8 m long channel would be required to fully mix the two colors (Stroock et al. 2002). Partial mixing was achieved when one channel wall contained herringbone grooves (52% green pixels), but for complete mixing in the length of the herringbone grooves (3.3 cm), two faces with herringbone grooves were required (100% green pixels). This result is in line with previous studies on improved mixing efficiencies with multiple walls of herringbone grooves compared to a single set (Choudhary et al. 2011;Hossain et al. 2011). As the three sets of herringbones deliver complete mixing, we anticipate most solvents and solvent mixtures would also mix satisfactorily through the current design. The fabrication method is designed as such to be able to tune the device to the application (i.e., choice of tape to define the channel height or number of herringbones). The speed and degree of mixing can be tuned with these parameters depending on the miscibility and viscosity of the reagents and the application.
Manufacturing microfluidic devices with herringbone grooves in the top and bottom channel walls is challenging using traditional micromachining methods. Typically, microfluidic devices are made with a channel in one material, like PDMS, which is bonded to a base substrate, like a glass coverslip, using plasma treatment. It can be challenging to make herringbone grooves in PDMS since these devices typically require photolithography as a fabrication process (Stroock et al. 2002;Williams et al. 2008). Our proposed device is made of two glass slides bonded with double-sided adhesive. The herringbone grooves are etched into each slide with a laser cutter, and the channel height and width are controlled by the placement and height of double-sided adhesive. Our device features a channel width of 3 mm and height of 50 µm, representing a hydraulic diameter of 98 µm. The use of wide channels with small heights is fairly common in microfluidic devices, in channel flow cells with electrochemical sensing (Channon et al. 2015). In addition, the flow and mixing of small sample volumes (ca. 40 µL) fit the definition of a microfluidic device. Given the benefits of the fabrication method (fast, easy, inexpensive), we envision these glass/tape herringbone devices as disposable mixers for point-of-care applications.
As proof of concept to demonstrate the device scope, two representative chemical and biological application reactions were chosen. First, the detection of Ni(II) in the environment, which is important due to the environmental abundance of Ni(II) and potential health issues such as reduced lung function and cancer (Denkhaus and Salnikow, 2002). A common colorimetric assay for Ni(II) detection is the reaction with dimethylglyoxime (DMG) in ethanol:H 2 O (Fig. S4) (Cate et al. 2015;Mentele et al. 2012). This reaction produces a reddish-pink precipitate which is insoluble in water, as shown in Fig. 3b (ESI video 'Ni-DMG herringbone'). Note, in the absence of the herringbones, no precipitate is observed (Fig. 3a, ESI video 'Ni-DMG control'). Since the DMG is dissolved in ethanol, the Ni-DMG reaction also demonstrates that the device applicability for reactions in non-aqueous solvents, which are commonplace in the industry. Many microfluidic materials (polydimethylsiloxane, PDMS, paper) are limited by incompatibility with non-aqueous solvents (Channon et al. 2016), but with the herringbone device being made out of glass, it is compatible with organic solvents as there were no visible adverse effects like degradation or delamination.
The mixing device can also be used for biological assays such as the enzymatic reaction of hydrogen peroxide (H 2 O 2 ) and horseradish peroxidase (HRP), using o-dianisidine (OD) as a colorimetric indicator (Fig. S5). The reaction of HRP with H 2 O 2 is important for many bioassays where HRP is a label (Hempen and Karst 2006). H 2 O 2 is turned over by the HRP, oxidizing the OD which turns from colorless to dark brown (Fig. 4a, ESI video 'HRP herringbone'). Without the herringbone grooves, mixing only occurs at the interface of the two liquids due to diffusion (Fig. 4B, ESI video The reagents are fully mixed in B to produce brown o-dianisidine. c Lineweaver-Burk plot of the enzymatic turnover of hydrogen peroxide 'HRP control'). To demonstrate that the herringbone device gives similar results to traditional microfluidic devices and bulk solution reactions, Michaelis-Menten kinetics were determined for this reaction using the micromixing device (details in the ESI, Fig. S6,S7). In short, videos with various concentrations of H 2 O 2 were analyzed and the brown color intensity was measured in ImageJ over the course of each video. This yielded the reaction rate, and its inverse was plotted against the inverse of H 2 O 2 concentration, generating a Lineweaver-Burk plot (Fig. 4c) (Boehle et al. 2018). Values for K m and k cat were determined to be 0.119 mM and 7.61 s -1 , respectively, and are comparable to those found in the literature (Ugarova et al. 1981). The similarity in measured values indicates the viability of this simple mixing system for carrying out kinetic reactions. In addition, the herringbone device could easily be used for other biological applications, for example, investigating the binding kinetics between fluorescently tagged antibodies and antigens without needing surface plasmon resonance measurements (Lago et al. 2018).
Previous herringbone devices have been fabricated using standard photolithography methods and a SU-8 photoresist, forming the channel when PDMS is plasma-bonded to a glass slip (Stroock et al. 2002;Williams et al. 2008). This method is time-intensive and expensive in comparison to the described approach of laser ablating the glass slides and sealing with tape. Each device costs ~ $0.40 and takes ~ 7 min to make by hand. Since the design is printed from a CAD file, changing the design to prototype new devices is fast and simple in comparison to making a new photoresist mold. Glass also has high solvent stability compared to other common materials used to make microfluidic devices. For example, PDMS swells in nonpolar solvents such as hydrocarbons, toluene, and dichloromethane (Lee et al. 2003), resulting in deformed channels and altered dimensions, which changes the fluid velocity and mixing capacity as well as affecting any sensing elements. Typically, paper-based devices use wax to form hydrophobic channels on the paper (Altundemir et al. 2017). Wax is not compatible with organic solvents, leaving the device without any barriers to direct fluid flow. Conversely, glass is compatible with a wider range of solvents (Lee et al. 2003), allowing this herringbone device to be used for a variety of reactions without compromising the integrity of the device.
A new herringbone microfluidic mixer is presented that can achieve complete mixing of small volumes over short times and distances. Crucially, these devices employ passive pumping (capillary action) and are capable of transporting both aqueous and organic solvents which are typically challenging to use in self-pumping devices. Laser cutting the asymmetric staggered herringbone grooves into glass slides as the channel walls delivers a fast, simple, inexpensive, and easily modifiable fabrication that has not been previously demonstrated as well as achieving fast flow rates and rapid mixing. The flow rate can easily be changed by altering the channel height through the choice of doublesided tape thickness or the number of tape layers (commercially available tapes range from 50 to 230 µm thicknesses) (Channon et al. 2019). The degree of mixing can be simply altered through the number of herringbone grooves, or the depth of the grooves (through the number of lasering rasters) (Choudhary et al. 2011;Hossain et al. 2011;Stroock and McGraw 2004;Williams et al. 2008). The device can be used for a range of biological and environmental applications, including inorganic, organic, and enzymatic reactions. Furthermore, relevant data like enzyme kinetics can be easily deduced from the device through colorimetric image analysis and our devices are amenable to other microfluidic sensing formats e.g., fluorescence and electrochemical detection. This device has great potential as a disposable mixer for point-of-care applications. Future work will seek to expand the application of these devices to a wider range of reactions in non-aqueous solvents, as well as to conduct finite element modeling of the device to understand the mixing in a capillary flow-driven asymmetric staggered herringbone mixer and optimize the device design (Choudhary et al. 2011;Hama et al. 2018;Hossain et al. 2010).