3D-Printed Microchip Electrophoresis Device Containing Spiral Electrodes for Integrated Capacitively Coupled Contactless Conductivity Detection

Abstract

In this work we demonstrate for the first time the design and fabrication of microchip electrophoresis devices containing cross-shaped channels and spiral electrodes around the separation channel for microchip electrophoresis and capacitively coupled contactless conductivity detection. The whole device was prepared in a digital light processing based 3D-printer in poly(ethylene glycol) diacrylate resin. Outstanding X-Y resolution of the customized 3D printer ensured the fabrication of 40 μm cross-section channels. The spiral channels were filled with melted gallium to form conductive electrodes around the separation channel. We demonstrate the applicability of the device on the separation of sodium, potassium, and lithium cations by microchip electrophoresis.

1. INTRODUCTION

Additive manufacturing (AM) has revolutionizing the way objects are produced [1, 2]. In fact, from sub-micron structures to very large objects can be prepared with a degree of complexity that is hard to achieve by using subtractive technologies. 3D printing uses the concept of AM and is becoming popular since many printers are now available at accessible prices. Different processes can be used in 3D printing, with the most common ones being Fusion Deposition Modelling (FDM), Inkjet (PolyJet or Multi-jet) 3D printing (i3DP), Stereolithography (SLA), Selective Laser Sintering (SLS) and Digital Light Processing (DLP) [1, 2]. These approaches differ basically in the way the objects are built layer-by-layer.

Although AM has been extensively used in different fields to produce relatively large objects, the use of 3D printing in the production of microfluidic devices has gained increased attention only in the last few years [3, 4]. In practice, most 3D printers offer resolution of hundreds of microns which allows the construction of fluidic systems on the millimeter scale with excellent repeatability. It is worthwhile to mention that layer-to-layer sealing steps typical for planar micromachining are not needed in 3D printing, and the fabrication can be done without cleanroom facilities, which allied to the low cost of the materials makes 3D printing very attractive for prototype development.

Efforts toward producing real microfluidic devices with channel features smaller than one hundred microns can be found [3, 5–8]. For applications such as microchip capillary electrophoresis (MCE), it is of utmost importance to keep the channel dimensions as small as possible, to avoid deleterious effects introduced by Joule heating. The Nordin and Woolley groups have pushed the limits of the DLP technology to produce microfluidic devices with channels as small as 18 × 20 μm in a custom DLP 3D printer [9–11]. They also demonstrated the production of high density of printed elements, such as integrated valves, in a single step production of microdevices [12, 13]. Conversely, it is still difficult to combine more than one material by using DLP technology, and more complete MCE devices could be produced by incorporating conductive materials to serve as electrodes for electric field application and for electrochemical detection. 3D printers that allow multi-material deposition still present high instrumental cost (i3DP) or suffer from insufficient resolution (FDM, for example).

In view of the advantages presented by the introduction of conductive parts in microdevices, such as resistive heaters, electrodes for detection and electric field application, and electrical connections, it is desirable to establish protocols for the fabrication of more complex microfluidic devices. Indeed, fluidic devices containing conductive parts can be created in a single step by multimaterial 3D printing. Despite this benefit, multimaterial FDM-based 3D printers cannot currently compete with the resolution of structures produced by a custom DLP-based 3D printer. Alternatively, a multimaterial DLP-based 3D printer uses several resin baths, requiring cleaning steps at each resin exchange, and each layer requires multiple exposures in each resin, making the process complex and inefficient [14]. To overcome this drawback of fabrication, some groups found creative and simple alternatives to assemble complex fluidic devices with integrated conductive parts. In 2017, Yu et al. presented a method to incorporate metal connections in FDM-based structures, aiming at the construction of 3D circuit boards [15]. In their approach, a eutectic alloy of gallium and indium was injected in void channels created during the printing to form conductive wires. Khoshmanesh et al. published a comprehensive review on the use of liquid metal in microfluidics [16]. In the same direction, Khondoker and Sameoto reviewed the applications of gallium alloys and described different techniques for structurally depositing these conductive materials [17]. In work not involving 3D printing, Gaudry et al. [18] and Thredgold et al. [19] reported electrophoretic microchips containing in-plane electrodes of molten metal alloy and liquid gallium, respectively. Both groups demonstrated the electrode performance using a capacitively coupled contactless conductivity system to detect inorganic cations.

Capacitively coupled contactless conductivity detection (C⁴D) [20, 21] is an interesting electrochemical detection mode due to its low cost, easy miniaturization and intrinsic decoupling of the separation electric field, which makes it suitable for MCE applications [22, 23]. C⁴D is a well-established detection technique and has been successfully applied to capillary electrophoresis and MCE, and many applications can be found [24]. Many configurations of the detection cell geometry have been proposed in MCE, which usually involve the deposition of thin metal films on the top or bottom of the separation channel, although some authors proposed the insertion of electrodes on the side walls [24]. The sensitivity of the detector and frequency response are directly related the cell geometry, and better results have been obtained in capillary electrophoresis by using tubular electrodes, whose positioning is facilitated by the cylindrical form of capillaries [20, 21]. Conversely, the conventional protocols for microfabrication of MCE devices make it difficult to construct detection cells that completely surround the separation channels. In this sense, 3D printing offers an effective way to construct C⁴D detection cells directly integrated in the microchip in a single fabrication procedure.

In this work, we demonstrate a novel approach to integrate electrodes into MCE devices for C⁴D detection. We develop a simple way to assemble a fully microfluidic device using liquid metals and DLP 3D printing technology. The suitability of the microfluidic device is demonstrated through a separation of alkaline cations in microchip electrophoresis. To the best of our knowledge, this is the first report on C⁴D detection in a DLP-based 3D-printed microfluidic device.

2. MATERIALS AND METHODS

2.1. Reagents and materials

All reagents used were of analytical grade and were used as received. The solutions were prepared with deionized water of 18.2 MΩ·cm. We used 2-(N-morpholino)ethanesulfonic acid (MES) and histidine (Sigma-Aldrich) to prepare the background electrolyte (BGE) buffer solutions. Potassium chloride, sodium chloride and lithium chloride from Sigma-Aldrich were utilized to prepare stock standard solutions. For 3D printing, poly(ethylene glycol) diacrylate (PEGDA, MW 250) and phenylbis (2,4,6-trimethylbenzoyl) phosphine oxide (Irgacure 819) was acquired from Sigma-Aldrich and 2-nitrophenyl phenyl sulfide (NPS) was purchased from TCI Chemicals (Portland, OR). Metallic gallium was purchased from Sigma-Aldrich.

2.2. Preparation of Microchips

We designed microfluidic devices in OpenSCAD (openscad.org), a free computer aided design software program (see Supplementary Information Fig. SI1). The standard cross channel MCE layout, with channels 45 μm wide and 50 μm tall, was adapted from one we published earlier [25]. Briefly, the simple t-device had an injection channel of 5.3 mm length and a separation channel 13 mm long, with an effective separation distance of 9.5 mm (Fig. 1a). We added to this design two independent spiral channels for C⁴D, which were wrapped around the separation channel near the detection point (Fig. 1b). Each coil channel radius was 100 μm, with a spiral amplitude of 300 μm and pitch of 300 μm, and an 11 mm total length of the gallium wire after filling. The coil channel cross section was larger than for the separation channels to facilitate gallium filling and reduce the likelihood of air encapsulation which would create a break in the wire. The amplitude and pitch parameters were set for coil channel proximity to the separation channel and limiting the overall length of the channel addressed by the coil, while also maintaining bulk 3D printed material integrity. The gap between the two coils was 750 μm, selected to address the tradeoff between increased sensitivity but decreased resolution as detector length is increased.

Fig. 1.

(a) Rendered CAD view of the microchip. (B1) Detail of the serpentine flow channel used to inject gallium. (B2) Inlet and outlet reservoirs connected to the serpentine channel for metal injection. (B3) Buffer waste reservoir. (B4) Sample reservoir. (B5) Buffer reservoirs. (b) 3D serpentine flow channels surrounding the separation channel in a DLP-fabricated microfluidic device (A1) without melted gallium and (A2) filled with melted gallium to form spiral electrodes for C⁴D. (c) Printed device showing the connections with the spiral electrodes and dimensions of the microchip.

Devices were created using a custom-built 3D printer we have described previously [12, 26]. In brief, this DLP stereolithography 3D printer uses 385 nm light to sequentially polymerize 10 μm thick patterned resin layers. As in prior work [12, 25, 26], the resin consisted of 97% PEGDA, 2% NPS UV absorber, and 1% Irgacure 819 photoinitiator. Once a device was polymerized, the channels were cleared of excess resin under negative pressure and next rinsed with isopropyl alcohol. The microfluidic device was then exposed to a UV postcure for 20 min to polymerize any excess resin [11]. After microchip preparation, gallium was melted by heating at about 35 °C and injected in the spiral channels by filling one channel end with gallium while negative pressure was applied to the other channel end using a plastic syringe to form metallic electrodes around the separation channel (Fig. 1c). Copper wires were used for electrical contact to gallium electrodes and the detection circuitry. To keep the electrical connections in place, it is advisable that the system temperature during operation does not exceed 25 °C.

2.3. Instrumentation and separation

Electropherograms were obtained using a custom built C⁴D system described previously [21]. Platinum electrodes (1 mm o.d.) connected with copper wires were placed into the reservoirs to apply the voltage for electrokinetic sample injection and MCE separation. The MCE device and electrical cables were arranged in an aluminum case to act as a Faraday cage, avoiding external electromagnetic interferences (see Supplementary Information Fig. SI2). One spiral electrode was connected to the signal generator whilst the other spiral electrode was connected to the input of the C⁴D circuitry (see Supplementary Information Fig. SI3) by copper wires. This detector was used and operated with a sinusoidal excitation signal with frequency of 250 kHz and 3.5 V_p (Peak voltage). Amounts (10 μL) of standard or buffer solutions were added to sample reservoir, sample waste, buffer reservoir, and buffer waste using a micropipette. The analytes present in the sample plug were then separated according to their mobilities and migrated toward the sensing electrodes under electrokinetic transport. The voltages for introduction and MCE separation were applied by a four-channel high voltage power supply ER430 (eDaq, Sydney, Australia). Data acquisition and electrokinetic control were performed using Labview and QuadSequencer (eDAQ), respectively.

3. RESULTS AND DISCUSSION

Achieving complex layouts, in this case spiral channels around a central separation channel, would be very complicated using conventional prototyping of microfluidic devices, such as lithography and soft lithography owing to its subtractive characteristic. Also, most AM techniques lack the desired resolution needed to produce truly microfluidic devices. Thus, SLA and DLP are the most promising approach to use in this work to create complex microfluidic structures due to its high resolution. First, we tried to reproduce a classical tubular design of electrodes to perform C⁴D detection. However, for this electrode design, it was laborious to fill the channels with liquids in a homogeneous way due to trapped air. In this context, we created a spiral design, for which the performance can be comparable to a tubular electrode since it surrounds the channel as well. Furthermore, the electrode channels can be easily filled with liquid metals without compromising the electrode connection or even its integration.

3.1. Detector characterization and parameter optimization

In C⁴D, the excitation voltage and operating frequency are parameters that directly affect the analytical response. Thus, both parameters were optimized prior to maximizing the detector sensitivity. The C⁴D signal was investigated using frequency values ranging from 84 to 628 kHz keeping the excitation voltages constants at 3.0, 5.0 and 7.0 V_p values (see Supplementary Information Fig. SI4) while the separation channel was filled with 5 mmol L⁻¹ MES/histidine BGE.

According to our previous work [21] on C⁴D development, the best frequency of operation depends mostly on the geometric parameters of the detection cell, although the conductivity of the BGE also exerts some influence. Moreover, the measurement of the baseline output voltage directly correlates with the peak area, being a more convenient method for frequency optimization [21]. As can see in Fig. SI4, by increasing the frequency of operation, the output voltage increases up to maximum value and from this point the output voltages decreases. This is due to the capacitances of the detection cell: at low frequencies, the wall channel capacitance dominates and blocks most of the excitation signal, while at high frequencies stray capacitances (including the input capacitance of the transimpedance amplifier) offer alternative paths for the signal. Usually, the frequency is set at this maximum for the best response. Moreover, this maximum is shifted to lower frequencies by increasing the amplitude of the signal, as can be seen in Fig. 2. The maximum values measured at 3.0, 5.0, and 7.0 V_p were obtained for the MCE-C⁴D were 293, 238, and 207 kHz, respectively. In summary, for a given amplitude and BGE composition the frequency is scanned and set at the maximum for further experiments.

Fig. 2.

Frequency response for the C4D-MCE device at different amplitudes of the excitation signal for three microdevices. (a) 3.0 Vp, (b) 5.0 Vp, and (c) 7.0 Vp. Black circles (●) and red triangles (▲) represent microdevices prepared in 2019 and blue squares (■) represent a device made in 2021. Separation channel was filled with a buffer solution containing 5 mmol L-1 of 2-(N-morpholino)ethanesulfonic acid and histidine. Error bars are too small for visualization. Lines do not represent the curve fitting of data. Instead, they were added using the B-Spline option on Origin 8.1 software.

We also compared the response in frequency for different MCE-C⁴D devices as shown in Fig. 2. Two microchips prepared in the same batch in 2019 presented very similar behavior. For example, at 5.0 V_p (Fig. 2b) the maximum responses occur at frequencies of 275 and 288 kHz, which represents a percent variation of 4.6 %. The variation was more pronounced when one compares different batches and we found 25.9 % variation by comparing a device prepared in 2019 and 2021. One might note that this variation is acceptable since it involves different resins and completely different printers (the device printed in 2021 used a DLP printer with better resolution). Considering the frequency response curves, the profiles agreed for the two batches produced, with a slight shift for lower frequencies for the 2021’s batch. An anomalous profile was observed for the 2019’s batch for the 3.0 V_p excitation signal (Fig. 2a), where a shoulder appeared around 500 kHz. Although we do not have an explanation for this phenomenon, it did not compromise the proper function of the detector.

To study the analytical performance of the MCE-C⁴D as well as its use for monitoring ionic species, the operating frequency and excitation voltage values were then fixed at 250 kHz and 3.5 V_p, respectively. We next sought to evaluate the operating frequency and excitation voltage values for the determination of the best separation conditions for K⁺, Na⁺ and Li⁺ concerning buffer concentration and the voltage applied. A standard mixture of these inorganic cations was selected as a model analyte system since they are commonly used ions to provide information about C⁴D system behavior, due to lack of complexity of the system [27, 28]. Several different concentrations of MES/histidine buffer (2, 20, 50 mM) were evaluated; we had issues with background noise using 50 mM buffer and low resolution between Na⁺ and Li⁺ peaks using 2 mM buffer. Regarding the DLP-based microchips, the best results were achieved by using BGE composed of 20 mmol L⁻¹ MES/histidine (pH = 6.0) at 25 °C. Standard solutions were injected inside microchannels using a gated injection protocol (See Supplementary Information Fig. SI5). The separation voltage of 800 V, injection voltage of 500 V, and injection time of 3 s were selected for the separation of K⁺, Na⁺ and Li⁺ with adequate efficiency and peak resolution (Fig. 3).

Fig. 3.

Electropherograms obtained from the introduction of standard solutions containing K⁺, Na⁺ and Li⁺ (25–200 μmol L⁻¹). Respective calibration curves are also shown. Gated injection using +500 V and +800 V at sample and buffer reservoirs, respectively, and injection time of 3 s. C⁴D detector operating at 250 kHz and 3.5 V_p.

In addition, linearity of the method was evaluated using standard solutions of K⁺, Na⁺ and Li⁺ (25–200 μmol L⁻¹). Adequate linearity (r > 0.996) was achieved for all calibration plots (Fig. 3), which shows a suitable response of the detector to concentration of the analytes. This is better performance compared to other detectors fabricated by liquid metals and previously reported in the literature. For example, Thredgold et al. [19], reported a gallium electrode with an in-plane design for C⁴D. This electrode provided a linear correlation coefficient of 0.9729, 0.9794 and 0.9982 for K⁺, Na⁺ and Li⁺, respectively. Even though the electrodes were made of the same material, the electrode spiral geometry shown here offered a better linear response than in-plane electrode design. This improvement in the performance regarding the linear response is expected for electrodes which mimic a tubular electrode, as can be explained using the theory of the electrode effective width [29]. Briefly, the dependence of the detection cell impedance on the width of the electrodes goes through a minimum which determines the electrode effective width because the electric current flows through the detection cell along the trajectory characterized by the lowest impedance value. For tubular and semi-tubular electrodes, detection occurs in a cell whose effective dimensions (width) are smaller than the geometric ones. For planar electrodes, the effective width essentially equals their geometric width, and thus detection takes place inside a substantially longer cell, which may affect the detector response.

4. CONCLUSION

In this work, we have demonstrated DLP-based 3D-printing of microfluidic devices with 40 μm channels for MCE analysis of reactive nitrogen species in biological samples. For the first time, we presented a DLP-based 3D-printed microfluidic device with C⁴D detection. These microfluidic devices have potential for use in ion analysis. Further work will be done to evaluate these devices in anion separations aiming the determination of nitrogen reactive species and other applications.