Ring Electrode Geometry for Microfluidic Electrochemistry

Abstract

We developed a ring electrode sensor for down-stream electrochemical sensing of microfluidic ELISA assay. The sensor is designed to easily integrate into the flow environment. Noble metal inks on Low Temperature Co-fired Ceramic provide straight forward fabrication at the mesoscale yielding robust sensors with very low sensing volumes. Two different sensor geometries were modeled. The best design was fabricated and tested in both static and flowing solutions. The sensor exhibits both macroelectrode and microelectrode behavior depending on fluid flow rate. Sensors of this type may be ideal for applications where electrochemical detection is desired in a flowing solution since the ring electrode has low propensity for bubble trapping.

1.1. Introduction

Electrochemical sensing of chemical/biochemical analytes is desirable in Point-of-Injury diagnostics. Sensors can be inexpensive (disposable), less complex, and hence more field compatible than optical sensors. These benefits coupled with Lab on Chip developments permit the transition of biomarker detection by ELISA from bench top instruments to point-of-injury applications, the end goal for this research.[1, 2] Microfluidics is a foundational technology to device miniaturization, as exemplified by its ubiquitous use in μTAS and hand-held/lateral flow diagnostic devices. The move to a microfluidic flow format permits lower volumes of both reagents and sample. Because of the higher surface to volume ratio conferred by channel dimensions in microfluidics chemical and biochemical reactions can be performed more efficiently—i.e., faster reactions with lower volume of reagents. This scale-dependent (microfluidic and nanofluidic) efficiency gain has not only been achieved in many biochemical assays, but also realized in electrochemical sensors. [3-7] However electrochemical sensing in a flowing solution is challenging. Air bubbles are a persistent source of error in microfluidic research. Air bubbles trapped on the electrode surface effectively reduce the surface area of the sensing electrode, causing reduced signal. In addition, bubbles may move along the electrode surface under the influence of fluid flow, causing increased signal variability. Many existing microelectrode designs employ a flat electrode with a gap/cavity above through which the analyte flows. We have noticed that this geometry tends to trap air bubbles on the electrode surface, most commonly in the corners of the flow cell. To solve this bubble trapping problem many try to make the surfaces hydrophilic, but we set out to make electrodes with a geometry that decreased the likelihood of bubble trapping. Our premise was that cylindrical ring electrode design can overcome this bubble trapping issue.

Tubular electrodes have been studied as early as the 1960's and 70's. Initially, the tubular electrodes were formed by sealing glass tubes around the ends of a platinum tube that had been cut to a specific length.[8, 9] In several experiments, the tubular platinum electrode was coated with mercury.[10, 11] Another assembly method was to drill a hole through a piece of platinum or gold, then clamp the electrode in an insulating housing containing a reference electrode.[12-15] Reference electrodes accessed the analyte typically through frit membranes via holes drilled in the nonconductive housing. More recently annular electrodes have been used for capillary electrophoresis with electrochemical detection (CEEC). Electrode construction for these applications ranges from sputter deposition to drilling holes through LaF3 crystals.[16-18]

Despite difficulty of fabrication, annular geometry permits analytical solutions of the steady state current in flowing solution where the current is given by:

$$i=2.01nF\pi C_b{D}^{\frac{2}{3}}{R}^{\frac{2}{3}}{X}^{\frac{2}{3}}{V}_0^{\frac{1}{3}}$$ Eq 1

where n is the number of electrons per molecule of the electroactive species, F is the Faraday constant, Cb is the concentration of the bulk solution, D is the diffusion coefficient for the electroactive species, R is the internal radius of the electrode, X is the length of the electrode and V₀ is the axial flow rate of the solution.[19] This current is derived by solving the 2-D steady-state diffusion equation. To arrive at the above equation several assumptions were made such as: a linear approximation of the Poiseuille velocity profile, a linear diffusion process in the radial direction, and axial diffusion is neglected. More simply, the current is composed of two parts; a current at zero flow rate, and a current dependent on flow rate. The total current takes the form:

$$i_T=i_{ind}+{\mathit{kv}}_f^{{}^1∕_3}$$ Eq 2

Where i_ind is the current independent of flow rate, and ${\mathit{kv}}_f^{{}^1∕_3}$ is a restatement of the flow rate dependent current from Eq. 1 where all non-flow rate variables are represented by the constant k.[14]

Since the initial studies with annular electrodes, great progress has been made in fabrication of microstructures. For the annular ring geometry, we believed the best fabrication method would be the use of Low Temperature Co-fire Ceramic (LTCC) methods. LTCC fabrication can be classified as a meso-scale fabrication technique where critical features can be on the order of 100 μm while the entire structure can be several cm² without great cost.

Fabrication of electrochemical sensors in Low Temperature Co-fired Ceramic (LTCC) has been reported earlier. In some cases the electrodes served the dual purpose of detection as well as a site for antibody immobilization.[20] The electrodes were oriented lengthwise along the sides of a rectangular channel. LTCC and screen printed gold electrodes on the sides of rectangular channels have also been demonstrated for magnetohydrodynamic studies.[21] Another design permitted diffusion through a membrane for enzymatic gas sensing.[5] The goal of this study was to model, design, fabricate, and test an annular electrode for adaptation in microfluidics.

2. Material and Methods

2.1. Sensor Design

LTCC fabrication is carried out by screen printing conductive inks onto a silica and alumina “green” sheet. These sheets are then laminated together under high pressure. The assembled product is then heated to a temperature of 200-500 C, during which the organic binding agents used in the ink and green tape are burned off. After this baking step, the LTCC assembly is heated to a peak temperature of 850-900 C during which the metal ink is fused into a conductor and the silica is sintered leaving behind a hard, coherent structure.

An electrode diameter of 500 μm (20mil) was chosen for study. To determine which specific geometry to pursue two annular geometries were examined by simulation. To reduce computation time 2-dimensional simulations were performed. Cross sections of these are shown in Figure 1A and 1B. The first geometry (Figure 1A) is a stepped design. The benefit of this design is the relaxed alignment of the central lumen and a relatively larger electrode area than in design 2 (Figure 1B). Design 2 had more stringent requirements on alignment but presented the lowest opportunity for bubble trapping. In addition, design 2 required gold ink to evenly coat the inner sides of the flow path. Design 1 only required the top surface of the green sheets to be coated. These two different annular geometries were examined with COMSOL computer modeling to determine which would be the most sensitive. Both designs were given the same initial concentration of species, 0 M. Flow was simulated as pressure driven with an inlet pressure of 30 Pa and outlet pressure of 0 Pa. A no slip boundary condition was used for the walls, and the physics was transport of dilute species and laminar flow. The concentration of incoming species was 1 M and the species concentration after a simulated 0.6 seconds was noted.

Figure 1A:

Results of COMSOL simulation of plug flow characteristics flowing through a stepped-wall designed electrochemical sensor.

Figure 1B:

Results of COMSOL simulation on plug flow characteristics through a smooth walled electrochemical sensor.

The stepped wall design shown in Figure 1A has three different electrode areas. The dimensions of the electrodes were chosen based on dimensions that could readily be achieved and aligned. Electrode area of the first electrode is simply the area of a disk with radius of 15 mil (381 μm) minus the area of a 10 mil (254 μm) radius disk.

$$A_i=\pi (381\mu m{)}^2-\pi (254\mu m{)}^2=0.253m{m}^2$$ Eq 3

$$A_{ii}=\pi (508\mu m{)}^2-\pi (381\mu m{)}^2=0.355\mathrm{m}{\mathrm{m}}^2$$ Eq 4

$$A_{iii}=\pi (635\mu \mathrm{m}{)}^2-\pi (508\mu m{)}^2=0.4563m{m}^2$$ Eq 5

This step design places constraints on the assignment of electrodes in such a device. The Counter electrode area should be equal to or larger than the working electrode area. Placing the reference electrode upstream of the working and counter electrodes prevents potential shifts due to the changing chemistry of the solution that occurs after the working and counter electrodes. The calculated volume in design 1 was 547 nL. The concentration of species after a simulated time of 0.6 seconds is shown in Figure 1A, with dark blue equivalent to 0 M concentration and red equal to 1 M. After the allotted time, the incoming species do not fully cover the reference electrode (right most); much less the working (center), and counter (left most).

The other geometry studied is shown in Figure 1B. This smooth walled design has three electrodes with identical sizes given by:

$$A_{\mathit{band}}=(2\pi rh)=(2\pi \ast 0.254mm)\ast 0.2mm=0.319m{m}^2$$ Eq 6

Having all three electrodes with the same area is less than ideal, but such a sensor design can be made to work. Additionally, it is easy to add additional layers to increase the electrode areas. As modeled this geometry had a volume of 162 nL or about 30% of the volume of design 1. Figure 1B shows the concentration profile after a simulated 0.6 seconds. It is evident that the concentration profile is better developed with high concentration solution touching all three of the electrodes.

Based on these computer simulations the stepped side wall design was not pursued further due to the persistent occurrence of low concentration stagnation zones around the electrodes and the increased volume to fill the detector. Even when the flow progressed for a simulated time of 2 seconds to account for the larger volume, the middle (working) electrode was only in contact with 0.5M solution. (not shown) Only more prolonged flow of the analyte would produce an electrochemical signal indicative of analyte concentration. In the smooth side wall design, only the flow boundary layer separates the electrodes from the bulk concentration. It was evident that the smooth walled design was favored for quick response with lowest volume of analyte despite the increased difficulty of alignment and electrode metallization.

2.2. Reagents and Materials

Fabrication occurred at the University of Arkansas High Density Electronics Center (HiDEC), (Fayetteville, AR). LTCC green tape was from DuPont (951PX).[22] Inks were likewise from DuPont, silver via fill (6141), silver screen printing (6142D), and gold screen printing ink (TC502). Ferro CN30-080M gold ink was used where noted in the Results section. Equipment consisted of a Blue M DCC-256C oven for preconditioning the green tape, a PTC APS-8718 Automated Puncher, AMI MSP-485 Screen printer, PTC IL-4008 Isostatic Laminator, and Fisher 750 Draft Furnace for final firing.

Potassium Hydroxide (P25-500) and Potassium Chloride (P217-500) were from Fisher Scientific, Hydrogen peroxide was from VWR (VW3690-5), KN0₃ was from Alpha Aesar (A14527), Ruthenium Hexamine (RuHex) was from Strem Chemicals (44-0620). Distilled deionized water was from Macron Fine Chemicals (H453-09). Potassium Ferricyanide (244023-5G) and Potassium Ferrocyanide (227684-5G) were from Sigma Aldrich. Electrochemical data was recorded with a Gamry Reference 600 potentiostat well as CHI 6204C and CHI1030 Flow was generated by a Harvard Apparatus Pump 11 Elite syringe pump for high flow rate measurements. Low flow rate measurements (<100 nL/min) were performed with an SFC Fluidics ePump Model 190, an electrochemiosmotic pump that offers precise, pulse-free flow at extremely low flow rates, compact size, and low power requirements, ideal for field use.[23, 24]

2.3. Fabrication

Each green sheet was punched with a 580 μm diameter via that served as the lumen through which fluid will later pass. In addition, 12 mil diameter electrical conduction vias were punched above the contact leads from lower levels to the top of the structure. A schematic of the side view of the structure is shown in Figure 2. A top-down view of the screen print pattern for the three electrodes is shown in Figure 3. Each layer has connection pads for the layers beneath it. The bottom electrode in the lamination was screen printed in the pattern of Figure 3A. The black area defines where the ink was printed, with the hole in the black disk defining the location of the sensing lumen for reference only. Screen printed ink covers the entirety of the black area and flows into the lumen. Atop this layer an LTCC sheet was placed with the main lumen as well as small electrical connection vias were punched and filled. The second electrode in the lamination was printed with the design in Figure 3B. In Figure 3B the left connection pad connects to the bottom electrode through the electrical connection vias punched and filled in the interposing insulation layer. The process was continued for the final electrode whose screen print pattern is shown in Figure 3C.

Figure 2:

Side view sketch of LTCC assembly. The electrode assembly is comprised of seven layers of LTCC green tape.

Figure 3:

Mask design for screen printing of LTCC green sheet layers. Via through hole connectors are not shown. The hole at the center of the electrode represents where the LTCC layer is punched through.

At the outset, it was unknown if the screen printed ink would evenly coat the inside of the lumen, so test structures were fabricated utilizing lower cost silver ink (DuPont 6142D). The layers were conditioned at 100°C for 20 minutes, punched, and screen printed. Vertical metallization was accomplished by use of vacuum stone to pull ink through the large lumen. After printing the LTCC sheets were placed on filter paper and then placed on a vacuum stone normally used to hold the LTCC tape during screen printing. The vacuum action prevented ink from clogging the large lumen. After screen printing the sheets were dried for 10 minutes at 70°C. Lamination was performed by stacking the LTCC sheets in alternating direction, then proceeded to use a PTC IL-4008 isostatic laminator with a pressure of 3,000 p.s.i., for 10 minutes at 70°C. Finally the laminated structures were fired according to protocols published by DuPont.[25] A cross sectional microscopy image of a test structure fabricated with silver ink is shown in Figure 4. The ink flows into the lumen and forms a well-defined annular electrode with the thickness of the LTCC sheet defining the length of the annular electrode. Tests performed with the silver ink chips confirmed that the design yielding the best results in simulation could be realized in practice. All electrochemical tests were performed with electrodes made of gold ink. Vertical metallization of gold ink was performed in a similar manner, and all fabrication parameters were the same.

Figure 4:

Cut away image of central lumen in fabricated electrochemical sensor. Each electrode is a separate ring sharing a common central axis. The width of each electrode is one LTCC sheet thickness. There is an insulating LTCC sheet between each electrode.

DuPont literature for the 951 LTCC green tape specifies a shrinkage rate of 12.7% in the x-y plane after lamination and firing. To account for this the central via was designed and punched with a diameter of 580 μm to yield a diameter of 500 μm in the finished part. After lamination and firing, the central lumen was found to be 445 ± 6 μm in diameter which is attributable to 23.5% shrinkage during firing. The increased shrinkage may be attributable to the low amount of ink printed per sheet, or the high curvature around the lumen. An additional factor could be deformation during lamination.[26] No sacrificial material was used to prevent deformation during lamination. In any case, fabrication yields an electrode area of approximately 0.03 mm² based on a post firing layer thickness of 200 μm.

2.4. Electrochemical Measurements

Electrode cleaning was performed similar to the procedure outlined elsewhere.[27] Electrodes were first placed in a solution of 50mM KOH, 30% H₂O₂ at 80°C for 10 minutes. The chips were then washed in DI water and then placed into 50mM KOH solution and the voltage was swept from 400 mV to −1.1V at 500 mV/s repeatedly until the CV's were superimposable. After cleaning, the chips were rinsed with DI water and CV's were taken in 5mM RuHex, 0.5M KNO₃. The CV response of the electrodes was tested by two methods. The first used Silver/Silver Chloride reference electrode and a Platinum rectangular flag counter electrode. This setup permitted the confirmation of screen printed electrode cleanliness and stability. For measurements with flow, the on-chip counter and gold ink pseudo-reference electrodes were used. For high flow rates cyclic voltammetry was performed with flow generated by a Harvard Apparatus Pump 11. For the study of low flow rate, a CHI 1030 was used for measurement with −1.0 V applied to the working electrode, with flow generated by an ePump 190. The flow rate was confirmed by pumping through the ring electrodes onto a Sartorius SE2 ultramicrobalance and measuring the rate of change in mass on the balance. For this test 0.1M PBS buffer was used with no electroactive species present.

3. Results and Discussion

After cleaning as outlined above, the electrodes were tested in static solution composed of 5mM Ruthenium hexamine with 0.5M KN0₃ used as the supporting electrolyte. CV's were performed on three separate chips at the scan rates of 25, 50, 100, 250, 500, 1,000, and 2,000 mV/sec. For the first test, a Silver/Silver Chloride reference electrode (saturated KCL) and a platinum flag counter electrode (Area = 300 mm²) were used. The CV's were performed at each scan rate 5 times and the average peak current and standard deviation of the current were calculated. Figure 5 shows the combined data of average current and standard deviations of the three chips at each scan rate. The linearity and low standard deviation indicate that chip-to-chip variation in performance was low. Furthermore, the theoretical current is in good agreement with measurement and is proportional to the square root of scan rate indicating that in static solution the electrode is operating in the linear diffusion regime.[28]

$$i_p=(2.69X{10}^5)n^{{}^3∕_2}A{D}^{{}^1∕_2}{v}^{{}^1∕_2}{C}^{\ast }$$ Eq 7

Figure 5:

Graph of current vs. square root of scan rate for LTCC sensor in static solution. Data taken with Ag/AgCl reference electrode and Pt flag counter electrode

After testing with Ag/AgCl reference electrode and a Pt flag counter, the electrodes were tested with on-chip gold pseudo reference and counter electrodes in the same 5mM Ruthenium hexamine, 0.5M KN0₃ solution used for the previous test. The same scan rates and number of runs were used as before. The data are shown in Figure 6. The peak current again is in good agreement with calculations and is again linear with the square root of scan rate. There is, however, higher variation at the highest scan rate under these conditions. The slope of the best fit line is similar to the best fit line from data taken with Ag/AgCl reference and Pt counter electrode in Figure 5 indicating the on-chip pseudo reference and counter electrode perform acceptably.

Figure 6:

Graph of current vs. square root of scan rate for LTCC sensor in static solution. Data taken with on chip reference and counter electrode.

To serve as a “down-stream” detector of assay products we expected the sensor would be used for chronoamperometry. The signal from ELISA depends on many factors; coating efficiency, incubation time, temperature, and substrate age to name a few. To test only the sensor, plugs of 2 mM Ferricyanide + 2mM Ferrocyanide in 0.1M KCl (4 mM FeCn) were used. Sensors were cleaned by CV sweeps in 50mM KOH then rinsed with 0.1 M KCl. Briefly, a 20 μL plug of FeCn was loaded upstream of the detector. The detector was operated in CA mode with working voltage of −0.4 Volts with respect to on-chip pseudo reference. Recording was started and after 30 seconds flow was initiated from Harvard Apparatus Pump 11 at 20 μL/min. The pump was fitted with a 1 mL glass syringe filled with 0.1 M KCl. A typical signal with 4 mM FeCn is shown in Figure 7. A slight current increase is seen when flow is initiated. When the plug of FeCn reaches the electrode, the current rises followed by a long tail as the FeCn is pushed through by 0.1 M KCl from the syringe pump. Tests were performed with 5 sensors with concentrations of 4 mM FeCn, 3 mM FeCn, 2 mM FeCn, and 1 mM FeCn in triplicate for each sensor. Peak height was measured by averaging the current for 10 seconds at the top of the peak and subtracting 10 second average current at the base. Average and standard deviation of all 15 measurements at each concentration are plotted in the inset of Figure 7. Sensors for this test were made as previously described with the exception that Ferro CN30-080M Au was used to define the electrodes.

Figure 7:

Graph of current from flowing 20 μL plug of 4 mM FeCn through sensor at 20 μL/min. Data showing average current and standard deviation for different concentrations of FeCn flowed through the sensor.

Of note is the performance of the electrode in flowing solution. Under static solution conditions the electrode exhibits macroelectrode behavior, while under flow conditions the electrode exhibits microelectrode characteristics. This change of behavior is shown in Figure 8. The data in Figure 8A was taken in static solution, while the data in Figure 8B was taken with a flow through the lumen of 25 μl/min. The same electrode was used for both data sets with a scan rate of 25 mV/s. The peak current was measured as the difference between the bottom plateau and the top plateau for microelectrode behavior. In addition to the transition from macro to microelectrode behavior, there is also an increase in the current under dynamic flow conditions as expected due to increased flux.

Figure 8:

Cyclic voltammograms taken at 25mV/sec sweep rate. Top figure is with on-chip pseudo reference electrode in static solution. Bottom graph is with solution flow rate of 25 μl/min.

Additional flow rates were studied, the data is shown in Figure 9. An increase in current is seen between data taken in static solution and data taken with a flow rate of 25 μl/min. However, no additional increase was seen at higher flow rates. A flow rate of 25 μl/min equates to a linear flow rate of 2.7 mm/sec through the central lumen of the electrodes. With an interelectrode spacing of 0.2 mm, the solution flows from the working to the counter in approximately 150 ms. For the electrodes to exhibit macroelectrode behavior under flow conditions the electrode potential needs to be able to switch from reducing to oxidizing before the reduced species can exit the electrode region. For the potential window from 250 mV to −500 mV used in this study, the scan rate would need to be > 5,000 mV/s, significantly above the highest scan rate (2000 mV/s) chosen for this study. Alternatively, studying the electrodes at lower flow rates could be used to probe the transition from macro to microelectrode behavior.

Figure 9:

Data showing peak current vs. square root of scan rate for several different flow rates. Note the rise in peak current once flow is initiated, but then no further increase in current at increasing flow rates.

To test the accuracy of Eq. 1 at very low flow rates it was decided to measure the current generated at the electrode surface without the presence of electroactive species. The data from this experiment is shown in Figure 10. This data was recorded by a CHI 1030 with a bias of −1.0V applied to the working electrode with on-chip reference electrode. Flow is generated by an ePump Model 190. The smooth flow from the ePump (lack of pulses) was advantageous in this study since fluctuations in the flow rate generate variations in the current measured by the electrode. Flow rate was verified by a Sartorius SE2 ultramicrobalance (0.1 μg resolution). Briefly, solution flowed through the electrode and into a cup positioned on the balance. The draft shield was replaced with a similar shield with a 360 μm diameter fused silica tube (idex health sciences) which protrudes down into a small fluid reservoir. The water in the reservoir is covered with high purity mineral oil (PML microbiologicals) to eliminate measurement error due to evaporation. The tube extends through the oil into the water without touching the reservoir sides. By measuring the change in mass of the solution, the flow rate of the pump was confirmed. Figure 10 shows the current measured by the electrode as well as the flow rate confirmed by the ultramicrobalance in relation to the intended flow rate plotted on the x-axis. Based on measurements with the ultramicrobalance, the ePump 190 can produce flow rates as low as 10 nL/min with both high precision and accuracy. Since no electrochemistry is being performed in the sensor, i_ind = 0 in Eq. 2, the current measured is proportional to the rate at which charge carriers (K⁺, Na⁺, Cl⁻) are carried to the electrode. The sensitivity for the electrochemical measurements is less than the balance since the current is proportional to v ^1/3 according to Eq. 1, while mass is proportional to v. However, the electrode requires much less space and requires much simpler measurement electronics.

Figure 10:

Graph of Current measured by chronoamperometry (right axis) at low flow rates. Flow rate is generated by an ePump 190 and confirmed by ultramicrobalance. (left axis)

The current measured in the annular electrodes is of the form predicted by Eq. 2, where:

$$i_T=k{v}^{{}^1∕_3}$$

According to Eq. 1 the constant k should be:

$$k=2.01nF\pi C_b{D}^{{}^2∕_3}{R}^{{}^2∕_3}{X}^{{}^2∕_3}$$

Where C_b =0.140 mol/L, D=1×10⁻⁷ cm²/sec, R=0.025 cm, X= 0.02 cm. For our device this yields:

$$i_T=3X{10}^{-9}A+1.8X{10}^{-5}\left(C\cdot {\mathit{sec}}^{{}^{-2}∕_3}\cdot c{m}^{-1}\right)\cdot v^{{}^1∕_3}$$ Eq 8

A line with this value is plotted in Figure 9 in green, while experimentally determined values are plotted as red squares, which shows good agreement down to these very low flow rates.

4. Conclusions

Annular ring electrodes were studied extensively in the 1960's and 1970's, with some work continuing with sensors for capillary electrophoresis. With new fabrication techniques such as LTCC, it is now possible to easily fabricate annular electrode designs. In this paper, we have designed, fabricated and tested an annular electrode geometry that can be used in axisymmetric microfluidic flows. The flow through the electrode lumen was studied with computational methods to glean an insight into the best design, which led to the choice of smooth sidewalls. Smooth sidewalls permit the rapid exchange of electroactive species around the electrode surfaces enabling detection of smaller volumes of analyte. Sensors were fabricated by means of LTCC. After fabrication, the electrodes were tested and shown to operate in the linear diffusion regime under static solution conditions. By introducing flow of electroactive species through the electrode, the nature of the CV response can be switched from macroelectrode to microelectrode behavior. Steady state current under flow conditions was studied under very low flow conditions and found to follow v^1/3 behavior. Annular ring electrodes are now easy to fabricate and deserve a second look by the microfluidics community.

Highlights.

• Computer modeled electrochemical sensor designed for sensing of flowing solutions.

• LTCC fabrication permits noble metal electrodes in mesoscale electrochemical sensor.

• Annular ring electrodes allow chemical sensing without trapping bubbles.

• Sensor performs as either macro or micro scale sensor based on fluid flow rate.