Electrochemical detection techniques in micro- and nanofluidic devices

Abstract

Electrochemical techniques are widely used in microfluidic and nanofluidic devices because they are suitable for miniaturization, have better sensitivity compared to optical detection techniques, and their components can be reliably microfabricated. In addition to the detection and quantification of analytes, electrochemical techniques can be used to monitor processes such as biological cell death and protein/DNA separations/purifications. Such techniques are combined with micro- and nanofluidic devices with point-of-care (POC) applications in mind, where cost, footprint, ease of use, and independence from peripheral equipment are critical for a viable design. A large variety of electrode materials and device configurations have been employed to meet these requirements. This review introduces the reader to the major electrochemical techniques, materials, and fabrication methods for working and reference electrodes, and to surface modifications of electrodes to facilitate electrochemical measurements, in the context of micro- and nanofluidic devices. The continuing development of these techniques holds promise for the next-generation lab-on-a-chip devices, which can realize the goals of this technology such as POC clinical analysis.

1 Introduction

Microfluidic and nanofluidic devices have been defined in a number of ways, based on fluid volumes, dimensions of fluidic channels, and phenomena involved. For example, Bruus defined microfluidics as “fluids and of suspensions in submillimeter-sized systems influenced by external force” and mentioned scaling down laboratory steps to the 100 μm scale (Bruus et al. 2008). Whitesides (2006) defined microfluidics as “the science and technology of systems that process or manipulate small (10⁻⁹–10⁻¹⁸ L) amounts of fluids, using channels with dimensions of tens to hundreds of micrometers”. Nanofluidic systems have also been defined based on size, such as “the study of fluids in channels with nanometer-scale—ideally less than 50 nm—dimension” (Whitesides 2006). This review focuses on the application of electrochemical techniques in microfluidic and nanofluidic systems, and a broad definition of micro- and nanofluidic systems is preferred to reflect the variety of applications that have been developed. This review article considers systems with microfluidic channels and pumping mechanisms (an example is shown in Fig. 1), devices that involve very small volumes of static fluid, and devices where fluid flow is through paper or similar matrices.

Fig. 1

Top view schematics of two similar microfluidic device designs. “Type I” (a) has a single working electrode (WE), while “Type II” (b) has multiple WE surfaces connected to a single lead. The novel auxiliary flow channel was used to measure liquid plug volumes, and syringe pumps were used to drive fluid through both channels. c Detailed views of the electrodes used for coulometric H₂O₂ detection. Only part of WE is shown for “Type II” design. d Side view schematic of the electrode region in “Type II” design, showing a liquid plug entering. Channel depth did not change in the “Type I” design [adapted with permission from Sassa et al. 2010, Copyright (2010) American Chemical Society]

Detection is a key aspect of analytical micro- and nanofluidic devices design, whether it refers to quantifying analyte concentrations on the order of nM or to determining the mere absence or presence of an analyte. Parameters affected by the detection technique include limit of detection (LOD), sensitivity, and selectivity (Bunyakul et al. 2009; Moeller and Fritzsche 2005; Connelly and Baeumner 2012; Dungchai et al. 2009; Gubala et al. 2012). LOD is the smallest concentration of analyte which can be detected (Armbruster and Pry 2008); sensitivity is the ratio of the change in measured signal to the change in analyte concentration (Inczèdy et al. 1998); and selectivity is the ability of an analytical method to detect the analyte separately from other substances (Vessman et al. 2001). However, in the literature, the word “sensitivity” is often used to convey the meaning of LOD, and care must be taken when reading published works to determine the intended meanings of these terms. Other factors affected by the detection technique are the simplicity of device operation, required peripherals, and cost (Bunyakul et al. 2009; Connelly and Baeumner 2012; Pohlmann et al. 2009; Castano-Alvarez et al. 2007). These factors are very important in micro- and nanofluidic devices, because interfacing the miniaturized components with macroscale peripherals such as pumps, optics, and cameras is a technological bottleneck (Moeller and Fritzsche 2005; Pohlmann et al. 2009; Gubala et al. 2012; Castano-Alvarez et al. 2007; Sin et al. 2011; Mariella 2008; Liu et al. 2011). Moreover, such devices are often developed with point-of-care (POC) applications in mind, where cost, footprint, ease of use, and independence from peripheral equipment are critical for a viable design (Nyholm 2005; Fu et al. 2011; Bunyakul et al. 2009; Moeller and Fritzsche 2005; Connelly and Baeumner 2012; Altinier et al. 2001).

There are four classes of detection methods used in microfluidic devices: mechanical, magnetic, optical, and electrochemical (Zhang et al. 2010; Grieshaber et al. 2008; Noh and Kim 2013; Sasso et al. 2012; Kuswandi et al. 2007; Nyholm 2005; Dahlin et al. 2012; Connelly and Baeumner 2012; Koydemir et al. 2011). Optical and electrochemical detection methods are far more common than mechanical and magnetic methods (Yi et al. 2006; Sin et al. 2011). This review focuses on electrochemical detection techniques employed in micro- and nanofluidic devices. Figure 1 shows a typical example of a microfluidic device built on a glass substrate, with working, reference, and auxiliary electrodes. The device quantifies hydrogen peroxide concentration with coulometric detection (LOD: 1.3 μM for “Type I” design, 410 nM for “Type II” design) (Sassa et al. 2010).

First, the advantages and disadvantages of using electrochemical versus optical detection are discussed. Then, the major electrochemical techniques used in micro- and nanoscale devices are described, followed by the variety of electrode materials used in these techniques, and a discussion of electrode surface modifications. Specific aspects of electrochemical detection techniques related to micro- and nanofluidic devices are also discussed. These include the main approaches in increasing the portability of analytical micro- and nanodevices, the research thrust into the nanoscale, and the most commonly detected analytes. Most of the work cited in this review employs electrochemical detection techniques in a micro- or nanofluidic device.

2 Electrochemical versus optical detection

Optical detection is very widely used, and it has important advantages such as very low LODs, easier multiplexing compared to electrochemical detection, and low-cost disposable components (Swinney and Bornhop 2000; Myers and Lee 2008). Some applications are very simple, such as lateral flow tests where the readout is colorimetric and can be done with the naked eye (Mark et al. 2010). On the other hand, precise alignment of components has hindered POC applications, and miniaturization of optical components has been problematic until recently (Myers and Lee 2008), while electrochemical detection has been regarded as being easier to miniaturize, and thus having cost and portability advantages over optical detection (Castano-Alvarez et al. 2007; Nyholm 2005; Ronkainen et al. 2010). However, this gap between electrochemical and optical detection has closed recently. This is because the miniaturization problems in optical detection have been addressed in recent years with cost reductions in optoelectronic components and developments in the fabrication of microoptical components (Myers and Lee 2008). Examples include applications that use LEDs as excitation sources (Chang et al. 2011; Irawan et al. 2007), and microfabrication techniques used to create microlenses (Ho et al. 2007), lasers, and photodiodes (Balslev et al. 2006). Chang et al. (2011) built a 7.5 cm × 5 cm × 5 cm unit using a LED excitation source. Balslev et al. (2006) fabricated on a single SU-8 layer the microfluidics components, an excitation laser, waveguides, and photodiodes for transduction. Detailed information on optical detection techniques can be found in a review by Kuswandi et al. (2007). In optical detection methods, miniaturization reduces the signal-to-noise ratio, since the amount of analyte and light path through the sample is reduced with decreasing sample volume (Kuswandi et al. 2007). For absorbance-based techniques, the Beer–Lambert law describes this limitation simply. In fluorescence-based techniques, the signal is reduced as the number of fluorescing analyte molecules decreases with decreasing sample volume. Nevertheless, single-molecule detection has been achieved with laser-induced fluorescence microscopy (LIF) (Lee et al. 1994), but adaptation of this technology to a portable microdevice system substantially reduces the sensitivity and practical LOD.

An advantage of electrochemical techniques with regard to miniaturization is that electrochemical signals are likely to improve, rather than deteriorate, when the electrode is miniaturized. This is unlike other detection mechanisms, and it is because the diffusion of the analyte to the electrode surface changes as the electrode size goes from macro to micro, increasing the collection efficiency, and therefore the signal-to-noise ratio (Bange et al. 2005). Charging current (contributes to noise) decreases with the area of the electrode, while the Faradaic current (signal) decreases with the radius, resulting in a boost to the signal-to-noise ratio as the electrode area decreases. Additionally, steady state is reached in less time as the electrode size decreases, allowing the analysis of more rapid processes (Michael and Wightman 1996). However, measured currents also decrease with electrode size, increasing the need for shielding (Bange et al. 2005). At the nanoscale, quantitative data acquired with electrodes 50 nm or smaller in size may not be reliable. At this size range, analyte molecule sizes and electrical double layer thickness are comparable to the electrode size. This, and altered fluid density and viscosity, causes electrodes smaller than 50 nm to carry much smaller currents (therefore produce weaker signal) than predicted by theory (Michael and Wightman 1996). In addition, fluid flow over the electrode has a significant effect on electrochemical measurements. In cyclic voltammetry and chronoamperometry experiments, signal-to-noise ratios as well as measured currents of recessed microelectrodes were found to be dependent on convection rate over the electrode (Henry and Fritsch 1998). This is an important factor to consider in microfluidic devices, as it suggests that accurate measurements require accurate knowledge of fluid flow rates.

Electrochemical detection has been touted as having great potential for low-cost, portable applications (Nyholm 2005; Daniels and Pourmand 2007; Ronkainen et al. 2010). Other advantages of electrochemical detection techniques include the ability to analyze turbid or colored analytes (Bange et al. 2005) and that devices do not need to be built using optically transparent materials (Nyholm 2005). Disadvantages include cumbersome electrode handling processes, frequent calibration (Swinney and Bornhop 2000), and electrode fouling problems (Manica et al. 2003).

3 Electrochemical techniques used in micro- and nanofluidic devices

A vast majority of electrochemical detection in microfluidic devices have included amperometric detection (AD) coupled to capillary electrophoresis (CE). The first report of CE–AD in a microfluidic system, a seminal 1998 paper by Woolley et al. (1998), which was the first report of CE–AD in a microfluidic system, has been cited more than 400 times, and the first report of CE and AD combined has been cited over 425 times since 1987 (Wallingford and Ewing 1987). Conductivity-based detection (CD), which is very similar to AD, potentiometric techniques such as cyclic voltammetry (CV), and electrochemical impedance spectroscopy (EIS) have also been adapted to micro- and nanofluidic devices. This section describes the use of these and other techniques as they are applied to the micro- and nanoscale. Table 1 lists the literature where electrochemical techniques were used for characterizing or monitoring species of interest rather than quantifying them. For example, Derrington et al. used conductivity measurements at a nanopore to sequence DNA. The conductivity change as each nucleotide passed through the nanopore depended on the nucleic acid species, enabling DNA sequencing (Derrington et al. 2010). In another study, percentages of resting and activated platelets in a sample population were measured (Evander et al. 2013). In these cases, common metrics such as LOD were not given, nor necessarily relevant. In other cases, available metrics were not comparable with the literature, as in the case of a microfluidic grid where the movements of nematodes were monitored through their detection by an electrodes array. The authors presented the spatial resolution afforded by their device, which does not have a counterpart in devices where the analyte concentration is quantified at the working electrode surface (Liu et al. 2013). Likewise, a potentiometry-based cytometer was reported with a throughput of 48,000 cells/s (Kim et al. 2013). Finally, Table 1 summarizes other works where electrochemical means were used to manipulate rather than quantify a system, such as electrochemical cell lysis (Jha et al. 2009). Even though these are not strictly in the realm of detection, they were included as representative of the capabilities that are made possible by the merging of micro- and nanofluidic devices with electrochemical techniques.

Table 1 Electrochemical applications in microfluidic devices other than analyte quantification

3.1 Amperometric detection

AD is the first electrochemical technique adapted to the microscale. In AD, a small potential difference is applied to an electrode pair, causing the analytes to undergo a redox reaction. The resulting current indicates the presence and quantity of analyte species (Wallingford and Ewing 1987). Most microfluidic AD applications involve analyte separation by CE prior to detection, since AD cannot distinguish the redox currents of multiple analytes simultaneously. Figure 2a shows the CE–AD scheme, which also applies to conductivity-based detection (Dawoud et al. 2007a). This approach combines the high separation efficiency and resolution of CE with the high sensitivity of AD (Nyholm 2005). CE and AD were combined first by Wallingford and Ewing (1987), and the first microfluidic, LOC, CE–AD application was introduced by Woolley et al. (1998).

Fig. 2

a Separation–detection scheme for CE–AD or CE–CD microfluidic devices. High voltage along the length of the capillary separates the analytes and drives them toward the detection electrodes. A constant potential is applied with a potentiostat to the detection electrodes. As each species passes, a change in current or conductivity between the detection electrodes is detected. The scheme shows in-channel detection, as the detection electrodes are within the channel. b CE–AD data showing separate peaks for dopamine (1) and catechol (2). Figure 1b is from [reprinted from Dawoud et al. 2007a, copyright (2007), with permission from Elsevier]

In CE–AD, interference from the strong electric field applied for separation is a key challenge (Lin et al. 2008). End-channel, in-channel, off-channel, and the very recent parallel dual detecting electrode configurations have been used to address this issue (Lin et al. 2008; Fischer et al. 2009; Martin et al. 2002; Dorris et al. 2012). In end-channel configuration, the detection electrodes are placed 5–20 μm beyond the separation channel (Fischer et al. 2009; Lin et al. 2008). In in-channel and off-channel configurations, the detection electrode pair is inside the separation channel. In in-channel detection, the detection electrodes are between the CE electrodes, while in off-channel detection, they are placed just beyond the CE ground electrode (Fischer et al. 2009; Martin et al. 2002). The CE ground is called a decoupler in this configuration. In-channel configuration requires an electrically isolated potentiostat, developed by Martin et al. (2002). It has no clear advantage over other configurations despite this requirement and has not been used in many applications. Parallel dual detecting electrodes introduced redox cycling and analyte identification mode (Dorris et al. 2012). CE–AD has also been simplified by using the CE electrodes for AD (Schwarz et al. 2001).

Neurotransmitters and catechol are the most common analytes in AD-based microfluidic devices (Fischer et al. 2009; Martin et al. 2002; Woolley et al. 1998; Kong et al. 2006; Huang et al. 2004; Wang et al. 1999, 2006, 2008; Pai et al. 2009; Dawoud et al. 2007a). Catechol can interfere with dopamine measurements, so dopamine–catechol mixtures are often used to test separation resolution as well as the LOD and sensitivity of detection (Fischer et al. 2009; Martin et al. 2002; Pai et al. 2009; Wang et al. 2008; Chua and Pumera 2013). Dopamine–catechol mixtures have become a de facto standard for benchmarks and proof of concept in research aimed at improving the CE–AD technique itself, and a typical electropherogram produced by CE–AD of this mixture is shown in Fig. 2b (Dawoud et al. 2007a). Other analytes quantified by AD include amino acids (Wang et al. 2003b, 2005b), phenols (Chu et al. 2008; Schwarz et al. 2001; Liu et al. 2006a; Wang et al. 2003a), pharmaceutical compounds (Bani-Yaseen 2009; Liu et al. 2006a), pesticides (Liu et al. 2012b), insulin and glucose (Wang et al. 2003c), explosives and nerve agents (Wang et al. 2003a), and food-packaging contaminants (Ha et al. 2009).

When AD is not coupled to CE, either an alternate separation method or an analyte-specific recognition element is needed. Liquid chromatography (Carvalhal et al. 2010), immunochromatography (Inoue et al. 2007), field-amplified sample stacking (Lee et al. 2009), micellar electrokinetic chromatography (Shiddiky et al. 2006), and microchip gel electrophoresis (Shiddiky and Shim 2007) have been used as separation steps (Carvalhal et al. 2010; Shiddiky et al. 2006). The separation technique does not affect how AD is carried out. Specific AD of biomolecules has been achieved by modifying electrode surfaces with enzymes and antibodies to capture a specific analyte. Charge transfer on electrodes modified this way relies on a specific reaction, and LODs in the low nM range are typically achieved (Chikkaveeraiah et al. 2009, 2011; Guan et al. 2005; Liu et al. 2012b; Mir et al. 2008).

Pulsed amperometric detection (PAD) has also been used in microfluidic systems, and it is especially effective in cases where electrode fouling occurs rapidly. In PAD, the electrode is first cleaned by a strong positive potential, then reactivated with a strong negative potential, and finally, the analyte is detected with a moderate potential (Garcia and Henry 2003). Mixtures containing of similar molecules, such as those of underivatized carbohydrates, amino acids, and sulfur-containing antibiotics, have been separated and analyzed with this technique (Garcia and Henry 2003).

3.2 Conductivity-based detection

Conductivity-based detection (CD) is also frequently coupled with CE. It is simple, sensitive, and applicable for electrochemically inactive species, so it is sometimes referred to as a “universal technique” (Nyholm 2005). Therefore, it is very attractive for the detection of inorganic species (Zemann 2001). CD has the same instrumentation as AD, and its operation is very similar (Fig. 2a). A constant potential is applied between the sensing electrodes, and conductivity changes are recorded as analytes passes between them (Grieshaber et al. 2008). A review of CE–CD written by Zemann in 2001 is a good primer for readers new to the topic, despite its age (Zemann 2001).

CE–CD has two modes, contact and contactless. In contact mode, the detection electrodes contact the buffer directly, while in contactless mode they are separated from the buffer by a thin dielectric layer. In contact mode, electrode stability is reduced due to fouling and interference from the CE field. The latter can be addressed by decoupling the electric field through electrode positioning choices, similarly to AD (Klett et al. 2001). Contactless conductivity does not have either problem, and offers simpler fabrication and electrode alignment. It is therefore generally preferred to contact mode detection (Xu et al. 2009).

CD was used to detect analytes such as small ions (Guijt et al. 2001; Liu et al. 2012a; Mai et al. 2013), explosives (Wang and Pumera 2002), DNA (Liang and Chou 2008; Dawoud et al. 2007b), and antibodies (Abad-Villar et al. 2004). It has similarly been used to count circulating tumor cells in whole blood samples (Dharmasiri et al. 2011; Adams et al. 2008). CD has also been used in numerous unique monitoring and measurement applications, which are presented in Table 1 (Kowalczyk et al. 2010; Cui et al. 2001; Liu et al. 2001, 2013). Noteworthy themes among these are cell monitoring, DNA analysis, and investigations of concentration distributions in microfluidic environments. An EOF measurement method based on current monitoring by Huang et al. (1988) can also be considered such an application. One useful approach is to combine AD and CD. Since AD only detects electroactive species while CD detects electroactive and nonelectroactive species, they can be used together to analyze mixtures containing both types of compounds. Two research groups have done this, one with peroxynitrite degradation products and the other with mixtures of ionic and nitroaromatic explosives (Vazquez et al. 2010; Wang and Pumera 2002). This has been reported to provide peak confirmation and improve reproducibility (Wang and Pumera 2002).

3.3 Voltammetric detection

In voltammetric techniques, the potential across an electrode pair is varied over time and the current change is measured as a function of the potential (Grieshaber et al. 2008). Current peaks are observed at potentials where electrochemically active species undergo redox reactions. The potential–current response can be used to identify analytes and determine concentration (Bard and Faulkner 2001; Xu et al. 2009). The best known voltammetric technique is cyclic voltammetry (CV), but there are several voltammetric techniques defined by the shape of the applied waveform (Bard and Faulkner 2001). In micro- and nanofluidic devices, voltammetric detection is less common than AD and CD, partly because many applications contain a separation step, reducing the need to distinguish different analytes based on their potential–current responses (Xu et al. 2009). However, voltammetric detection can be used to detect multiple analytes simultaneously, unlike AD and CD. For example, a DNA probe and its target were simultaneously detected by functionalizing them with different electroactive moieties and using voltammetry (Swami et al. 2005).

Voltammetric techniques detect analytes through the current associated with their electrochemical reactions. The main difference between different voltammetric techniques is the potential signal applied in each case. Figure 3 compares the signals applied for cyclic voltammetry (Fig. 3b) and square wave voltammetry (Fig. 3c) to that applied for AD or CD (Fig. 3a). Moreover, they distinguish reactions taking place at different electrode potentials. This means that they can identify different reactions taking place in the same system, and multiple analytes whose redox potentials are different enough can be simultaneously detected. For these reasons, voltammetric techniques have been used in a large number of microfluidic monitoring applications in addition to analyte detection.

Fig. 3

Signals applied for some common electrochemical detection schemes. a In AD and CD, a constant potential is applied, and current or conductivity is monitored. b In CV, a triangular potential signal is applied to alternately reduce and oxidize analytes, whereby the reduction and oxidation currents are measured. c In square wave voltammetry, the potential alternates between reducing and oxidizing potentials abruptly. The reduction or oxidation current is measured some time after the potential change (shown by dashed lines), after the charging current has decayed

The most common voltammetric technique is cyclic voltammetry (CV), which uses a triangular potential wave. Well-known, practical microfluidic applications include glucose and cholesterol sensors using enzyme-functionalized electrodes (Li et al. 2003; Guan et al. 2005). Cyclic voltammetry (CV) has also been used in microfluidic devices to detect specific DNA targets and base pair mismatches in DNA without labeling, which is significant because labeling is the most time-consuming and complex part of conventional DNA target detection (Mir and Katakis 2008). Chemical processes such as cell death (El-Said et al. 2009) and polymerase chain reaction (PCR) progress (Defever et al. 2009) can also be monitored by CV. Electric field measurements (Forry et al. 2004), analysis of pL scale fluid droplets (Neugebauer et al. 2004), and electrocatalyst characterization (Dumitrescu et al. 2012) are other examples of CV applications in microfluidic devices.

Even in devices employing other electrochemical techniques, CV has been used to characterize the performance of electrodes of different materials or with different surface modifications (Chikkaveeraiah et al. 2009; Tang et al. 2010; Lee et al. 2008; Gonzalez et al. 2009; Yan et al. 2003; Zhang et al. 1999; Lim et al. 2003; Hao et al. 2008; Goluch et al. 2009; Wolfrum et al. 2008; Inoue et al. 2007; Dam et al. 2007), and to determine suitable potentials for other techniques such as AD (Sun and Gillis 2006; Lim et al. 2003; Zevenbergen et al. 2009; Inoue et al. 2007). However, CV is not the most suitable technique for quantifying analytes for a number of reasons. Perhaps, the most important is the presence of a charging current due to the constantly changing applied potential, which reduces the signal-to-noise ratio and whose subtraction from the measured signal can be difficult. It has been argued that small overestimations of charging current could lead to significant overcorrection and large apparent reduction in currents at fast scan rates (Henry and Fritsch 1998). It is important that the charging current is stable over a relevant length of time for this to be viable. Stability over several seconds may be enough for a single measurement, but much longer stability periods are necessary to observe physiological changes or continuous measurements (Robinson et al. 2003). There are other factors that make it difficult to quantify analytes with CV. Convection has a significant effect on the measured currents, especially at slow scan rates. Amatore et al. (2009) found this effect to be present in Pt disk electrodes with diameters of 12.5–250 μm, although the effect became less important with smaller electrodes. Analyte adsorption can also make CV analysis challenging, and missing information about the occurrence of adsorption and its characteristics can lead to misinterpretation of data. For a given species, diffusion or adsorption may dominate, or both may be significant; adsorption tends to be more significant at low concentrations. Reduction of solute molecules diffusing to the electrode surface and reduction of adsorbed molecules have different reduction/oxidation potentials (Brown and Sandifer 1986). Thus, adsorption can be detected as a separate peak from diffusion, and thus separately measured, but if a species adsorbs weakly, the reduction potentials for solute and adsorbed species can be close enough that the peaks are superimposed (Brown and Sandifer 1986). In addition, an adsorption peak may not appear when the scan rate is very slow and a diffusion peak may not appear when the scan rate is very high. As a result, when adsorbing species are present, adsorption kinetics must be well understood, and experiments must be designed and interpreted very carefully (Brown and Sandifer 1986). In microfluidic devices, there is an additional complication similar to the interference problem found in AD. The high electric fields applied across channels reduce the potential difference between the working electrode (WE) used for CV, causing peak shifts. Forry et al. (Forry et al. 2004) studied this in a cross-shaped channel design commonly used for CE. In addition to the shifts caused by the applied electric field, they found that the peak potentials were prone to shift depending on emersion depth of the WE inserted into the microfluidic channel.

Fast scan cyclic voltammetry (FSCV) is a CV mode, which is simply CV carried out with a very high-potential scan rate. Peak heights increase with increasing potential scan rate, but so does charging current, which is a serious limitation. As discussed in Sect. 2, decreasing electrode size into the microscale mitigates the charging current problem due to the small capacitance and surface area of the electrodes (Michael and Wightman 1996). High-potential scan rates can cause peaks associated with slow kinetics to shift, and for this reason, FSCV can be more selective than CV when species have different kinetic rates. This is the case for nitric oxide, which interferes with dopamine but can be resolved by increasing the potential scan rate (Robinson et al. 2003). FSCV can also be used to investigate systems where conditions change rapidly, such as biochemical processes (Michael and Wightman 1996). An example of this is the analysis of a 150-pL droplet between its dispensation and evaporation by Neugebauer et al. (2004). Forry et al. (2004) performed FSCV with the Ru(bpy) ⁺²₃ /Ru(bpy) ⁺³₃ redox couple and observed that the peak potentials shifted with the electric field strength inside the microfluidic channel, then used this as a means of measuring electric field strength locally. However, there are very few examples of FSCV in microfluidic systems in the literature (Sinkala et al. 2012; Neugebauer et al. 2004; Forry et al. 2004).

Square wave voltammetry (SWV) gives similar information to CV, but with lower background since charging current is largely eliminated. SWV is most suitable for studying electrode processes over a large potential range in a short time. It uses an electrical signal that is a square wave where the minimum and maximum potentials increase by the same amount at each period. In other words, the signal consists of the sum of a square wave and a staircase signal (Fig. 2c). Current is measured at the end of each maximum or minimum potential step, after the charging current has decayed (Bard and Faulkner 2001). In microfluidic devices, it has been used to detect pesticides in environmental water (Liu et al. 2012b), PCR products (Fang et al. 2009), toxic metabolites (Wasalathanthri et al. 2011), and dopamine (Lindsay et al. 2007). Liu et al. (2007) have combined an immunochromatographic test strip with SWV, demonstrating the technique with IgG.

Micro- and nanofluidic devices involve small amounts of analyte and minute sample volumes, which is complemented by stripping voltammetry (SV), a sample concentration method, to detect trace amounts of analytes, especially heavy metal ions (Wang 2002; Vandaveer and Fritsch 2002; Liu et al. 2007; Belmont et al. 1996; Kounaves et al. 1994). Some other electrochemical techniques have been adapted to microfluidic devices. Rahimi and Mikkelsen (2011) demonstrated cyclic biamperometry on a thin film electrode configuration where CV could not be performed due to electrode corrosion. Pulse voltammetry and differential pulse voltammetry were used in cell counting sensors to detect circulating tumor cells. LOD for the pulse voltammetry application was reported as “100 circulating tumor cells in a 40 μL sample” (2,500 cells/mL) (Ivanov et al. 2013). LOD for differential pulse voltammetry was reported merely as “125 cells” (Moscovici et al. 2013).

3.4 Electrochemical impedance spectroscopy

Electrochemical impedance spectroscopy (EIS) is an AC electrochemical technique that determines the Faradaic impedance associated with an electrochemical reaction, which can be used to quantify analyte concentrations. Under a very small amplitude AC signal, an electrochemical cell can be represented as an impedance to the signal, and the current changes with the same frequency ω as the potential, since the current–overpotential relationship is linear for small amplitudes (Bard and Faulkner 2001).

Figure 4 shows the layout of a three-electrode electrochemical cell (Fig. 4a) and the equivalent circuit diagram for it (Fig. 4b). Figure 4c is a photograph of a paper-based microfluidic device with three three-electrode cells (Dungchai et al. 2009). The electrical double layer on the electrode/solution interface has a capacitance, which is represented by C _DL on both WE and CE. At the same time, analytes undergo Faradaic processes at the electrode/solution interface, which is why Faradaic impedance Z _f, represented in Fig. 4b by R _s and C _s, is parallel to C _DL. Beyond this interface is the Ohmic resistance of the bulk solution (R _Ω ). If the cell and control electronics are designed well (i.e., properly positioned electrodes), it is possible to only consider the WE section of the equivalent circuit diagram (outlined with a dashed line in Fig. 4b) (Kissinger and Ridgway 1996). The total impedance of the working electrode portion of the equivalent circuit, comprising of C _DL,WE and (R _S,WE + C _S,WE) in parallel, and R _Ω,WE in series, can be measured experimentally over a range of frequencies. Both the imaginary and real parts of this total impedance are frequency-dependent. Plotting the imaginary part of this impedance against the real part over a range of frequencies allows determination of electrochemical information such as analyte concentrations or reaction rates. This is possible because the equivalent circuit components, such as R _s, are related to analyte concentrations and kinetic parameters. The total impedance of the equivalent circuit outlined with the dashed line in Fig. 4b (Z _T) can be written as (Bard and Faulkner 2001):

$$Z_{\text{T}} = Z_{\text{Re}} + jZ_{\text{Im}}$$

(1)

$$Z_{\text{Re}} = R_{\varOmega } + \frac{{R_{\text{S}} }}{{(C_{\text{DL}} \sigma \omega^{0.5} + 1)^{2} + \omega^{2} C_{\text{DL}}^{2} R_{\text{S}}^{2} }}$$

(2)

$$Z_{\text{Im}} = \frac{{\omega C_{\text{DL}} R_{\text{S}}^{2} + \sigma \omega^{ - 0.5} (\omega^{0.5} C_{\text{DL}} \sigma + 1)}}{{(C_{\text{DL}} \sigma \omega^{0.5} + 1)^{2} + \omega^{2} C_{\text{DL}}^{2} R_{\text{S}}^{2} }}$$

(3)

$$\sigma = \frac{1}{nFA\sqrt 2 }\left( {\frac{{\beta_{\text{O}} }}{{D_{\text{O}}^{0.5} }} - \frac{{\beta_{\text{R}} }}{{D_{\text{R}}^{0.5} }}} \right)$$

(4)

In Eqs. (1–4), “WE” has been dropped from the subscripts for convenience. Z _Re and Z _Im are the real and imaginary parts of total impedance Z _T, j is $\sqrt { - 1}$, ω is frequency, n is the number of electrons required to reduce or oxidize one molecule of the analyte, A is electrode surface area, and D is diffusion coefficient. Subscripts O and R denote the oxidized and reduced forms of a redox system. β terms are kinetic parameters, which depends on the specific redox process. For a one-electron (n = 1), one-step process, ${\text{O}} + {\text{e}}\,{\leftrightarrows}\,{\text{R}}$, β _O and β _R are (Bard and Faulkner 2001):

$$\beta_{\text{O}} = \frac{RT}{{FC_{\text{O}}^{*} }};\quad \beta_{\text{R}} = -\frac{RT}{{FC_{\text{R}}^{*} }}$$

(5)

In Eq. (5), $C^{*}$ terms indicate bulk concentration of O and R.

Fig. 4

a Layout of a three-electrode electrochemical cell. Potential of the WE is set against that of the RE, and the measured current passes between the WE and the CE. b Equivalent electrical circuit of a three-electrode cell. C _DL represents the capacitance of the electrical double layer at the electrode surface, and R _s and C _s represent series resistance and pseudocapacity, respectively. R _Ω represents Ohmic resistance presented by the bulk solution. In properly designed cells, it is possible to only consider the WE portion of the system (dashed line). c Photograph of three three-electrode cells on a paper substrate, with each WE functionalized with an enzyme specific to a different analyte. Connection to the potentiostat (not shown) is through the silver epoxy contact pads. c is from [reprinted with permission from Dungchai et al. 2009, copyright (2009) American Chemical Society]

By applying a frequency sweep and plotting the real and imaginary parts of Z _t against each other on a Nyquist plot, information can be obtained about charging on surface films or the double layer, Ohmic conduction and charge transfer. All such processes change the electrochemical circuit and therefore the time constant (Ratnakumar et al. 2002). Figure 5 shows the Nyquist plots obtained with electrochemical impedance spectroscopy of a surface-modified electrode at each stage of electrode surface modification and upon analyte binding to the recognition element (Fan et al. 2013). Each data point in a Nyquist plot represents impedance measurement at a different frequency, and the different Z _Re and Z _Im values come from their frequency dependence. Thus, EIS is able to provide information about individual electrochemical processes or individual steps of chemical interactions. Another advantage of EIS is that it is affected less by environmental disturbances relative to other electrochemical techniques (Ronkainen et al. 2010; Choi et al. 2011).

Fig. 5

Modification of an Au working electrode surface for selective detection of the analyte acetamiprid via specific analyte–aptamer binding. Electrochemical impedance spectra are shown for each step of modification, starting with the bare Au surface spectrum (leftmost) and progressing to analyte binding on the fully functionalized electrode (rightmost). The extent of the shift in EIS response (two final spectra) upon analyte binding is a function of analyte concentration [reprinted from Fan et al. 2013, copyright (2013), with permission from Elsevier]

Complications that must be considered include that the solution may not strictly display Ohmic resistance (Kissinger and Ridgway 1996). This is especially true when the solution is within a matrix, such as the paper-based microfluidic device shown in Fig. 4c. In that case, the “solution” represented by R _Ω in the equivalent circuit diagram is in fact a solution that is soaked into the solid paper matrix. Also, the description of the double layer by a simple capacitance overlooks its dependency on the potential applied to the electrode (Kissinger and Ridgway 1996).

Most microfluidic devices involving EIS rely on a specific binding event between the analyte and a recognition element (e.g., enzyme, DNA probe, and antigen), which form a surface layer on the detection electrode, bound directly or through a linker molecule, as seen in Fig. 5. In Fig. 5, the electrochemical impedance spectra for each step of modification are shown, starting with the spectrum for the bare Au surface (leftmost). The spectra changed as the surface was functionalized with Au nanoparticles, aptamer, and a blocking agent to block nonspecific binding sites on the aptamer molecule. Analyte binding to the functionalized electrode also changed the electrochemical impedance spectrum (rightmost). The extent of the shift in EIS response enabled quantification of analyte concentration (Fan et al. 2013). Although specific binding of analyte to a recognition element is not a principle behind EIS, it contributes to improved selectivity. By contrast, AD simply returns a current value over time, to which all electrode processes contribute. However, preventing nonspecific reactions between nonanalyte species and modified electrode surfaces can be a challenge, which means selectivity may not be easy to achieve in EIS (Ronkainen et al. 2010).

Disadvantages of EIS include difficult data interpretation and longer analysis times (Mir et al. 2008). The difficulty of data interpretation is best explained by contrasting with the simpler AD data. AD data consist of a time–current plot at constant potential. EIS data include real and imaginary parts of the impedance, measured over a range of frequencies. These data are plotted in different forms such as Nyquist and Bode plots, because no one format easily provides all the information. In addition, correctly interpreting the data requires an equivalent electrical circuit that most accurately describes the electrode surface considering any surface chemistries, adsorbed species, charge transfer through each as well as the analyte. Mir et al. gave a great example of the trade-off between EIS and AD. For thrombin, they achieved a LOD of 5 nM with AD and 30 fM with EIS. However, EIS measurements took 600 s compared to 120 s with AD, and EIS data were more difficult to analyze than AD data (Mir et al. 2008). A different limitation of EIS noted by Choi et al. is charge screening in micro- and nanofluidic systems. Changes to charging states of surface layers cannot be detected if the charges transferred between the surface layer and analyte are screened. This restricts the buffer concentrations to much lower values (e.g., 0.005 × phosphate-buffered saline solution) than those of clinically relevant fluids (Choi et al. 2011).

EIS was used to detect target DNA sequences with electrodes functionalized with DNA capture probes, where hybridization changed the potential–impedance response (Mir and Katakis 2008; Kafka et al. 2008; Gebala et al. 2009; Vamvakaki and Chaniotakis 2008). This eliminates chemical labeling of DNA, which is the most complex and time-consuming part of fluorescence-based techniques (Mir and Katakis 2008; Kafka et al. 2008). However, there is no clear advantage since electrodes need to be functionalized instead. Other analytes detected by EIS include antibiotics (Takhistov 2004), antigens (Yu et al. 2006), enzymes (Barton et al. 2008), and bacteria (Oliver et al. 2006; Gomez et al. 2001). Reported LODs are in the low pg/mL or the 100 pM–1 μM range (Choi et al. 2011; Yu et al. 2006; Barton et al. 2008). EIS can also detect single base pair mismatches (Mir and Katakis 2008; Kafka et al. 2008; Gebala et al. 2009) and sense bacteria viability (Gomez et al. 2001). EIS has also been used to characterize charge transfer characteristics of electrodes used for other applications, such as CE–AD (Liu et al. 2006b). Similarly, enzyme-based biosensors are often characterized using EIS (Guan et al. 2004; Liu et al. 2006b). Guan et al. (2004) reviewed recognition elements used in EIS-based micro- and nanofluidic devices.

3.5 Other electrochemical detection techniques

Other electrochemical techniques are less common in micro- and nanofluidic devices, but some applications have been reported. Chronocoulometry (CC) is identical to AD, except that the measured current is integrated over time, so the total charge accumulation is measured instead of the current. This inherently increases signal strength and has been proposed as a way to achieve sensitivities at the nanoscale not possible with AD (Sassa et al. 2010). In a comparison of AD and CC data from a set of detection experiments, CC was found to have a better coefficient of variation than AD (Bunyakul et al. 2009). The same study found CC to have over sixfold decrease in LOD compared to fluorescence microscopy experiments, with identically prepared samples. However, the major advantage of CC, the integration of signal over time, was not applied to fluorescence microscopy in this case. The authors of this study stated that integrating the fluorescence signal may improve the reliability of the fluorescence signal (Bunyakul et al. 2009).

Redox cycling has been employed with nanoscale electrodes in microfluidic devices (Zevenbergen et al. 2007). Redox cycling is employed in nanoscale devices, where each molecule cycles between the sensing electrodes and undergoes redox reactions multiple times. This is feasible at the nanoscale because the analyte diffuses very short distances. In one study, the electrochemical signal was increased by a factor of ~400 with this technique and shown to be sensitive to fluctuations of 70 molecules (Zevenbergen et al. 2007). While promising, the technique is only practiced by a small circle of researchers thus far (Kaetelhoen et al. 2010; Goluch et al. 2009; Wolfrum et al. 2008; Zevenbergen et al. 2007). Examples of other techniques that are not commonly used in micro- and nanofluidic devices are summarized in Table 2. Some important techniques such as scanning electrochemical microscopy (SECM) are listed in Table 2; this is because they have not been widely adapted to micro- and nanofluidic devices.

Table 2 Examples of less common electrochemical techniques in micro- and nanofluidic devices

4 Electrode materials

Various electrode materials are used in electrochemical detection in micro- and nanofluidic devices. The material selection for the working electrodes (WE), where the analytes are detected, is different than for the reference electrodes (RE), which is used to set the potential applied to the working electrode. The function of counter electrodes (CE) is to complete the electrochemical cell; they tend to be simple (e.g., bare Pt) to not alter the conditions near the WE or RE. This translates into microfluidic device designs having the CE sufficiently far from the other electrodes with a large surface area compared to the WE (Bard and Faulkner 2001). It is also important that the CE material be inert under the experimental conditions. Working electrodes (WE) surface modifications are usually not applied to counter and reference electrodes. In some devices, a single electrode is used as both RE and CE (Cortes-Salazar et al. 2010) and some devices have electrodes that are a part of both separation and detection schemes (Schwarz et al. 2001; Mai et al. 2013). The following sections discuss WE and RE materials. CEs are not discussed further.

Several electrode fabrication techniques are available. Some are relatively low tech. For example, electrodes can be patterned by a stencil (Shinwari et al. 2010), or an inkjet printer (Hu et al. 2012). Other techniques require specialized equipment and access to a cleanroom. Photolithography is used to pattern the electrodes, similar to patterning microfluidic channels on Si wafers. A uniformly applied photoresist is lithographically patterned with the electrode design and then the substrate is coated with a thin film. The photoresist can then be lifted off with sonication in acetone, leaving the electrodes patterned. Figure 6 shows thin film deposition, pyrolized carbon, and screen-printing methods, which are commonly used for electrodes made of metal, pyrolized carbon, and carbon inks or pastes, respectively. Generally, there is a trade-off between the simplicity of fabrication and the precision of microscale, precisely positioned electrodes. Table 3 lists examples of traditional techniques along with other novel techniques that have been used for electrode fabrication. Entries in Table 3 are either novel with regard to the electrode material (e.g., polyaniline), the form of the material (e.g., Au layer on commercial CDs), or some aspect of electrode fabrication (e.g., electrode on PDMS substrate).

Fig. 6

a Thin film deposition: a positive photoresist is spin-coated on a glass slide (1–2). The electrode areas are selectively exposed to UV light via masking, and the UV-exposed photoresist is developed away (3–4). The whole substrate is coated first with a thin film adhesion layer, typically Cr or Ti, then with the electrode material, such as Au, Pt, or ITO (5–6). Photoresist is lifted off by sonication in acetone, leaving the thin film electrode patterned (7). b Pyrolized carbon: a negative photoresist (i.e., SU-8) is spin-coated on a glass slide (1–2). The electrode areas are selectively exposed to UV light and the photoresist not exposed to UV is developed away (3–4). The remaining photoresist is pyrolized into a glassy carbon electrode. c Screen-printing: a mesh is soaked in carbon ink or paste. A stencil is placed between the mesh and substrate, with the electrode areas defined by openings in the stencil. Pressure is applied on the mesh to transfer the carbon ink or paste onto the substrate

Table 3 Examples of microfluidic devices with novel electrode materials or electrode fabrication approaches

4.1 Working electrode materials

Noble metals (Guan et al. 2004; Xie et al. 2004; El-Said et al. 2009; Castano-Alvarez et al. 2007; Wang et al. 2008; Vazquez et al. 2010; Oliver et al. 2006; Radke and Alocilja 2005), other metals (Yan et al. 2003; Kikura-Hanajiri et al. 2002; Vazquez et al. 2010; Gamboa et al. 2009; Paixao et al. 2007; Corbo and Bertotti 2002; Martin et al. 2001), indium tin oxide (ITO) (Sun and Gillis 2006; Ha et al. 2009; Ghanim and Abdullah 2011), carbon-based electrode materials, and conductive polymers (Erlandsson and Robinson 2011; Janin et al. 2011; Mannerbro et al. 2008) have been used in microfluidic devices. Mercury is not used in microfluidic devices because it presents practical difficulties as a liquid at room temperature, and its toxicity makes it incompatible with bioanalytes. Carbon-based materials include carbon ink, carbon paste (Wang et al. 1999, 2003b; Fischer et al. 2009; Spence et al. 2004; Guan et al. 2005; Martin et al. 2001; Kovarik et al. 2004; Mecker and Martin 2008; Dossi et al. 2012), carbon nanotubes (Guan et al. 2005; Qureshi et al. 2009; Liu et al. 2012b; Kong et al. 2000; Zhao et al. 2002; Trojanowicz 2006; Wang et al. 2004b), diamond (Wang et al. 2003a, 2004a; Ghanim and Abdullah 2011; Wu et al. 1996; Zhou and Zhi 2009), graphene (Robinson et al. 2012), and glassy carbon produced by pyrolysis (Wang et al. 2005a; Schueller et al. 1999; Lee et al. 2008). The most commonly used materials are Au, Pt, and carbon inks and pastes, which are discussed first followed by metals and forms of carbon electrodes, as well as different materials such as polymers.

Noble metals are the most commonly used electrode materials in microfluidic devices. Their advantages include established fabrication methods, excellent charge transfer kinetics, and the ability to be functionalized with a wide variety of materials (Dawoud et al. 2007a, b; El-Said et al. 2009; Schwarz et al. 2001; Kuo-Hoong et al. 2007; Radke and Alocilja 2005; Cui et al. 1998). However, they are expensive, and fabrication may require microfabrication techniques, which are complex, time-consuming, and add to the expenses. Thin film deposition is the most common fabrication approach for noble metals, but metal wires are also used. Wire electrodes simplify fabrication greatly and are replaceable (Chikkaveeraiah et al. 2009; Gencoglu et al. 2011; Castano-Alvarez et al. 2006; Kong et al. 2001). However, the electrode area and position cannot be as precisely defined as with thin films. Fabrication methods that are more precise than metal wires but still simpler than thin film deposition have been reported (Daniel and Gutz 2003; Hao et al. 2009; Yan et al. 2003; Kong et al. 2006) and are summarized in Table 3. Some of these novel techniques may lower cost and lower entry barriers to the field by becoming a “rapid prototyping” method similar to soft lithography with PDMS. Complementing rapid prototyping of microfluidics with rapid prototyping of electrodes is an exciting concept, but thus far these novel fabrication techniques have remained novelties. Technique adoption combined with practical uses may make microdevice fabrication cheaper and easier.

Several types of carbon electrodes, especially carbon inks, carbon pastes, and glassy carbon, are commonly available in electrochemical micro- and nanofluidic devices. Reviews exist that focus on carbon nanotube (CNT) (Zhao et al. 2002; Qureshi et al. 2009), diamond, and diamond-like carbon electrodes (Zhou and Zhi 2009; Qureshi et al. 2009). Carbon ink has been used to simply, cheaply, and rapidly fabricate electrodes. It can be inkjet-printed (Chikkaveeraiah et al. 2011; Bani-Yaseen 2009; Fischer et al. 2009; Shabani et al. 2008; Kovarik et al. 2004; Mecker and Martin 2008) or screen-printed (Barton et al. 2008; Castano-Alvarez et al. 2006; Dungchai et al. 2009). Screen-printing is schematically shown in Fig. 6c. A stencil to define the electrode area is placed between the substrate and a carbon ink-soaked mesh. Carbon ink is transferred to the substrate by pressure. This is usually done by hand, and the structures defined by the stencil do not have the resolution of photolithography masks (Shinwari et al. 2010). Glassy carbon, or pyrolized carbon, was first developed by Schueller et al. (1999), and is used widely (Lee et al. 2008; Bani-Yaseen 2009; Fischer et al. 2009; Dumitrescu et al. 2012; Liu et al. 2012b; Li et al. 2003; Schueller et al. 1999; Hebert et al. 2003). Fabrication is via photoresist patterning where the electrode structure is first realized and then pyrolized (Fig. 6b) (Schueller et al. 1999). Fischer et al. (2009) compared pyrolized carbon to carbon ink, carbon fiber and Pd, and found that under identical conditions, pyrolized carbon gave the lowest LOD for dopamine. Wang et al. (2005a) have reported that pyrolized, aka glassy carbon, electrodes can be reversibly charged/discharged with Li ions. Although they are cheap and relatively simple to fabricate, glassy carbon electrodes suffer from high background currents (Qureshi et al. 2009).

Nonprecious metals such as Cu (Gamboa et al. 2009) and Ti (Guijt et al. 2001), as well as ITO (Sun and Gillis 2006; Ha et al. 2009; Ghanim and Abdullah 2011) are also used as electrodes in microfluidic devices. They can be fabricated with the same techniques as noble metals. ITO is grouped with metals despite being a metal oxide because of the similarities in fabrication and use. It also has high conductivity and a wide electrochemical window (Sun and Gillis 2006). Its main advantage is optical transparency, which is useful when concurrent visual or optical information is needed, for example to observe cell adhesion on electrodes (Sun and Gillis 2006). Another nonprecious metal, Cu, is cheap and available in forms that enable more fabrication innovation than alternatives. Thus, various forms of Cu are used as electrodes, despite its susceptibility to corrosion and its toxicity (Lin et al. 2008; Abad-Villar et al. 2004; Gamboa et al. 2009; Corbo and Bertotti 2002). Lin et al. used a length of Cu wire as a WE, which had a very short lifetime. They “replenished” the electrode by pulling the wire through the device, repeatedly moving fresh surface in contact with the sample. Some examples of Cu electrodes can be found in Table 3.

Graphene and carbon nanotubes (CNTs) are another type of carbon-based material, which are used in microfluidic devices both directly as an electrode material (Ang et al. 2011; Heldt et al. 2013) and as a material for electrode surface modification (Chua et al. 2011). They are biocompatible, mechanically stiff, and have high thermal and electrical conductivity (Lu et al. 2012). A drawback is that graphene dissolves and agglomerates in aqueous environments (Lu et al. 2012). Although graphene and CNTs appear to be excellent materials for electrodes and as linkers between electrodes and analyte capture elements, they should not be assumed to be an ideal choice. Chua and Pumera compared graphite, graphite oxide, graphene oxide, and three forms of reduced graphene (reduced with different mechanisms) to glassy carbon electrodes. They used electrodes made of these materials for AD of catechol and dopamine, finding that glassy carbon gave the largest analyte peaks. However, graphite oxide and electrochemically reduced graphene had better analyte resolution than glassy carbon. They also noted that bulk graphene performance was not an indicator of microfluidic device performance because small sample and electrode sizes, fluid flow, and applied electric fields changed analyte/graphene interactions (Chua and Pumera 2013). The same researchers compared graphite and graphene to determine that graphene did not offer any sensitivity or resolution advantages (Chua et al. 2011). Graphene field-effect transistors have been used by multiple research groups for flow velocity measurement (Newaz et al. 2012; He et al. 2012) and single-cell detection from flowing, malaria-infected red blood cells (Ang et al. 2011). CNTs are less commonly used directly as the electrode material for these applications (Sansuk et al. 2013).

Synthetic diamonds, another class of carbon electrodes, are good WE materials thanks to their conductivity, inertness, biocompatibility, and the ability of biomolecules to covalently attach to diamond surfaces (Qureshi et al. 2009; Wang et al. 2003a). Diamonds are reported to have background currents that are 10 times lower than that of Au and 100 times lower than that of glassy carbon (Qureshi et al. 2009). Their surface can be hydrophobic or hydrophilic, depending on whether it is H-terminated or OH-terminated (Qureshi et al. 2009). Wang et al. (2003a) used chemical vapor deposition to fabricate a boron-doped diamond (BDD) electrode. The same group also analyzed a mixture of five purines with CE–AD using a boron-doped diamond electrode and found that it provided a superior signal-to-noise ratio (Wang et al. 2004a).

Finally, conducting polymers are a more recent and promising class of electrode materials. Polymers have been widely used in electrochemical techniques to modify electrodes, but electrodes themselves can also be fabricated out of conducting polymers. Conducting polymers present a variety of fabrication options, and they have tunable chemistry. For example, three novel polypyrroles were developed as DNA detecting electrodes, and functionalized polymers with unsaturated side chains performed better than those with saturated side chains (Peng et al. 2007). Electrodes composed of polymers are a promising class of materials with many potential future applications.

4.2 Reference electrode materials

Reference electrodes have a known, stable potential against which the potential of a working electrode is set. The working electrode potential is expressed in V against the reference electrode potential (e.g., V vs. Hg) (Kahlert 2010). The primary requirement for a reference electrode is that its potential remains constant. Therefore, reference electrodes consist of two phases in chemical equilibrium (Bard and Faulkner 2001). In conventional electrochemical cells, Ag/AgCl electrodes, saturated calomel electrodes (SCE), and Hg droplets are popular reference electrodes, but SCEs and Hg droplets are physically not suited to adaptation to microfluidic devices. The toxicity of Hg is also undesirable with biological analytes. Shinwari et al. (2010) have written a review of reference electrodes in microfluidic systems and pointed out that this subject has not been researched adequately.

Ag/AgCl is very popular in both conventional and micro- and nanoscale electrochemical systems. The equilibrium reaction for this electrode is (Bard and Faulkner 2001):

$${\text{AgCl}} + {\text{e}}{ \,\leftrightarrows \,}{\text{Ag}} + {\text{Cl}}^{ - } \quad E^{0} = 0.2223\,{\text{V}}/{\text{NHE}}$$

(6)

For a true reference electrode, Cl⁻ must be replenished. In a microfluidic device, this can be accomplished with a flow-through channel that supplies Cl⁻ ions to the electrode from a saturated KCl solution (Shinwari et al. 2010). This complicates device fabrication and operation. Another option is to deposit an Ag thin film and chlorinate its surface. This thin film can be used as a pseudoreference electrode, which makes fabrication relatively simple, but electrode lifetime is short since the small amount of AgCl on the electrode surface can be depleted (Shinwari et al. 2010). However, several device designs are available to replenish such Ag/AgCl electrodes (Shinwari et al. 2010).

A new type of RE was recently developed and consists of an ionic liquid-doped membrane with a macroporous solid contact. The absence of a salt bridge alleviated some limitations of existing reference electrodes, especially contamination between electrolyte solution and sample (Zhang et al. 2012).

Pseudoreference electrodes, also referred to as quasi-reference electrodes, simplify fabrication and do not require replenishment. They consists of a metal, such as Au, Pt, Ag, or ITO, that does not sustain an electrochemical equilibrium reaction (Lee et al. 2008; Kasem and Jones 2008; Ha et al. 2009; Neugebauer et al. 2004; Inoue et al. 2007). As a result, their potential may be unstable during an experiment and inconsistent across experiments (Kahlert 2010). A two-electrode CE–AD system, where an Au or Cu counter electrode doubled as a reference electrode, was reported to be at least as good as the conventional designs in terms of noise and sensitivity (Schwarz et al. 2001). Two-electrode systems where the CE/RE was Ag/AgCl were also reported (Yu et al. 2006).

4.3 Electrode surface modifications

Electrode surfaces are commonly modified with other materials to protect electrodes, improve performance, or for analyte selectivity. The most common surface modification is a recognition element with which the analyte interacts, changing charge transfer across the electrode. These interactions include enzyme reactions (Liu et al. 2000; Grieshaber et al. 2008), DNA hybridization (Pohlmann et al. 2009; Ivanov et al. 2013), antigen–antibody interactions (Barton et al. 2008; Yu et al. 2006), and analyte–aptamer interactions (Du et al. 2011; He et al. 2013; Liu et al. 2012c; Swensen et al. 2009; Lubin and Plaxco 2010). They can be used to detect specific gene sequences, organisms, or toxins. Further, surface modifications can decrease LOD. For example, Prussian blue-modified electrodes have very good LODs and sensitivities, and the lowest LODs for H₂O₂ have been reported with Prussian blue-modified electrodes (Ricci and Palleschi 2005). Similarly, using carbon nanotube linkers to functionalize electrodes with glucose oxidase enzyme was reported to improve LOD over electrodes functionalized directly with glucose oxidase (Guan et al. 2005). Electrodes can also be functionalized with biological cells, but this approach is for studying or detecting the cells themselves (Guan et al. 2004).

Recognition elements that react with the analyte specifically are usually not directly bound to the electrode. Instead, the electrode surface is first functionalized with another surface layer, and the recognition elements bind to this layer. Graphene and CNTs are used often as linkers between the electrode surface and receptor molecules (Lin et al. 2012; Lu et al. 2012; Liu et al. 2012b; Guan et al. 2005) because they increase electron transfer rates between the electrode and the analyte, enhance electrochemical reactivity of bioanalytes, and alleviate fouling (Xu et al. 2009; Lin et al. 2012). Figure 7 shows an example of a graphene layer used to modify a carbon ink electrode (Lu et al. 2012). Gold nanoparticles used as linkers between metal electrodes and recognition elements such as aptamers increase the available electrode surface area. This increases recognition elements deposited on the electrode yielding increases in the electrochemical signal. Au nanoparticles have the same material advantages as Au thin films, such as biocompatibility and high rates of electron transfer (Chikkaveeraiah et al. 2009; Du et al. 2011; Fan et al. 2013). Fan et al. (2013) found that the available surface area (as determined by chronocoulometry) of an Au electrode increased by a factor of 4 after deposition of Au nanoparticles.

Fig. 7

Modification of a carbon ink electrode with a layer of graphene, followed by Au nanoparticles (AuNPs), ssDNA as a recognition element (S1), 6-mercapto-l-hexanol (MCH), nanoporous gold (NPG), and the electroactive species thionine. MCH ensured proper S1 alignment, while NPG and thionine were required for electrochemical signal transduction [reprinted from Lu et al. 2012, copyright (2012), with permission from Elsevier]

Electrode surfaces can also be modified with a protective dielectric material coating; multiple groups have successfully substituted Saran wrap for dielectric thin films (Abdelgawad and Wheeler 2008; Chuang and Fan 2006). Similarly, Teflon compounds, which are costly and require a legal agreement with DuPont, have been replaced as hydrophobic coatings with a cheap water repellant, Rain-X, used on windshields (Abdelgawad and Wheeler 2008).

5 Current trends and future directions in electrochemical detection

Current electrochemical detection trends in micro- and nanofluidic systems include improvements to device portability, simplicity for POC applications (Choi et al. 2011; Mannerbro et al. 2008; Liu and Crooks 2011), nanometer-scale electrochemical systems (Zevenbergen et al. 2011; Goluch et al. 2009; Cui et al. 2001; Wolfrum et al. 2008), use of integrated circuits, and the use of novel electrode materials (Guan et al. 2005; Qureshi et al. 2009; Ha et al. 2009; Liu et al. 2012b; Erlandsson and Robinson 2011; Zhou and Zhi 2009; Robinson et al. 2012) and modifications, such as functionalization with specific recognition elements (Tang et al. 2010; Shiddiky et al. 2005; Shabani et al. 2008; Ha et al. 2009; Mir and Katakis 2008; Yu et al. 2006; Lee et al. 2009; Cui et al. 1998; Brett et al. 1999; Kulesza et al. 2001).

Portability and simplicity of use are key for POC applications, but development of useful real-world applications has been disappointingly slow, and there is significant unmet need for POC medical diagnostic devices (Fu et al. 2011; Heikali and Di Carlo 2010). There are two common approaches to this challenge. The first is integrating all components of an analytical technique in a single, portable unit. In some examples, the sensor is disposable, typically the size of a glass slide, and inserted into the base unit responsible for other functions such as logic control, power supply, and readout (Erickson and Li 2004; Fu et al. 2011; Ordeig et al. 2012). A microfluidic nucleic acid sensor with an integrated, battery-powered, 85 mm × 60 mm potentiostat based on a commercially available microcontroller was developed with this approach (Kwakye et al. 2006). Similarly, CE–AD was performed on a 2 in. × 2 in. disposable element, which was used with an in-house built base unit. The dimensions of the base unit were 4 in. × 6 in. × 1 in. and it weighed 0.35 kg (Jackson et al. 2003). The microfluidic chip and main device unit are shown in Fig. 8.

Fig. 8

a Photograph of the base unit of a battery-operated portable microfluidic device that performs CE–AD. a Dual source power supply, b Interface circuit, c AD circuit. b Photograph of the microfluidic chip where CE–AD is performed. Detection chamber electrodes are used for AD. The other electrodes are used for CE and to move fluid between the buffer, sample and waste reservoirs [reprinted with permission from Jackson et al. 2003, copyright (2003) American Chemical Society]

Reviews have specifically discussed microfluidic device integration and portability. A review by Erickson and Li (2004) specifically focused on integrated microfluidic device development with many examples involving electrochemical detection. The focus was on prototype-stage work, which was not necessarily portable. Another review by Lewis et al. (2013) focused on in situ CE systems with many portable examples. These works describe electrochemical detection integrated into microfluidic chips, although relatively few are truly portable; most require connections to large peripherals in the laboratory. These newly developed applications and technical progress suggests similar hurdles to developing portable electrochemical sensing devices and portable microfluidic devices (Erickson and Li 2004). Most microfluidic devices already require a power supply and electronic control elements. Electrochemical detection requires the addition of a potentiostat circuit to this ensemble. Fluid flow and electrokinetic processes such as CE require high-voltage power supplies, while electrochemical detection requires smaller power sources (Lewis et al. 2013). Fernandez-la-Villa et al. (2012) have assembled these key components in a portable device. They built a 150 mm × 165 mm × 90 mm CE–AD unit that included a bipotentiostat and a high-voltage power supply capable of applying 3 kV potential. The microfluidic channels and electrodes were in a removable, reusable cassette. The unit was also battery-powered and capable of >8-h operation. Operational lifetime is as important a challenge as construction of a small, portable device. Electrode fouling is a limiting factor for lifetime of electrochemical detection (Lewis et al. 2013). Disposable chips circumvent this problem as well as avoiding cross-contamination. Another useful approach is PAD, which consists of electrode cleaning, electrode reactivation, and detection steps (Garcia and Henry 2003). Martin et al. (2000) used a reversible seal so that the microfluidic device could be disassembled between uses to clean the electrodes. Material choice for the best lifetime is not straightforward. Glass is more prone to breakage and similar problems, while polymers are more susceptible to surface contamination (Lewis et al. 2013). Therefore, the material, operation procedure, and expected field conditions for an individual instrument must be considered. Finally, recognition elements and reagents can limit the lifetime of a microfluidic device with electrochemical detection. Many microfluidic devices make use of electrodes functionalized with proteins, aptamers, and DNA to detect analytes. The lifetimes of such compounds, and their resilience against environmental conditions, can limit device lifetimes. For reagents that are mixed into the fluid phase, one solution is to design systems where these analytes are only dissolved on demand during operation (Bange et al. 2005).

The second approach to improve portability and simplicity of micro- and nanofluidic systems is to simplify the analytical techniques to eliminate components whose miniaturization is difficult, costly, and not scalable (Liu et al. 2007). A good example is a lateral flow assay test strip, similar to a pregnancy test, which was combined with SWV. Using inexpensive, simply fabricated, screen-printed electrodes functionalized with antibodies, this disposable device was able to detect 10 pg/mL prostate-specific antigen with 20-min immunoreaction time. A schematic of the device is shown in Fig. 9 (Liu et al. 2007). These devices tend to trade sensitivity for lower cost and manufacturing volume. Paper-based microfluidic devices, some with colorimetric detection, are examples of this approach and they are a promising new trend. Microfluidic channels can be defined by printing wax channel walls and then heating to melt the wax into the paper. The hydrophobic wax confines aqueous media (Carrilho et al. 2009; Zhao and van den Berg 2008). This simplifies fabrication because wax can be applied by an inkjet printer or even by hand if the device is only expected to reveal analyte presence or absence (Carrilho et al. 2009; Dossi et al. 2012). Channel closing as well as multilayered devices can be accomplished by folding the paper (Lu et al. 2012; Liu and Crooks 2011, 2012). Devices have also been sealed by lamination with polymers (Dossi et al. 2012). Electrodes are often simply and inexpensively screen-printed or inkjet-printed on the paper (Dossi et al. 2012). Electrochemical detection in a paper-based microfluidic device was first accomplished by Dungchai et al. in 2009, using screen-printed carbon ink as the WE and CE, and Ag/AgCl ink for the RE (Dungchai et al. 2009). Mannerbro et al. (2008) had investigated inkjet printing of electrodes on paper substrates the year before. Recent examples include devices with self-powered electrochemical detection and colorimetric readouts. Device assembly was achieved by cutting and folding the paper substrate, as shown in Fig. 10 (Liu and Crooks 2011, 2012). Dossi et al. (2012) developed a paper-based gas sensor that used ionic liquid medium. The electrochemical cell consisted of screen-printed electrodes inside a wax circle, and the device was laminated against gas and electrolyte leaks. These examples illustrate that electrodes printed on paper have great potential to enable electrochemical microfluidic devices that can be fabricated easily, cheaply, and fast (Carvalhal et al. 2010; Dossi et al. 2012; Dungchai et al. 2009). These paper-based devices do sacrifice sensitivity and accuracy and are not very reliable quantitatively. In addition, fluid flow in paper is complex and not completely understood. Similarly, modeling the electrical behavior of paper-based devices with an equivalent circuit is challenging. Thus, it is unlikely that the progress made with nanofluidic devices can be adapted to paper-based platforms (Zhao and van den Berg 2008).

Fig. 9

Schematic of a disposable microfluidic device that combines immunochromatography with electrochemical detection. Liquid sample is introduced from the sample application pad and moves toward the absorbent pad. An inexpensive screen-printed electrode at the capture antibody pad uses SWV to quantify the captured analyte [reprinted with permission from Liu et al. 2007, copyright (2007) American Chemical Society]

Fig. 10

Schematic (a) and photograph (b) of a paper-based microfluidic device with electrochemical detection. The two paper reservoirs are connected with an ITO electrode, whose transparency is required for readout. The battery section powers the device upon contact with the urine sample. The sensor section features an electrode functionalized with an enzyme. In the presence of the enzyme-specific analyte, the electrode facilitates the Fe(CN) ⁴⁻₆ /Fe(CN) ³⁻₆ reaction, which causes a color change that can be seen with the naked eye [reprinted with permission from Liu and Crooks 2012, copyright (2012) American Chemical Society]

Nanometer-scale systems are becoming more common in miniaturized analytical devices because they offer unique capabilities, which include redox cycling (Zevenbergen et al. 2007), stretching and sequencing single DNA molecules using nanopores (Kowalczyk et al. 2010; Liang and Chou 2008; Derrington et al. 2010; Harrell et al. 2006; Levy and Craighead 2010), in vivo measurements in environments such as brain tissue (Goluch et al. 2009; Rocchitta et al. 2012), and handling and probing single molecules (Goluch et al. 2009; Zevenbergen et al. 2007). Reviews dedicated to specific aspects of nanoscale electrochemical techniques were written, including nanofluidic detection of terrorist agents (Liu et al. 2010), DNA stretching and transport into and through nanopores (Levy and Craighead 2010), and nanopore preparations (Rhee and Burns 2007). Interesting applications of electrochemical techniques, which do not aim to detect and quantify an analyte, are summarized in Table 1.

In nanofluidic systems, fabrication and alignment of nanoelectrodes, nanogaps, and nanopores is an active research field. One example is a redox-gated nanojunction made by polymerizing aniline in a microscale gap to form a nanojunction (Janin et al. 2011). B-doped silicon nanowires were functionalized with amines or oxidized to create field-effect transistors to sense pH or streptavidin (Cui et al. 2001). Etching sacrificial layers (Goluch et al. 2009), combining hot embossing with photolithography (Heyderman et al. 2003), focused ion beam (FIB) (Lillo and Losic 2009), and high-resolution electron beam lithography combined with atomic layer deposition are some of the approaches that have been implemented (Ahmadi et al. 2010).

Sophisticated electrochemical sensors employing integrated circuits such as complementary metal–oxide–semiconductor (CMOS) have been developed (Hwang et al. 2009; Welch and Christen 2013; Huang and Mason 2013). Hwang et al. (2009) developed a 32 × 32 electrode array of Au-coated Al electrodes with an integrated potentiostat. They showed that each electrode in the array could detect K₃Fe(CN)₆ by CV. The array of 1,024 electrodes occupied an area less than 1 mm², and the electrodes were individually addressable, although CV could only run at one electrode at a time. The ultimate goal is screening applications, which use a functionalized array with different recognition elements. However, their work focused on the electrode fabrication and measurement at the electrode surface without a true integration of microfluidic channels with an electrochemical detection system (Hwang et al. 2009). Welch and Christen (2013) used flip chip bonded on a CMOS electrochemical sensing device to a flexible printed circuit board (PCB). The space between the CMOS device and the PCB acted as a microfluidic channel with a burst pressure of 175 kPa. However, they did not carry out any analytical experiments. Huang and Mason (2013) bonded a more conventionally formed microfluidic layer to a CMOS device. They made open microfluidic channels of SU-8 and bonded these to the CMOS layer to form closed microfluidic chambers with lateral tubing connections for sample loading. The CMOS layer consisted of a CMOS pad inside a die well. Since CMOS and carrier die needed very good vertical alignment, they used a leveling process and verified their vertical alignment with profilometry. The reported detailed fabrication technique steps underscore how complicated fabrication can be, depending on the type of technology to be used for micro- and nanofluidic electrochemical devices. Multilayered CMOS elements are therefore more difficult, time-consuming, and expensive to fabricate than thin film metal electrodes (Hwang et al. 2009; Welch and Christen 2013), although these are likely a cornerstone technology for integrated micro- and nanofluidic devices with electrochemical detection systems.

6 Conclusions

Electrochemical detection techniques in micro- and nanofluidic devices have matured, and their transition from the relatively large-scale conventional electrochemical cells to the micro- and nanofluidic scale has only required adaptation of existing techniques. The challenges have been electrode fabrication and maintaining good signal-to-noise detection ratios despite shrinking sample sizes and electrode proximity. Advantages of adapting electrochemical detection to micro- and nanoscale include better sensitivity, the ability of a direct electronic readout, simpler methods, more portable and cheaper systems, and less chemical modification of analytes compared to optical detection methods.

AD is by far the most popular electrochemical detection method used in microfluidic devices. Many analytes have been detected by EC–AD. EIS has proven to be a powerful analytical tool in detection of biomolecules such as antigens and DNA fragments, which bind specifically to a recognition element. EIS and CV have been used to characterize both electrodes and analytes.

Electrochemical microfluidic devices can be based on various electrode materials, depending on the relative importance of great sensitivity and selectivity, or cheap and fast fabrication. Different surface modifications such as dielectric coatings and conducting polymers increase the options to tailor a device for a specific purpose. Integrated devices that combine all components of an electrochemical micro- or nanofluidic device are being developed, as well as devices with simpler microfabrication. Both of these trends bring the existing electrochemical micro- and nanofluidic analysis techniques closer to real-world applications.