Determination of inorganic ions using microfluidic devices

The separation and detection of inorganic ions on microfluidic devices has received little attention since the ‘lab‐on‐a‐chip’ concept has revolutionised the field of electrokinetically driven analysis. This review presents a summary and discussion of the published literature on inorganic analysis using microfluidic devices and includes sections on electromigration separation methods, namely isotachophoresis (ITP), capillary electrophoresis (CE), and hyphenated ITP‐CE, together with a brief account of flow injection analysis. The review concludes with the authors' perspective on future directions for inorganic analysis on microfluidic devices.


Introduction
The ingenious concept of total analytical systems (TAS), in which all the necessary steps in an analysis are integrated, including sample preparation, analytical separation, detection and signal evaluation and quantification, was introduced in the 1980s [1]. Through the 1990s, this concept was developed into a miniaturised form using a microfluidic chip platform, and has been described by several different terms such as 'miniaturised total analytical systems' (mTAS), 'microTAS', 'Lab-on-a-chip', or 'CE-on-achip' [2][3][4][5]. The potential benefits of the lab-on-a-chip approach are enormous, following the general principle that small-scale processes consume less time, samples, and reagents [6]. These benefits include cost savings, increased throughput from shorter run times and the possibility of parallel processing, reduced quantities of hazardous materials and associated waste and improved portability as a result of smaller feature size and lower power demands. Such chips usually incorporate a separation technique and consist of an interconnected network of micrometer-sized microfluidic channels filled with liquids. To date, most separations are electrically driven because of the ease of fluid manipulation that exploits the electroosmotic flow (EOF). At this stage, controlled pressuredriven flow in microfluidic devices is technologically too challenging to permit practical implementation of chromatographic techniques.
The majority of work carried out on microfluidic devices has involved the life sciences. Applications have included the separation of proteins and amino acids [7], highthroughput DNA analysis [8], cell culture and handling [9], clinical diagnostics [4], immunoassays [10], and pharmaceutical analysis [11]. Several lab-on-a-chip devices have been commercialised, such as for DNA, RNA, and protein analysis. In addition to the applications in the life sciences, chip-based electrophoresis is gaining acceptance in environmental analysis [12] and there is increasing interest in the analysis of warfare-related chemical agents [13].
As mentioned above, most research activity on microfluidic devices has been associated with biological applications. In comparison, the determination of inorganic ions on chips has attracted considerably less attention. Most of the published work deals with isotachophoresis (ITP) or ITP hyphenated with capillary electrophoresis (ITP-CE). In recent years, there has been a significant number of reviews dealing with electrokinetically driven separations on microfluidic devices, but none having a primary focus on the determination of inorganic ions on chips. However, several reviews have covered this area to a limited extent. In 2003, the review of Kaniansky et al. [14] focused specifically on electrophoretic separations on chips with hydrodynamically closed separation systems. In 2004, Vilkner et al. [15] reviewed recent developments in mTAS but had limited coverage of the analysis of inorganic ions. Previous reviews involving electrokinetically driven inorganic determinations have tended to focus on electrochemical detection, which appears to be the method of choice for the detection of inorganic ions given the limited choice of suitable alternative detection methods [16][17][18][19][20]. While reviews [16] and [17] are restricted to CE, reviews [18][19][20] each have a small section devoted to the determination of inorganic ions on chips.
The present review focuses on determinations of inorganic ions using electrokinetically driven separations on microfluidic devices. The group of inorganic ions will be termed 'target analytes' throughout this review. It should be pointed out that the analytical potential of the determination of anions is far more significant than that of cations, for which there are numerous competing alternative techniques (mainly spectroscopic in nature) available. For this reason, while the general discussion of individual separations is arranged in chronological order, the corresponding summary tables have been arranged in the order of determination of anions, followed by cations and finally simultaneous determination of both.

Specific aspects of inorganic determinations on chips 2.1 Type of microdevices used
Chips or microfluidic devices are planar devices usually made using microfabrication technologies applied to glass or polymers [6]. The devices are normally constructed from two plates, which are bonded together and have dimensions typically from 2-10 cm in both width and length and heights between tens of mm and several mm. Usually the plates contain open microchannels and wells, and sometimes also electrodes for electrochemical detection. After bonding of the plates, sealing of the channels occurs and liquids are introduced through the wells. The electrodes for performing the high-voltage separations are normally positioned in the wells. The most common separation mode is CE, but electrochromatography (CEC) is becoming more common as a direct outcome of recent advances in the synthesis of monolithic stationary phases [21,22].
Readers requiring coverage of some basics and the current state of the art for microfluidic chips can consult some excellent reviews. Reyes et al. [6] have reviewed the history, theory of miniaturisation, and technologies associated with mTAS. Becker and Locascio [23] have discussed the material properties, fabrication techniques, and applications of microfluidic devices made from polymers. Tanyanyiwa et al. [20] and Vandaveer et al. [24] have published works describing electrochemical detection. The former focuses on conductometric and potentiometric detection, while the latter deals with amperometry.
The choice of substrate for the manufacture of microfluidic chips suitable for the determination of inorganic ions should reflect their specific chemical properties, together with pragmatic considerations of cost, availability, and ease of fabrication. The target analytes are typically hydrophilic ions, and they all exhibit ion-exchange interactions to varying degrees [25]. Therefore, the presence of charged silanol groups on glass can introduce ion-exchange adsorptions onto the separation channel wall, which could lead to problems such as broadening and tailing of analyte zones, poor reproducibility, or memory effects [26]. Microfluidic devices were initially made from glass or quartz using photolithographic techniques that are expensive and labour-intensive [27]. More recently, there has been a strong trend to choose polymers, especially poly(dimethylsiloxane) (PDMS), and poly(methylmethacrylate) (PMMA). Once a master template had been prepared, multiple low-cost, highly reproducible copies can be manufactured using methods such as casting for PDMS and hot-embossing or injection moulding for thermoplastic polymers. So for reasons of analyte properties and pragmatic considerations of cost and ease of manufacture, polymer devices are preferred over glass chips for inorganic determinations.

Detection methods
Electrochemical detection methods, including amperometry, potentiometry and conductometry are, in principle, all amenable to chip-based determinations of inorganic ions. Amperometry, which involves measuring a current originating from a redox reaction of the analyte on a suitable electrode, is limited to the detection of redox-active species only. Potentiometric detection, which involves measuring the Nernstian potential as the analyte passes a suitable sensing electrode, is typically instrumentally demanding and so far has not been established as a routine detection method in CE or on microfluidic chips. Conductometric detection, which involves measuring the conductivity of the solution containing the analyte, is a universal detection technique requiring a relatively simple experimental setup. Recently conductivity detection has become an attractive method for inorganic determina-tions on a chip [20]. Conductometric detection has been used both in the 'contact' mode (having the electrodes in physical contact with the solution [28,29]), and the 'contactless' mode (where the electrodes are not in physical contact with the solution [30,31]).
Optical detection methods for the determination of inorganic ions have not been widely used on chips. Typically, the target analytes do not absorb in the UV region, neither do they exhibit fluorescence. Laser-induced fluorescence is often used on chips but only for the detection of analytes that exhibit native fluorescence or are easily labeled with fluorescent tags. Unfortunately, this is not applicable for most inorganic ions. Indirect fluorescence has been used in CE with varying degrees of success, but in principle it is unfavourable to use fluorescence under conditions where a relatively high background is present, as the lowest possible noise and consequently the best possible signal to noise ratio is obtained for a nonfluorescent background. Given that most inorganic ions are UV transparent, direct photometric detection is not an option. Indirect photometric detection is widely used for the detection of the target analytes in capillaries, but in chips integrated photometric detection has not been sufficiently developed to permit it to be applied on a routine basis [32]. The bottleneck is in the integration of miniaturised light sources and detectors on the microfluidic device and although this has been achieved by a limited number of groups, the process is expensive to implement [33].

Determination of ions by ITP on chips
ITP is an electroseparation technique which is historically older than other methods and in many analytical applications has been replaced by CE [34]. ITP is unique in that the sample is placed in a discontinuous electrolyte system. As a result of the varying electrical field strength in the electrolyte, analyte ions are separated into zones with sharp boundaries which travel at the same speed, hence the term isotachophoresis [35]. Zones are usually detected by sharp changes in the signal response [36]. The main advantage of ITP is its ability to preconcentrate trace amounts of analytes and to separate these from undesirable components of the sample matrix. Therefore, usually very little sample preparation is needed [37].
The degree of concentration that occurs during the ITP process is given by the Kohlrausch regulating function, where c A is the final concentration of the analyte, c L is the concentration of the coion in the leading electrolyte and m A , m L , and m R are the mobilities of the analyte ion, co-ion in the leading electrolyte, and counter-ion in the leading electrolyte, respectively [35]. In practice, the concentration of dilute ions may be increased by factors of 1000-fold [34]. Conversely, ions in the sample matrix with concentrations higher than that of the leading ion will tend to be diluted.
The first work involving ITP on a chip was performed by Boč ek et al. [38]. They were able to overcome some of the previously reported problems associated with excessive void volumes and with Joule heating in strong electrical fields by performing the separation on a thermostatically controlled Perspex block. A rectangular 200 mm deep channel was milled into the block and sealed with poly-(tetrafluoroethene) (PTFE) foil. The device was clamped with the foil side onto a thermostatted metallic plate for cooling. Potential gradient detection was performed by monitoring the potential drop over two detection electrodes in the channel using a high input-resistance voltmeter, an arrangement often used in ITP. Potential gradient detection has the advantage of being a universal detection method, but it has only a relatively low detection sensitivity. Separations were reported involving anions of some carboxylic acids, chlorate, bromate and a range of phosphorus-containing species in times less than 5 min for concentrations between 0.1 and 0.8 mM.
A more sophisticated microfluidic device using ITP was reported by Fielden et al. [39] in 1998. The planar device was made by casting PDMS on an etched copper mould, and bonding of the PDMS slab to a circuit board, containing gold-plated electrodes for conductivity detection.
Hydroxyethylcellulose was added to the electrolyte to reduce the EOF. The chip was used to demonstrate the separation of three transition metal ions at a concentration of 10 ppm. The same group also reported the separation of anionic dyes and fluoride using a similar device at the same conference [36]. In 1999, Prest et al. [40] developed another PDMS chip with single-electrode conductivity detection and demonstrated its use for the separation of a mixture containing Na 1 and K 1 ions. Later, the performance of the PDMS device with integrated single-electrode conductivity detector was compared with that of a conventional capillary ITP system [41]. They found that there was good agreement between the relative step heights for the metal ions: Li 1 , La 31 , Dy 31 , and Yb 31 measured for the chip and the capillary device, but that the capillary system gave better reproducibility and sharper steps in the isotachopherogram. The reduced sensitivity in the microfluidic device was probably due to lower sensitivity of the single-electrode detector compared with the two-electrode conductivity detector used in the capillary setup. The manual injection procedure followed for injections on the chip probably affected the efficiency and reproducibility of the separation system. However, the chip was able to perform the analysis in less than half the time of the capillary system due to its shorter separation channel and better heat dissipation. The same group later developed a PMMA device for bidirectional ITP, illustrated in Fig. 1A, which was used for simultaneous determination of anions and cations migrating in opposite directions. The chip incorporated two-electrode conductivity detectors at either end of the ITP channel and valves to control the hydrodynamic transport of solutions, addressing the issues with the system that had been used before. The pressure injection made it much simpler to introduce the sample than if electrokinetic transport had been used; otherwise only ions with a positive or negative charge would have been introduced.
In the first step, the anolyte (anode electrolyte) was introduced into reservoir A and the catholyte (cathode electrolyte) into C. These flowed along the separation channel until they met at the injection cross where they eventually moved to the waste. After the sample was introduced at D, electrode A was made positive and electrode C negative. Contact conductivity detection electrodes near A and C registered the zones of anions and cations, respectively. Figure 1B shows the separation of ions from a mixture of (NH 4 ) 2 SO 4 , NH 4 NO 3 , NaF, and Li 2 CO 3 . Critical to the success of this technique was the selection of the anolyte and catholyte. The anion of the anolyte served as the leading electrolyte (LE) for the separation of anions and its cation served as the terminating electrolyte (TE) for the cation separation. In the same way, the cation of the catholyte served as the LE for the cations and its counteranion as the TE for the anions. The group used 10 mM HCl buffered to a pH of 3.6 with glycylglycine as the anolyte and 12.5 mM Cs 2 CO 3 as the catholyte. This enabled them to separate a mixture containing 10 mM of each of the above-mentioned salts [42].
More recently, Prest et al. [43] developed a technique for the determination of the inorganic arsenic III (arsenite, AsO 2

2
) and arsenic V (arsenate, AsO 3   2 ). The analysis was performed at pH 9 so as to maximise the effective charge of AsO 2 2 (for HAsO 2 , pK a 9.23). Glycine, which has a lower mobility than AsO 2 2 , was used as the TE [44]. They also developed a method of determining Se IV O 3 2- (Table 3). It seems that most research groups who have conducted ITP on a chip have used it as a stepping-stone to an inline hyphenated ITP-CE. In this way, the sample preconcentration and matrix elimination benefits of ITP can be combined with the high separation efficiency and peak capacity of CE.  The electrolyte was at a low pH to enable the separation of Se(IV) and Se(VI) which have different degrees of dissociation of their acids.

Hyphenated ITP and CE on chips
As an analytical technique, ITP has experienced a revival as a sample in-line pretreatment hyphenated with CE. Figure 2 is a schematic diagram that compares ITP, CE, and ITP-CE in order to demonstrate the advantages of the hyphenated technique for the separation of anions. In ITP the sample is injected between the leading electrolyte (LE) and the terminating electrolyte (TE). When the high voltage is applied, the ions in the sample arrange themselves into Preliminary work on single electrode conductor. HEC was used to suppress EOF. [40] HIBA, 2-hydroxyisobutyric acid a) Concentration of HEC is given as % w/v. Gly-Gly, glycylglycine; ADA, N-(2-acetoimido)iminodiacetic acid a) Concentration of HEC is given as % w/v. migrating zones dependent on their electrophoretic mobilities (m). In Fig. 2, the target analytes A, B, and C migrate in the sequence shown because m C . m B . m A . m TE . As shown in the plot of the detector response for B, if a zone is very narrow, the analyte may be difficult to quantify accurately. In CE, the sample is injected into a single electrolyte, optimally having a higher concentration than the analytes. The analytes undergo sample stacking and peak focusing unless this process is impeded by higher concentration such as various matrix ions. As shown in Fig. 2, the matrix ions can mask the analyte peaks and prevent their detection. In the hyphenated ITP-CE system shown, the separation channel divides into two parts at the T-intersection. The ITP channel continues at a right angle to the original channel and the CE channel continues straight after the intersection. A detector (labelled 'ITP detector' in Fig. 2) is located just before this intersection. Once the first analyte of the ITP separation is detected at this point, the high voltage is switched from the ITP channel to the CE channel. In this way only the analyte zones continue to the CE channel where they separate into sharp peaks according to their mobility. It is possible to use the same electrolyte for the TE and the background electrolyte (BE) but their concentrations need not be the same.
There are two major benefits associated with this hyphenation when compared with either technique used separately. First, ITP generally possesses a much greater sample loading capacity, typically in the order of mL rather than the nL sample volumes used in CE. Second, the concentration of analytes can be increased by a factor of up to 10 4 in the ITP step [34,46,47] while the concentration of matrix ions may be decreased. The progress of these analytes through the ITP capillary or channel on a chip device may be monitored using conductivity detection. By judicious timing and switching of the electric field it is possible to divert the analytes of interest into the CE column for further separation and detection. This is particularly useful when the analytes are present at a much lower concentration than the sample matrix, which is often the case in practical samples, such as food and beverages [48]. Hyphenated ITP-CE can therefore be used to achieve sample preconcentration and cleanup (matrix removal), leading to dramatically decreased limits of detection (LODs).
The group of Kaniansky [49] has recognized that improved precision could be achieved in ITP by eliminating both the EOF and the hydrodynamic flow (HDF) caused by siphoning effects resulting from slight differences in pressure. HDF is a particular problem in channels with larger cross-sectional areas, which have a lower resistance to flow. Suppression of EOF was achieved by adding methylhydroxyethylcellulose to the electrolyte or by using it to coat the channel walls. HDF was eliminated by designing a chip with electronically controlled valves and driving electrodes that were in contact with a semipermeable membrane. This prevented the entrance of bubbles produced by electrolysis into the separation channels. Both factors improved the reproducibility by a factor of 30. Their second innovation was to couple an ITP channel with a CE channel on the same chip. The ITP column was used to preconcentrate the sample constituents and to separate them from the more concentrated ions present in the LE. As explained earlier, the sample matrix and LE were "shunted" while the sample analytes continued to the CE channel where further separation and detection were performed. This technique of column coupling of the separation channels provided much lower LODs (around 10 mM for a number of inorganic and organic anions) than had been reported previously for ITP on a chip. Figure 3A shows a schematic diagram of the hyphenated ITP-CE chip [50]. The reservoirs are filled with the appropriate electrolytes using the electronically controlled peristaltic pumps. The CE separation channel, SC2 is filled with the background electrolyte. Excess reagents are allowed to flow to the waste. Similarly, the ITP separation channel, SC1, is filled with the leading electrolyte before the sample is introduced into the sample injection channel. All the valves are closed during the electrophoretic separations. The high voltage is applied to an electrode in the end of the TE channel and the LE reservoir is grounded. ITP zones migrate towards the LE until the first analyte is detected by the contact conductivity detector CD1. The reservoir for the background electrolyte BE is now grounded and the analytes migrate along the CE separation channel SC2 and are detected by conductivity detector CD2. Figure 3B shows an electropherogram of a mixture of anions using the column-coupled chip. Masar et al. [50] used a similar column-coupled device to determine the organic acid anions and inorganic anions present in red and white wines. The chip was used to resolve three inorganic anions: SO , and 11 carboxylate ions in less than 14 min. Bodor et al. [51] applied the same approach to determine BrO 3 2 in drinking water and achieved a very impressive limit of detection of just 20 nM. The devices were also used for the separation and detection of NO 2 2 and PO 4 3-. Recently, Kaniansky et al. [14] have published a review of electrophoretic separations on chips with hydrodynamically closed separation systems, which specifically focused on the optimisation of ITP and ITP-CE systems on column-coupled microfluidic devices of the type shown in Fig. 3. In addition to reviewing their earlier work, new data were presented on the ITP-CE separation of NO 2 2 , F 2 and PO 4 3in the presence of 12 organic ions in just over 7 min, and on the ITP-CE separation of BrO 3 2 , ClO 2 2 , PO 4 3-, F 2 and mono-, di-, and trichloroacetate in chlorinated tap water.
Baldock et al. [52] used integrated conducting polymer electrodes in microfluidic devices made from Zeonor (an amorphous norborene-ethene copolymer) and from polystyrene to separate the anions NO 2 2 and F 2 and the cations Mg 21 , Ca 21 , Mn 21 , Ni 21 , Zn 21 , La 31 , Nd 31 , and Gd 31 . The electrodes were made from nylon or poly(styrene) impregnated with 40% graphite. Tables 4 and 5 provide summaries of published methods using hyphenated ITP-CE for the determination of anions and cations, respectively.  Review article on hydrodynamically closed separation systems that also includes new data. [14] MHEC, methylhydroxyethylcellulose; His, histidine; BTP, bis-trispropane; HDF, hydrodynamic flow; n.a., not available a) Concentration of MHEC and HEC given as % w/v was used to suppress EOF. a) Concentration of HEC given as % w/v was used to suppress EOF.

Capillary electrophoresis on chips
From the 1980s onwards, CE has been established as a powerful separation method [53]. In 1983, Gebauer et al. [54] demonstrated the feasibility of CE in a Perspex ITP device with 2 mm wide, 0.2 mm deep rectangular channels. They successfully separated a mixture of 1 mM nitrate, chloride, sulfate, and nitrite in the 20 cm long channel and then performed a similar analysis using a 0.50 mL sample of drinking water. Harrison et al. [55] introduced CE in a chip format and the first determination of inorganic ions on a microfluidic device was performed by Jacobson et al. [56] in 1995. The ions, Zn 21 , Cd 21 , and Al 31 were first complexed with 8-hydroxyquinoline-5-sulfonic acid (HQS) before the separation was carried out in an electrolyte that also contained the complexing agent. They used laser-induced fluorescence (LIF) to achieve detection limits of less than 1 mM for both Zn 21 and Cd 21 .
In 1997, Kutter et al. [57] coupled CE separation and fluorescent derivatisation of Mg 21 and Ca 21 on a glass chip. The ions were also complexed with HQS but this time the reaction occurred after the electrophoretic separation, about two-thirds of the way along the separation channel, before being detected by LIF. LODs for Mg 21 and Ca 21 of and 0.02 mM and 0.045 mM were achieved. Liu et al. [58] used post-separation chemiluminescence detection to monitor the separation of a mixture of transition metal ions. Luminol was added to the separation channel via a conjoined channel. The separated cations catalysed the oxidation of luminol, producing light that was detected using a photomultiplier tube mounted on an inverted microscope. They achieved a LOD of 0.493 mM for Co 21 .
An unconventional microfluidic device, consisting of a fused-silica capillary mounted on top of a microscope slide, was used by Huang et al. [59] for the CE separation of Co 21 and Cu 21 followed by end-column chemiluminescence detection. Luminol was present in the background electrolyte, whereas H 2 O 2 was added in the reservoir. The presence of Co 21 and Cu 21 in the capillary effluent catalysed the luminescent reaction of luminol with peroxide, allowing end-column detection of these analytes.
Most of the research carried out on determination of inorganic ions on chips has involved electrochemical detection. There have been three major research groups that have worked in this area using CE since the pioneering work by Jacobson et al. One group, working at the Delft University and the University of Twente have investigated the microfabrication of glass microfluidic devices, the fabrication of contact-and contactless-conductivity detection systems and testing of the finished product [28,[60][61][62][63]. Another group at the University of New Mexico has taken particular interest in the use of contactless conductivity detection for the analysis of warfare agents [30,[64][65][66]. Finally, a group at the University of Basel has undertaken considerable research in the area of contactless conductivity detection [20,67].
Guijt et al. [28] reported a number of innovations for electrophoretic chip technology. Two new microfabrication techniques were trialled for the fabrication of CE chips. Miniaturised transparent insulating channels (mTIC) with a wall thickness of just 380 nm were constructed using silicon-rich nitride. These had excellent thermal properties and very low EOF, but tended to be very fragile. Powder blasting was used as an alternative to wet etching of channels in Pyrex  glass. Both the mTICs and powderblasted devices were equipped with integrated conductivity detectors consisting of two Pt electrodes and used to analyse mixtures of alkali metal ions [63]. The conductivity detection system proved successful but the efficiency of the CE separation was limited. Traditional wet etching methods were used for the fabrication of the microchannel in a glass microdevice for 4-electrode capacitively coupled contactless conductivity detection (CCD), used by Guijt et al. [60]. The detection electrodes were deposited in recesses, and covered with a 30 nm thick film of silicon carbide to isolate the electrodes from the separation voltage using a sophisticated microfabrication process [61]. This provided a better LOD than for the previous devices. To demonstrate the superiority of the 4-electrode conductivity detection, the same device was tested using the detector in the 2-and 4-electrode mode. The latter had a more stable frequency response and gave improved sensitivity. LODs from 5-15 mM were obtained for the alkali metal ions tested [62]. Further improvements in fabrication methods and read-out electronics enabled each of the alkali metals to be detected at levels of 5 mM [61] .
Pumera et al. [30] were the first to demonstrate the determination of anions on electrophoretic chips. Their solution for isolating the electrodes from the electrical field was to glue them onto the cover of a PMMA microdevice. This greatly simplified the manufacture and allowed them to use much stronger electrical fields to analyse ions in as little as 10 s. They were able to achieve LODs of down to 5.1 mM for anions and 1.2 mM for cations. Determinations of oppositely charged ions could be carried out sequentially. Similar PMMA chips were used to carry out fast analyses of low explosive ionic components. Wang et al. [64] recognized that the low EOF for the PMMA device facilitated simultaneous detection of anions and cations using the same background electrolyte. This was achieved simply by switching the polarity of the field once all the cations had been detected. As one would expect, they found that as the electrical field strength was increased, there was a decrease in the time taken to separate both anions and cations and an improvement in the number of theoretical plates. Wang and Pumera [65] developed a dual contactless conductivity and amperometric detection system on a glass chip. The conductivity detector was constructed as described previously and the amperometric system used a thick film carbon electrode, which was screen-printed onto the end of the channel [68] (see Fig. 4A). The two detection methods worked quite independently and made it possible to simultaneously analyse inorganic explosives and organic explosives. Figure 4B shows the electropherogram of a mixture of the inorganic and organic analytes. It was also pointed out that the two independent signals could act as a fingerprint for electroactive ions to help in identification of explosives. This was illustrated using a mixture of nitrophenols. The run-to-run injection times were deliberately changed to demonstrate the independence of the response ratio (amperometric signal/conductivity signal) for p-aminophenol.
More recently, Wang et al. [66] developed an electrophoretic microchip with dual-opposite injection for simultaneous measurements of anions and cations. As illustrated in Fig. 5A, dual injection was carried out by giving the sample reservoirs at opposite ends of the channel different polarities. When the sample was loaded at the two injection crosses, the voltages were switched, and the separation potential applied over the separation channel (see Fig. 5B). The anode was placed at the cation injection side of the channel, and the cathode at the anion injection side. The cations in the sample injected at the anodic side, and the anions injected at the cathodic side migrated in opposite direction through the channel. Detection of ions was carried out using a movable conductivity detector placed near the middle of the separation channel where both the cations and anions would pass. The position of the detector could be adjusted to optimise the apparent detector selectivity for a simultaneous separation of cations and anions. This is required to prevent overlapping peaks originating from cations and anions passing the detector simultaneously in opposite direction. Figure 5C shows an electropherogram for a mixture of anions and cations separated on this chip. The detection electrodes were constructed separately to the microfluidic device and this reduced the costs associated with microfabrication of the device.  that could be moved to sections where washing, flushing, filling or surface treatment of the channels could be carried out. Electrophoresis was conducted in the final posi-tion. Contact between the chip and external devices such as pumps or the high-voltage source were made with purpose-built arms at each position. The PMMA chip was characterized using a mixture of alkali metal ions.
Tanyaniwa et al. [20] transferred their CCD technology from capillaries to microfluidic devices. They found that the sensitivity of CCD could be improved by using a high excitation voltage for the actuator electrodes. They demonstrated this technique on a PMMA chip for a range of inorganic cations and also improved the CCD technique further by mounting their chip in a holder and attaching the electrodes to the holder (see Fig. 6). This greatly simplified production of the chips [67]. The efficiency of the new design was demonstrated for a range of inorganic and organic anions, alkali metal ions, transition metal ions and amino acids.
A range of further interesting approaches to CE of inorganic ions on a chip have appeared. Da Silva et al. [70] presented a novel method of fabricating microfluidic devices based on laminating laser-printed polyester films. They demonstrated its performance using a mixture of alkali metal ions. Kikura-Hanajiri et al. [71] developed a technique for the indirect detection of nitric oxide production using a glass microfluidic device. NO (g) is unstable in biological systems and is rapidly oxidized to a mixture of NO 2 2 and NO 3

2
. The reaction can be monitored by on-chip reduction of the NO 3 2 to NO 2 2 and then using amperometric detection following the CE run. Cd particles were placed in the sample reservoir, flushed with CuSO 4 , and used for the reduction of NO 3

2
. The NO 2 2 produced from this reaction was detected amperometrically with a LOD of 1.0 mM. Recently, Girault and co-workers [72] described the use of passive conductivity detection on a poly(ethlyleneterephthalate) (PET) chip. Rather than isolating the detector from the field, the separation field was used to induce a potential difference across two electrodes placed perpendicular to the separation channel. This voltage was measured using a high-resistance voltmeter, and is proportional to the resistance of the electrolyte between the electrodes. Electrophoretic zones produce corresponding drops in the measured potential difference. The characteristics of this device were modelled mathematically and tested using a range of alkali metal ions. Passive conductivity detection, however, is less sensitive than active conductivity detection employing an AC signal applied over the detection cell. Table 6 summarises publications describing CE separations of inorganic anions performed on a chip, Table 7 the determination of cations by CE on a chip and Table 8 the simultaneous determination of cations and anions.  [67] Reprinted from [67], with permission.

Ion chromatography on a chip
Murrihy et al. [73] have been the only researchers to date to perform ion chromatography (IC) on a microfluidic de- vice. An open-tubular microseparation column was fabricated by coating the wall of the channel with functionalised anion-exchange latex nanoparticles. These positively charged nanoparticles adhered strongly to the wall due to electrostatic interactions with ionised silanol groups on the channel wall. Injection and detection of a mixture of NO 2 2 , NO 3 2 , and I 2 was performed off chip. Connections to an HPLC pump and spectrophotometric detector were made using fused-silica capillaries, which were inserted into the ends of the channel. The chip per-  Separation of tryptophan, phenylalanine, threonine, and tyrosine is also presented.
[67]  , NO 3 -and two nerve agent degradation products also presented [66] formance was compared with that of a similarly coated fused-silica capillary. The chip had a higher ion-exchange capacity due to the narrower depth of the channels, which enabled greater interactions between the analytes and the coated stationary phase. The LODs were comparable to those achieved by CE on microfluidic devices. Details of this method are summarised in Table 9.

Continuous-flow analysis on a chip
Continuous-flow analysis (CFA) and flow-injection analysis (FIA) are not classified as separation techniques, but have been used for determination of inorganic ions on microfluidic devices. In CFA, sample and reagents are mixed in a continuous flow until a chemical equilibrium or a so-called steady-state condition is obtained. At the detector (optical, electrochemical, etc.), this steady state is monitored as a plateau. In order to distinguish between different samples, air bubbles were introduced to segment the flow in between different samples [74]. The introduction of FIA in 1975 revolutionised CFA by injecting a well-defined sample volume into the carrier flow [75].
The reproducibility and precise timing of the sample introduction and subsequent manipulations in the system result in 'controlled' dispersion of the sample zone. This results in a reproducible sample concentration gradient at the point of detection, eliminating the need for achievement of complete chemical equilibrium before detection. The recorded detector signal is a peak of which the Connections from the inlet and outlet to the pump and detector were made using capillaries. [73] area corresponds to the analyte concentration in the sample [74]. A brief description of CFA and FIA applications for the analysis of inorganic ions on microfluidic devices is included for the sake of completeness of this review.
In 1995, Daykin et al. [76] introduced a planar microfluidic device for FIA of orthophosphate using colorimetric detection of the blue PO 4 3molybdenum complex. A schematic drawing of the microfluidic device is given in Fig. 7. Once channel AB has been filled electrokinetically with molybdate and ascorbate, a discrete sample volume is loaded into channel AB by a potential difference between C and D. Once the field between A and B is reestablished, the phosphomolybdate complex is formed. Optical fibres were used to irradiate the channel with light from an light-emitting diode (LED), and measure the absorbance using a photodetector. A similar device was used by Greenway et al. [77] for the determination of NO 2 2 . Here, detection was based on the reaction of nitrite with sulphanilamide to for a diazonium salt, that yields an azo dye when coupled with N-(1-naphtyl) ethylene diamine. The two reagents were mixed in reservoir A, channel AB was electrokinetically filled with the reagent mix prior to introduction of sample by a potential applied between C and D. After re-establishement of the potential difference between A and B, the NO 2 2 introduced with the sample reacted with the reagents, and the formed product was monitored using a microspectrophotometer.
Vahey et al. [78] used multiwavelength grating light reflection spectroscopy (GLRS) to determine PO 4 3in a buffer solution. This technique monitors the light reflected from a transmission diffraction grating in contact with the sample and relies upon the attenuation of the sampling beam and the loss of coherence of the penetrating eva- . Initially, a potential difference is applied between A and B, resulting in migration of molybdate and ascorbic acid. After 2 min, molybdate and ascorbic acid meet between D and B. A potential is applied between C and D to load a discrete sample amount in channel AB. Subsequently, the field between A and B is restored. The orthophosphate present in the sample reacts with the two reagents, and the light absorption of the phosphomolybdate complex is monitored using an LED and photodetector. Drawing made based on information in [76]. nescent wave relative to the wave that originates at the grating. Evanescent waves are transmitted waves associated with total internal reflection that travel along the interface between the two media [79]. The GLRS signal is a combination of responses based on refractive index and absorbance and is both concentration-dependent and path-length-independent, making it an attractive detection technique for mTAS. Sensitivity has been demonstrated to be on the order of 2610 26 refractive index units (RIU), depending on the detection system resolution. The LOD for PO 4 3- , the bright yellow vanadomolybdophosphoric acid is formed with an absorbance maximum at approximately 370 nm. The microdevice was equipped with a UV-LED which had its maximum output at a very similar wavelength. The mixed reagent was shown to be stable over a 12-month period and a LOD of less than 4 ppm was achieved for PO 4 3- [83]. The same group presented a similar system using laser ablated PMMA devices [84]. Here, a LED and a silicon photodiode were used for detection, resulting in a small, portable system.
Recently, Tyrell et al. [85] designed a PMMA microfluidic manifold for the chemiluminescent detection of Cu(II). The Cu 21 ions catalysed the oxidation of 1,10-phenanthroline by H 2 O 2 in the presence of CTAB and NaOH. By optimising the concentrations of each reagent, an LOD of less than 0.02 mM was achieved. Fujii et al. [86] developed a method for the simultaneous detection of SO 3 2and NO 2 2 on a microfluidic device using a novel fluorescent detection unit. The derivatised samples were irradiated with a continuous range of wavelengths and a fluorimeter was used to detect the different signals coming from the analytes. LODs were 1.0 mM and 0.4 mM for SO 3 2-and NO 2 2 , respectively.
The Berthelot reaction was used by Daridon et al. [87,88] and by Tiggelaar et al. [89]  Sato et al. [92] used thermal lens microscopy as a novel detection method for CFA to achieve the ultrasensitive determination of nonfluorescent ions. Sample and o-phenanthroline were simultaneously introduced into a Y-shaped channel. The product of the chelation reaction was monitored using photothermal microscopy. A calibration plot for Fe 21 concentrations between 4 and 20 mM was presented. Using a similar device, Hisamoto et al. [93] introduced the use of a neutral ionophore for ion-pair extraction. An organic phase containing the ionophore and a lipophilic pH indicator dye and a nondissociated acid and an aqueous phase containing the sample are pumped into the two arms of the Y-shaped channel. If the two flows meet, the laminar flow forces the organic and an aqueous phase to flow alongside each other. Ions are selectively extracted from the aqueous phase, and replaced by H 1 ions formed by dissociation of the acid. The complexation of the ionophore with the dissociated dye was measured by thermal lense microscopy. The same research group extended this research by having intermittent zones of two different ionophores and >empty' organic phase [94]. This way, Na 1 and K 1 could be analysed simultaneously.
The same group presented a microfluidic device for continuous flow wet analysis of Co 21 including a chelating reaction, solvent extraction, and purification [95]. Detection was again carried out using thermal lens microscopy, and the samples and reagents were introduced using syringe pumps. Co 21 from the sample forms a chelate with 2nitroso-1-naphthol and is extracted together with the other metal chelates into the organic phase. The excess of 2nitroso-1-naphthol is removed with the aqueous phase. A stream of HCl is added at one side of the organic flow, and a stream of NaOH at the other side. The Co 21 chelate is the only stable chelate at low and high pH, and will therefore be the only remaining complex that can be detected. A review on the CFA on microfluidic devices by the Kitamori group can be found in [96]. Tables 10 and 11 show details of chipbased CFA methods for anions and cations, respectively.

Outlook
A number of general conclusions can be drawn from this review. First, plastic microfluidic devices will have a bright future due to their favourable physical properties and their ease of fabrication, including applications to determination of inorganic ions. Second, until significant technical improvements occur in miniaturised optics, electrochemical detection will remain the preferred detection scheme for the target analytes.   [86] NED, N-(1-naphtyl)diamine; NAM, N-(9-acridimyl)maleimide; DAN, diaminonapthalene to become routine for quality control and environmental monitoring. Contactless conductivity detection using external electrodes mounted on the microfluidic device holder will be recognised as a simpler and more cost-efficient method of microfabrication. Third, FIA on microfluidic devices has already made a significant contribution to the field of inorganic determinations on a chip. It appears to be particularly suited to the detection of indi- DD16C5, 4a,20a:11a,15a-dibutano-17H-dibenzo[b,k] [1,4,7,10,13]pentaoxacyclohexadecin,tetradecahydro-(9CI); DB18C6, dibenzo 18-crown-6; KD-A3, N-2,4dinitro-6-octadecyloxyphenyl-2',4'-dinitro-6'trifluoromethylphenylamine vidual analytes rather than to mixtures of unknown composition. Finally, there are significant technical challenges in performing integrated IC on chips and given that column IC is a well-established and robust technique offering excellent performance, it will be interesting to see if miniaturised IC becomes a competitive option.
The determination of inorganic ions on microfluidic devices will be significant into the future for a number of reasons. First, from a historical perspective, the target analytes have frequently been used to characterise new devices in their developmental stages and advances have led to wider applications. Second, increasing concern for state of the environment and for quality control in a number of industries will see the need for miniaturised fully integrated portable devices capable of rapid and sensitive detection of inorganic species on site and in the field. Finally, the increasing needs of forensic analysis and security issues will demand high-speed, portable, and reliable analytical methods. Developments in chemical analysis on microfluidic devices will continue to strive to meet each of these challenges.
Financial support from the Australian Research Council is gratefully acknowledged. RMG would also like to acknowledge financial support from the Dutch Foundation for Science and Technology STW (project DPC 6168).