Critical Review Development and Operating Characteristics of Micro Flow Injection Analysis Systems Based on Electroosmotic Flow A Review S. J. Haswell School of Chemistry, University of Hull, Hull, UK HU6 7RX Summary of Contents Introduction Fabrication Pump Design Injector Design Reactor Design Detector Design Future Trends Conclusion References Keywords: Micro flow injection; electroosmotic flow; miniaturisation; micro total analytical system ( mTAS) Introduction Over the past 50 years, the design of instrumentation for the measurement of chemical, biological and physical parameters has brought analytical chemistry to a highly automated and technologically advanced science.It would be naive, however, to believe that the chemistries and physics associated with present day analytical measurements, are in any way limited by current technology. On the contrary, laboratory based analysis has been almost exclusively designed, for the ergonomic needs of the human worker.There can be little doubt that this preoccupation with the physical size and operation of equipment, for laboratory based measurements, has significantly influenced the teaching and practice of analytical chemistry. If one considers the fundamental chemical and physical phenomena on which measurements are based, then it is apparent that one only requires a few molecules or atoms to be present for quantitative measurement to occur. In short, the physics and engineering of measurement science can be considered as oversized and in many cases over engineered to meet the needs of the chemical measurement.Whilst instrumental design has relentlessly moved towards automation, the realization that the quality of analytical data can be profoundly influenced by factors such as sampling, sample storage, sample pretreatment and matrix interferences has led to the concept of the total analytical approach. This more holistic view of analysis, in which traceability and uncertainty prediction can be carried out, is often embodied in the concept of the ‘total analytical system’ (TAS).1 There are, however, numerous disadvantages to TAS, including slow sample processing, lack of selectivity and high reagent consumption;2 however, many such problems may be overcome through system miniaturisation.The term micro TAS (mTAS) was first used at the Transducers 89 Conference.3 In its simples sense, mTAS involves the miniaturisation of all the functions found in an analytical method for example, pumps, valves, flow manifolds, mixing and reaction chambers, phase and analyte separation, detectors, control and communication electronics.One of the more exciting prospects of the mTAS concept is the suggestion that the entire chemical measurement laboratory could be miniaturised onto a device of a few square centimetres.4 This type of miniaturisation has become possible largely through the adoption of microfabrication techniques developed in general by the microelectronics industry, and one can consider miniaturisation of chemical reactors and their corresponding instrumental requirements to be in a similar position to the microelectronics industry 30 years ago.Interestingly, it was from the heart of the microelectronics industry that Terry et al.5 were one of the first to demonstrate that integrated circuit (IC) technology could be used to fabricate a miniature GC instrument using a 5 in silicon wafer.Although the work of Terry and co-workers served to point the way, it was not until the advent of capillary electrophoresis (CE)1,6–8 and in particular, the exploitation of electroosmotic flow (EOF), that the realization of mTAS came about some 10 years after the miniature GC work was first reported. Stephen J. Haswell is a Senior Lecturer in Analytical Chemistry at the University of Hull, UK. Following a period of five years working in the food and plastics industries, he undertook a Ph.D.in the area of atomic spectroscopy graduating in 1983 from the University of Plymouth. His current research activities are in the areas of mFIA, microwave enhanced reaction chemistry, trace elemental speciation and chemometrics. He is author of over 100 research papers, books and patents and presenter of numerous national and international lectures on his research activities. The subject of this present review has become a great interest of the author over the past five years.Analyst, January 1997, Vol. 122(1R–10R) 1RA significant research base has now built up in the area of mTAS9–14 and the development of mFIA over the past 5 years can be attributed to key research carried out by groups based in Basle, Switzerland,14 Texas Tech. University, USA,15 University of Alberta, Canada,16 the Oak Ridge National Laboratory, Tennessee, USA,17 and in a very modest way in the author’s own laboratory.18 What follows is a review of the features present in a relatively small, but rapidly expanding research base, which can be exploited to develop micro flow injection analysis (mFIA) systems, based on electroosmotic flow, to produce reliable sensor-type devices which incorporate mechanical and chemical robustness.Fabrication The fabrication of mFIA systems has attracted various approaches in recent years, ranging from manifolds based around CE-type fused-silica capillaries,1 to modular micro pump systems.19,20 This particular section, however, will focus specifically on fabrication methodology which leads to the production of a single integrated device, designed to use electroosmotic mobility for sample and reagent pumping.18 The discussion will focus almost exclusively on micro machined monolithic device fabrication, which adopts standard IC photolithographic, wet etching and bonding techniques to produce a planar structure.21,22 As the techniques used are based essentially on IC technology, they represent well established methodology, outlined schematically in Fig. 1. If one considers the current ‘state-of-the-art’ in IC fabrication, then it is clear that the challenge of producing mFIA manifolds can in no way be considered as technologically demanding. From what follows, it will become clear that mFIA device fabrication can be relatively simple and offer a reasonably high degree of flexibility in systems design. The choice of a suitable substrate into which the channels of a mFIA manifold can be etched is closely related to factors such as the fabrication method, the analytical chemistries involved and the proposed pumping, injection and detection systems.The obvious candidate given the close relationship with IC technology is silicon,23 which is amenable to the fabrication of structures in the nano- and micro-metre range. This crystalline material can be obtained in a very pure form, at relatively low cost, and offers good mechanical and chemical properties.Whilst silicon may seem an ideal microengineering candidate for mFIA fabrication, it is by nature a conductor and so surface modifications will be required if electroosmotic pumping of samples and reagents is to be achieved. This can be carried out by coating the silicon with SiO2 or Si3N4 to produce the required surface chemistries. Harrison et al.16 described the fabrication of such a manifold, in which the properties of silicon and oxide/nitride-modified layered surfaces were evaluated.They reported that the operating voltage achievable with such devices will be limited by the quality of the oxide insulating film produced. Nevertheless, the results are sufficient to demonstrate that silicon-based devices using a sandwich structure (i.e., oxide–nitride–oxide) can sustain potentials in the range 400–1200 V before dielectric breakdown, which is certainly sufficient to generate EOF mobility for mFIA systems. Although silicon does offer the attraction of producing high precision engineered mFIA manifolds, with well characterised surfaces and the potential to integrate control and detection on to one substrate, other materials such as glass24 and silica (quartz)25 are also amenable to IC based fabrication techniques and will produce surfaces eminently suitable for EOF pumping.Judging from the reports in the literature, glass and silica have proved to be the most popular materials for device fabrication to date, but metal, plastic and ceramic substrates are also possible candidates for the manufacture of mFIA systems.26,27 In general, three basic concepts are required for the preparation of mFIA systems: (i) a suitably prepared substrate, (ii) photolithographic equipment and (iii) wet etching and bonding facilities.28 Prepared substrates based on glass, silica and silicon can be obtained commercially from photomask producers (e.g., Alignrite, Wales, UK); these come ready coated with a metal film (0.1 mm) such as chromium over which is spun a positive photoresist layer (0.7 mm). These plates, which are typically 152.4 mm square (5 3 5 in) and 3 mm thick, can be used directly for manifold pattern transfer using photolithographic methods. A mask or negative of the final channel patterns required for the mFIA system can be produced using CAD computer software.29,30 The mask pattern can be transferred to the photoresist film on the substrate, using basic photographic development equipment.For channels larger than 1 mm, visible or UV light can be used to transfer the pattern from the mask to the photoresist (exposure times being approximately 5 ms); however, to obtain good line definition and submicro patterning, X-ray or electron-beam (e-beam) photolithography will be required. Where facilities permit, e-beam photolithography offers at present one of the most flexible ways of producing mFIA manifold designs; however, for most applications simple basic photographic equipment will suffice.Once the pattern has been transferred to the photoresist film on the substrate, the plate can be developed and wet etched. If one uses the large plates as described, then multiple devices can be prepared; in our laboratory, for example, 25 manifolds of various geometries and channel widths are produced from one plate (Fig. 2).18 Workers in the field tend to prepare their own plates rather than using a commercial source and, although this adds additional steps to the fabrication process, it is not too technically demanding.29,30 As indicated previously, these are typically made from silica or glass on to which a thin metal film (0.05–0.1 mm) of chromium, gold or a combination is produced, using sputtering or chemical vapour deposition (CVD) methods.The metal film is then spin coated with a layer of positive photoresist (approximately 0.5–2 mm). Fig. 1 Schematic diagram of the steps involved in the anistropic etching technique for the fabrication of mFIA devices. The arrow in Step 1 represents radiation. 2R Analyst, January 1997, Vol. 122Following the photolithographic transfer of the manifold pattern, the exposed photoresist and corresponding metal film are removed using commercially available reagents. The prepared plates may then be placed in an oven at 120 °C for 72 h to harden the photoresist and remove solvent residues from the plate.18 It should be noted that although the presence of a metal film on the glass is important for controlling the degree of surface etching that will occur, the exposed substrate will experience undercutting or sideways etching, which will proceed at a rate of approximately 2 : 1.Thus, as the etch goes down to 1 mm it spreads laterally 2 mm at the edges, producing channels, for which the width always exceeds the depth.31 Clearly, for a given substrate, the choice of the initial photomask line width and the time of etch will influence the size of the final channels produced.Various reagents have been used for the wet etching of the substrate, but in general for glass and silica, hot (70 °C) dilute HF–NH4F (1% HF + 5% NH4F in water)18,29 will give an etch rate of approximately 0.3–0.5 mm min21, whereas dilute HF– HNO3 produces an etch rate of between 0.5 and 0.8 mm min21 in Pyrex.30 Silicon etching requires the use of reagents such as ethylenediaminepyrocatechol (115 °C, etch rate 2.5 mm min21) or KOH.16 Although indicative figures for etch rates can be obtained,31,32 it is strongly recommended that sample strips of the substrate are tested in the selected etchant to establish the rate of etching for a given system.It should be stressed that the final size of the mFIA channels will depend on the initial width of the lines on the photomask, the substrate material and mode of etching. During the etching process, agitation of the substrate or etchant is advisable otherwise asymmetric channel formation, i.e., uneven etching on the bottom edge or side of the channel, may occur.It is usual, however, to obtain a channel profile that is wider at the top than the bottom. This effect, associated with undercutting, will vary as a function of the etch time, substrate type and reagents used. The surface quality of the etch is generally related to the types of material used. Good quality silica, for example, gives a well defined etch, whereas Pyrex or borosilicate glass can produce a rougher surface owing to the crystalline structure of the material.The real effect or significance of the surface properties in mFIA channels has not yet been fully characterised, but clearly the more controlled the etch, the more precise and smaller will be the channels that one can produce. What is not apparent from experimental results is the real influence that surface topography may have on the mobility and reactivity of reagents in a mFIA system.Rough or poorly defined surfaces and intersections, of one or more channel, will increase the effect of turbulence and in turn promote dispersion, which in some instances may prove to be an advantage where mixing is required. This factor is particularly important in capillary systems where the Reynolds number is lower than the transitional values of 2000 or 2300 indicating laminar flow, thus minimal mixing will dominate.33–35 However, research indicates that at low Reynolds numbers microfluidic mixing cannot be simply classified as a laminar or turbulent model.36 What is clear is that as channel sizes become smaller, surface effects will inevitably become more significant, which in turn will have an impact on fabrication and material specifications.Manz and Simon37 demonstrated that for a 10-fold decrease in channel size, a 1000-fold decrease in reagent consumption and a 100-fold decrease in related time variables would be obtained. Pressure requirements of such a system will, however, increase by a factor of 100, but as indicated later this does not effect the voltage requirements for EOF.Once an etched base has been produced, the top of the channels need to be sealed. Various bonding methods have been suggested, including glueing, low temperature bonding or annealing, high temperature fusion and anodic bonding.38 The use of an adhesive in such systems can pose problems due to channel plugging and, although this may be overcome to some extent by using photosensitive dry films laminated on to the substrate surface as a protective coating,39 the technique has found little real use in mFIA.Thermal bonding or annealing of top plates represents one of the simplest and therefore most widely used methods in device fabrication. These methods usually involve heating the substrate and top plate, which may be under slight pressure,30 in an oven or furnace to near the upper annealing temperature.For glass and silica this is usually between 500 and 600 °C, whereas silicon will require higher temperatures of around 900 °C. The period of heating varies between 48 and 96 h, after which the device is slowly cooled (48–72 h) to minimize physical stress. For glass substrates, hydrolysing the surface with dilute NH3–H2O2 followed by heating at 500 °C for 24 h29 or 575 °C for 96 h18 has been demonstrated to achieve good reliable bonding.Slightly more complex temperature programmes have been suggested, but these are essentially slight modifications to the same basic method.30 In order to keep the fabrication of a mFIA manifold as simple as possible and to increase the bonding success rate, it is advisable to use the same material (i.e., similar thermal expansion coefficients) for both the base and top of the device, so minimizing stress features. The fabrication of multilayer devices can become complex and techniques using intermediate layers have been described in which more complex structures such as pumps and valves are required.40,41 One technique which has found particular favour with workers fabricating microvalves, micropumps and silicon devices generally is anodic bonding.42,43 In this process, ions such as sodium and oxygen are thought to migrate at elevated temperatures to electrodes placed on the outer surface of one of the layers, so increasing the electrostatic surface charge.The temperature and applied voltages used will depend on the materials in question, but for glass these are found to be 200–400 °C at 30–300 V and for silica 700–800 °C at 30 V applied for a period of 45 min. It might be necessary when employing anodic bonding to treat the surface with HF prior to bonding. Holes or ports made in the top plate are commonly used as reagent or sample feeds into the manifold and can be generated either before18 or after29,30 the bonding process.The holes can simply be produced by using mechanical18 or ultrasonic16 drilling of the substrate and vary in diameter from 0.5–2 mm. At present, laser ablation, which offers an attractive method for hole production, has not been employed, but no doubt the technique will find its way into the literature before long. Plastic Fig. 2 Example of the photomask design used to produce five mFIA manifolds designs with variable line or channel widths.The line widths shown are A and E, 50, B, 30 and C, 10 mm. Analyst, January 1997, Vol. 122 3Rreservoirs are usually glued to hold liquids and support the platinum electrodes when required. A photograph of a completed device used for phosphate and nitrite determination is shown in Fig. 3. To date, only planar mFIA systems have been reported based on the fabrication techniques described. The fabrication of three dimensional systems is, however, an attractive prospect and no doubt stacked systems will be produced as more sophisticated chemistries are exploited.Such systems based on micropumps have been reported in which multi-layers or modules are clamped or stuck together to produce a complete device.19,20 Once a complete mFIA manifold has been fabricated for use with EOF pumping, the success of the bonding process can be checked by filling the channels with a buffer44 or weak acid18 and plotting the current–voltage relationship.This plot is generally found to be linear up to 10 kV in glass and silica, after which point dielectric breakdown of the substrate occurs. Any deviation in linearity at lower applied voltages is indicative of a device failure, associated with incomplete bonding. Pump Design In order to achieve reliable flow dynamics in an FIA manifold, which may contain numerous reagent streams, pulse free, constant flow characteristics must be maintained. It is not surprising therefore, to find that mFIA systems also call for pulse free, variable nl min21–ml min21 flow control. Although a direct syringe or piston pump can be used with mFIA systems,45,46 the closest equivalent to a conventional peristaltic FIA pump is the so called micropump, which usually takes the form of a pulsating one-way valve driven typically by a piezoelectric device.36,45,47–50 Based on microengineering technology, micropumps have been employed in mFIA manifolds consisting of channels 100 mm wide, 10 mm deep and 4 cm in total length.47 When such small channel dimensions are employed, the hydrodynamics of the system can produce a pressure drop of 0.7 atm at a flow rate of 0.5 ml min21.However, diaphragm devices, commonly using microchemical silicon membranes, are only reliable up to backpressures of 0.2 atm and so high pressure pumps or large channel dimensions are therefore required. High pressure pump designs using nickel rather than silicon check values are now being developed using low temperature bonding techniques, incorporating intermediate photoresist layers to adhere the valve to the body of the pump.47 These devices, which use a piezoelectric disc glued to the outer surface of the pump housing, operate at many hundred Hertz, requiring around 300 V to produce flow rates in the range 20–300 ml min21.In an elaborate example, four such pumps have been used to drive reagents through a three-dimensional system constructed from silicon and glass at a flow rate of 1 ml min21, the channel dimensions being 600 mm wide and 200 mm deep in order to accommodate pressure effects.19 Diaphragm micropumps do have an important role to play in applications where electroosmotic pumping is not applicable, owing to sample or reagent chemistries.It is essential, however, that micropumps remain chemically inert to the reagents and samples with which they may come into contact. In addition to piezoelectrically driven pumps, ultrasonic51 and other more exotic electrically activated pumps have also been described.36,52–54 Electrophoresis embodies two basic electrokinetic components, electroosmotic and electrophoretic flow.These two components combine to give the total flow or velocity (n) in the following way: total flow rate = n = (meo + mep)E (1) where meo = electroosmotic mobility, mep = electrophoretic mobility and E = applied electric field. At relatively low electric field strengths, electroosmotic mobility represents the larger component of the two processes; however, as the field strengths are increased the influence of the electrophoretic component (migration of analyte ions) increases also.For mFIA, where bulk mobility is desired, one finds that field strengths around 300–400 V cm21 are usually adequate; however, if separation is required, as in CE, then field strengths greater than 1 kV cm21 are typically required. It is common for mFIA systems based on EOF to operate with field strengths of the order of 80–300 V cm21, at which little or no analyte separation is observed The generation of an EOF to pump reagents and samples through a mFIA system is subject, however, to certain physico-chemical limitations.Firstly, in order to generate the EOF one must use a material which will yield negatively charged groups on the surface or walls of the channels when placed in contact with an appropriate liquid. Secondly, the liquid phase must dissociate to some extent in order to generate counter positive ions (notably H+ ions).The combination of the negatively charged surface (typically SiO2) and the H+ ions in solution will form a diffuse double layer (sometimes referred to as the Gouy or Helmholtz layer). This diffuse double layer acts as a parallel-plate electric capacitor whose plates are d cm apart each carrying a charge e per cm2. The zeta potential (x) will be the potential difference between the plates, given by the general equation x = 4ped/er (2) where er is the relative permittivity of the medium between the hypothetical plates. If an electrical field is now applied through the liquid phase, ions will migrate to their respective electrodes dominated by the positive ions (H+) moving to the negative or ground electrode.As the ions move, they induce a drag on the bulk of the liquid, which in turn results in a corresponding net transfer of the solution to the negative electrode.Thus, in EOF the mobility of the solution is from the anode (positive electrode) to the cathode (negative electrode) or ground. In practice, the formation of the double layer is limited to the pH range 4–10. At lower pH values the cationic population becomes so high that the EOF is overrun by conductive flow and at pH values greater than 10 the cation population becomes too low to sustain the double layer. The zeta potential that is generated when the double layer forms will be influenced by changes in pH and ionic strength of the solutions in the channel, and this can affect the corresponding flow rate.Clearly, as one of the main attributes of a pump in FIA is to maintain a reproducible flow rate, the pH and ionic strength of the reagents Fig. 3 A fully assembled mFIA device showing the mounted plastic reservoirs which act as wells for the reagents and samples and support the platinum electrodes required for EOF pumping. The device is shown mounted in a rig which allows fibre optics (seen either side of the block) to be coupled into relevant sections or channels of the manifold. 4R Analyst, January 1997, Vol. 122used in a mFIA manifold need to be controlled, through the use of buffer systems. As the ionic strength increases, for example, a counter-ion effect will prevent the ions from migrating through the solution independent of the EOF. In general, it is therefore preferable to keep the total ionic strength of reagents and samples as low as possible.Once an EOF has been generated, the flow establishes a flat, trapezoidal profile notably different from the parabolic (bullet) shape commonly associated with conventional FIA. Closer investigation of the mFIA profile reveals that the edges of the profile (i.e., closest to the wall surface) are slightly in advance of the bulk owing to the drag effect and the extent of this effect will be a function of the solution viscosity.55 The fact that the flow profile, which offers minimal band broadening, does not contain the parabolic flow characteristics of a pressure or hydrodynamically pumped system, does have implications with respect to the mixing of reagents in the mFIA reactor, where some dispersion will be necessary to achieve the desired chemical reactions.Thus EOF-pumped mFIA can be considered to be more akin to segmented flow analysis rather than conventional FIA, in its flow characteristics.In an EOF-pumped mFIA system, the flow rate can be effectively controlled by varying the applied voltage as follows: n flow rate v = m (3) L where n = applied voltage (V), L = length of the channel (m) and m = sum of the electroosmotic and electrophoretic mobility component under different conditions of pH and ionic strength. The EOF and hence the flow rate in mFIA, will be unaffected by capillary diameter up to channel sizes approaching 250 mm, after which drag effects of the bulk solution may cause serious disruption to the EOF.The EOF produced in a channel can be considered, in the electrical sense, as a resistor, thus Joule heat will occur in the capillary. However, the almost negligible bulk properties of components in a channel, relative to the substrate, will ensure an effective dissipation of heat, which rarely poses a serious problem in mFIA. If, however, the resistance is allowed to increase, for example when a highly viscous or an immobile solvent is present in a channel, heating can occur.It should be stressed that an EOF can only be established if the double layer is formed, and this requires ions or dipoles to be present in the liquid stream. In the case, for example, where organic compounds or solvents may be present, such conditions may not be met. For example Zheng and Dasgupta55 described some initial studies, using silica capillaries, to evaluate the suitability of an EOF in a mFIA system for carrying out in-line analyte phase separation or solvent extraction.They described the use of a 50 cm 3 7.5 mm id polyimide coated silica capillary to investigate various aqueous ionic complexes, based on well characterised cationic, anionic and neutral ion-pair chemistries. A quaternary ammonium salt, tetrabutylammonium perchlorate, was added to chloroform as a supporting electrolyte to assist in the mobilization process56 and to catalyse phase transfer.57 The results clearly indicated that organic solvent pumping was possible, when modified with the quaternary ammonium salt, but that migration of the ammonium ion occurred, creating a positive front end to the solvent slug.Bleeding of ions into the secondary aqueous buffer from the organic phase was also reported. Furthermore, they suggest that a thin interfacial layer of buffer is generated between the organic solvent and the capillary walls which enables the EOF to be generated. Investigations into solvent extraction indicated that the ion-pair complexes studied were either extracted into the organic phase or accommodated in the aqueous phase ahead of the organic slug interface.Not only were the findings of Zheng and Dasgupta a significant contribution in terms of demonstrating phase extraction and organic solvent mobility using EOF in a mFIA system, but they also clearly indicate that the migration of ions within a solvent under the influence of an electric field could lead to the development of gradient and separation techniques, complementary to CE, such as selective solvent extractions and matrix modification. Interestingly, the use of micellar electrokinetic capillary chromatography, described by Moore et al.,58 suggests that multiphase systems could be developed for mFIA applications, thus offering significant advantages for organic solvent based chemistries.Although it is possible to use EOF as the primary pumping mechanism in a mFIA manifold, it may be preferable or even necessary to isolate the pumping mechanism from the injector, reactor and detector components of the system.Such occasions might be, for example, when the chemistry of the method or detector system is not compatible with the electrical field required for direct EOF pumping. One such system has been described by Dasgupta and Liu45 in which a section of polyimide-coated fused-silica tubing (40 cm 3 75 mm id) was connected to a second capillary, via an isolating membrane. In the first channel a 2 mm borax buffer is pumped by direct EOF, with a field strength of around 40 V cm21.As the flow in the pumped capillary was electrically isolated from the second capillary it produced a hydrodynamic effect, sufficient to create a flow in the second channel, which could be varied between 1 nl min21 and 100 ml min21. In this case the hydrodynamic effect of the EOF is exploited in a mFIA manifold without the need for a direct electrical field. Injector Design The introduction or injection of a sample into an FIA manifold can generally be classified into two general types.The first is the timed or gated injection in which a sample is drawn into the FIA manifold, usually through a sampling probe, for a controlled period of time, after which the flow is switched back to the carrier stream. In this way, a variable volume can be injected into a manifold as a function of time. The second and more usual method of injection in FIA, is the introduction of a constant sample or reagent volume into a mobile carrier in a way that affords minimal flow dispersion.The most common approach for such systems is to use a rotary or slide valve, which contains a sample loop of a defined volume. It would therefore seem appropriate to have both these forms of sample injection available in mFIA systems, and this is indeed found to be the case. The injection of a sample into a mFIA manifold can be performed using hydrodynamic/pneumatic pressure control,46 miniaturised valves1 or electrokinetic mobility based on EOF.15,18,59–62 Even the lowest volumes obtainable with traditionally based rotary-type valves63 are clearly too large for practical use in mFIA.However, Liu and Dasgupta1 recently described the use of commercially available valves with an injection volume of 60 nl. It should be stressed, however, that most of the work relating to the introduction of small sample volumes ( < 20 nl) into capillary systems has been focused on CE methodology rather than addressing mFIA systems.Of particular relevance to mFIA systems is the development of the so-called valveless injection method, which uses EOF control in conjunction with specific capillary channel geometries.64,65 The two simplest approaches to sample injection based on EOF are first to pump a sample for a given period of time into a flow channel and second to fill a defined volume (equivalent of a sample loop) built into the mFIA manifold.In the former case, Zheng and Dasgupta55 described a modified CE system in which an organic solvent was loaded for a given period of time into a mFIA system consisting of a 50 cm 3 75 mm id polyimide Analyst, January 1997, Vol. 122 5Rcoated fused-silica capillary. The solvent was introduced into the system by placing the capillary electrode in the solvent reservoir at 15 kV for 10 s. Following the injection step, the capillary electrode was switched to a borate buffer which subsequently pumped the solvent through the capillary to the detector.Using this approach, the authors reported an RSD of 0.51% associated with migration effects and 1.7% for the peak, suggesting that surface tension, viscosity and flow rate will influence this particular mode of injection. It should be noted that in this work, the authors were using the method described for introducing a solvent into the mFIA system for the purpose of solvent extraction.The second approach to sample introduction, based on defined volumes using EOF, falls into two categories, the Xand Z-type injections (Fig. 4). Of the two injection geometries, the X or cross design, has been most widely investigated. Depending on the geometry and electrical field used, the sample injection can be classified as ‘floating’ (gated) or ‘pinched’ (discrete), with the former occasionally being referred to as the continuous mode.59,60,64,65 Experimentally, the simplest of these two modes is the ‘floating’ method, as it only requires one pair of electrodes (Fig. 5). In this case the sample is pumped by an EOF from the sample reservoir (A the positive electrode) to the sample waste reservoir (B the negative electrode) along channel AB crossing part of channel CD. Note that the channel AB will have been filled prior to the sample introduction with a suitable buffer. The reservoirs C and D contain no electrodes and therefore have no applied field, hence their potential is floating relative to the field in channel AB.As the sample passes across the intercept of AB and CD, diffusion and eddy effects will allow some of the sample to migrate in the direction of both C and D [Fig. 5(a)]. Subsequently, on placing the electrodes in reservoirs C and D the field can be made to run in the CD direction, so any sample molecules in the intercept will be pumped towards D, the grounded reservoir [Fig. 5(b)], allowing the sample to pass via a detector on its way. Clearly, we can see that the leakage or migration of the sample solution into channel CD during injection and the possibility of dragging the stationary sample from channels A and B into D as a function of flow [Fig. 5(c)] will lead to an uncontrolled volume injection. Examples of this effect have been reported.59,60,64 The leakage observed into the main channel may not be a serious problem for certain applications and clearly it offers a simple method of operation, in which only one pair of electrodes are required. However, the quality of etch, surface topography and geometry of the channels concerned will influence the degree of diffusion using the ‘floating’ injection method.The need, especially in CE, to have more precise control over the sample volume injected has lead to the development of the so-called ‘pinch’ method (Fig. 5). In this case, the sample is once again pumped under an EOF from reservoir A to B, but this time reservoirs C and D are not electrically floating, but like A are given a positive potential, relative to the waste reservoir B [Fig. 5(d)]. The effect of this is to draw buffer from channels C and D into B along with the sample from A. Under these flow dynamics, the sample gains a pinched or trapezoidal shape at the intercept of channels AB and CD. When the flow is redirected towards D [Fig. 5(e)], the sample volume, which may be only a few picolitres, is of a more precisely defined volume, less affected by diffusion effects.Jacobson et al.60 have reported improvements in the RSD from 2.7 to 1.7% for a 90 pl injection volume using the ‘pinch’ method, compared with the ‘floating’ approach. In this particular study, the applied voltages based on a 1 kV power supply were reservoir A 90, B 0, C 90 and D 100%, giving channel field strengths of 270, 20, 400 and 690 V cm21 in A, B, C and D, respectively.Clearly, operating the manifold in a multi-electrode configuration will reduce dispersion effects at intersections and allow improved control of the sample injection volume. The concept of the ‘push-back distance’ associated with surface diffusion has been used66 to define the migration rate and shape of flow patterns at X-type intercepts. This approach to sample injection has considerable potential in controlling sample volumes as a function of applied field. The larger volumes (9–22 nl) of the Z-type injection (Fig. 6) are introduced by pumping the sample between two reservoirs across a defined volume in a Z geometry. In the example shown, the sample is pumped from reservoir C or D to the waste reservoir D or E, so filling a defined volume in channel AX. The loaded sample can then be reacted with reagents in channel AX and moved to a detector or be monitored in situ. The Z mode of injection produces a larger volume than that obtained by the X mode, which in turn will be less influenced by the diffusional effects described previously.18,67 Recent work in the author’s laboratory using the Z injection technique has indicated that Fig. 4 Discrete volume injection using (a) the X and (b) the Z approach. Fig. 5 Flow and dispersion characteristics of floating and pinched modes of injection. 6R Analyst, January 1997, Vol. 122pinch control offers no improvement in precision over the floating method.68 The ability to matrix modify a sample in an injector system could be an attractive option in mFIA and techniques such as synchronized cyclic capillary electrophoresis22,69 or sample fractionation67 may offer considerable scope for such methodology.In these systems selected, analyte fractions can be separated from a more complex and possibly interfering matrix, owing to the variation in migration and flow direction under electrophoretic conditions. The switching of electrodes can be used to move components along channels and across channel intersections akin to shunting railway trucks, until the analytes required become isolated from matrix or interfering components. This aspect of sample manipulation is potentially very exciting for mFIA system development and will no doubt be exploited in future applications.Reactor Design Having satisfactorily injected a sample into an FIA system, the next objective is usually to present the sample to a detector in an appropriate form as quickly as possible, using tube or channel geometries that minimize the dispersion of the sample slug.In mFIA systems sample diffusion is known to be a function of the square root of the time after the injection of a sample.69 There remains one additional and often complex step, however, that of achieving the appropriate chemistry or biochemistry necessary for the detection of the analyte of interest. This process typically requires the addition of at least one reagent and can include more complex steps such as in-line solid-phase extraction and multiphase separations.70 Indeed, the complexity of the FIA manifold is only limited by the imagination or ingenuity of the analyst.The first and perhaps the most important step in reactor design is a consideration of the physical and chemical characteristics that occur when two solutions combine at a channel junction in a mFIA manifold. From the previous section, it is clear that even at Reynolds numbers less than 2000 some turbulent mixing takes place at channel intercepts and that the electrical potentials at which the channels are held can significantly influence the interfacial or mixing zone. Turbulent mixing, for example, has been reported in channels 5.2 mm deep and 57 mm wide.71 The characterization of channel intersections indicates that leakage or diffusion of reagents from a ‘floating’ side stream into a channel in which EOF is present is around 2–5%, depending on the viscosity and flow rates.64 This bleed occurs as a result of hydrodynamic effects, in which the molecules in the pumped stream entrain the stationary side channel molecules owing to frictional and eddy properties in a Venturi- or Bernoulli-type effect.What is interesting with such systems is that the flow characteristics in the main channel will have a flat, trapezoidal profile whereas the side channel, whose flow will be pressure driven, will have a parabolic profile.Hence the profiles of the merging streams may have a significant effect on the mixing characteristics at such an interface. As one of the main objectives of reagent addition in an FIA system is to induce mixing and so achieve the required reaction, some consideration needs to be given to this aspect of pumping in mFIA systems. In a valuable study, Seiler et al.44 demonstrated that the individual electrical resistances in a series of interconnecting channels can be used to predict the flow characteristics, in a similar way to hydrodynamic estimations in conventional FIA.The basic concept considers the intercepting channels as electrical resistors and evokes Kirchhoff’s rules,72 to predict the net flow of current which can be attributed to the flow dynamics of the device. One of the important findings of the work by Seiler et al.44 is that by controlling channel voltages one is able to manipulate sample or buffer dilutions by adjusting flow rates, and this offers considerable scope for stopped-flow or reverseflow operations. This effect was also recently demonstrated in our laboratory using a manifold previously described for phosphate analysis.18 Figure 6(a) illustrates how a sample of orthophosphate, held in reservoir D, is injected into channel AX using the Z technique, by pumping the sample to reservoir E.Thus the slug in channel AX defined by the intercepts DE represents the injected sample which reacts with ammonium heptamolybdate (0.01 m) in the presence of ascorbic acid (0.05 m) to produce molybdenum blue.Detection of the molybdenum blue is achieved by measuring the absorbance at 744 nm in a spectrophotometer coupled to the manifold by a fibre-optic system attached along channel AX (optical path 2 cm). Through the selection of different injection volumes [i.e., C–D (9 nl), D– E (13 nl) and C–E (22 nl)] and variation in the positive voltage applied at reservoirs A and B relative to D or C, the absorbance signal was observed to decrease as the intercept ‘pinch’ potential applied across A–E and B–E became sufficient to dilute the formation of the coloured complex at each end of the injection slug.68 In effect, the flow rate in channel A–B was being increased as a function of voltage, so diluting the flow from channels D–E, C–E or C–D into channel AX.Results obtained using this approach (Fig. 7) indicated that calibration based on one standard is possible and that the technique will have an important role to play in the development of future methodology.Following the mixing of the reagent and analyte streams, some period of time is usually required to produce sufficient product, either for further reactions or for subsequent detection. Conventionally, this hold time is effected by using a coil or knotted reactor and/or a stopped-flow mode, in order to minimize dispersion whilst achieving the required reaction chemistries.In mFIA it may be that owing to the more efficient interfacial nature of the mixing, i.e., the need for less bulk mixing, reactions rates may increase; however, reactions are still likely to require a finite period of time for product production. A serpentine-type pattern or structure offers a very simple, but effective, use of space for extending channel length and is compatible with photolithiographic fabrication techniques. The flow characteristics of serpentine channels have been studied and whilst disruption to the electroosmotic flow Fig. 6 Controlled dilution experiments using a pinch voltage. Analyst, January 1997, Vol. 122 7Rwas observed at the 90° bends, no significant band broadening has been found.60 The particular design used had a total channel length of 17 cm operating with a field strength of 700 V cm21 and was used to perform electrophoretic separations. It would seem that using such an approach, time-dependent mFIA reactions could be accommodated in a physically small area with minimal dispersion through the use of serpentine or spiral channels.61 Modification to the surface properties to induce selective flow characteristics, given the limitations of producing a double layer and hence EOF, has attracted some interesting approaches. These include the use of reversed-phase hydrophobic polymers, 73 surfactants,74 silanals and quaternary ammonia groups.75 In addition, the covalent bonding or immobilisation of glucose oxidase has been used to develop a glucose mFIA method based on a serpentine geometry using a 260 cm long 3 100 mm wide 3 70 mm diameter reactor.76 One of the interesting extensions of the mFIA technology described is the development of electrochromatographic (EC) separations in which capillaries are packed with small particles (approximately 3 mm) on which efficient separation can be achieved.77 Using EOF as the primary pumping mechanism, EC can be mediated with reference to the zeta potential and hence separation is potentially possible through selection of the appropriate surface properties of a packing material.Thus the walls of the capillary act to facilitate the primary pumping mechanism and the packed chromatographic material offers enhanced electrophoretic separations. In addition to surface modification, the physical effect of field flow fractionation may also be incorporated as a separation process in capillary systems.78 In this case, an external force, e.g., magnetism or gravity, is used to pull fractions to the wall of a channel; on removing the force, a gradual diffusion of the fractions in the sample occurs back into the flowing stream, usually based on molecular size.79,80 As yet field flow fractionation has not been widely studied in mFIA systems, but the possibility of using field flow fractionation and modifying the zeta potential using a magnetic field might be an interesting subject for future research. Detector Design Much of the research reported to date in the area of mFIA has focused on characterising the physical processes of reagent mobilisation and mixing in microchannel manifolds.Such work has relied heavily on imaging techniques, using, for example, charged coupled device (CCD) cameras60 and microscopybased techniques.30,44,61,81 In this section, consideration is given to the design of detectors suitable for direct integration into EOF based mFIA systems. The major detector systems used in conventional FIA are based on optical absorption/emission and electrochemical techniques,70 and this is also found to be the case for mFIA systems.82,83 Clearly, if electroosmotic based pumping is employed in the mFIA manifold, then some care is required with the design of electrochemical detectors, but this does not pose any real serious limitations.83 Not surprisingly, optical detection has proved to be the most attractive approach to on-device detection and various examples can be found in the literature,15,40,45,67 mainly associated with CE detectors.84 One of the less obvious advantages of miniaturisation is the ability to introduce detector systems not readily available to conventional larger flow systems, such as mass-selective devices of the surface acoustic wave type.85 Clearly, the design of a flow cell in terms of physical volume and optical geometry is very important in miniature systems if sensitivity, reliability and robustness are to be achieved. The incorporation of fibre optics as part of an optical detection system has proved to be of value, particularly with the advent of fibres with diameters of the order of 0.1 mm.86 In the area of optical detectors, some effort has gone into developing axially rather than perpendicularly oriented measurement cells.87 Increasing the volume of the flow cell in a perpendicular viewing axis has been used to increase the optical pathlength,88 whilst the use of multiple reflections axially in a flow cell has also been evaluated.89 The first example led to severe dispersion effects and a subsequent loss in sensitivity and the second approach, based on a silicon device, suffered from signal loss due to scatter and absorption of light at the silicon surface.One interesting approach to absorbance measurements is to consider the mFIA channel as an extension of a fibre optic system.18,90 In this way, total internal reflection of light may be achieved and any photoactive species spatially present in the channel will undergo interaction with the photons present, to produce either a direct absorption measurement or a subsequent fluorescence effect.As with most absorption methods, it would be preferable to use a dual-channel system to improve stability and reduce scatter. Although a long pathlength is appealing for absorption measurements, emission-based techniques would benefit from a small spatial volumes in which the emission effect can be concentrated, and volumes less than 1 nl have been suggested.89 The size of the detector cell is clearly related to the concentration and sensitivity of the method in question and a flow cell of 15 ml has been reported to be adequate for chemiluminescent measurements of glucose and lactate in human serum.91 More recently Liang, et al.92 have described a UV cell with a parallel flow optical path of 120–140 mm for absorbance and fluorescence detection at the end of a CE column, which would be most suitable for mFIA applications.Sequential detector arrays are an attractive approach in sensor design and lend themselves well to mFIA systems.82,93 One such system82 has been described, based on electrochemical detectors, for liquid chromatographic separations of catecholamines and consisted of four photolithographically prepared ISFET sensors aligned sequentially in a 5 mm silicon channel 100 mm wide and 70 mm deep.The total volume of the detector was 20 ml. Another device,93 using a peristaltic pump to control flow rates, consisted of nine 5 mm ISFET sensors housed in a 15 ml cell, used for pH, potassium and calcium determinations in biological fluids. Although neither of these systems was used in mFIA manifolds, they do illustrate the ability to develop array detectors compatible with mFIA technology. One of the most exciting prospects for mFIA detector design is the ability to incorporate their fabrication into one integrated device.For example, miniature mass spectrometers (3 3 3 3 3 mm) have been produced based on fabrication technology that would be ideally suited for mFIA.94 Miniature spectroscopic systems are also becoming available95,96 and future developments in spectrometers incorporating opto-electronic systems Fig. 7 Calibration based on 100 ppb PO4 with dynamic applied pinch voltage. The equivalent absorbance values for 100, 50 and 25 ppb phosphate are indicated on the x-axis. 8R Analyst, January 1997, Vol. 122will undoubtedly bring valuable complementary technology for future mFIA detector design. Future Trends The preceding sections have tried to focus on the basic concepts and developments relating to mFIA systems based on EOF. Some indication has been given of the likely areas where current and future research may prove to be of great value. These include the fabrication of devices where there is considerable potential for the construction of stacked or three-dimensional systems, possibly using cold bonding and direct laser etching techniques.The mobilisation of reagents and analytes, based on EOF, requires more complete characterisation if flows and mixing effects at intersecting channels are to be effectively exploited. The most important development if mFIA technology is to be fully realised is the fabrication of self contained operational systems with proven application robustness.The close relationship of mFIA technology with separation techniques such as capillary electrophoretic and micro-electrochromatographic separations may well lead to some form of hybride system in the near future. The area perhaps where the greatest advances in mFIA-based technology will be most readily realised is the biotechnology sector, where methodology and applications are complementary to miniature systems. Already examples are emerging of applications in DNA fragment analysis97 and immunoassay methodology.98 Developments, however, need to be focused not only on the integrated device which may be encapsulated and equipped with telecommunication for remote operation, but also to include interfacing to existing measurement systems such as MS or NMR, which would benefit from some form of sample pretreatment.One area which has not yet been considered in mFIA systems is a return to the early gasphase work started by Terry et al.5 Clearly there are some exciting possibilities in gas and multiphase systems yet to be realised.Conclusion Where has mFIA and more generally mTAS got to? In 1991, Manz et al.13 questioned whether the developments in mTAS were ‘a look into next century’s technology or just a fashionable craze’. In their concluding remarks, the authors made some general comments relating to the uptake or acceptance of mTAS technology, ‘namely that changes are required in the political and cultural opinions of analytical work if appropriate financial support is to be forthcoming and that the research base must grow worldwide to foster both competitive and collaborative research’.Further, they identified that ‘the market acceptance of the technology must be embraced through the design and production of mTAS concepts’. Clearly, these requirements have been only partially fulfilled. An examination of the literature indicates that there has been a positive growth in the research base and this will obviously support the fundamental development of the science.What is less obvious is the political and more importantly the economic will to support the development of the technology, and this may yet be seen to be the most seriously limitation to the growth of the science. Although miniaturisation is conceptually appealing, there remain some important technical obstacles to overcome, such as the introduction of ‘real’ samples and the ability to deal with suspended particles.These limitations are not beyond the scope of present day membrane technology, and should pose no serious hindrance to the advancement of the science. Developments in the field of mTAS since 1991 clearly point to the technique forming the basis of future methodology applicable to a wide range of applications ranging from measurement science to chemical synthesis in which mFIA based on EOF flow will play a significant role.The limitation in releasing the considerable potential that micro reactor technology can offer resides not in the technological challenge but in the imagination of our minds and only requires us to realize it. References 1 Liu, S., and Dasgupta, P. K., Anal. Chim. Acta, 1993, 283, 739. 2 Bogue, R., Lab. Equip. Dig., 1995, February, 14. 3 Manz, A., Effenhauser, C. S., Burggraf, N., Verpoorte, E. J. M., Raymond, D. E., and Widmer, H. M., Analusis Mag., 1994, 22, M25. 4 Craston, D., and Cowen, S., Chem.Br., 1996, October, 31. 5 Terry, S. C., Jarman, J. H., and Angell, J. B., IEEE Trans. Electron. Devices, 1979, ED-26, 1880. 6 Capillary Electrophoresis: Theory and Practice, ed. Grossman, P. D., and Colburn, J. C., Academic Press, New York, 1992. 7 Mikkers, F. E. P., Everaerts, F. M., and Verheggen, Th. P. E. M., J. Chromatogr., 1979, 169, 11. 8 Jorgenson, J. W., and Lukacs, K. D., Anal. Chem., 1981, 53, 1298. 9 Whalen-Pechersen, E. K., and Jurs, P.C., Anal. Chem., 1981, 53, 2184. 10 Manz, A., Graber, N., and Widmer, H. M., Sens. Actuators B, 1990, 1, 244. 11 Advances in Chromatography, ed. Brown, P. R., and Grushka, E., Marcel Dekker, New York, 1993, pp. 1–66. 12 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Kluwer, Dordrecht, 1995. 13 Manz, A., Fettinger, J. C., Verpoorte, E. J. M., L�udi, H., Widmer, H. M., and Harrison, D. J., TrAC, Trends Anal. Chem. (Pers. Ed.), 1991, 5, 144. 14 Manz, A., Harrison, D.J., Verpoorte, E. M. J., Fettinger, J. C., L�udi, H., and Widmer, H. M., Chimia, 1991, 45, 103. 15 Lui, S., and Dasgupta, P. K., Anal. Chim. Acta, 1992, 268, 1. 16 Harrison, D. Glavina, P. G., and Manz, A., Sens. Actuators B, 1993, 10, 107. 17 Jacobson, S. C., Hergenr�oder, R., Koulng, L. B., and Ramsey, J. M., Anal. Chem., 1994, 66, 2369. 18 Daykin, R. N. C., and Haswell, S. J., Anal. Chim. Acta, 1995, 313, 155. 19 Verpoorte, E. J. M., Van der Schoot, B. H., Jeanneret, S., Manz, A., and de Rooij, N.F., in Interfacial Design and Chemical Sensing, ACS Symposium Series, American Chemical Society, Washington, DC, 1994, ch. 21, p. 244. 20 Schomburg, W. K., Vollmer, J., B�ustgens, J., Fahrenberg, J., Hein, H., and Manz, W., J. Micromech. Microeng., 1994, 4, 1. 21 Sensors: a Comprehensive Survey, ed. Grandke, T., and Ko, W. H., VCH, Weinheim, 1989, vol. 1, p. 107. 22 Manz, A., Verpoorte, E. M. J., Effenhauser, C. S., Burggraf, N., Raymond, D.E. and Widmer, H. M., Fresenius’ J. Anal. Chem., 1994, 348, 567. 23 Woffenbuttel, R. F., Sens. Actuators A, 1992, 30, 109. 24 Altmann, J. P., Das Neue Lehrbuch de Glus�alzerei, Gentner, Stuttgart, 1963, p. 29. 25 Daniel, J. S., and Delapierre, G., J. Micromech. Microeng., 1991, 1, 187. 26 Becker, E. W., Ehrfeld, W., Hagmann, P., Maner, A., and M�uchmeyer, D., Microelectron. Eng., 1986, 4, 35. 27 Knapp, J., Andreae, G., and Petersohn, D., Sens. Actuators A, 1990, 21–23, 1080. 28 Kovacs, G. T. A., Petersen, K., and Albin, M., Anal. Chem., 1996, 68, 407A. 29 Jacobson, S. C., Koutny, L. B., Herenr�oder, R., Moore, A. W., Jr., and Ramsey, J. M., Anal. Chem., 1994, 66, 3472. 30 Fan, Z., and Harrison, D. J., Anal. Chem., 1994, 66, 177. 31 Elliot, D. J., Integrated Circuit Fabrication Technology, McGraw- Hill, New York, 1982. 32 Thin Film Processes, ed. Vossen, J. L., and Keren, W., Academic Press, San Diego, 1978, ch. V-1, p. 401. 33 van der Linden, W. E., TrAC, Trends Anal.Chem. (Pers. Ed.), 1987, 6, 37. 34 Tyssen, R., Anal. Chim. Acta, 1980, 114, 71. 35 Hungerford, J., Thesis, University of Washington, 1986. 36 Gravesen, P., Branebjerg, J., and Jensen, O. S., J. Micromech. Microeng., 1993, 3, 168. 37 Manz, A., and Simon, N., Anal. Chem., 1987, 59, 74. Analyst, January 1997, Vol. 122 9R38 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Klewer, Dordrecht, 1995, p. 165. 39 Shoji, S., and Esashi, M., Sens.Actuators B, 1992, 8, 205. 40 Field, L. A., and Muller, R. S., Sens. Actuators A, 1990, 21–23, 935. 41 Fung, C. D., Cheung, P. W., and Fleming, D. G., Microachining and Micropacking of Transducers, Elsevier, Amsterdam, 1985. 42 Harz, M., J. Micromech. Microeng., 1992, 2, 161. 43 Esashi, M., Nakano, A., Shoji, S., and Hebiguchi, H., Sens. Actuators A, 1990, 21–23, 931. 44 Seiler, K., Fan, Z. H., Fluri, K., and Harrison, D. J., Anal. Chem., 1994, 66, 3485. 45 Dasgupta, P.K., and Liu, S., Anal. Chem., 1994, 66, 1792. 46 Li, S. F. Y., Capillary Electrophoresis, Elsevier, New York, 1992. 47 Shaji, S., Esashi, M., van der Schoot, B., and de Rooij, N. F., Sens. Actuators A, 1992, 32, 335. 48 Horowitz, P., and Hill, W., The Art of Electronics, Cambridge University Press, Cambridge, 2nd edn., 1989, p. 368. 49 van der Schoot, B. H., Jeanneret, S., van den Berg, A., and de Rooij, N. F., Anal. Methods Instrum., 1993, 1, 38. 50 Stewart, K. K., Beecher, G.R., and Hare, P. E., Fed. Proc., Fed. Am. Soc. Exp. Biol., 1974, 33, 1434. 51 Moriney, R. M., White, R. M., and Howe, R. T., Proc. MEMS’91, Nara, Japan, 1991, p. 277. 52 Richter, A., Plettner, A., Hofmann, K. A., and Sandmeier, H., Proc. MEMS ’91, Nara, Japan, 1991, p. 271. 53 Osada, Y., Okuzzaki, H., and Hari, H., Nature (London), 1992, 355, 242. 54 Fuhr, G., Fiedler, S., M�uller, T., Schnell, T., Glasser, H., Lisec, T., and Wagner, B., Sens. Actuators A, 1994, 41–42, 230. 55 Zheng, H. J., and Dasgupta, P. K., Anal. Chem., 1994, 66, 3997. 56 Adams, R. N., Electrochemistry at Solid Electrodes, Marcel Dekker, New York, 1969. 57 Masson, D., Magdassi, S., and Sasson, Y., J. Org. Chem., 1990, 55, 2714. 58 Moore, A. W., Jr., Jacobson, S. C., and Ramsey, J. M., Anal. Chem., 1995, 67, 4184. 59 Jacobson, S. C., Hergenr�oder, R., Koutny, L. B., and Ramsey, J. M., Anal. Chem., 1994, 66, 1114. 60 Jacobson, S. C., Hergenr�oder, R., Koutny, L. B., Warmack, R. J., and Ramsey, J.M., Anal. Chem., 1994, 66, 1107. 61 Seiler, K., Harrison, D. J., and Manz, A., Anal. Chem., 1993, 65, 1481. 62 Manz, A., Effenhauser, C. S., Burggraf, N., Harrison, D. J., Seiler, K., and Fluri, K., J. Micromech. Microeng., 1994, 4, 257. 63 Tseida, T., Mizuno, T., and Akiyama, J., Anal. Chem., 1987, 59, 799. 64 Harrison, D. J., Manz, A., Fan, Z., L�udi, H., and Widmer, H. M., Anal. Chem., 1992, 54, 1926. 65 Manz, A., Harrison, D. J., Verpoorte, E. M. J., Fellinger, J.C., Paulus, A., L�udi, A., and Widmer, H. M., J. Chromatogr., 1992, 593, 253. 66 Effenhauser, C. S., Paulus, A., Manz, A., and Widmer, H. M., Anal. Chem., 1994, 66, 2949. 67 Effenhauser, C. S., Manz, A., and Widmer, H. M., Anal. Chem., 1995, 67, 2284. 68 Hasser, A., MSc Thesis, University of Hull, 1996. 69 Burggraf, N., Manz, A., Verpoorte, E. M. J., Effenhauser, C. S., Widmer, H. M., and de Rooij, N. F., Sens. Actuators B, 1994, 20, 103. 70 Ruzicka, J., and Hansen, E.H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 71 Halliday, D., and Resnick, R., Physics, Wiley, New York, 1960. 72 Jacobson, S. C., Koutny, L. B., Hergenr�oder, R., Moore, A. W., Jr., and Ramsey, J. M., Anal. Chem., 1994, 66, 3472. 73 Monning, C. A., and Jorgenson, J. W., Anal. Chem., 1991, 63, 802. 74 Yao, X. W., Wu, D., and Regnier, F. E., J. Chromatogr., 1993, 636, 21. 75 Smith, J. T., and Rassl, Z. E., J. High Resolut. Chromatogr., 1992, 15, 573. 76 Marakami, Y., Suda, M., Yokogama, K., Tukeuchi, T., Tamiya, E., and Kurube, I., Microchem., 1994, 49, 319. 77 Tsuda, T., Nomura, K., and Nakagawa, G., J. Chromatogr., 1982, 248, 241. 78 Giddings, J. C., Science, 1993, 260, 1456. 79 Widmer, H. M., and Koch-Kellner, L., Chimia., 1989, 43, 320. 80 Koch, L., Koch, T., and Widmer, H. M., J. Chromatogr., 1990, 517, 395. 81 Raymond, D. E., Manz, A., and Widmer, H. M., Anal. Chem., 1994, 66, 2858. 82 Sudfa, M., Sakuhara, R., and Karube, I., Appl.Biochem. Biotechnol., 1994, 41, 3. 83 Haber, C., Sclvestri, I., B�o�osli, S., and Seman, W., Chimia, 1991, 45, 117. 84 Chervet, J. P., Ursem, M., Salzmann, J. P., and Vannoort, R. W., LC– GC, 1989, 7, 515. 85 Wohltjen, J., Sens. Actuators, 1984, 5, 305. 86 Ton, W., Zhang-You, S., Smith, S., Birnbaum, D., and Kopelman, R., Science., 1992, 258, 778. 87 Xi, X., and Yeing, F. S., Anal. Chem., 1990, 62, 1580. 88 Xue, Y., and Yeung, E. S., Anal. Chem., 1994, 66, 3575. 89 Verpoorte, E.M. J., Manz, A., L�udi, H., Bruno, A. E., Maystre, F., Kruttiger, B., Widmer, H. M., van der Schoot, B. H., and de Rooij, N. F., Sens. Actuators B, 1992, 6, 66. 90 Weigl, B. H., and Woelfbeis, O. S., Anal., Chem., 1994, 66, 3323. 91 Suda, M., Sakuchara, T., and Karube, I., 2nd Bioengineering Symposium, Tsukuba, Japan, 1992, p. 113. 92 Liang, Z., Chiem, N., Ocvirk, G., Tang, T., Fluri, K., and Harrison, D. J., Anal. Chem., 1996, 68, 1040. 93 Arquint, P., Koudelka-Hep, M., van der Schoot, B.H., van der Wal, P., and de Rooij, N. F., Clin. Chem., 1994, 40, 1805. 94 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Kluwer, Dordrecht, 1995, p. 299. 95 Conzen, J. P., B�urch, J., and Ache, H. J., Appl. Spectrosc., 1993, 47, 753. 96 Ache, H. J., Chem. Ind., 1993, 1, 40. 97 Jacobson, S. C., and Ramsey, J. M., Anal. Chem., 1996, 68, 720. 98 Koutny, L. B., Schmalzing, D., Taylor, T. A., and Fuchs, M., Anal. Chem., 1995, 68, 18. Paper 6/06289J Received September 16, 1996 Accepted November 18, 1996 10R Analyst, January 1997, Vol. 122 Critical Review Development and Operating Characteristics of Micro Flow Injection Analysis Systems Based on Electroosmotic Flow A Review S. J. Haswell School of Chemistry, University of Hull, Hull, UK HU6 7RX Summary of Contents Introduction Fabrication Pump Design Injector Design Reactor Design Detector Design Future Trends Conclusion References Keywords: Micro flow injection; electroosmotic flow; miniaturisation; micro total analytical system ( mTAS) Introduction Over the past 50 years, the design of instt of chemical, biological and physical parameters has brought analytical chemistry to a highly automated and technologically advanced science.It would be naive, however, to believe that the chemistries and physics associated with present day analytical measurements, are in any way limited by current technology. On the contrary, laboratory based analysis has been almost exclusively designed, for the ergonomic needs of the human worker. There can be little doubt that this preoccupation with the physical size and operation of equipment, for laboratory based measurements, has significantly influenced the teaching and practice of analytical chemistry.If one considers the fundamental chemical and physical phenomena on which measurements are based, then it is apparent that one only requires a few molecules or atoms to be present for quantitative measurement to occur.In short, the physics and engineering of measurement science can be considered as oversized and in many cases over engineered to meet the needs of the chemical measurement. Whilst instrumental design has relentlessly moved towards automation, the realization that the quality of analytical data can be profoundly influenced by factors such as sampling, sample storage, sample pretreatment and matrix interferences has led to the concept of the total analytical approach.This more holistic view of analysis, in which traceability and uncertainty prediction can be carried out, is often embodied in the concept of the ‘total analytical system’ (TAS).1 There are, however, numerous disadvantages to TAS, including slow sample processing, lack of selectivity and high reagent consumption;2 however, many such problems may be overcome through system miniaturisation. The term micro TAS (mTAS) was first used at the Transducers 89 Conference.3 In its simples sense, mTAS involves the miniaturisation of all the functions found in an analytical method for example, pumps, valves, flow manifolds, mixing and reaction chambers, phase and analyte separation, detectors, control and communication electronics. One of the more exciting prospects of the mTAS concept is the suggestion that the entire chemical measurement laboratory could be miniaturised onto a device of a few square centimetres.4 This type of miniaturisation has become possible largely through the adoption of microfabrication techniques developed in general by the microelectronics industry, and one can consider miniaturisation of chemical reactors and their corresponding instrumental requirements to be in a similar position to the microelectronics industry 30 years ago.Interestingly, it was from the heart of the microelectronics industry that Terry et al.5 were one of the first to demonstrate that integrated circuit (IC) technology could be used to fabricate a miniature GC instrument using a 5 in silicon wafer. Although the work of Terry and co-workers served to point the way, it was not until the advent of capillary electrophoresis (CE)1,6–8 and in particular, the exploitation of electroosmotic flow (EOF), that the realization of mTAS came about some 10 years after the miniature GC work was first reported.Stephen J. Haswell is a Senior Lecturer in Analytical Chemistry at the University of Hull, UK.Following a period of five years working in the food and plastics industries, he undertook a Ph.D. in the area of atomic spectroscopy graduating in 1983 from the University of Plymouth. His current research activities are in the areas of mFIA, microwave enhanced reaction chemistry, trace elemental speciation and chemometrics. He is author of over 100 research papers, books and patents and presenter of numerous national and international lectures on his research activities.The subject of this present review has become a great interest of the author over the past five years. Analyst, January 1997, Vol. 122(1R–10R) 1RA significant research base has now built up in the area of mTAS9–14 and the development of mFIA over the past 5 years can be attributed to key research carried out by groups based in Basle, Switzerland,14 Texas Tech. University, USA,15 University of Alberta, Canada,16 the Oak Ridge National Laboratory, Tennessee, USA,17 and in a very modest way in the author’s own laboratory.18 What follows is a review of the features present in a relatively small, but rapidly expanding research base, which can be exploited to develop micro flow injection analysis (mFIA) systems, based on electroosmotic flow, to produce reliable sensor-type devices which incorporate mechanical and chemical robustness. Fabrication The fabrication of mFIA systems has attracted various approaches in recent years, ranging from manifolds based around CE-type fused-silica capillaries,1 to modular micro pump systems.19,20 This particular section, however, will focus specifically on fabrication methodology which leads to the production of a single integrated device, designed to use electroosmotic mobility for sample and reagent pumping.18 The discussion will focus almost exclusively on micro machined monolithic device fabrication, which adopts standard IC photolithographic, wet etching and bonding techniques to produce a planar structure.21,22 As the techniques used are based essentially on IC technology, they represent well established methodology, outlined schematically in Fig. 1. If one considers the current ‘state-of-the-art’ in IC fabrication, then it is clear that the challenge of producing mFIA manifolds can in no way be considered as technologically demanding. From what follows, it will become clear that mFIA device fabrication can be relatively simple and offer a reasonably high degree of flexibility in systems design.The choice of a suitable substrate into which the channels of a mFIA manifold can be etched is closely related to factors such as the fabrication method, the analytical chemistries involved and the proposed pumping, injection and detection systems. The obvious candidate given the close relationship with IC technology is silicon,23 which is amenable to the fabrication of structures in the nano- and micro-metre range.This crystalline material can be obtained in a very pure form, at relatively low cost, and offers good mechanical and chemical properties. Whilst silicon may seem an ideal microengineering candidate for mFIA fabrication, it is by nature a conductor and so surface modifications will be required if electroosmotic pumping of samples and reagents is to be achieved. This can be carried out by coating the silicon with SiO2 or Si3N4 to produce the required surface chemistries.Harrison et al.16 described the fabrication of such a manifold, in which the properties of silicon and oxide/nitride-modified layered surfaces were evaluated. They reported that the operating voltage achievable with such devices will be limited by the quality of the oxide insulating film produced. Nevertheless, the results are sufficient to demonstrate that silicon-based devices using a sandwich structure (i.e., oxide–nitride–oxide) can sustain potentials in the range 400–1200 V before dielectric breakdown, which is certainly sufficient to generate EOF mobility for mFIA systems.Although silicon does offer the attraction of producing high precision engineered mFIA manifolds, with well characterised surfaces and the potential to integrate control and detection on to one substrate, other materials such as glass24 and silica (quartz)25 are also amenable to IC based fabrication techniques and will produce surfaces eminently suitable for EOF pumping.Judging from the reports in the literature, glass and silica have proved to be the most popular materials for device fabrication to date, but metal, plastic and ceramic substrates are also possible candidates for the manufacture of mFIA systems.26,27 In general, three basic concepts are required for the preparation of mFIA systems: (i) a suitably prepared substrate, (ii) photolithographic equipment and (iii) wet etching and bonding facilities.28 Prepared substrates based on glass, silica and silicon can be obtained commercially from photomask producers (e.g., Alignrite, Wales, UK); these come ready coated with a metal film (0.1 mm) such as chromium over which is spun a positive photoresist layer (0.7 mm).These plates, which are typically 152.4 mm square (5 3 5 in) and 3 mm thick, can be used directly for manifold pattern transfer using photolithographic methods. A mask or negative of the final channel patterns required for the mFIA system can be produced using CAD computer software.29,30 The mask pattern can be transferred to the photoresist film on the substrate, using basic photographic development equipment.For channels larger than 1 mm, visible or UV light can be used to transfer the pattern from the mask to the photoresist (exposure times being approximately 5 ms); however, to obtain good line definition and submicro patterning, X-ray or electron-beam (e-beam) photolithography will be required. Where facilities permit, e-beam photolithography offers at present one of the most flexible ways of producing mFIA manifold designs; however, for most applications simple basic photographic equipment will suffice.Once the pattern has been transferred to the photoresist film on the substrate, the plate can be developed and wet etched. If one uses the large plates as described, then multiple devices can be prepared; in our laboratory, for example, 25 manifolds of various geometries and channel widths are produced from one plate (Fig. 2).18 Workers in the field tend to prepare their own plates rather than using a commercial source and, although this adds additional steps to the fabrication process, it is not too technically demanding.29,30 As indicated previously, these are typically made from silica or glass on to which a thin metal film (0.05–0.1 mm) of chromium, gold or a combination is produced, using sputtering or chemical vapour deposition (CVD) methods. The metal film is then spin coated with a layer of positive photoresist (approximately 0.5–2 mm).Fig. 1 Schematic diagram of the steps involved in the anistropic etching technique for the fabrication of mFIA devices. The arrow in Step 1 represents radiation. 2R Analyst, January 1997, Vol. 122Following the photolithographic transfer of the manifold pattern, the exposed photoresist and corresponding metal film are removed using commercially available reagents. The prepared plates may then be placed in an oven at 120 °C for 72 h to harden the photoresist and remove solvent residues from the plate.18 It should be noted that although the presence of a metal film on the glass is important for controlling the degree of surface etching that will occur, the exposed substrate will experience undercutting or sideways etching, which will proceed at a rate of approximately 2 : 1.Thus, as the etch goes down to 1 mm it spreads laterally 2 mm at the edges, producing channels, for which the width always exceeds the depth.31 Clearly, for a given substrate, the choice of the initial photomask line width and the time of etch will influence the size of the final channels produced.Various reagents have been used for the wet etching of the substrate, but in general for glass and silica, hot (70 °C) dilute HF–NH4F (1% HF + 5% NH4F in water)18,29 will give an etch rate of approximately 0.3–0.5 mm min21, whereas dilute HF– HNO3 produces an etch rate of between 0.5 and 0.8 mm min21 in Pyrex.30 Silicon etching requires the use of reagents such as ethylenediaminepyrocatechol (115 °C, etch rate 2.5 mm min21) or KOH.16 Although indicative figures for etch rates can be obtained,31,32 it is strongly recommended that sample strips of the substrate are tested in the selected etchant to establish the rate of etching for a given system.It should be stressed that the final size of the mFIA channels will depend on the initial width of the lines on the photomask, the substrate material and mode of etching.During the etching process, agitation of the substrate or etchant is advisable otherwise asymmetric channel formation, i.e., uneven etching on the bottom edge or side of the channel, may occur. It is usual, however, to obtain a channel profile that is wider at the top than the bottom. This effect, associated with undercutting, will vary as a function of the etch time, substrate type and reagents used. The surface quality of the etch is generally related to the types of material used.Good quality silica, for example, gives a well defined etch, whereas Pyrex or borosilicate glass can produce a rougher surface owing to the crystalline structure of the material. The real effect or significance of the surface properties in mFIA channels has not yet been fully characterised, but clearly the more controlled the etch, the more precise and smaller will be the channels that one can produce.What is not apparent from experimental results is the real influence that surface topography may have on the mobility and reactivity of reagents in a mFIA system. Rough or poorly defined surfaces and intersections, of one or more channel, will increase the effect of turbulence and in turn promote dispersion, which in some instances may prove to be an advantage where mixing is required. This factor is particularly important in capillary systems where the Reynolds number is lower than the transitional values of 2000 or 2300 indicating laminar flow, thus minimal mixing will dominate.33–35 However, research indicates that at low Reynolds numbers microfluidic mixing cannot be simply classified as a laminar or turbulent model.36 What is clear is that as channel sizes become smaller, surface effects will inevitably become more significant, which in turn will have an impact on fabrication and material specifications. Manz and Simon37 demonstrated that for a 10-fold decrease in channel size, a 1000-fold decrease in reagent consumption and a 100-fold decrease in related time variables would be obtained.Pressure requirements of such a system will, however, increase by a factor of 100, but as indicated later this does not effect the voltage requirements for EOF. Once an etched base has been produced, the top of the channels need to be sealed. Various bonding methods have been suggested, including glueing, low temperature bonding or annealing, high temperature fusion and anodic bonding.38 The use of an adhesive in such systems can pose problems due to channel plugging and, although this may be overcome to some extent by using photosensitive dry films laminated on to the substrate surface as a protective coating,39 the technique has found little real use in mFIA.Thermal bonding or annealing of top plates represents one of the simplest and therefore most widely used methods in device fabrication.These methods usually involve heating the substrate and top plate, which may be under slight pressure,30 in an oven or furnace to near the upper annealing temperature. For glass and silica this is usually between 500 and 600 °C, whereas silicon will require higher temperatures of around 900 °C. The period of heating varies between 48 and 96 h, after which the device is slowly cooled (48–72 h) to minimize physical stress. For glass substrates, hydrolysing the surface with dilute NH3–H2O2 followed by heating at 500 °C for 24 h29 or 575 °C for 96 h18 has been demonstrated to achieve good reliable bonding.Slightly more complex temperature programmes have been suggested, but these are essentially slight modifications to the same basic method.30 In order to keep the fabrication of a mFIA manifold as simple as possible and to increase the bonding success rate, it is advisable to use the same material (i.e., similar thermal expansion coefficients) for both the base and top of the device, so minimizing stress features. The fabrication of multilayer devices can become complex and techniques using intermediate layers have been described in which more complex structures such as pumps and valves are required.40,41 One technique which has found particular favour with workers fabricating microvalves, micropumps and silicon devices generally is anodic bonding.42,43 In this process, ions such as sodium and oxygen are thought to migrate at elevated temperatures to electrodes placed on the outer surface of one of the layers, so increasing the electrostatic surface charge.The temperature and applied voltages used will depend on the materials in question, but for glass these are found to be 200–400 °C at 30–300 V and for silica 700–800 °C at 30 V applied for a period of 45 min. It might be necessary when employing anodic bonding to treat the surface with HF prior to bonding. Holes or ports made in the top plate are commonly used as reagent or sample feeds into the manifold and can be generated either before18 or after29,30 the bonding process.The holes can simply be produced by using mechanical18 or ultrasonic16 drilling of the substrate and vary in diameter from 0.5–2 mm. At present, laser ablation, which offers an attractive method for hole production, has not been employed, but no doubt the technique will find its way into the literature before long. Plastic Fig. 2 Example of the photomask design used to produce five mFIA manifolds designs with variable line or channel widths. The line widths shown are A and E, 50, B, 30 and C, 10 mm.Analyst, January 1997, Vol. 122 3Rreservoirs are usually glued to hold liquids and support the platinum electrodes when required. A photograph of a completed device used for phosphate and nitrite determination is shown in Fig. 3. To date, only planar mFIA systems have been reported based on the fabrication techniques described. The fabrication of three dimensional systems is, however, an attractive prospect and no doubt stacked systems will be produced as more sophisticated chemistries are exploited.Such systems based on micropumps have been reported in which multi-layers or modules are clamped or stuck together to produce a complete device.19,20 Once a complete mFIA manifold has been fabricated for use with EOF pumping, the success of the bonding process can be checked by filling the channels with a buffer44 or weak acid18 and plotting the current–voltage relationship.This plot is generally found to be linear up to 10 kV in glass and silica, after which point dielectric breakdown of the substrate occurs. Any deviation in linearity at lower applied voltages is indicative of a device failure, associated with incomplete bonding. Pump Design In order to achieve reliable flow dynamics in an FIA manifold, which may contain numerous reagent streams, pulse free, constant flow characteristics must be maintained.It is not surprising therefore, to find that mFIA systems also call for pulse free, variable nl min21–ml min21 flow control. Although a direct syringe or piston pump can be used with mFIA systems,45,46 the closest equivalent to a conventional peristaltic FIA pump is the so called micropump, which usually takes the form of a pulsating one-way valve driven typically by a piezoelectric device.36,45,47–50 Based on microengineering technology, micropumps have been employed in mFIA manifolds consisting of channels 100 mm wide, 10 mm deep and 4 cm in total length.47 When such small channel dimensions are employed, the hydrodynamics of the system can produce a pressure drop of 0.7 atm at a flow rate of 0.5 ml min21. However, diaphragm devices, commonly using microchemical silicon membranes, are only reliable up to backpressures of 0.2 atm and so high pressure pumps or large channel dimensions are therefore required.High pressure pump designs using nickel rather than silicon check values are now being developed using low temperature bonding techniques, incorporating intermediate photoresist layers to adhere the valve to the body of the pump.47 These devices, which use a piezoelectric disc glued to the outer surface of the pump housing, operate at many hundred Hertz, requiring around 300 V to produce flow rates in the range 20–300 ml min21. In an elaborate example, four such pumps have been used to drive reagents through a three-dimensional system constructed from silicon and glass at a flow rate of 1 ml min21, the channel dimensions being 600 mm wide and 200 mm deep in order to accommodate pressure effects.19 Diaphragm micropumps do have an important role to play in applications where electroosmotic pumping is not applicable, owing to sample or reagent chemistries.It is essential, however, that micropumps remain chemically inert to the reagents and samples with which they may come into contact.In addition to piezoelectrically driven pumps, ultrasonic51 and other more exotic electrically activated pumps have also been described.36,52–54 Electrophoresis embodies two basic electrokinetic components, electroosmotic and electrophoretic flow. These two components combine to give the total flow or velocity (n) in the following way: total flow rate = n = (meo + mep)E (1) where meo = electroosmotic mobility, mep = electrophoretic mobility and E = applied electric field.At relatively low electric field strengths, electroosmotic mobility represents the larger component of the two processes; however, as the field strengths are increased the influence of the electrophoretic component (migration of analyte ions) increases also. For mFIA, where bulk mobility is desired, one finds that field strengths around 300–400 V cm21 are usually adequate; however, if separation is required, as in CE, then field strengths greater than 1 kV cm21 are typically required.It is common for mFIA systems based on EOF to operate with field strengths of the order of 80–300 V cm21, at which little or no analyte separation is observed The generation of an EOF to pump reagents and samples through a mFIA system is subject, however, to certain physico-chemical limitations. Firstly, in order to generate the EOF one must use a material which will yield negatively charged groups on the surface or walls of the channels when placed in contact with an appropriate liquid.Secondly, the liquid phase must dissociate to some extent in order to generate counter positive ions (notably H+ ions). The combination of the negatively charged surface (typically SiO2) and the H+ ions in solution will form a diffuse double layer (sometimes referred to as the Gouy or Helmholtz layer). This diffuse double layer acts as a parallel-plate electric capacitor whose plates are d cm apart each carrying a charge e per cm2.The zeta potential (x) will be the potential difference between the plates, given by the general equation x = 4ped/er (2) where er is the relative permittivity of the medium between the hypothetical plates. If an electrical field is now applied through the liquid phase, ions will migrate to their respective electrodes dominated by the positive ions (H+) moving to the negative or ground electrode. As the ions move, they induce a drag on the bulk of the liquid, which in turn results in a corresponding net transfer of the solution to the negative electrode. Thus, in EOF the mobility of the solution is from the anode (positive electrode) to the cathode (negative electrode) or ground.In practice, the formation of the double layer is limited to the pH range 4–10. At lower pH values the cationic population becomes so high that the EOF is overrun by conductive flow and at pH values greater than 10 the cation population becomes too low to sustain the double layer.The zeta potential that is generated when the double layer forms will be influenced by changes in pH and ionic strength of the solutions in the channel, and this can affect the corresponding flow rate. Clearly, as one of the main attributes of a pump in FIA is to maintain a reproducible flow rate, the pH and ionic strength of the reagents Fig. 3 A fully assembled mFIA device showing the mounted plastic reservoirs which act as wells for the reagents and samples and support the platinum electrodes required for EOF pumping. The device is shown mounted in a rig which allows fibre optics (seen either side of the block) to be coupled into relevant sections or channels of the manifold. 4R Analyst, January 1997, Vol. 122used in a mFIA manifold need to be controlled, through the use of buffer systems. As the ionic strength increases, for example, a counter-ion effect will prevent the ions from migrating through the solution independent of the EOF.In general, it is therefore preferable to keep the total ionic strength of reagents and samples as low as possible. Once an EOF has been generated, the flow establishes a flat, trapezoidal profile notably different from the parabolic (bullet) shape commonly associated with conventional FIA. Closer investigation of the mFIA profile reveals that the edges of the profile (i.e., closest to the wall surface) are slightly in advance of the bulk owing to the drag effect and the extent of this effect will be a function of the solution viscosity.55 The fact that the flow profile, which offers minimal band broadening, does not contain the parabolic flow characteristics of a pressure or hydrodynamically pumped system, does have implications with respect to the mixing of reagents in the mFIA reactor, where some dispersion will be necessary to achieve the desired chemical reactions.Thus EOF-pumped mFIA can be considered to be more akin to segmented flow analysis rather than conventional FIA, in its flow characteristics.In an EOF-pumped mFIA system, the flow rate can be effectively controlled by varying the applied voltage as follows: n flow rate v = m (3) L where n = applied voltage (V), L = length of the channel (m) and m = sum of the electroosmotic and electrophoretic mobility component under different conditions of pH and ionic strength. The EOF and hence the flow rate in mFIA, will be unaffected by capillary diameter up to channel sizes approaching 250 mm, after which drag effects of the bulk solution may cause serious disruption to the EOF.The EOF produced in a channel can be considered, in the electrical sense, as a resistor, thus Joule heat will occur in the capillary. However, the almost negligible bulk properties of components in a channel, relative to the substrate, will ensure an effective dissipation of heat, which rarely poses a serious problem in mFIA.If, however, the resistance is allowed to increase, for example when a highly viscous or an immobile solvent is present in a channel, heating can occur. It should be stressed that an EOF can only be established if the double layer is formed, and this requires ions or dipoles to be present in the liquid stream. In the case, for example, where organic compounds or solvents may be present, such conditions may not be met. For example Zheng and Dasgupta55 described some initial studies, using silica capillaries, to evaluate the suitability of an EOF in a mFIA system for carrying out in-line analyte phase separation or solvent extraction.They described the use of a 50 cm 3 7.5 mm id polyimide coated silica capillary to investigate various aqueous ionic complexes, based on well characterised cationic, anionic and neutral ion-pair chemistries. A quaternary ammonium salt, tetrabutylammonium perchlorate, was added to chloroform as a supporting electrolyte to assist in the mobilization process56 and to catalyse phase transfer.57 The results clearly indicated that organic solvent pumping was possible, when modified with the quaternary ammonium salt, but that migration of the ammonium ion occurred, creating a positive front end to the solvent slug.Bleeding of ions into the secondary aqueous buffer from the organic phase was also reported. Furthermore, they suggest that a thin interfacial layer of buffer is generated between the organic solvent and the capillary walls which enables the EOF to be generated.Investigations into solvent extraction indicated that the ion-pair complexes studied were either extracted into the organic phase or accommodated in the aqueous phase ahead of the organic slug interface. Not only were the findings of Zheng and Dasgupta a significant contribution in terms of demonstrating phase extraction and organic solvent mobility using EOF in a mFIA system, but they also clearly indicate that the migration of ions within a solvent under the influence of an electric field could lead to the development of gradient and separation techniques, complementary to CE, such as selective solvent extractions and matrix modification.Interestingly, the use of micellar electrokinetic capillary chromatography, described by Moore et al.,58 suggests that multiphase systems could be developed for mFIA applications, thus offering significant advantages for organic solvent based chemistries.Although it is possible to use EOF as the primary pumping mechanism in a mFIA manifold, it may be preferable or even necessary to isolate the pumping mechanism from the injector, reactor and detector components of the system. Such occasions might be, for example, when the chemistry of the method or detector system is not compatible with the electrical field required for direct EOF pumping. One such system has been described by Dasgupta and Liu45 in which a section of polyimide-coated fused-silica tubing (40 cm 3 75 mm id) was connected to a second capillary, via an isolating membrane.In the first channel a 2 mm borax buffer is pumped by direct EOF, with a field strength of around 40 V cm21. As the flow in the pumped capillary was electrically isolated from the second capillary it produced a hydrodynamic effect, sufficient to create a flow in the second channel, which could be varied between 1 nl min21 and 100 ml min21. In this case the hydrodynamic effect of the EOF is exploited in a mFIA manifold without the need for a direct electrical field.Injector Design The introduction or injection of a sample into an FIA manifold can generally be classified into two general types. The first is the timed or gated injection in which a sample is drawn into the FIA manifold, usually through a sampling probe, for a controlled period of time, after which the flow is switched back to the carrier stream. In this way, a variable volume can be injected into a manifold as a function of time.The second and more usual method of injection in FIA, is the introduction of a constant sample or reagent volume into a mobile carrier in a way that affords minimal flow dispersion. The most common approach for such systems is to use a rotary or slide valve, which contains a sample loop of a defined volume. It would therefore seem appropriate to have both these forms of sample injection available in mFIA systems, and this is indeed found to be the case.The injection of a sample into a mFIA manifold can be performed using hydrodynamic/pneumatic pressure control,46 miniaturised valves1 or electrokinetic mobility based on EOF.15,18,59–62 Even the lowest volumes obtainable with traditionally based rotary-type valves63 are clearly too large for practical use in mFIA. However, Liu and Dasgupta1 recently described the use of commercially available valves with an injection volume of 60 nl.It should be stressed, however, that most of the work relating to the introduction of small sample volumes ( < 20 nl) into capillary systems has been focused on CE methodology rather than addressing mFIA systems. Of particular relevance to mFIA systems is the development of the so-called valveless injection method, which uses EOF control in conjunction with specific capillary channel geometries.64,65 The two simplest approaches to sample injection based on EOF are first to pump a sample for a given period of time into a flow channel and second to fill a defined volume (equivalent of a sample loop) built into the mFIA manifold.In the former case, Zheng and Dasgupta55 described a modified CE system in which an organic solvent was loaded for a given period of time into a mFIA system consisting of a 50 cm 3 75 mm id polyimide Analyst, January 1997, Vol. 122 5Rcoated fused-silica capillary. The solvent was introduced into the system by placing the capillary electrode in the solvent reservoir at 15 kV for 10 s.Following the injection step, the capillary electrode was switched to a borate buffer which subsequently pumped the solvent through the capillary to the detector. Using this approach, the authors reported an RSD of 0.51% associated with migration effects and 1.7% for the peak, suggesting that surface tension, viscosity and flow rate will influence this particular mode of injection. It should be noted that in this work, the authors were using the method described for introducing a solvent into the mFIA system for the purpose of solvent extraction.The second approach to sample introduction, based on defined volumes using EOF, falls into two categories, the Xand Z-type injections (Fig. 4). Of the two injection geometries, the X or cross design, has been most widely investigated. Depending on the geometry and electrical field used, the sample injection can be classified as ‘floating’ (gated) or ‘pinched’ (discrete), with the former occasionally being referred to as the continuous mode.59,60,64,65 Experimentally, the simplest of these two modes is the ‘floating’ method, as it only requires one pair of electrodes (Fig. 5). In this case the sample is pumped by an EOF from the sample reservoir (A the positive electrode) to the sample waste reservoir (B the negative electrode) along channel AB crossing part of channel CD. Note that the channel AB will have been filled prior to the sample introduction with a suitable buffer.The reservoirs C and D contain no electrodes and therefore have no applied field, hence their potential is floating relative to the field in channel AB. As the sample passes across the intercept of AB and CD, diffusion and eddy effects will allow some of the sample to migrate in the direction of both C and D [Fig. 5(a)]. Subsequently, on placing the electrodes in reservoirs C and D the field can be made to run in the CD direction, so any sample molecules in the intercept will be pumped towards D, the grounded reservoir [Fig. 5(b)], allowing the sample to pass via a detector on its way.Clearly, we can see that the leakage or migration of the sample solution into channel CD during injection and the possibility of dragging the stationary sample from channels A and B into D as a function of flow [Fig. 5(c)] will lead to an uncontrolled volume injection. Examples of this effect have been reported.59,60,64 The leakage observed into the main channel may not be a serious problem for certain applications and clearly it offers a simple method of operation, in which only one pair of electrodes are required.However, the quality of etch, surface topography and geometry of the channels concerned will influence the degree of diffusion using the ‘floating’ injection method. The need, especially in CE, to have more precise control over the sample volume injected has lead to the development of the so-called ‘pinch’ method (Fig. 5). In this case, the sample is once again pumped under an EOF from reservoir A to B, but this time reservoirs C and D are not electrically floating, but like A are given a positive potential, relative to the waste reservoir B [Fig. 5(d)]. The effect of this is to draw buffer from channels C and D into B along with the sample from A. Under these flow dynamics, the sample gains a pinched or trapezoidal shape at the intercept of channels AB and CD.When the flow is redirected towards D [Fig. 5(e)], the sample volume, which may be only a few picolitres, is of a more precisely defined volume, less affected by diffusion effects. Jacobson et al.60 have reported improvements in the RSD from 2.7 to 1.7% for a 90 pl injection volume using the ‘pinch’ method, compared with the ‘floating’ approach. In this particular study, the applied voltages based on a 1 kV power supply were reservoir A 90, B 0, C 90 and D 100%, giving channel field strengths of 270, 20, 400 and 690 V cm21 in A, B, C and D, respectively.Clearly, operating the manifold in a multi-electrode configuration will reduce dispersion effects at intersections and allow improved control of the sample injection volume. The concept of the ‘push-back distance’ associated with surface diffusion has been used66 to define the migration rate and shape of flow patterns at X-type intercepts. This approach to sample injection has considerable potential in controlling sample volumes as a function of applied field.The larger volumes (9–22 nl) of the Z-type injection (Fig. 6) are introduced by pumping the sample between two reservoirs across a defined volume in a Z geometry. In the example shown, the sample is pumped from reservoir C or D to the waste reservoir D or E, so filling a defined volume in channel AX. The loaded sample can then be reacted with reagents in channel AX and moved to a detector or be monitored in situ.The Z mode of injection produces a larger volume than that obtained by the X mode, which in turn will be less influenced by the diffusional effects described previously.18,67 Recent work in the author’s laboratory using the Z injection technique has indicated that Fig. 4 Discrete volume injection using (a) the X and (b) the Z approach. Fig. 5 Flow and dispersion characteristics of floating and pinched modes of injection. 6R Analyst, January 1997, Vol. 122pinch control offers no improvement in precision over the floating method.68 The ability to matrix modify a sample in an injector system could be an attractive option in mFIA and techniques such as synchronized cyclic capillary electrophoresis22,69 or sample fractionation67 may offer considerable scope for such methodology. In these systems selected, analyte fractions can be separated from a more complex and possibly interfering matrix, owing to the variation in migration and flow direction under electrophoretic conditions.The switching of electrodes can be used to move components along channels and across channel intersections akin to shunting railway trucks, until the analytes required become isolated from matrix or interfering components. This aspect of sample manipulation is potentially very exciting for mFIA system development and will no doubt be exploited in future applications. Reactor Design Having satisfactorily injected a sample into an FIA system, the next objective is usually to present the sample to a detector in an appropriate form as quickly as possible, using tube or channel geometries that minimize the dispersion of the sample slug.In mFIA systems sample diffusion is known to be a function of the square root of the time after the injection of a sample.69 There remains one additional and often complex step, however, that of achieving the appropriate chemistry or biochemistry necessary for the detection of the analyte of interest.This process typically requires the addition of at least one reagent and can include more complex steps such as in-line solid-phase extraction and multiphase separations.70 Indeed, the complexity of the FIA manifold is only limited by the imagination or ingenuity of the analyst. The first and perhaps the most important step in reactor design is a consideration of the physical and chemical characteristics that occur when two solutions combine at a channel junction in a mFIA manifold. From the previous section, it is clear that even at Reynolds numbers less than 2000 some turbulent mixing takes place at channel intercepts and that the electrical potentials at which the channels are held can significantly influence the interfacial or mixing zone.Turbulent mixing, for example, has been reported in channels 5.2 mm deep and 57 mm wide.71 The characterization of channel intersections indicates that leakage or diffusion of reagents from a ‘floating’ side stream into a channel in which EOF is present is around 2–5%, depending on the viscosity and flow rates.64 This bleed occurs as a result of hydrodynamic effects, in which the molecules in the pumped stream entrain the stationary side channel molecules owing to frictional and eddy properties in a Venturi- or Bernoulli-type effect.What is interesting with such systems is that the flow characteristics in the main channel will have a flat, trapezoidal profile whereas the side channel, whose flow will be pressure driven, will have a parabolic profile.Hence the profiles of the merging streams may have a significant effect on the mixing characteristics at such an interface. As one of the main objectives of reagent addition in an FIA system is to induce mixing and so achieve the required reaction, some consideration needs to be given to this aspect of pumping in mFIA systems. In a valuable study, Seiler et al.44 demonstrated that the individual electrical resistances in a series of interconnecting channels can be used to predict the flow characteristics, in a similar way to hydrodynamic estimations in conventional FIA.The basic concept considers the intercepting channels as electrical resistors and evokes Kirchhoff’s rules,72 to predict the net flow of current which can be attributed to the flow dynamics of the device. One of the important findings of the work by Seiler et al.44 is that by controlling channel voltages one is able to manipulate sample or buffer dilutions by adjusting flow rates, and this offers considerable scope for stopped-flow or reverseflow operations.This effect was also recently demonstrated in our laboratory using a manifold previously described for phosphate analysis.18 Figure 6(a) illustrates how a sample of orthophosphate, held in reservoir D, is injected into channel AX using the Z technique, by pumping the sample to reservoir E.Thus the slug in channel AX defined by the intercepts DE represents the injected sample which reacts with ammonium heptamolybdate (0.01 m) in the presence of ascorbic acid (0.05 m) to produce molybdenum blue. Detection of the molybdenum blue is achieved by measuring the absorbance at 744 nm in a spectrophotometer coupled to the manifold by a fibre-optic system attached along channel AX (optical path 2 cm). Through the selection of different injection volumes [i.e., C–D (9 nl), D– E (13 nl) and C–E (22 nl)] and variation in the positive voltage applied at reservoirs A and B relative to D or C, the absorbance signal was observed to decrease as the intercept ‘pinch’ potential applied across A–E and B–E became sufficient to dilute the formation of the coloured complex at each end of the injection slug.68 In effect, the flow rate in channel A–B was being increased as a function of voltage, so diluting the flow from channels D–E, C–E or C–D into channel AX.Results obtained using this approach (Fig. 7) indicated that calibration based on one standard is possible and that the technique will have an important role to play in the development of future methodology. Following the mixing of the reagent and analyte streams, some period of time is usually required to produce sufficient product, either for further reactions or for subsequent detection. Conventionally, this hold time is effected by using a coil or knotted reactor and/or a stopped-flow mode, in order to minimize dispersion whilst achieving the required reaction chemistries. In mFIA it may be that owing to the more efficient interfacial nature of the mixing, i.e., the need for less bulk mixing, reactions rates may increase; however, reactions are still likely to require a finite period of time for product production.A serpentine-type pattern or structure offers a very simple, but effective, use of space for extending channel length and is compatible with photolithiographic fabrication techniques.The flow characteristics of serpentine channels have been studied and whilst disruption to the electroosmotic flow Fig. 6 Controlled dilution experiments using a pinch voltage. Analyst, January 1997, Vol. 122 7Rwas observed at the 90° bends, no significant band broadening has been found.60 The particular design used had a total channel length of 17 cm operating with a field strength of 700 V cm21 and was used to perform electrophoretic separations.It would seem that using such an approach, time-dependent mFIA reactions could be accommodated in a physically small area with minimal dispersion through the use of serpentine or spiral channels.61 Modification to the surface properties to induce selective flow characteristics, given the limitations of producing a double layer and hence EOF, has attracted some interesting approaches. These include the use of reversed-phase hydrophobic polymers, 73 surfactants,74 silanals and quaternary ammonia groups.75 In addition, the covalent bonding or immobilisation of glucose oxidase has been used to develop a glucose mFIA method based on a serpentine geometry using a 260 cm long 3 100 mm wide 3 70 mm diameter reactor.76 One of the interesting extensions of the mFIA technology described is the development of electrochromatographic (EC) separations in which capillaries are packed with small particles (approximately 3 mm) on which efficient separation can be achieved.77 Using EOF as the primary pumping mechanism, EC can be mediated with reference to the zeta potential and hence separation is potentially possible through selection of the appropriate surface properties of a packing material.Thus the walls of the capillary act to facilitate the primary pumping mechanism and the packed chromatographic material offers enhanced electrophoretic separations. In addition to surface modification, the physical effect of field flow fractionation may also be incorporated as a separation process in capillary systems.78 In this case, an external force, e.g., magnetism or gravity, is used to pull fractions to the wall of a channel; on removing the force, a gradual diffusion of the fractions in the sample occurs back into the flowing stream, usually based on molecular size.79,80 As yet field flow fractionation has not been widely studied in mFIA systems, but the possibility of using field flow fractionation and modifying the zeta potential using a magnetic field might be an interesting subject for future research.Detector Design Much of the research reported to date in the area of mFIA has focused on characterising the physical processes of reagent mobilisation and mixing in microchannel manifolds. Such work has relied heavily on imaging techniques, using, for example, charged coupled device (CCD) cameras60 and microscopybased techniques.30,44,61,81 In this section, consideration is given to the design of detectors suitable for direct integration into EOF based mFIA systems.The major detector systems used in conventional FIA are based on optical absorption/emission and electrochemical techniques,70 and this is also found to be the case for mFIA systems.82,83 Clearly, if electroosmotic based pumping is employed in the mFIA manifold, then some care is required with the design of electrochemical detectors, but this does not pose any real serious limitations.83 Not surprisingly, optical detection has proved to be the most attractive approach to on-device detection and various examples can be found in the literature,15,40,45,67 mainly associated with CE detectors.84 One of the less obvious advantages of miniaturisation is the ability to introduce detector systems not readily available to conventional larger flow systems, such as mass-selective devices of the surface acoustic wave type.85 Clearly, the design of a flow cell in terms of physical volume and optical geometry is very important in miniature systems if sensitivity, reliability and robustness are to be achieved.The incorporation of fibre optics as part of an optical detection system has proved to be of value, particularly with the advent of fibres with diameters of the order of 0.1 mm.86 In the area of optical detectors, some effort has gone into developing axially rather than perpendicularly oriented measurement cells.87 Increasing the volume of the flow cell in a perpendicular viewing axis has been used to increase the optical pathlength,88 whilst the use of multiple reflections axially in a flow cell has also been evaluated.89 The first example led to severe dispersion effects and a subsequent loss in sensitivity and the second approach, based on a silicon device, suffered from signal loss due to scatter and absorption of light at the silicon surface.One interesting approach to absorbance measurements is to consider the mFIA channel as an extension of a fibre optic system.18,90 In this way, total internal reflection of light may be achieved and any photoactive species spatially present in the channel will undergo interaction with the photons present, to produce either a direct absorption measurement or a subsequent fluorescence effect.As with most absorption methods, it would be preferable to use a dual-channel system to improve stability and reduce scatter.Although a long pathlength is appealing for absorption measurements, emission-based techniques would benefit from a small spatial volumes in which the emission effect can be concentrated, and volumes less than 1 nl have been suggested.89 The size of the detector cell is clearly related to the concentration and sensitivity of the method in question and a flow cell of 15 ml has been reported to be adequate for chemiluminescent measurements of glucose and lactate in human serum.91 More recently Liang, et al.92 have described a UV cell with a parallel flow optical path of 120–140 mm for absorbance and fluorescence detection at the end of a CE column, which would be most suitable for mFIA applications.Sequential detector arrays are an attractive approach in sensor design and lend themselves well to mFIA systems.82,93 One such system82 has been described, based on electrochemical detectors, for liquid chromatographic separations of catecholamines and consisted of four photolithographically prepared ISFET sensors aligned sequentially in a 5 mm silicon channel 100 mm wide and 70 mm deep.The total volume of the detector was 20 ml. Another device,93 using a peristaltic pump to control flow rates, consisted of nine 5 mm ISFET sensors housed in a 15 ml cell, used for pH, potassium and calcium determinations in biological fluids. Although neither of these systems was used in mFIA manifolds, they do illustrate the ability to develop array detectors compatible with mFIA technology. One of the most exciting prospects for mFIA detector design is the ability to incorporate their fabrication into one integrated device.For example, miniature mass spectrometers (3 3 3 3 3 mm) have been produced based on fabrication technology that would be ideally suited for mFIA.94 Miniature spectroscopic systems are also becoming available95,96 and future developments in spectrometers incorporating opto-electronic systems Fig. 7 Calibration based on 100 ppb PO4 with dynamic applied pinch voltage. The equivalent absorbance values for 100, 50 and 25 ppb phosphate are indicated on the x-axis. 8R Analyst, January 1997, Vol. 122will undoubtedly bring valuable complementary technology for future mFIA detector design. Future Trends The preceding sections have tried to focus on the basic concepts and developments relating to mFIA systems based on EOF. Some indication has been given of the likely areas where current and future research may prove to be of great value.These include the fabrication of devices where there is considerable potential for the construction of stacked or three-dimensional systems, possibly using cold bonding and direct laser etching techniques. The mobilisation of reagents and analytes, based on EOF, requires more complete characterisation if flows and mixing effects at intersecting channels are to be effectively exploited. The most important development if mFIA technology is to be fully realised is the fabrication of self contained operational systems with proven application robustness.The close relationship of mFIA technology with separation techniques such as capillary electrophoretic and micro-electrochromatographic separations may well lead to some form of hybride system in the near future. The area perhaps where the greatest advances in mFIA-based technology will be most readily realised is the biotechnology sector, where methodology and applications are complementary to miniature systems. Already examples are emerging of applications in DNA fragment analysis97 and immunoassay methodology.98 Developments, however, need to be focused not only on the integrated device which may be encapsulated and equipped with telecommunication for remote operation, but also to include interfacing to existing measurement systems such as MS or NMR, which would benefit from some form of sample pretreatment.One area which has not yet been considered in mFIA systems is a return to the early gasphase work started by Terry et al.5 Clearly there are some exciting possibilities in gas and multiphase systems yet to be realised. Conclusion Where has mFIA and more generally mTAS got to? In 1991, Manz et al.13 questioned whether the developments in mTAS were ‘a look into next century’s technology or just a fashionable craze’. In their concluding remarks, the authors made some general comments relating to the uptake or acceptance of mTAS technology, ‘namely that changes are required in the political and cultural opinions of analytical work if appropriate financial support is to be forthcoming and that the research base must grow worldwide to foster both competitive and collaborative research’.Further, they identified that ‘the market acceptance of the technology must be embraced through the design and production of mTAS concepts’. Clearly, these requirements have been only partially fulfilled. An examination of the literature indicates that there has been a positive growth in the research base and this will obviously support the fundamental development of the science.What is less obvious is the political and more importantly the economic will to support the development of the technology, and this may yet be seen to be the most seriously limitation to the growth of the science. Although miniaturisation is conceptually appealing, there remain some important technical obstacles to overcome, such as the introduction of ‘real’ samples and the ability to deal with suspended particles.These limitations are not beyond the scope of present day membrane technology, and should pose no serious hindrance to the advancement of the science. Developments in the field of mTAS since 1991 clearly point to the technique forming the basis of future methodology applicable to a wide range of applications ranging from measurement science to chemical synthesis in which mFIA based on EOF flow will play a significant role.The limitation in releasing the considerable potential that micro reactor technology can offer resides not in the technological challenge but in the imagination of our minds and only requires us to realize it. References 1 Liu, S., and Dasgupta, P. K., Anal. Chim. Acta, 1993, 283, 739. 2 Bogue, R., Lab. Equip. Dig., 1995, February, 14. 3 Manz, A., Effenhauser, C.S., Burggraf, N., Verpoorte, E. J. M., Raymond, D. E., and Widmer, H. M., Analusis Mag., 1994, 22, M25. 4 Craston, D., and Cowen, S., Chem. Br., 1996, October, 31. 5 Terry, S. C., Jarman, J. H., and Angell, J. B., IEEE Trans. Electron. Devices, 1979, ED-26, 1880. 6 Capillary Electrophoresis: Theory and Practice, ed. Grossman, P. D., and Colburn, J. C., Academic Press, New York, 1992. 7 Mikkers, F. E. P., Everaerts, F. M., and Verheggen, Th. P. E. M., J. Chromatogr., 1979, 169, 11. 8 Jorgenson, J. W., and Lukacs, K. D., Anal. Chem., 1981, 53, 1298. 9 Whalen-Pechersen, E. K., and Jurs, P. C., Anal. Chem., 1981, 53, 2184. 10 Manz, A., Graber, N., and Widmer, H. M., Sens. Actuators B, 1990, 1, 244. 11 Advances in Chromatography, ed. Brown, P. R., and Grushka, E., Marcel Dekker, New York, 1993, pp. 1–66. 12 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Kluwer, Dordrecht, 1995. 13 Manz, A., Fettinger, J. C., Verpoorte, E. J. M., L�udi, H., Widmer, H.M., and Harrison, D. J., TrAC, Trends Anal. Chem. (Pers. Ed.), 1991, 5, 144. 14 Manz, A., Harrison, D. J., Verpoorte, E. M. J., Fettinger, J. C., L�udi, H., and Widmer, H. M., Chimia, 1991, 45, 103. 15 Lui, S., and Dasgupta, P. K., Anal. Chim. Acta, 1992, 268, 1. 16 Harrison, D. J., Glavina, P. G., and Manz, A., Sens. Actuators B, 1993, 10, 107. 17 Jacobson, S. C., Hergenr�oder, R., Koulng, L. B., and Ramsey, J. M., Anal. Chem., 1994, 66, 2369. 18 Daykin, R.N. C., and Haswell, S. J., Anal. Chim. Acta, 1995, 313, 155. 19 Verpoorte, E. J. M., Van der Schoot, B. H., Jeanneret, S., Manz, A., and de Rooij, N. F., in Interfacial Design and Chemical Sensing, ACS Symposium Series, American Chemical Society, Washington, DC, 1994, ch. 21, p. 244. 20 Schomburg, W. K., Vollmer, J., B�ustgens, J., Fahrenberg, J., Hein, H.,z, W., J. Micromech. Microeng., 1994, 4, 1. 21 Sensors: a Comprehensive Survey, ed. Grandke, T., and Ko, W. H., VCH, Weinheim, 1989, vol. 1, p. 107. 22 Manz, A., Verpoorte, E. M. J., Effenhauser, C. S., Burggraf, N., Raymond, D. E. and Widmer, H. M., Fresenius’ J. Anal. Chem., 1994, 348, 567. 23 Woffenbuttel, R. F., Sens. Actuators A, 1992, 30, 109. 24 Altmann, J. P., Das Neue Lehrbuch de Glus�alzerei, Gentner, Stuttgart, 1963, p. 29. 25 Daniel, J. S., and Delapierre, G., J. Micromech. Microeng., 1991, 1, 187. 26 Becker, E. W., Ehrfeld, W., Hagmann, P., Maner, A., and M�uchmeyer, D., Microelectron.Eng., 1986, 4, 35. 27 Knapp, J., Andreae, G., and Petersohn, D., Sens. Actuators A, 1990, 21–23, 1080. 28 Kovacs, G. T. A., Petersen, K., and Albin, M., Anal. Chem., 1996, 68, 407A. 29 Jacobson, S. C., Koutny, L. B., Herenr�oder, R., Moore, A. W., Jr., and Ramsey, J. M., Anal. Chem., 1994, 66, 3472. 30 Fan, Z., and Harrison, D. J., Anal. Chem., 1994, 66, 177. 31 Elliot, D. J., Integrated Circuit Fabrication Technology, McGraw- Hill, New York, 1982. 32 Thin Film Processes, ed.Vossen, J. L., and Keren, W., Academic Press, San Diego, 1978, ch. V-1, p. 401. 33 van der Linden, W. E., TrAC, Trends Anal. Chem. (Pers. Ed.), 1987, 6, 37. 34 Tyssen, R., Anal. Chim. Acta, 1980, 114, 71. 35 Hungerford, J., Thesis, University of Washington, 1986. 36 Gravesen, P., Branebjerg, J., and Jensen, O. S., J. Micromech. Microeng., 1993, 3, 168. 37 Manz, A., and Simon, N., Anal. Chem., 1987, 59, 74. Analyst, January 1997, Vol. 122 9R38 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Klewer, Dordrecht, 1995, p. 165. 39 Shoji, S., and Esashi, M., Sens. Actuators B, 1992, 8, 205. 40 Field, L. A., and Muller, R. S., Sens. Actuators A, 1990, 21–23, 935. 41 Fung, C. D., Cheung, P. W., and Fleming, D. G., Microachining and Micropacking of Transducers, Elsevier, Amsterdam, 1985. 42 Harz, M., J. Micromech. Microeng., 1992, 2, 161. 43 Esashi, M., Nakano, A., Shoji, S., and Hebiguchi, H., Sens. Actuators A, 1990, 21–23, 931. 44 Seiler, K., Fan, Z. H., Fluri, K., and Harrison, D. J., Anal. Chem., 1994, 66, 3485. 45 Dasgupta, P. K., and Liu, S., Anal. Chem., 1994, 66, 1792. 46 Li, S. F. Y., Capillary Electrophoresis, Elsevier, New York, 1992. 47 Shaji, S., Esashi, M., van der Schoot, B., and de Rooij, N. F., Sens. Actuators A, 1992, 32, 335. 48 Horowitz, P., and Hill, W., The Art of Electronics, Cambridge University Press, Cambridge, 2nd edn., 1989, p. 368. 49 van der Schoot, B. H., Jeanneret, S., van den Berg, A., and de Rooij, N. F., Anal. Methods Instrum., 1993, 1, 38. 50 Stewart, K. K., Beecher, G. R., and Hare, P. E., Fed. Proc., Fed. Am. Soc. Exp. Biol., 1974, 33, 1434. 51 Moriney, R. M., White, R. M., and Howe, R. T., Proc. MEMS’91, Nara, Japan, 1991, p. 277. 52 Richter, A., Plettner, A., Hofmann, K. A., and Sandmeier, H., Proc. MEMS ’91, Nara, Japan, 1991, p. 271. 53 Osada, Y., Okuzzaki, H., and Hari, H., Nature (London), 1992, 355, 242. 54 Fuhr, G., Fiedler, S., M�uller, T., Schnell, T., Glasser, H., Lisec, T., and Wagner, B., Sens. Actuators A, 1994, 41–42, 230. 55 Zheng, H. J., and Dasgupta, P. K., Anal. Chem., 1994, 66, 3997. 56 Adams, R. N., Electrochemistry at Solid Electrodes, Marcel Dekker, New York, 1969. 57 Masson, D., Magdassi, S., and Sasson, Y., J. Org. Chem., 1990, 55, 2714. 58 Moore, A. W., Jr., Jacobson, S. C., and Ramsey, J. M., Anal. Chem., 1995, 67, 4184. 59 Jacobson, S. C., Hergenr�oder, R., Koutny, L. B., and Ramsey, J. M., Anal. Chem., 1994, 66, 1114. 60 Jacobson, S. C., Hergenr�oder, R., Koutny, L. B., Warmack, R. J., and Ramsey, J. M., Anal. Chem., 1994, 66, 1107. 61 Seiler, K., Harrison, D. J., and Manz, A., Anal. Chem., 1993, 65, 1481. 62 Manz, A., Effenhauser, C. S., Burggraf, N., Harrison, D. J., Seiler, K., and Fluri, K., J. Micromech. Microeng., 1994, 4, 257. 63 Tseida, T., Mizuno, T., and Akiyama, J., Anal. Chem., 1987, 59, 799. 64 Harrison, D. J., Manz, A., Fan, Z., L�udi, H., and Widmer, H. M., Anal. Chem., 1992, 54, 1926. 65 Manz, A., Harrison, D. J., Verpoorte, E. M. J., Fellinger, J. C., Paulus, A., L�udi, A., and Widmer, H. M., J. Chromatogr., 1992, 593, 253. 66 Effenhauser, C. S., Paulus, A., Manz, A., and Widmer, H. M., Anal. Chem., 1994, 66, 2949. 67 Effenhauser, C. S., Manz, A., and Widmer, H. M., Anal. Chem., 1995, 67, 2284. 68 Hasser, A., MSc Thesis, University of Hull, 1996. 69 Burggraf, N., Manz, A., Verpoorte, E. M. J., Effenhauser, C. S., Widmer, H. M., and de Rooij, N. F., Sens. Actuators B, 1994, 20, 103. 70 Ruzicka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 71 Halliday, D., and Resnick, R., Physics, Wiley, New York, 1960. 72 Jacobson, S. C., Koutny, L. B., Hergenr�oder, R., Moore, A. W., Jr., and Ramsey, J. M., Anal. Chem., 1994, 66, 3472. 73 Monning, C. A., and Jorgenson, J. W., Anal. Chem., 1991, 63, 802. 74 Yao, X. W., Wu, D., and Regnier, F. E., J. Chromatogr., 1993, 636, 21. 75 Smith, J. T., and Rassl, Z. E., J. High Resolut. Chromatogr., 1992, 15, 573. 76 Marakami, Y., Suda, M., Yokogama, K., Tukeuchi, T., Tamiya, E., and Kurube, I., Microchem., 1994, 49, 319. 77 Tsuda, T., Nomura, K., and Nakagawa, G., J. Chromatogr., 1982, 248, 241. 78 Giddings, J. C., Science, 1993, 260, 1456. 79 Widmer, H. M., and Koch-Kellner, L., Chimia., 1989, 43, 320. 80 Koch, L., Koch, T., and Widmer, H. M., J. Chromatogr., 1990, 517, 395. 81 Raymond, D. E., Manz, A., and Widmer, H. M., Anal. Chem., 1994, 66, 2858. 82 Sudfa, M., Sakuhara, R., and Karube, I., Appl. Biochem. Biotechnol., 1994, 41, 3. 83 Haber, C., Sclvestri, I., B�o�osli, S., and Seman, W., Chimia, 1991, 45, 117. 84 Chervet, J. P., Ursem, M., Salzmann, J. P., and Vannoort, R. W., LC– GC, 1989, 7, 515. 85 Wohltjen, J., Sens. Actuators, 1984, 5, 305. 86 Ton, W., Zhang-You, S., Smith, S., Birnbaum, D., and Kopelman, R., Science., 1992, 258, 778. 87 Xi, X., and Yeing, F. S., Anal. Chem., 1990, 62, 1580. 88 Xue, Y., and Yeung, E. S., Anal. Chem., 1994, 66, 3575. 89 Verpoorte, E. M. J., Manz, A., L�udi, H., Bruno, A. E., Maystre, F., Kruttiger, B., Widmer, H. M., van der Schoot, B. H., and de Rooij, N. F., Sens. Actuators B, 1992, 6, 66. 90 Weigl, B. H., and Woelfbeis, O. S., Anal., Chem., 1994, 66, 3323. 91 Suda, M., Sakuchara, T., and Karube, I., 2nd Bioengineering Symposium, Tsukuba, Japan, 1992, p. 113. 92 Liang, Z., Chiem, N., Ocvirk, G., Tang, T., Fluri, K., and Harrison, D. J., Anal. Chem., 1996, 68, 1040. 93 Arquint, P., Koudelka-Hep, M., van der Schoot, B. H., van der Wal, P., and de Rooij, N. F., Clin. Chem., 1994, 40, 1805. 94 Micro Total Analysis Systems, ed. van den Berg, A., and Bergveld, P., Kluwer, Dordrecht, 1995, p. 299. 95 Conzen, J. P., B�urch, J., and Ache, H. J., Appl. Spectrosc., 1993, 47, 753. 96 Ache, H. J., Chem. Ind., 1993, 1, 40. 97 Jacobson, S. C., and Ramsey, J. M., Anal. Chem., 1996, 68, 720. 98 Koutny, L. B., Schmalzing, D., Taylor, T. A., and Fuchs, M., Anal. Chem., 1995, 68, 18. Paper 6/06289J Received September 16, 1996 Accepted November 18, 1996 10R Analy