MINIATURISATION FOR CHEMISTRY, BIOLOGY & BIOENGINEERING Development of a micro-fluidic manifold for copper monitoring utilising chemiluminescence detection Éadaoin Tyrrell,a Ceri Gibson,b Brian D. MacCraith,b David Gray,b Pat Byrne,b Nigel Kent,b Conor Burkeb and Brett Paull*a a National Centre for Sensor Research, School of Chemical Sciences, Dublin City University, Glasnevin, Dublin 9, Ireland. E-mail: Brett.Paull@dcu.ie; Fax: 00353 (0)1 7005503; Tel: 00353 (0)1 7005060 b National Centre for Sensor Research, School of Physical Sciences, Dublin City University, Glasnevin, Dublin 9, Ireland Received 16th January 2004, Accepted 31st March 2004 First published as an Advance Article on the web 26th April 2004 The progressive development of a micro-fluidic manifold for the chemiluminescent detection of copper in water samples, based on the measurement of light emitted from the Cu(II) catalysed oxidation of 1,10-phenanthroline by hydrogen peroxide, is reported. Micro-fluidic manifolds were designed and manufactured from polymethylmethacrylate (PMMA) using three micro-fabrication techniques, namely hot embossing, laser ablation and direct micro-milling. The final laser ablated design incorporated a reagent mixing channel of dimensions 7.3 cm in length and 250 3 250 mm in width and depth (triangular cross section), and a detection channel of 2.1 cm in length and 250 3 250 mm in width and depth (total approx. volume of between 16 to 22 mL). Optimised reagents conditions were found to be 0.07 mM 1,10-phenanthroline, containing 0.10 mM cetyltrimethylammonium bromide and 0.075 M sodium hydroxide (reagent 1 delivered at 0.025 mL min21) and 5% hydrogen peroxide (reagent 2 delivered at 0.025 mL min21). The sample stream was mixed with reagent 1 in the mixing channel and subsequently mixed with reagent 2 at the start of the detection channel. The laser ablated manifold was found to give a linear response (R2 = 0.998) over the concentration ranges 0–150 mg L21 and be reproducible (% RSD = 3.4 for five repeat injections of a 75 mg L21 std). Detection limits for Cu(II) were found to be 20 mg L21. Selectivity was investigated using a copper selective mini-chelating column, which showed common cations found in drinking waters did not cause interference with the detection of Cu(II). Finally the optimised system was successfully used for trace Cu(II) determinations in a standard reference freshwater sample (SRM 1640). DOI: 10.1039/b400805g Introduction 384 The determination of trace elements in environmental samples requires sensitive and selective analytical techniques. For measurements of trace concentrations of metals in various matrices there are a number of complex instrumental methods widely used, including graphite furnace atomic absorption spectroscopy (GF-AAS), inductively coupled plasma techniques with atomic emission spectroscopy (ICP-AES) or mass spectrometry (ICP-MS).1 However, there are also many sensitive simpler alternatives to these complex techniques. The use of chemiluminescence for the determination of dissolved trace metals has become a popular choice over recent years as it requires only simple inexpensive instrumentation compared to the above methods and is often very sensitive and linear over a wide dynamic range. In addition, chemiluminescent techniques can be highly selective for certain species.2 Chemiluminescence is commonly used in flow injection analysis (FIA) where it delivers low limits of detection, high precision and fast sample throughput. Cu(II) is an important element for the metabolism of many living organisms, but, as with many other metals, at high concentrations it is toxic. Cu(II) compounds are commonly used in agriculture, for preservatives and for water treatment. The Environmental Protection Agency (EPA) has stated that drinking water should not contain more than 1.3 mg L21 of Cu(II),3 with levels above this producing astringent tastes. The recommended daily allowance (RDA) for adults is 1.1 mg per day.4 Cu(II) can be determined sensitively using chemiluminescence. A number of different chemiluminescent methods have been proposed for the determination of Cu(II) in different sample matrices. Yamada and Suzuki determined trace amounts of Cu(II) using a chemiluminescent reaction based on a flavin mononucleotide–hydrogen peroxide– phosphate buffer system.5 A second method, developed by Yan and Worsfold, determined Cu(II) by its catalytic effect upon the reaction Lab Chip, 2004, 4, 384–390 between luminol (5-amino-2,3-dihydrophtalazine-1,4-dione) and hydrogen peroxide.6 More recently, a method based on the quenching effect of Cu(II) on the chemiluminescent reaction of dichlorofluorescein with hydrogen peroxide was developed by Safavi and Baezzat, and subsequently applied to the determination of Cu(II) in blood samples.7 Some methods have included an additional pre-treatment step prior to detection to improve selectivity. One such example used a micro-column of immobilised 8-hydroxyquinoline (8-HQ) for analyte preconcentration and removal of matrix interferences in seawater analysis,8 although later work by Zamzow et al.9 showed how the chemiluminescent reaction involved could, under certain conditions, be applied directly to the analysis of seawater without the use of such a column. The reaction involved was a well established reaction for Cu(II) determinations, based upon the formation of a complex between Cu(II) and 1,10-phenanthroline.8–12 The chemiluminescent reaction, which emits between 445–450 nm, results from the oxidative destruction of the Cu(II)–1,10-phenanthroline complex by hydrogen peroxide at an alkaline pH, with increased sensitivity obtained through the addition of a cationic surfactant.10 Chemiluminescence is a promising method of detection for micro-fluidic analytical systems due to its high sensitivity and the simplicity of the measurement technique. The combination of chemiluminescence detection, with its simple instrumentation requirements, and micro-fluidic systems, with low reagent consumption and portability, is proving very successful, with a number of chemical and biological flow sensor systems having already been developed.13–15 This promising combination is aided by recent progress in the development of fabrication methods for plastic micro-fluidic manifolds made from polymers such as poly(methylmethacrylate) (PMMA) or poly(dimethylsiloxane) (PDMS).16 In the work described here, the performance of PMMA microfluidic manifolds fabricated using a variety of micro-fabrication This journal is © The Royal Society of Chemistry 2004 techniques has been investigated. In combination with chemiluminescence detection, these manifolds have been applied to the determination and monitoring of Cu(II) in water samples. The aim of the study was to develop an analytical system, capable of low mg L21 determinations of Cu(II) in water samples, and which could be applied to ‘on-line’ monitoring with minimal reagent consumption. Experimental Preliminary Studies—stage I Preliminary (stage I) studies took place using standard FIA instrumentation (see Fig. 1(a)) in order to optimise the chemistry of the system. A peristaltic pump (Gilson Minipuls 312, Villiers, France) was employed to deliver the sample and reagents through 0.8 mm id poly(ether ether ketone) (PEEK) tubing, at a total flow rate of 2.1 mL min21. The manifold included a manual sample injection valve, model 7125, (Rheodyne, Cotati, CA, USA) fitted with a 120 mL PEEK injection loop for the introduction of the samples. The injected sample and carrier stream merged with the 1,10-phenanthroline solution at a T-piece connection and were mixed in a 100 cm long mixing coil of PEEK tubing prior to meeting with the hydrogen peroxide solution through a second Tpiece connector. The distance between the second T-piece and the detector flow cell was kept as short as possible (approximately 50 mm) as the chemiluminescence reaction occurred almost instantaneously. The flow cell itself consisted of 65 cm of 0.6 mm id transparent, flexible polyethylene (PE) tubing which was spiralled to a diameter of 25.4 mm and fixed to the window of a photomultiplier tube (PMT) (detailed below) and housed within a light tight box. Other details have been described previously by Coale et al.8 Micro-column A 0.8 cm long, 0.3 cm id micro-column packed with PRP-X800 itaconic acid functionalised 20 mm PS-DVB resin (Hamilton Company, Reno, NV, USA) was used for Cu(II) selectivity studies. The column was incorporated in the various manifolds prior to the introduction of the sample into the reagent streams. Back-pressure from the 0.8 cm column was sufficiently low not to affect performance of the peristaltic pumps used. Reagents In this study, all reagents were of analytical reagent grade and contained negligible concentrations of trace metals unless stated otherwise. Water obtained from a Milli-Q (Millipore) water purification system was used throughout this work. All the solutions were degassed using sonication and filtered through a 0.45 mm nylon membrane filter from Gelman Laboratories (Michigan, USA) prior to use. Under optimal final conditions the following reagents were prepared; Reagent 1; Deionised (Milli-Q) water containing 0.07 mM 1,10-phenanthroline made from a 12 mM stock solution, which was found to be stable for several days (obtained from BDH Laboratory Supplies, Poole, England and further purified by recrystalisation with nitric acid), 0.10 mM cetyltrimethylammonium bromide (CTAB) (BDH Laboratory Supplies, Poole, England) and 0.075 M sodium hydroxide (Sigma Aldrich Ltd., Dublin, Ireland). This solution was made up fresh each day. Reagent 2; A 5% w/v hydrogen peroxide (Sigma Aldrich Ltd., Dublin, Ireland) solution was prepared daily. All Cu(II) standard solutions were prepared freshly each day from stock solutions (atomic absorption spectroscopy standard solution (Sigma Aldrich Ltd., Dublin, Ireland)) (1000 mg L21) stored in 1% nitric acid. Calcium chloride (obtained from Sigma Aldrich Ltd., Dublin, Ireland) and magnesium chloride (Sigma Aldrich Ltd., Dublin, Ireland) were used to make up a pseudo sample matrix to investigate selectivity. Other metals used for interference studies included zinc, manganese, lead, nickel, cadmium and cobalt. All of these were atomic absorption spectroscopy standard solutions stored in 1% nitric acid and were obtained from Sigma Aldrich Ltd., Dublin, Ireland. Normal precautions for trace metal analysis were taken including acid washing of all glassware and plastic containers. Micro-fluidic manifold fabrication Fig. 1 Schematic diagrams and dimensions of the FIA and micro-fluidic manifolds developed. Chip dimensions = 7.0 3 7.0 3 0.8 cm. The micro-fluidic manifolds were designed using CAD 3D Excalibur software. Following this, the manifold could be fabricated in one of three different ways; hot embossing from a micromilled brass master, direct micro-milling into the PMMA chip or laser ablation. The first technique used a brass master that was fabricated by high precision micro-milling the channels of various dimensions. This was accomplished using a Datron 3-D M6 MicroMachining Centre, (Datron Technology Ltd., Milton Keynes, UK). The resultant master was then used to hot emboss the PMMA chips using a Model Hex02 Hot Embosser (Jenoptik Mikrotechnik, Germany). In this case, the brass master was heated to 138 °C and the PMMA was heated to 125 °C, just above the glass transition temperature (Tg) of PMMA (105 °C). Using a controlled force (1000 N) the master was used to emboss the PMMA. This procedure typically took 15 min. A top plate of PMMA of equal width was then drilled to make the reagent inlet holes and this was bonded to the embossed plate at 130 °C and 500 N in a Class 1000 clean room. The second technique used involved direct milling of the PMMA substrate. In this case, the micro-milling system was used to cut the micro-fluidic channels (200 3 200 mm) directly into the PMMA. Although direct micro-milling removed the additional hot embossing fabrication step, the surface of the channels was not as smooth Lab Chip, 2004, 4, 384–390 385 as that produced by the hot embossing method. The device was then bonded to a pre-drilled top plate as previously described above. The third method of fabrication used in this study was laser ablation. A Excimer laser (Optec Micro-Master System, Optec S.A., Frameries, Belgium) was used to make the micro-fluidic channels in the PMMA (The laser mask size was set to 250 mm, with a laser energy density of 1358 mJ cm22 and a frequency of 50 Hz, resulting in a machining speed of 0.250 mm s21). The manifold was bonded as before. 730 mm in length with an internal width and depth of 250 mm 3 250 mm (triangular cross-section), and 210 mm in length and 250 mm 3 250 mm in width and depth for the reaction/detection channel. This meant the theoretical approximate internal volume of the reaction/ detection channel was between 16 and 22 mL. Chemiluninescent detection was monitored with a PMT as with the stage II instrument described above. Micro-fluidic manifold and instrumentation—stage II Preliminary studies—stage I For the initial micro-fluidic manifolds (stage II) both reagents and sample carrier were again delivered using a standard peristaltic pump (Gilson Minipuls 312, Villiers, France) delivering the sample and reagents into the plastic manifold via connecting 0.8 mm inner diameter poly(ether ether ketone) (PEEK) tubing, glued into reagent inlet holes (1 mm diameter) on the underside of the plastic chips, at individual flow rates of 0.06 mL min21 (0.18 mL min21 total). The sample was introduced into the carrier stream using a manual sample injection valve, model 7125, (Rheodyne, Cotati, CA, USA) fitted with a 3 mL PEEK injection loop. The manifold itself had a mixing channel of 1000 mm in length, < 900 mm in depth and 1000 mm in width, and a detection channel 190 mm in length and again < 900 mm in depth and 1000 mm width. Chemiluminescence was continuously measured using a PMT (Hamamatsu Photonics HC135–01 series with R1924 bi-alkali tube). The PMT was approximately 135 mm in height and the window of the detector was 25.4 mm inner diameter. In this work, the chemiluminescence occurred between 445 and 450 nm, and the PMT spectral range was from 300 nm to 650 nm (absorbance of PMMA is negligible above 400 nm). The PMT operated in photon counting mode and the signal was recorded on a computer and data processed using Microsoft Excel. The PMT was placed over the detection channel and was secured tightly to prevent stray light from reaching the detector. The complete micro-fluidic manifold and PMT assembly was wrapped several times in aluminium foil to reduce the dark count rate and enclosed in a light tight box. Fig. 1(b) shows the schematic diagram of this initial system. Initial studies took place using standard FIA in order to optimise the chemistry of the system and to ascertain the analytical performance characteristics that could be later compared to the micro-fluidic method. Using this system three of the reaction variables were optimised, the concentration of hydrogen peroxide, 1,10-phenanthroline and the concentration of surfactant. Micro-fluidic manifold and instrumentation—stage III For the final micro-fluidic system (stage III) developed, three micro-peristaltic pumps with 0.51 mm id peristaltic tubing (manufactured by BVT Technologies, Euro-link Associates, Tyne & Wear, England) were used to drive the reagents and sample streams through the micro-fluidic manifold via connecting 0.13 mm id PEEK tubing, glued into reagent inlet holes (1 mm diameter) on the underside of the plastic chips. Each micro-pump weighed only 34 g and was 60 mm in height and only 16 3 16 mm at the base. These pumps were operated using a DC supply at 3 V and were connected together using a custom made interface driven by one PC connection. When connected in this manner it was possible to set each flow rate independently for each pump in the range of 0.5 to 200 mL min21. Under optimal conditions for this study, individual flows rates of 25 mL min21 (75 mL min21 total) were used. The final version of the micro-fluidic manifold design is illustrated schematically as Fig. 1(c). The micro-fluidic manifold itself was manufactured from PMMA. A number of different designs and dimensions were investigated but the basic design consisted of two inlets for the sample and 1,10-phenanthroline solutions, which were subsequently mixed together in the mixing channel. This provided the mixing necessary for the reaction between the Cu(II) in the sample and the 1,10-phenanthroline to take place. A third inlet at the end of the mixing channel prior to the detection/reaction channel was used for the introduction of the hydrogen peroxide. At the end of the detection channel there was an outlet leading to waste. For the final laser ablated version of the micro-fluidic manifold the dimensions of the mixing channel were 386 Lab Chip, 2004, 4, 384–390 Results and discussion Optimisation of hydrogen peroxide Concentrations of 1,10-phenanthroline, CTAB and NaOH were kept constant at 0.06 mM, 0.1 mM and 0.075 M respectively, while the concentration of hydrogen peroxide was varied. From this work it was found that a 5% hydrogen peroxide solution resulted in a higher analytical signal for a Cu(II) standard than higher hydrogen peroxide concentrations, with little difference in background noise, even though in some of the literature a 10% hydrogen peroxide solution was used.9 The actual signal to noise ratio for the 5% hydrogen peroxide was 13.2 compared to 8.4 for the 10% peroxide solution. The peroxide solution was prepared daily to avoid a decrease in sensitivity that resulted from reagent instability over time. No significant reductions in detector response when using a single hydrogen peroxide solution for a period of 12 h. Optimisation of 1,10-phenanthroline Using the optimised 5% hydrogen peroxide concentration, the concentration of 1,10-phenanthroline was then optimised. The concentrations of CTAB and sodium hydroxide again remained constant at 0.1 mM and 0.075 M, respectively. A 0.03 mM 1,10-phenanthroline solution was found to be optimal, producing a significant reduction in background noise compared to higher levels. The signal itself due to Cu(II) was also reduced but the signal to noise ratio for the 0.03 mM 1,10-phenanthroline solution was 27.63 compared to 8.86 for a 0.06 mM 1,10-phenanthroline solution. The 1,10-phenanthroline was also purified by recrystilisation with nitric acid as the Cu(II) signal was enhanced considerably (by approximately 40%) when the purified 1,10-phenanathroline was used and noise further decreased. Purified 1,10-phenanthroline was used in all future work. Optimisation of surfactant Previous work by Yamada and Suzuki10 investigated the effect of different surfactants on the Cu(II)–1,10-phenanthroline chemiluminescent reaction. It was found that an increase in signal resulted from the addition of a cationic surfactant, while anionic and nonionic surfactants had no significant effect. In this work, several surfactants including Triton X-100, sodium dodecylsulfate (SDS) and CTAB were investigated. It was found that Triton X-100 (non-ionic) and SDS (anionic) did not increase the signal and the background noise was dramatically increased. The addition of low concentrations ( < 0.1 mM) of the cationic surfactant CTAB was found to dramatically reduce some of the background noise whilst not affecting the analytical signal for Cu(II). A concentration of 0.05 mM CTAB produced a signal to noise ratio of 19.9 compared to 6.4 for a 0.10 mM CTAB solution. The optimal reaction pH was determined by Coale and coworkers to be between 9.8 and 10.18. In this work, the reaction pH for Cu(II) analysis under optimised conditions was found to be 10.35. Using this system, Cu(II) could be easily determined at levels as low as 1 mg L21 and the response was linear (n = 5, standards injected in triplicate) over the concentration range 1 to 50 mg L21, producing a R2 value of 0.995 (see Fig. 2(a)). Selectivity studies Previous work has shown the high selectivity of the Cu(II)– 1,10-phenanthroline chemiluminescent reaction, which shows little or no response to excess concentrations ( > 10 mg L21) of alkali metals or Cr(III) and (VI), Mn(II), Ni(II), Co(II), Cd(II), Pb(II), Al(III), and Fe(II) and Fe(III).7,10 However, it has been reported that high concentrations of Zn(II), Ca(II) and Mg(II) can cause minor interference at the above concentration. To further increase the selectivity of the developing method, a Cu(II) selective itaconic acid functionalised resin (20 mm resin size) micro-column was investigated. Itaconic acid is a dicarboxylic acid capable of acting as a weak cation exchanger and/or a strong chelating ion exchanger, which within a 0.3 3 0.8 cm micro-column at pH 2 to pH 4 completely retained Cu(II) whilst showing no retention of alkali and alkaline earth metal ions, or Mn(II), Cd(II), Co(II), Zn(II), Pb(II) and Ni(II). By selectively removing Cu(II) from the sample matrix using the above micro-column on-line, followed by its subsequent elution, it was possible to remove the above potential interferences. To illustrate this, a complete set of Cu(II) standards over the range 1 mg L21 to 1 mg L21 were made up in a sample matrix containing 10 mg L21 Ca(II) and 10 mg L21 Mg(II) (typical concentrations found in drinking waters), whilst a second set was made up in MilliQ water only. The standards containing Ca(II) and Mg(II) were each passed through the micro-column (column buffered to pH 4) and the Cu(II) was selectively retained. The Cu(II) was subsequently eluted with an equal volume of 100 mM nitric acid. Comparison of the eluted Cu(II) standards with those prepared and analysed directly in Milli-Q water would show how the column could be used for complete retention and elution of Cu(II) from samples containing high levels of Ca(II) and Mg(II). Fig. 3(a) shows the resultant peaks for the Cu(II) standards, which were separated from the sample matrix containing Ca(II) and Mg(II) by the itaconic acid column. The results for Cu(II) standards made up in Milli-Q water over the same concentration range of 1 mg L21 to 1 mg L21 are show as Fig. 3(b). It can be seen that there was effectively 100% separation and recovery of the copper from the sample matrix using the itaconic acid micro-column. This means the micro-column could also be used for preconcentration of Cu(II) from complex sample matrices should the need arise. From Fig. 3 it is also interesting to note that Milli-Q water itself gave a small response equivalent to < 1 mg L21 Cu(II) when analysed directly and that this response was eliminated after treatment with the micro-column. This would indicate that some potentially interfering species (other than Cu(II)) in the Milli-Q water itself was causing a slight positive signal within the blank. Initial micro-fluidic work—stage II The first micro-fluidic manifold in this study was manufactured in PMMA using the micro-milling and hot embossing facility. The first manifold fabricated had a mixing channel of 1000 mm in length, 900 mm in depth and 1000 mm in width. However, bonding the top plate to the lower design plate causes considerable reduction in the actual depth of the channels in the finished manifold, particularly with wide channels such as these, and so the exact channel depth in the bonded manifold could not be ascertained. This early design provided the same mixing channel length as the Fig. 3 Detector responses obtained for (a) Cu(II) standards prepared in a matrix of 10 mg L21 Ca(II) and Mg(II) and passed through itaconic acid mini-chelating column followed by elution with 100 mM HNO3, and (b) Cu(II) standards prepared in Milli-Q water only. Fig. 2 Detector responses to Cu(II) standards obtained using (a) standard FIA manifold compared to (b) initial hot embossed design (1000 3 900 mm channels) and (c) laser ablated manifold (250 3 250 mm channels). Standard concentrations shown in mg L21. Lab Chip, 2004, 4, 384–390 387 standard FIA method and was found to be more than adequate for the required complete mixing of the Cu(II) in the sample and the 1,10-phenanthroline. The inlet for the introduction of the hydrogen peroxide was placed immediately prior to the detection/reaction channel as the chemiluminescent oxidation reaction was known to be almost instantaneous. A series of experiments were again carried out in order to optimise the concentrations of 1,10-phenanthroline and CTAB under the new flow conditions. The experimental space was defined by varying systematically both the concentrations of 1,10-phenanthroline and CTAB from 0.01 mM to 0.10 mM. In all of the experiments, the following conditions were used; a total combined flow rate of 180 mL min21, sample injection volume of 3 mL, 5% hydrogen peroxide solution and 0.075 M sodium hydroxide. From this work the highest signal to noise ratio was achieved when using 0.07 mM 1,10-phenanthroline and 0.06 mM CTAB. Using these new conditions, linearity and response was again investigated, and the peaks obtained for a range of Cu(II) standards between 10 and 1000 mg L21 can be seen in Fig. 2(b). As can be seen from the figure shown, linearity (if based upon peak height) was restricted to < 100 mg L21 Cu(II). However, as can be seen, given the much reduced injection volume, analyte sensitivity was much improved, despite the large reduction in the size and volume of the reaction/detection coil. There was also very little change in background noise, giving a detection limit of approximately 6 mg L21 (using 3 3 baseline noise criterion). The repeat injection of a Cu(II) standard solution using this system (n = 6) gave a % RSD value of 2.2 based upon peak height. Investigation of mixing process In an attempt to improve method linearity and further miniaturise the manifold the impact of the dimensions of the mixing channel and reaction channels was investigated. Coloured dyes were used in place of the reagents to investigate the mixing process, in order to reduce the length of the mixing chamber. It was found that complete mixing of the two streams occurred after approximately 700 mm. It was decided, therefore, that the length could be decreased by nearly one third of the initial length (1000 mm to 700 mm). An investigation on the length of the reaction/detection channel was also carried out in order to determine the optimal length. Using a 100 mg L21 Cu(II) standard, the detection channel was increasingly masked from the PMT detector to ascertain the effect of shortening the reaction channel by 25, 50 and 75% (equivalent to 143, 95 and 48 mm respectively). The resultant signals for the 100 mg L21 standard can be seen in Fig. 4. During this experiment it was found that masking the last 25% of the reaction/detection channel, the chemiluminescent signal for the Cu(II) standard decreased (by more than half). It was also found that masking the last 75% of the channel led to the signal being totally lost. This meant that although the chemiluminescence reaction was thought to occur almost instantaneously, maximum emission took place in the latter part of the detector channel due to a short time delay to facilitate mixing of the hydrogen peroxide and the sample/ reagent flow. As a result, the length of the reaction channel could not be decreased and was subsequently increased (see stage III below). Further miniaturisation—stage III Taking the above results into account, the design for the final (stage III) micro-fluidic manifold was finalised. As above, a manifold was prepared using micro-milling to produce a brass master, followed by hot embossing into PMMA. The final design, as shown in Fig. 1(c), utilised three individual micro-pumps to supply the reagent stream, hydrogen peroxide stream and sample stream. To further simplify the system, the sample injection valve was removed and the sample fed directly into the mixing channel itself. This would result in a continuous detector response rather than the previous transient signal. For introduction of new sample and standard solutions the sample pump was stopped whilst solutions were changed. It was found that this manual procedure did not introduce air into the system (as is evidenced by Fig. 5), although for future on-line work a low-pressure switching valve could be readily incorporated into the system for this purpose. The mixing channel was reduced in length to 700 mm, while the detection channel was increased to 210 mm. In addition, the shape of the detection channel was optimised (fitted exactly to shape of PMT window) to maximise sensitivity and reduce background noise. The channels themselves were significantly narrower than the channels of the larger chip, being 200 mm wide by 200 mm deep (although again due to bonding the ultimate depth of the channels was significantly < 200 mm). As in each previous case the reagent concentrations were re-optimised to suit the new manifold design. The highest signal to noise ratio was achieved when using 0.07 mM 1,10-phenanthroline and 0.1 mM CTAB, with other conditions set at a total combined flow rate of 76 mL min21, 5% hydrogen peroxide solution and 0.075 M sodium hydroxide. Flow rates The flow rates used in the stage III micro-fluidic system were each controlled by a separate micro-pump, however all the reagents were pumped at the same flow rate. Briefly, the effect of flow rate on response was investigated using individual flow rates of 0.029, 0.025 and 0.021 mL min21 for each of the three streams (measuring in total 0.086, 0.076 and 0.064 mL min21). Results showed that the analytical signals were slightly increased by decreasing flow rates (see Fig. 5(a)). However, although the lowest flow rates (0.021 mL min21 for each stream) produced the highest absolute signal, the background was also noisier, and therefore individual flow rates of 0.025 mL min21 was found to be optimum. Linearity Using the optimised conditions, a series of measurements was performed with Cu(II) standard solutions ranging from 0 to 50 mg L21. A stepwise graph was produced over this concentration range and is shown as Fig. 5(b). The resulting signal heights were plotted as a function of concentration and the results were found to produce an excellent linear correlation within this given concentration range (R2 = 0.993, n = 6, standards measured in duplicate). Using the micro-fluidic manifold, Cu(II) could be easily determined at concentrations as low as 10 mg L21, with the S/N ratio of 3. Fig. 4 Detector responses obtained for a 100 mg L21 Cu(II) standard with increasingly reduced reaction/detection channel length (equivalent to 190, 143, 95 and 48 mm). 388 Lab Chip, 2004, 4, 384–390 Comparison of fabrication techniques Three alternative micro-fabrication techniques were investigated, namely hot embossing into PMMA, direct micro-milling of the channels into the PMMA itself and thirdly the use of laser ablation. Each of these techniques result in different channel profiles and surface morphologies. Laser ablation results in a characteristic shallow ‘V’ shaped channel and direct micro-milling into the PMMA results in a rougher surface than the hot embossed manifold. For direct comparison with the hot embossed manifold, detector response and linearity was determined on each manifold using identical reagent concentrations and flow rates. The directly micro-milled manifold resulted in the largest comparative unit response, with a calibration slope equal to ~ 751 counts s21 for each mg L21 of Cu(II) compared to ~ 378 counts s21 for each mg L21 Cu(II) for the above hot embossed manifold. However, background noise was also increased. The results using the directly micro-milled manifold were found to be linear up to 100 mg L21 (R2 = 0.983, n = 5, standards injected in duplicate). The laser ablated manifold was found to give the lowest unit response with the lowest calibration slope, only 61 compared to the above values. However the manifold also produced the most linear response over the greatest concentration range 0–150 mg L21 (R2 = 0.998, n = 5, standards injected in duplicate, see Fig. 2(c)). In addition, the laser ablated manifold also gave the lowest background noise, resulting in a detection limit of approximately 20 mg L21 Cu(II). The reason for the lower response and noise with this manifold is simply related to the fact that the laser ablation process results in the shallow ‘V’ shaped channels which effectively reduce the channel volumes considerably ( ~ 50%), compared to the alternative fabrication methods. To partially compensate for this profile the laser ablated channels were cut 250 mm wide by 250 mm deep, hence the reaction/detection channels of the ablated manifolds was calculated to be within the range 16–22 mL compared to 21–30 mL for the hot embossed manifold. Fig. 6(a) shows the comparative responses for each of the above micro-fluidic manifolds for increasing Cu(II) standard solutions (Milli-Q water was introduced between each standard reading). Fig. 6(b) shows the reproducibility of the laser ablated manifold with the repeated analysis (n = 5) of a 75 mg L21 Cu(II) standard. Sample carry over Fig. 6 shows recovery time (time taken for a return to baseline signal following removal of the sample) for each of the microfluidic manifolds was in the order of 2–3 min. This indicates some degree of adsorbance of the reagents onto the channel walls, which may be slowly washed off during the Milli-Q washing step. However, as Figs. 5(b) and 6(a) show, the system does return to the starting baseline when allowed sufficient recovery time, and although this may increase individual sample analysis time, for Fig. 5 Detector responses obtained using the hot embossed micro-fluidic manifold (200 3 200 mm channels). (a) Effect of flow rate upon response, (b) system linearity, and (c) on-line analysis of drinking water. All concentrations shown in mg L21. Fig. 6 (a) Comparison of detector response for three micro-fluidic manifolds, (1) laser ablated manifold (250 3 250 mm), (2) hot embossed manifold (200 3 200 mm), (3) direct micro-milled manifold (200 3 200 mm). (b) Repeat analysis of 75 mg L21 Cu(II) standard using laser ablated manifold. (c) Analysis of SRM 1640 standard reference freshwater sample. Milli-Q water blanks run between each sample/standard analysis. Lab Chip, 2004, 4, 384–390 389 Table 1 Analytical performance data for developed methods Fabrication method Flow rates/ mL min21 reagent Linear rangea/ mg L21 Slope Standard FIA Hot embossing 1000 3 900 mm Hot embossing 200 3 200 mm Micro-milling 200 3 200 mm Laser ablation 250 3 250 mm 0.7 0.06 0.025 0.025 0.025 1–50 < 100 20–100 40–100 25–150 726 — 751 378 61 a Approx. detection limit/mg L21 0.995 — 0.996 0.983 0.998 1–2% (n = 3) 2.2% (n = 6) 2–3% (n = 3) 4–5% (n = 2) 2–3.5% (n = 5) 1 6 10 10 20 Data based upon peak height measurements. longer term on-line monitoring purposes this should not be a significant problem. Table 1 shows the comparative analytical performance data determined for the three micro-fluidic manifolds, and those data obtained for standard FIA and the initial larger scale hot embossed manifold. The data shown were obtained under the specific reagent conditions used within this study only. Analysis of real samples on-line drinking water analysis Using the hot embossed micro-fluidic manifold and conditions described above, the on-line analysis of drinking water was undertaken. The drinking water from a laboratory tap was continuously fed via an in-line filter to the sample inlet of the manifold. Within the sample line a switching valve was placed which allowed the tap water to either by-pass or be passed through the Cu(II) selective itaconic acid micro-column (pre-buffered to pH 4) detailed earlier, allowing verification that the detector response was due to Cu(II) only. The results of this experiment can be seen in Fig. 5(c). As can be seen, the concentration of Cu(II) found within the laboratory tap supply was approximately 80 mg L21. On-line passage of the tap water through the itaconic acid column completely removed the chemiluminescent signal, indicating that the sample matrix was not causing significant interference in this application. Analysis of certified reference water sample To check for method accuracy a standard reference material (SRM) from the National Institute of Standards and Technology (NIST) was analysed for Cu(II). The sample (SRM 1640) was composed of natural fresh water (river) had been filtered and acidified with 0.5 M nitric acid. This sample was certified to contain 85.2 mg L21 ± 1.2 mg L21 Cu(II). The sample was neutralised using sodium hydroxide and was diluted by 50% with Milli-Q water. This sample was then analysed using the laser ablated manifold. Firstly, 50 and 100 mg L21 Cu(II) standards were analysed, with a Milli-Q water blank run between standards. The 50% dilution of the NIST standard was then analysed. The resultant signals can be seen in Fig. 6(c). This sample was found to contain approximately 40 mg L21 Cu(II) which corresponds to ~ 80 mg L21 in the SRM, within 390 R2 value Reproducibility % RSD (no. replicates) Lab Chip, 2004, 4, 384–390 ±6% of the true value, representing excellent accuracy for a microfluidic device when analysing a complex freshwater sample for trace Cu(II). Conclusions The development of a micro-fluidic manifold and analytical method for Cu(II) determinations has been described. The results have shown that the standard FIA method can be effectively reduced to the micro-fluidic format whilst maintaining acceptable linearity, precision and accuracy and with appropriate detection methods such as chemiluminescence, also maintaining excellent sensitivity. The micro-fluidic system developed can be readily made portable or be used for on-line monitoring with as little as 3.0 mL h21 total reagent consumption. References 1 D. Harris, Quantitative Chemical Analysis, Freeman, New York, 5th edn, 1998. 2 C. M. Sakamoto-Arnold and K. S. Johnson, Anal. Chem., 1987, 59, 1789. 3 http://www.epa.gov/safewater/dwh/c-ioc/copper.html. 4 http://193.120.54.7/publications/reports/recommended_dietary_allowances_Ireland_1999.pdf. 5 M. Yamada and S. Suzuki, Chem. Lett., 1982, 1747. 6 B. Yan and P. J. Worsfold, Anal. Chim. Acta, 1990, 236, 287. 7 A. Safavi and M. R. Baezzat, Anal. Lett., 2000, 33, 667. 8 K. H. Coale, K. S. Johnson, P. M. Stout and C. M. Sakamoto, Anal. Chim. Acta, 1992, 266, 345. 9 H. Zamzow, K. H. Coale, K. S. Johnson and C. M. Sakamoto, Anal. Chim. Acta, 1998, 377, 133. 10 M. Yamada and S. Suzuki, Anal. Lett., 1984, 17, 251. 11 W. G. Sunda and S. A. Huntsman, Mar. Chem., 1991, 36, 137. 12 J. M. Lin and M. Yamada, Analyst, 2001, 126, 810. 13 Y. Lv, Z. Zhang and F. Chen, Talanta, 2003, 59, 571. 14 G. M. Greenway, L. J. Nelstrop and S. N. Port, Anal. Chim. Acta, 2000, 405, 43. 15 A. M. Jorgensen, K. B. Mogensen, J. P. Kutter and O. Geschke, Sens. Actuators, B, 2003, 90, 15. 16 A. de Mello, Lab Chip, 2002, 2, 31.