摘要:

Highlight High density oligonucleotide and DNA probe arrays for the analysis of target DNA Michael Thompson* and L. Michelle Furtado Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario, Canada M5S 3H6. E-mail: mikethom@alchemy.chem.utoronto.ca Received 22nd April 1999, Accepted 8th June 1999 The acquisition of sequence, expression and other information concerning genetic material constitutes a crucial component of the modern revolution in molecular biology.One important advance in this area is the development of high density oligonucleotide/DNA microarrays which allows the rapid sequence analysis of genomic target samples in addition to diagnostic possibilities with respect to genetic and infectious disease. In the present article we review protocols for the design of such microarrays and their principles of operation. Together with a look at some recent applications we include brief remarks as to the possibilities for the future.Introduction The original goals of the Human Genome Project have been the construction of complete genetic and physical maps of the human genome, determination of the complete nucleotide sequence, the performance of similar studies on small organisms such as Escherichia coli, yeast and mouse and, finally, the localization of the estimated 100 000 genes within the human genome. Enormous effort and expense to the present time have resulted in considerable progress in achieving these aims, including the sequencing of the 3 billion or so base pairs of the human genome.The information generated by this research clearly constitutes a revolution in molecular biology particularly as it pertains to genetic diagnostic testing, detection of pathogens, study of the fundamentals of gene expression and examination of regulatory protein and small molecule interactions with DNA and RNA. Not surprisingly, the tremendous significance of the information generated by the Project has not been lost on the major pharmaceutical companies, who are becoming increasingly concerned with genomics, drug discovery and the development of human therapeutics.1 A major and critical component in obtaining genetic information is the ability to screen a DNA sequence.The basic experimental protocols employed in the past have been the electrophoresis gel assays of Sanger et al.2 and Maxam and Gilbert.3 Although there have been advances with respect to throughput in these strategies, such as the use of capillary zone electrophoresis, the separation-based protocol is still relatively time consuming and, therefore, rate limiting with regard to the overall sequencing effort.An elegant alternative to the above is the use of so-called gene microchips or, more accurately, oligonucleotide and DNA probe arrays. This approach represents an exquisite aggregation of the technologies of sequencing by hybridization (SBH), light-directed spatially addressable interrogation, combinatorial chemical synthesis, confocal fluorescence microscopy, robotics and polymerase chain reaction (PCR).The technology offers significant advantages in terms of avoidance of time-consuming protocols through multiplexing, the ability to produce probe sequences of any type, the future possibility to generate inexpensive re-usable chips and capability for incorporation into micro-fabrication and micro-device technology. In the present brief review we discuss the principles of the nucleic acid probe array method, and recent example applications together with some concise comments as to future possibilities.Sequencing by hybridization The principle of this approach is based upon the fact that any linear DNA sequence incorporating the four bases is composed of overlapping, shorter sequences. Using this concept as a basis, SBH employs hybridization of a set of oligonucleotide probes (e.g., 8- to 20-mers) with sub-sequences in a particular target DNA fragment or sample.4 In practical terms, the DNA target bound on a surface or present in solution is allowed to interact with a set of conveniently labelled probes which may also be surface-attached or reacting from solution.In order to achieve analytical specificity, complete discrimination in hybridization between fully complementary target–probe sequences compared to combinations involving mismatches must be effected.The number of possible probes is large for probe sequences over approximately heptamers (the possible number of probe sequences 4n, where n is the base length of the probe). As an example, if a 12-mer target DNA sequence AGCCTAGGTGAA, is allowed to perfectly hybridize with all 65 536 possible octamer probes only five of the latter, TCGGATCG, CGGATCGA, GGATCGAC, GATCGACT and ATCGACTT will combine with the target.5 Consideration of overlapping sequences reconstructs the assembly of bases complementary to the above-specified 12-mer(TCGGATCGACTT).Obviously, increasing probe length could lead to enhanced specificity on hybridization, but there is a limit in producing a domain size for the rapidly increasing number of probes required. Finally, it should be mentioned that it is not usually the case that all possible probes are employed in this type of study. The nature of the set of probes used, or their length, depends very much on the type of mapping or sequencing information that is required at the outset.Fabrication of single-strand DNA and oligonucleotide arrays The use of directed light to produce oligonucleotide arrays combines photolithographical and combinatorial chemical synthetic technologies.5–8 Although the early emphasis appeared to centre on the surface immobilization of peptides, recent times have seen greater concentration on oligonucleotide chemistry.6 The production of oligonucleotide arrays by the imposition of Analyst, 1999, 124, 1133–1136 1133(b) directed light beams involves, initially, the derivatization of a glass substrate with linking molecules, such as those based on multi-layer forming aminopropyltriethoxysilane, which are in turn capped with a photolabile protecting functionality.Removal of the latter by light directed through a mask generates an area of reactive hydroxy groups. These are then exposed to solutions of, for example, 5A-photolabile N-acyl-deoxynucleoside- 3A-O-phosphoramidite (oligonucleotide synthesis using phosphoramidites is reviewed in ref. 9). Next, light is directed to a different or adjacent area of the substrate, again through a photolithographic mask prior to analogous immobilization of a different nucleoside. This whole process can be repeated to produce any number of areas containing one of the four bases. These steps and a typical 5A-photolabile nucleoside are shown in Fig. 1. Subsequent steps in producing the oligonucleotide array involve turning of the substrate in a perpendicular fashion followed by appropriate photo-deprotection to effect round 2 of the process.In Fig. 2 we illustrate the whole procedure for the immobilization of the 256-possible tetramers. Clearly, although the number of probes associated with particular base sequences increases in a polynomial fashion in terms of oligo base length, the number of chemical cycles required is 4 3 n, i.e., for tetramers, 16 steps.Accordingly, even the generation of the 420 20-mers requires only 80 cycles and can be conducted in a matter of hours. The main points of this technology are that the precise location of a particular oligonucleotide, and of its sequence, is known accurately and, secondly, because of the use of the photolithographic process high density arrays can be achieved. The limit of resolution is obviously limited by the wavelength of the light, and its diffraction, resulting in approximate micrometer areas.If domains of 50 mm are envisaged, 40 000 oligonucleotide sites can be imposed per square centimeter. In reality, to avoid ‘cross-talk’ between domains and to produce distinct areas for confocal microscope detection, spaces are often placed between oligo domains. This photolithography-combinatorial chemistry strategy has been commercialized by Affymetrix Company.10 An alternative approach to the fabrication of single strand DNA arrays is that based on chemisorption of nucleic acid to the substrate surface.11 This strategy possesses the obvious advantage that significantly longer chemically pure probes can be employed, resulting in a higher level of probe specificity.The protocol varies but, essentially, solutions of double-stranded DNA can be applied in relatively small domains to poly-llysine- coated glass slides. Presumably, the forces binding the nucleic acid to the substrate are largely electrostatic in nature. In order to achieve application of DNA to small areas, sophisticated robotic ‘spotter’ instruments are employed that can supply extremely small volumes in an automated manner.The final step is to denature the DNA by increased temperature exposure of the wafer, resulting in a population of single-stranded nucleic acid. This approach has been pioneered and commercialized by Synteni Co.12 Hybridization of target DNA to probe arrays Whether the probe array has been manufactured by lightdirected chemical synthesis or robotic printing, hybridization methodologies and subsequent detection are somewhat analogous.For example, Affymetrix has developed a fluidics station that allows the reproducible hybridization of target DNA to oligonucleotides on the chip surface. With respect to this instrument it is crucial to control temperature in a uniform mixing protocol. Much of the appropriate steps involved in the processing of each microarray is fully automated and involves very low volumes of solutions for application and washing.Nearly all studies of completed hybridization of probe arrays have involved the use of confocal fluorescence microscopy to detect the level of duplex formation in a particular array cell. This protocol mandates the fluorescence-tagging of the target DNA. There are obvious assumptions here, namely that the species used for fluorescence-labelling does not compromise the hybridization reaction and that the detected level of duplex formation accurately reflects the nature of complementarity in each cell.In order to detect the fluorescence radiation emanating from each cell a confocal microscope capable of sensing multiple-emission wavelengths is generally employed. This instrument must be capable of the discrimination of fluorescence associated with labelled hybridization from other sources and pixel resolution at the several micrometer scale. Incident light from an argon ion laser excites the appropriate tag Fig. 1 (a) A typical 5A-photolabile phosphoramidite used for light-directed addition of bases to each cell of oligonucleotide probe arrays 5A-0A-(amethyl- 6-nitropiperonyloxycarbonyl)-N-acyl-2A-deoxynucleoside phosphoramidite, where ‘Base’ represents G, C, T or A). (b) Steps involved in photolithographic addition of oligonucleotide bases to a substrate surface. (P represents a photolabile protecting moiety). 1134 Analyst, 1999, 124, 1133–1136through a confocal optical system in a configuration whereby the chip is moved relative to the scanner.Emitted light is transferred to a storage structure ready for image processing. We include in the present paper two examples of images obtained from hybridization of DNA on glass slides, both based on robotic printing rather than light-directed synthesis. Fig. 3 depicts a section of a typical coloured image representing various levels of hybridization (red being the maximum) achieved in a set of array cells.Note that, on close examination, there is significant intensity variability within each cell, which can lead to problems in quantification. In the second example (Fig. 4) we show a black and white image from an attempt to hybridize variable-length probe ssDNA sequences associated with the yeast genome to sample DNA. The array is composed of 6400 probes (in duplicate form) ranging in size from 500 to 10 000 bases in length. The probes were attached to the glass substrate by robotic printing.Example applications of nucleic acid microarrays This section outlines a small selection of examples of where oligonucleotide microarrays have been employed in bioanalytical chemistry in recent times. Not surprisingly there has been considerable interest in the use of microarrays for the diagnosis of genetic disease. For example, blood obtained from b-thalassemia patients can be used to develop diagnostic tests for mutations within the first exon and first intron of the bglobulin gene.DNA samples were hybridized with 10-mers immobilized on a microchip having cell dimensions of 40 3 40 3 20 mm and 100 3 100 3 20 mm.13 The results indicated a significant difference in hybridization levels for matched and mismatched sequences. The authors of the work concluded that the identification of gene mutations in patients was unequivocal in yielding yes–no diagnostic answers. As an aside here, it is worth noting that various researchers have concluded that single-bases mismatches at the centre of the duplex can be detected with facility, whereas flanking bases often cause difficulty in terms of their mismatch distinction.Another example is that concerning the now well-known, early-onset breast cancer gene, BRCA1.14–16 The protein coding region of this gene contains 5,592 basepairs in 22 coding exons spread over 100 kB of genomic DNA. Because many mutations have been associated with the malfunctioning gene it has become necessary to screen the relatively large gene for all possible heterozygous mutations in patients.In elegant studies it has been shown that high-density arrays consisting of over 96 600 oligonucleotides 20-bases in length in tandem with 2-colour fluorescence analysis (use of 2 tagging agents) can successfully detect a number of mutations in a particular fragment of the gene, BRCA1. In the real screening of patients known to harbor Fig. 2 Combinatorial chemical synthesis of all possible 256 oligonucleotide tetramers.Note that each synthesis cycle in round 2 generates 4 dinucleotides, whereas in round 3 (by mask subdivision) each cycle produces 16 trinucleotides. Fig. 3 An example confocal fluorescence microscope image of hybridized target DNA on an approximate 1600 cell section of a probe microarray. Cy3. BPI image: brown lab specimen. Area 1.2 cm 3 1.2 cm. Fig. 4 Variable length DNA probes for the yeast genome hybridized with sample DNA.Spots are applied by robotic printing and are configured in duplicate experiments. Cy3 1 mA slide. Analyst, 1999, 124, 1133–1136 1135genes with mutations the success rate for microarray detection was over 90%. Other studies have been concerned with efforts to obtain basic sequence information. For example, the simultaneous analysis of the entire human mitochondrial genome using DNA arrays of about 135 000 probes has been achieved.17 To give the reader a feel for the accuracy and time involved in this type of analysis, the authors pointed out that the throughput of conventional gel-based sequences could result in reading two mitochondrial genomes a day at best.In contrast, using 5 microarrays per hour yields the possibility to examine 50 genomes in one day. There are also significant reductions in sample preparation time in concert with a highly reliable analysis. As for the study mentioned above, use of the 2-fluorophor protocol enhanced the detection of sequence polymorphisms significantly, and with single base resolution.Finally, we note that DNA microarrays have been employed successfully to monitor metabolic and genetic control of gene expression on a genomic scale. Brown et al.18 used a probe array to identify genes where expression was affected by deletion of a transcriptional co-repressor or overexpression of a transcriptional activator. This highly detailed work was conducted with respect to the organism, Saccharomyces cerevisiae, in terms of the metabolic shift associated with gene expression from fermentation to respiration.As a result of their work, the authors concluded, prophetically, that mutations in specific genes encoding candidate drug targets can serve as surrogates for an ideal chemical inhibitor or modulator of their activity. DNA microarrays can be employed to screen for patterns in the alteration of gene expression prior to assaying for chemicals that produce particular patterns.Concluding remarks 1. There is no doubt that the photolithography-combinatorial synthesis of oligonucleotide arrays is appealing in terms of its elegant simplicity, however, the configuration, by definition, is restricted to the use of relative short oligonucleotides. Despite the possibility to fabricate huge arrays, this factor can lead to a lack of specificity in the characterization of relatively large genomes. Other aspects of this technique are the lack of ability to produce arrays with variable base-length DNA, at least in a straightforward fashion, and the fact that increased chain length of oligos suffers from lack of probe purity. 2. Although some of these issues are resolved in the case of robotic printing, sensitivity may be an issue for this technique because of potential difficulties connected to the availability for hybridization of multi-point attached probes on the poly-llysine surface. In somewhat analogous work with the more traditional membrane-oligonucleotide probe arrays such lack of probe availability for reaction is known to occur.The effect does not constitute a real practical hindrance here, because of the extremely high sensitivity of 32P radiochemical labelling. 3. Both strategies described above generally involve the mandatory use of fluorescent tagging agents. The future might see attempts at label-free detection of hybridization events. A somewhat earlier attempt at this sort of approach was the atomic force microscopy study of protein-based microarrays.19 4.Surface chemistry, in particular, the immobilization of nucleic acids, is an area that clearly requires development. Better control of surface chemical aspects could lead to increased oligonucleotide coverage, enhanced orientation with respect to hybridization and optimization of packing density. Furthermore, at the present time microarrays are quite expensive to produce, which is reflected in the purchase price, and to a large degree are considered to be expendable devices.The future may see new approaches for the ‘regeneration’ of microarrays using chemical or other protocols. 5. Finally, in the new millennium we are undoubtedly going to witness a scaling down in size and, of course, cost of the whole DNA microchip configuration. We are some time away yet from having gene chip systems at the end of every molecular biologist’s bench top, however, the excitement generated by this technology (President Clinton mentioned it in his 1998 State of the Union address) will clearly spawn efforts to increase general availability and ease of use.Acknowledgements We are very grateful to Ted Dixon of BPI Co. Ltd. of Waterloo, Ontario for the provision of the colour microarray image and to Bryan McNeil of the Ontario Cancer Institute, Princess Margaret Hospital, Toronto for donating a photograph of an image connected to his work on the yeast genome, and for much helpful discussion.References 1 J. W. Hawkins, Genomics and Human Therapeutics Development, Drug and Market Development, Report No. 938, Scarborough, MA, 1997. 2 F. Sanger, S. Nicklen and R. Coulson, Proc. Natl. Acad. Sci., 1977, 74, 5463. 3 A. M. Maxam and W. Gilbert, Proc. Natl. Acad. Sci., 1977, 74, 560. 4 R. Drmanac, S. Drmanac, Z. Strezoska, T. Paunesku, I. Labat, M. Zerenski, J. Snoddy, W. K. Funkhouser, B. Koop, L. Hood and R.Crkvenjakov, Science (Washington, D.C.), 1993, 260, 1649. 5 A. Caviani Pease, D. Solas, E. J. Sullivan, M. T. Cronin, C. P. Holmes and S. P. A. Fodor, Proc. Natl. Acad. Sci., 1994, 81, 5022. 6 S. P. A. Fodor, J. L. Read, M. C. Perrung, L. Stryer, A. T. Lu and P. Solas, Science (Washington, D.C.), 1991, 251, 767. 7 T. Kreiner, Am. Lab., 1996, March, p. 39. 8 R. J. Lipshutz, D. Morris, M. Chee, E. Hubbell, M. J. Kozal, N. Shah, N. Shen, R. Yang and S. P. A. Fodor, Biotech., 1995, 19, 442. 9 M. Yang, M. E. McGovern and M. Thompson, Anal. Chim. Acta., 1997, 346, 259. 10 http:/www.affymetrix.com 11 M. Schena, D. Shalon, R. W. Davis and P. O. Brown, Science (Washington, D.C.), 1995, 270, 467. 12 www.synteni.org 13 G. Yershof, V. Barsky, A. Belgovskty, E. Kirrolov, E. Kreindin, I. Ivanov, S. Parinov, D. Gushin, A. Drobishev, S. Dubiley and A. Mirzabekov, Proc. Natl. Acad. Sci., 1996, 93, 4913. 14 J. G. Hacia, L. C. Brody, M. S. Chee, S. P. A. Fodor and F. S. Collins, Nature Genetics, 1996, 14, 441. 15 J. G. Hacia, W. Makalowski, K. Edgeman, M. R. Erdos, C. M. Robbins, S. P. A. Fodor, L. C. Brody and F. S. Collins, Nature Genetics, 1998, 18, 155. 16 J. G. Hacia, K. Edgeman, B. Sun, D. Stern, S. P. A. Fodor and F. S. Collins, Nucl. Acid. Res., 1998, 26, 3865. 17 M. Chee, R. Yang, E. Hubbell, A. Berno, X. C. Huang, D. Stern, J. Winkler, P. J. Lockhart, M. S. Morris and S. P. A. Fodor, Science (Washington, D.C.), 1996, 274, 610. 18 J. L. DeRisi, V. R. Iyer and P. O. Brown, Science (Washington, D.C.), 1997, 278, 680. 19 L. T. Mazzola and S. P. A. Fodor, Biophys. J., 1985, 73, 1653. Paper 9/04581C 1136 Analyst, 1999, 124, 1133–1136

ISSN:0003-2654    

DOI:10.1039/a904581c

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Aspects and applications of non-aqueous high temperature packed capillary liquid chromatography P. Molander, R. Trones,* K. Haugland and T. Greibrokk Department of Chemistry, University of Oslo, H.P.O. Box 1033, Blindern, Oslo, 0315 Norway. E-mail: roger.trones@kjemi.uio.no Received 17th May 1999, Accepted 16th June 1999 The effect of using high isothermal temperatures and temperature programming above ambient in liquid chromatography has been investigated by the use of relatively long packed reversed phase capillary columns and non-aqueous mobile phases.Efficiency measurements have been performed in the temperature interval 50– 175 °C, indicating that the best efficiency was achieved at 100 °C, with a corresponding optimum linear velocity of approximately 0.07 cm s21. The efficiency measurements revealed that the linear velocity had to be slightly increased in order to operate at the optimum reduced plate height at elevated temperatures. The contribution from extra-column volumes increased the plate height at temperatures above 100 °C where the solute had low retention.This effect was closely examined by introducing larger pre-column dead volumes, demonstrating the need to minimize dead volumes. Temperature programming has successfully been used for partial characterization and for purity testing of different polymer additives. The within and between day precision of retention times for the temperature-programmed separations gave a relative standard deviation of 0.3–2.9%.Aim of investigation Several studies have investigated the use of elevated temperatures in liquid chromatography (LC), after the early reports of Strain1 and Chang.2 Most of the early research was carried out on conventional packed columns with an internal diameter (id) of 4.6 mm, whereas to an increasing extent recent studies have been performed with packed or open tubular capillary columns. In addition, nearly all of these studies have been at moderate temperatures with aqueous mobile phases, with applications belonging within the traditional application area of reversed phase LC.Consequently, the main reported advantages have been reduced column back pressure, reduced analysis time and enhanced efficiency.3–10 Traditional solvent gradient elution has been reported to be difficult with columns of small id, due to the small flow rates required.11 However, the smaller dimensions of packed capillary columns make them especially well suited for temperature programming, due to their small thermal mass.12 Hence, temperature-programmed LC has in some studies proven to be an alternative to solvent gradient elution when utilizing packed capillary columns.13–16 Temperature programming is not just an alternative to solvent gradient elution, but is required if successful separations of a niche of applications are to be performed.Exploiting temperature programming can thus extend the application area of LC.Examples are some polymer additives and technical waxes, which often are not adequately separated by gas chromatographic (GC), supercritical fluid chromatographic (SFC) or conventional LC methods. The combination of temperature programming, non-aqueous mobile phases and relatively long, packed capillary columns has enabled the separation of these high molecular weight compounds, exploiting the increased solubility in liquid phases at elevated temperatures.17–21 The most important influence of increased temperature is a reduction in the viscosity of the mobile phase and an increase in diffusion rates.22,23 The diffusion coefficients are in general considered small in LC at ambient temperature, which has led to the conclusion that the contribution from axial diffusion to band broadening may be considered negligible.24 However, at elevated temperatures, and at low linear velocities in particular, the contribution from axial diffusion may become more important.Increased axial solute diffusion in the mobile phase will increase the contribution to band broadening from the B term in the van Deemter equation. The C term contribution should be reduced by elevated temperature due to increased mass transfer between the mobile phase and the stationary phase. The main objective of this study was to investigate the effect of temperature elevation and temperature programming on liquid chromatographic performance using packed capillary columns, non-aqueous mobile phases and high molecular weight test compounds in an extensive temperature interval.In addition, certain polymer additives are characterized, illustrating the potential of temperature-programmed high temperature LC. Experimental Instrumentation The equipment consisted of a Merck LaChrom L-7100 pump (Darmstadt, Germany), a model CI4W manually operated injection valve equipped with a 60 nL internal injection loop (Valco Instruments, Houston, TX, USA), a model 3400 gas chromatograph (Varian, Walnut Creek, CA, USA) and a UV 2000 detector (Thermo Separation System, Fremont, CA, USA).On-column detection took place in a fused silica capillary of 20 cm, connected to the column inside the oven (100 mm id, 375 mm od). A linear fused silica restrictor of approximately 25 cm (20 mm id, 375 mm od) was connected to the detection capillary to prevent the mobile phase from boiling. The mobile phase was covered with helium gas to avoid oxygen from entering the mobile phase.For the polymer additives, a modified evaporative light scattering detector (ELSD) was utilized (Varex Mark III, Alltech Associates, Deerfield, IL, USA).25 The fused silica capillary transfer line from the column outlet to the nebulizer acted as a restrictor (40 cm, 20 mm id, 375 mm od). Analyst, 1999, 124, 1137–1141 1137Column preparation The capillary columns (id = 0.32 mm) were packed with 3 mm or 5 mm porous Hypersil ODS particles (Hypersil, Shandon, UK), according to a procedure previously described.17 Materials The fused silica capillaries were purchased from Polymicro Technologies Inc.(Phoenix, AZ, USA). Unions, nuts and ferrules were produced by Valco. Carbon dioxide (99.998%) was purchased from AGA (Oslo, Norway) while HPLC grade of acetonitrile, ethyl acetate and N,N-dimethylformamide were obtained from Rathburn (Watherburn, UK). Triethylamine was purchased from Fluka Chemie AG (Buchs, Switzerland).Water was deionized and glass distilled. The following polymer additives were supplied by Borealis A/S (Stathelle, Norway): Irganox 1076 [octadecyl 3-(3,5-di-tert-butyl-4-hydroxyphenyl) propionate], Irganox 1010 [benzenepropanoic acid- 3,5-bis(1,1-dimethylethyl)-4-hydroxy-2,2-bis{[3-[3,5-bis(1,1- dimethylethyl)-4-hydroxyphenyl]-1-oxopropoxy]methyl}-1,3- propanediylester], Irgafos 168 [phenol-2,4-bis(1,1-dimethylethyl)- phosphite (3+1)], Irgafos P-EPQ [phosphonous acid [1,1A-biphenyl]-4,4A-diylbistetrakis-[2,4-bis(1,1-dimethylethyl)- phenyl] ester], Irgafos 168-phosphate, Anchor DNPD [N,NA-dib- naphthyl-p-phenylenediamine] and Chimassorb 119 [N,NA,N,NAB-tetrakis(4,6-bis(butyl-(N-methyl-2,2,6,6-tetramethylpiperidin- 4-yl)-amino)triazin-2-yl)-4,7-diazadecane- 1,10-diamine].Uracil and naphthalene were purchased from Fluka. Results and discussion Column efficiency at elevated temperatures In the first part of the study, the reduced plate height, h, of Irganox 1076 was measured at 50, 75, 100, 125, 150 and 175 °C at flow rates from 1 to 6 mL min21.This compound was strongly retained at temperatures below 50 °C. The column used was 0.32 3 700 mm, packed with 5 mm Hypersil ODS particles. The mobile phase consisted of acetonitrile–DMF (90 + 10, v/v). Three injection replicates were performed at each temperature and flow rate. Uracil was used as to-marker, and the UV detector was operated at 280 nm.The linear velocity increased linearly with temperature at all the flow rates investigated (r2 > 0.995), and the 30% increase when elevating the temperature from 50 to 175 °C was similar for all flow rates. When operating at a flow rate of 1.0 mL min21, the linear velocity increased from 0.03 to 0.04 cm s21 when the temperature was elevated from 50 to 175 °C. These values are definitely lower than the optimal linear velocity, uo, with regard to efficiency, and hence, the band broadening process is primarily controlled by axial solute diffusion, which increases with temperature.Accordingly, at a flow rate of 1 mL min21, the reduced plate height increased when elevating the temperature gradually from 50 to 175 °C. Similar behavior has been observed earlier by Yoo et al., using an aqueous mobile phase in the temperature interval 25–60 °C.13 Van Deemter plots of reduced plate height as a function of linear velocity at different temperatures are shown in Fig. 1. As illustrated, the optimal linear velocity with regard to efficiency increased with temperature, which is in accordance with results presented by Yoo et al.13 and Antia and Horvath.26 At optimal linear velocities and beyond, it was found that the reduced plate height decreased with temperature up to 100 °C. At these linear velocities mass transfer mainly dominates the contribution to band broadening, which should be positively influenced by temperature elevation.However, when increasing the temperature from 100 to 125 °C and above, an increase in the reduced plate height was observed. This behavior might be attributed to extra-column dead volume contributions to band broadening, having a higher effect at the higher temperatures where the solute had little retention, which will be discussed in detail in the next section. Yoo et al. reported that the reduced plate height at the optimal linear velocity was roughly invariant with temperature,13 in agreement with theoretical calculations by Antia and Horvath.26 However, Sheng et al.found that the reduced plate height at optimal linear velocity did vary in the temperature interval from 24 to 80 °C.27 Regarding the present study, the analyte was of rather high molecular weight (MW = 531 g mol21), with a lower diffusion in the mobile phase compared to smaller molecules. Consequently, large molecules are more influenced by operation at elevated temperatures, where increased diffusion coefficients are observed.28 Differences between the LC systems, such as dead volumes, the temperature interval investigated and choice of analyte, stationary and mobile phase, seem to be relevant to the effect of elevated temperature upon efficiency in LC.Contribution to band broadening from extra-column volumes The column end-fittings used in this experiment were not optimized with regard to dead volumes. The use of better suited, polymeric end-fittings was investigated, but they were not found suitable due to the high temperatures and the mobile phases used.In addition, the column was located in an oven, requiring rather long fused silica capillaries for connection to the injector and the detector, introducing extra dead volumes. The dead volumes remained constant at all temperatures, while the retention time on the chromatographic column decreased with temperature. When the temperature was elevated from 50 to 175 °C, the retention factor, k, decreased exponentially from approximately 2.90 to 0.15.The strongest reduction of the retention factor was observed in the temperature interval from 50 to 100 °C, while at temperatures above 125 °C the reduction was rather small, and Irganox 1076 became practically unretained. The temperature region where only small reductions on the retention factor were observed is identical to the temperature region where an increase in the optimal reduced plate height was observed.Consequently, the relative contribution to band broadening from extra-column Fig. 1 Van Deemter plots of Irganox 1076 performed at different temperatures. The mobile phase consisted of acetonitrile–DMF (90 + 10, v/ v), and the column was 0.32 3 700 mm, packed with 5 mm Hypersil ODS particles. On-column UV detection at 280 nm. 1138 Analyst, 1999, 124, 1137–1141dead volumes most certainly increased rapidly at temperatures where the solute had low retention.Horvath and Lin showed a relationship between the non-linear part of a curve, where peak widths, tw0.5, were plotted vs. the retention factor, and the presence of dead volumes in a LC system.29 Their experimental data were obtained at ambient temperature, utilizing a sample solution containing five homologue isomeric aromatic acids of different retention factors in the defined LC system. All other chromatographic conditions were kept constant during their experiment.The plots showed nonlinear behavior at low retention factors, if dead volumes were present. Alternatively to the use of five different compounds of different retention factors, temperature adjustments can yield different retention factors, utilizing one single compound. Fig. 2 shows a plot of peak width vs. retention factor of Irganox 1076 at different temperatures, using a flow rate of 3 mL min21. The extra pre-column dead volume in the system was calculated to be 0.5 mL.The shift in the curve was observed around 100 °C, indicating that at this point contributions to band broadening from extra-column dead volumes became critical, which is in accordance with the observed increase in the reduced plate height in this temperature region. By replacing the fused silica capillary from the injector to the column with a capillary of larger dimensions, the dead volume of the pre-column connecting capillary was increased from 0.5 to 7.0 mL. The plot of the peak width vs. the retention factor utilizing the capillary of 7.0 mL is shown in Fig. 2, as well, displaying that a measurable contribution to band broadening from extra-column volumes occurred already at 75 °C. Retention mechanism Operation at elevated temperatures, and at sub-ambient temperatures as well, has in certain studies proven to influence the retention mechanism between the solute and the stationary phase material, often resulting in selectivity changes. A van’t Hoff plot of the natural logarithm of the retention factor, ln k, vs.the reciprocal temperature, 1/T, is often performed when operating at various temperatures, with a non-linear plot indicating a shift in the retention mechanism involved.30 The enthalpy change, DH, of the interaction can be calculated from the slope of a linear van’t Hoff plot. A linear van’t Hoff plot, indicating that the retention mechanism is invariant with temperature, is essential with regard to the reliability of the plots shown in Fig. 2. Consequently, a van’t Hoff plot of the interaction involving Irganox 1076 and the stationary phase in the temperature interval 50–175 °C, at a flow rate of 3 mL min21, was performed, showing linear behavior with a correlation factor of 0.9998. From the slope value of the regression line the interaction enthalpy was calculated to be 229.3 kJ mol21, which is fairly high, due to the relatively high molecular weight of Irganox 1076.28 Typical enthalpy values for low molecular weight compounds in reversed phase LC systems are approximately 217 kJ mol21.31 Retention time; within and between day precision When operating at ambient temperature, at the relatively high pressures commonly used in LC, local zones of variable temperature are formed, due to friction between the mobile phase and the stationary phase as the solvent is pumped through the column.This causes local viscosity changes of the mobile phase, and thereby different mobile phase velocities for the various radial positions in the column.Poppe et al. predicted that the temperature dependence of the retention factor is about 2% per °C, and that viscous heat dissipation will affect the accuracy of the retention times by a few per cent.32 They suggested that the temperature fluctuations could be stabilized by operation at elevated temperatures, only implied by using small-bore columns. The relative standard deviations, RSD, observed in the present study, with regard to retention times in the temperature interval 50–175 °C, were less than 0.1% at all flow rates (n = 3).In addition, the within day precision (n = 3) and the between day precision (n = 2) with regard to retention times were determined in temperature programming mode, where the mobile phase viscosity is changed gradually during the chromatographic run. A typical temperature program for the separation of polymer additives using high temperature LC and non-aqueous mobile phases was utilized; 50 °C for 1 min, then 4 °C min21 to 150 °C.A test mixture consisting of naphthalene, Irganox 1010 and Irganox 1076 was chosen, with naphthalene eluting first, Irganox 1010 in the middle and Irganox 1076 at the end of the chromatogram. Other chromatographic conditions were kept constant. The RSD with regard to retention time within day precision ranged from 0.3 to 2.9%, as shown in Table 1, which is slightly poorer than the repeatability typical for solvent gradient elution separations.33 The between day precision RSD ranged from 2.0 to 2.6%, which is in the same area as the within day precision.Introduction of a dedicated instrument for temperature controlled m-LC would probably further improve the performance and retention time precision of temperature-programmed separations. Separation of polyolefin additives Fig. 3 shows the separation of two commonly used antioxidants in polyolefins, Irganox 1076 and Irgafos 168, and the oxidation product of Irgafos 168, Irgafos 168-phosphate, utilizing temperature- programmed LC.It is often of interest to determine the amount of these three compounds in extracts from polyolefins, in order to determine how far the oxidation process of the polymer has gone. Baseline separation of the three compounds was achieved, utilizing temperature-programmed packed capillary LC. Normally, both GC and traditional HPLC are used as complementary techniques to adequately separate and identify these compounds by retention time.34 Fig. 2 Plots of peak width at half the peak height (tw0.5) vs. retention factor (k) for determination of the presence of critical extra-column band broadening at various temperatures. The flow rate was 5 mL min21. Other conditions as given in Fig. 1. Analyst, 1999, 124, 1137–1141 1139Fig. 4 shows the temperature-programmed packed capillary LC characterization of Irgafos P-EPQ, which is a polymer antioxidant, containing five different components; three diphosphonites, one monophosphonite and one phosphite compound (Irgafos 168).Irgafos P-EPQ has previously been separated by the use of SFC, using open tubular Phenyl-5 columns.35 The SFC analysis revealed that the product contained several impurities or oxidation products. The packed capillary temperature-programmed LC method (Fig. 4) revealed the existence of several additional compounds, compared to the SFC analysis, indicating the high-resolution potential of the present technique.Fig. 5 shows the separation of Anchor DNPD. According to the authors’ knowledge, chromatographic characterization of this polyolefin antioxidant has previously never been published. Although the compound is a highly basic compound containing several amino groups, it was not necessary to add amines to the mobile phase or to operate at low pH. As illustrated in Fig. 5, the separation of Anchor DNPD resulted in two major peaks, and several minor, most probably impurities or oxidation products. Chimassorb 119 is a high molecular weight (MW = 2286 g mol21) hindered amine light stabilizer, added to polyolefins for long-term UV and thermal stability. This compound contains several basic amino groups, and is not easily separated using SFC, GC or traditional HPLC methods.Fig. 6 shows the separation of Chimassorb 119, utilizing temperature-programmed packed capillary LC. A mobile phase consisting of acetonitrile–ethyl acetate– triethylamine (50 + 40 + 10, v/v) was found appropriate for this separation, using a temperature program starting from 30 °C.One major peak was separated from the other smaller components. Conclusions Temperature-programmed packed capillary LC has shown to be a robust and accessible technique, and has the potential of making available the separation of complex samples containing relatively high molecular weight compounds of low water solubility.The variety of stationary and mobile phases available, in addition to the relatively easy way of adjusting the elution strength and the solubility of the solutes in the mobile phase utilizing temperature programming, make this technique Table 1 Within (n = 3) and between (n = 2) day precision for the temperature-programmed separation of naphthalene (Napht.), Irganox 1010 (Ix 1010) and Irganox 1076 (Ix 1076).Temperature program: 50 °C for 1 min, then 4 °C min21 to 150 °C.Other experimental conditions as in Fig. 1 Retention time/min Day 1 Day 2 Napht. Ix 1010 Ix 1076 Napht. Ix 1010 Ix 1076 #1 12.48 17.34 21.54 11.90 16.32 20.59 #2 12.24 17.34 21.97 12.29 16.50 20.75 #3 12.19 17.25 21.53 11.66 17.23 21.69 Average 12.30 17.31 21.68 11.95 16.68 21.01 Within day precision RSD (%) 1.26 0.30 1.16 2.66 2.89 2.83 Between day precision Napht. Ix 1010 Ix 1076 RSD (%) 2.04 2.62 2.21 Fig. 3 Temperature-programmed separation of Irgafos 168-phosphate (1), Irganox 1076 (2) and Irgafos 168 (3) dissolved in dichloromethane on a 0.32 3 400 mm column packed with 3 mm Hypersil ODS particles.The mobile phase consisted of 100% acetonitrile. Temperature program; 30 °C for 5 min, then 5 °C min21 to 120 °C. Detection was performed with a modified evaporative light scattering detector. The flow rate was 5 mL min21. Fig. 4 Temperature-programmed separation of Irgafos P-EPQ dissolved in dichloromethane on a 0.32 3 700 mm column packed with 5 mm Kromasil C18 particles.The mobile phase consisted of acetonitrile–DMF (90 + 10, v/v). Temperature program; 50 °C for 15 min, then 5 °C min21 to 150 °C. Detection was performed with a modified evaporative light scattering detector. The flow rate was 5 mL min21. Fig. 5 Temperature-programmed separation of Anchor DNPD dissolved in acetonitrile–DMF (95 + 5, v/v) on a 0.32 3 400 mm column packed with 3 mm Hypersil ODS particles. The mobile phase consisted of acetonitrile– water (90 + 10, v/v).Temperature program; 30 °C for 1 min, then 3 °C min21 to 100 °C. On-column UV detection at 280 nm. The flow rate was 3 mL min21. 1140 Analyst, 1999, 124, 1137–1141capable of filling the void between SFC, GC, high temperature size exclusion chromatography (SEC) and traditional HPLC, with regard to a niche of applications. Further development of this technique will focus on the separation of high molecular weight polymers of low solubility at ambient temperature and on the coupling to other detectors.Acknowledgements The Norwegian Research Council (Science and Technology) supported Molander and Trones. Dag Roar Hegna at Borealis A/ S (Stathelle, Norway) provided the polymer additives. Bert Ooms (Spark Holland) is greatly acknowledged for helpful discussions. References 1 H. H. Strain, Ind. Eng. Chem., Anal. Ed., 1946, 18, 605. 2 L. T. Chang, Anal. Chem., 1953, 25, 1235. 3 T. Takeuchi, Y. Watanabe and D. Ishii, J. High Resolut.Chromatogr., 1981, 4, 300. 4 Y. Hirata and E. Sumiya, J. Chromatogr., 1983, 267, 25. 5 G. Liu, N. M. Djordjevic and F. Erni, J. Chromatogr., 1992, 592, 239. 6 G. Liu, L. Sveson, N. M. Djordjevic and F. Erni, J. Chromatogr. A, 1993, 633, 25. 7 G. Liu, N. M. Djordjevic and F. Erni, J. Chromatogr., 1992, 598, 153. 8 N. M. Djordjevic, D. Stegehuis, G. Liu and F. Erni, J. Chromatogr., 1993, 629, 135. 9 G. Kura, E. Kitamura and Y. Baba, J. Chromatogr. A, 1993, 628, 241. 10 P.Molander, T. E. Gundersen, C. Haas, T. Greibrokk, R. Blomhoff and E. Lundanes, J. Chromatogr. A, in the press. 11 Z. Zhang and A. G. Marshall, J. High Resolut. Chromatogr., 1998, 12, 574. 12 D. Ishii, Introduction to Microscale High-performance Liquid Chromatography, VCH, New York, 1988, p. 7. 13 J. S. Yoo, J. T. Watson and V. L. McGuffin, J. Microcolumn Sep., 1992, 4, 349. 14 M. H. Chen and C. Horvàth, J. Chromatogr. A, 1997, 788, 51. 15 D. J. Miller and S. B. Hawthorne, Anal.Chem., 1997, 69, 623. 16 F. Houdiere, P. W. J. Fowler and N. M. Djordjevic, Anal. Chem., 1997, 69, 2589. 17 R. Trones, A. Iveland and T. Greibrokk, J. Microcolumn Sep., 1995, 7, 505. 18 P. Molander, E. Ommundsen and T. Greibrokk, J. Microcolumn Sep., in the press. 19 R. Trones, T. Andersen, D. R. Hegna and T. Greibrokk, J. Chromatogr. A, submitted. 20 R. Trones, T. Andersen, D. R. Hegna and T. Greibrokk, J. Microcolumn Sep., submitted. 21 P. Molander, E. Ommundsen and T. Greibrokk, Chromatographia, submitted. 22 J. J. van Deemter, F. J. Zuiderweg and A. Klinkenberg, Chem. Eng. Sci., 1956, 5, 27. 23 A. Einstein, Z. Electrochem., 1908, 14, 1908. 24 C. F. Poole and S. K. Poole, Chromatography Today, Elsevier Science B.V., Amsterdam, The Netherlands, 1991, p. 17. 25 R. Trones, T. Andersen and T. Greibrokk, J. High Resolut. Chromatogr., 1999, 22(5) 283. 26 F. D. Antia and C. Horvath, J. Chromatogr., 1988, 435, 1. 27 G. Sheng, Y. Shen and M. L. Lee, J. Microcolumn Sep., 1997, 9, 63. 28 H. Chen and C. Horvath, Anal. Methods Instrum., 1993, 1(4), 213. 29 C. Horvath and H. Lin, J. Chromatogr., 1978, 149, 43. 30 C. M. Bell, L. C. Sander and S. A. Wise, J. Chromatogr. A, 1997, 757, 29. 31 M. C. Gennaro, D. Giacosa, C. Abrigo and E. Marengo, J. Chromatogr. Sci., 1984, 33, 363. 32 H. Poppe, J. C. Kraak, J. F. K. Huber and J. H. M. van den Berg, Chromatographia, 1981, 14, 9. 33 H. E. Schwartz and V. V. Berry, LC Mag., 1985, 3, 110. 34 Personal communication from Borealis A/S, Stathelle, Norway. 35 T. Greibrokk, B. E. Berg, S. Hoffmann, H. R. Norli and Q. Ying, J. Chromatogr., 1990, 505, 283. Paper 9/03908B Fig. 6 Temperature-programmed separation of Chimassorb 119 dissolved in ethyl acetate on a 0.32 3 400 mm column packed with 3 mm Hypersil ODS particles. The mobile phase consisted of acetonitrile–ethyl acetate– triethylamine (50 + 40 + 10, v/v). Temperature program; 30 °C for 5 min, then 2 °C min21 to 120 °C. Detection was performed with a modified evaporative light scattering detector. The flow rate was 3 mL min21. Analyst, 1999, 124, 1137–1141 1141

ISSN:0003-2654    

DOI:10.1039/a903908b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Automated extraction chromatographic separations of actinides using separation-optimized sequential injection techniques Jay W. Grate,*a Oleg B. Egorova and Sandra K. Fiskumb a Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, WA 99352 USA b Radiochemical Processing Laboratory, Pacific Northwest National Laboratory, Richland, WA 99352 USA Received 3rd March 1999, Accepted 20th May 1999 A sequential injection (SI) separation system has been developed for automated analytical separations of actinides using an actinide specific extraction chromatographic material (TRU-resin, Eichrom Industries, Inc., USA).On-line liquid scintillation counting was used to observe eluting species during method development, and fraction collection and alpha energy analysis were used for quantification. Several procedures for individual and group actinide elution are demonstrated and discussed, including elution of actinides as a single group; elution as groups based on valence state; selective separation of Pu using on-column redox chemistry; selective separation of Th; and various sequential actinide elution schemes.Eluent solution compositions and reagent chemistries were investigated with regard to elution peak shapes, selectivity, recoveries, carryover, and suitability for rapid automated procedures. The SI separation methods described serve as the basis for an automated actinide separation work station.An automated actinide separation procedure has been applied towards the analysis of Am, Cm, and Pu isotopes in three types of aged nuclear waste samples. Results from automated analytical separations followed by quantification by alpha spectroscopy were in good agreement with results obtained using manual separation techniques. Introduction Environmental restoration of the radioactively contaminated sites and processing of stored radioactive wastes require reliable radioanalytical characterization methods.1 Determination of actinide isotopes in nuclear waste is important due to their long radioactive half-lives, high radiological toxicities, and criticality concerns.In addition, actinide isotope determinations are necessary for waste classification purposes.2 Actinide analyses typically require preconcentration and separation from excess inactive matrix components and highly radioactive fission products. Group and/or individual actinide separations are required because a number of important actinide isotopes have unresolvable alpha-energies (e.g., 241Am /238Pu) or have mass to charge ratios that are indistinguishable by low resolution mass spectrometric techniques (e.g. 241Am/241Pu, 238Pu/238U, etc.) Therefore, the development of improved analytical separation methods for actinide elements is of interest. Recently, various extraction chromatographic methods have been developed by Horwitz and co-workers at the Argonne National Laboratory (USA) to simplify and improve the chemical separations required in radiochemical analyses.3–8 Impregnation of macroreticular polymer beads with a solution of a neutral bifunctional organophosphorus complexant, octyl- (phenyl)-N,N-diisobutylcarbamoylmethylphosphine oxide (CMPO) in tri-n-butyl phosphate (TBP), yielded an actinide specific sorbent material now called TRU-resin (Eichrom Industries, Inc., USA).3,6,9 Nitrato complexes of tri-, tetra- and hexavalent actinides are strongly and nearly selectively extracted from nitric acid solutions by the CMPO–TBP stationary organic phase on TRU-resin, with retention increasing with greater aqueous phase nitric acid concentration. Chloro complexes of tetravalent and hexavalent actinides, but not trivalent actinides, are extracted from hydrochloric acid solutions.Actinides retained on TRU-resin columns can be recovered, individually or in groups, by eluting with acidic solutions, complexants, or redox reagents.TRU-resin can be utilized for a number of analytical purposes, including separation of the actinides as a group from a sample matrix; group actinide separations based on the valence state; individual separation of Am/Cm and Pu from each other and other actinides; and possibly the sequential separation of individual actinides on a single TRU-resin column.1,3,6,9–12 In a number of radiochemical procedures, TRU-resin has been utilized in combination with other extraction chromatographic materials such as Actinide-resin, UTEVA-resin, and TEVA-resin (Eichrom).5,7,12,13 The use of TRU-resin for analyses of nuclear waste samples, enviromental samples, and biological samples has been described. 1,3,6,7,9,14,15 In a typical manual extraction chromatographic separation, samples and various eluents are added to the top of an open column operated under gravity flow, and fractions are collected for subsequent radioactivity measurement or additional separations.Vacuum manifolds for solid phase extraction (SPE) are sometimes used to speed up the separation.7 Despite significant improvement over classical radiochemical separation procedures based on combinations of precipitation, solvent extraction, and ion exchange steps, the extraction chromatographic separation format remains somewhat tedious when performed manually, and the analyst is exposed to the open sources of radioactivity. It would be advantageous to perform these separations in an automated closed-column format.Though known primarily for automating simple chemical analyses, flow injection (FI) and sequential injection (SI) techniques provide a very versatile fluid handling approach that can be used to automate chemical separations.16–19 In several recent reports, FI and SI methods have been demonstrated as Analyst, 1999, 124, 1143–1150 1143useful approaches for automating extraction chromatographic separations of radionuclides prior to detection by inductively coupled plasma mass spectrometry (ICPMS) or radiometric methods.19–27 In one of our own recent reports, we described the use of a continuous-forward-flow FI instrument with an on-line liquid scintillation detector to investigate the separation of Am and Pu using a TRU-resin column.25 Particular attention was paid to the on-column redox chemistries involved in adjusting the Pu speciation from the trivalent to the tetravalent and back to the trivalent states.No nuclear waste samples were analyzed in this study. SI techniques represent a more recent approach to flow analysis18,28 that has been demonstrated for automated extraction chromatographic separations and analyses of 99Tc and 90Sr in aged nuclear waste samples.23,26,27 These separations involve loading the sample on an extraction chromatographic column, washing away matrix and interfering species while the desired analyte is retained, and then eluting the analyte with another solution. These procedures are most effectively carried out using a ‘separation-optimized’ SI methodology for solution handling.19,23,26 This approach supercedes an initial report where conventional ‘stacked-zone’ SI methods were used to load the required sample, wash and eluent solutions into a holding coil.22 In the separation-optimized SI method, each eluent solution is loaded into a wide-bore holding coil at a high flow rate, and then immediately dispensed prior to loading and dispensing the next solution.An air segment between each solution and the carrier fluid prevents dispersion. This method provides rapid manipulation of milliliter quantities of discrete solution compositions for column separation procedures. The separation-optimized SI approach is well suited for the automation of more challenging extraction chromatographic separations of actinide elements involving multiple eluent and reagent solutions. In this report we investigate and demonstrate a number of automated group and individual actinide separations on TRU-resin in a SI format.The SI separation methods described can serve as the basis for an automated actinide separation work station. On-line liquid scintillation counting was used during procedure development to observe eluting species. Fraction collection and alpha energy analysis were used for quantification. Alpha spectrometry is a standard method of analyzing separated actinides that provides isotopic information.Isotopic, as opposed to elemental information, is required because different isotopes have different half-lives, specific activities, and fissile properties. Activities of individual alphaemitting isotopes are required for waste classification, for bioassay purposes, and for assessing criticality issues. In this paper, the SI automated separation method is applied to the determination of Am, Cm, and Pu isotopes in three types of aged nuclear waste samples.Experimental Reagents, standards and nuclear waste samples All reagents used were of analytical grade. Deionized water (18 MWcm21) was obtained from a Barnstead E-Pure (Thermolyne Corporation, Dubuque, IA, USA) water purification system and was used for all dilutions. A 0.02 mol l21 Ti(iii) chloride reagent solution in 4 mol l21 HCl was prepared daily by an appropriate dilution of a 20% stabilized solution (EM Science, Gibbstown, NJ, USA). A 0.05 mol l21 Fe(ii) sulfamate–ascorbic acid reagent in 4 mol l21 HCl was prepared daily by dissolving required amount of FeCl2.4H20 (Sigma, St.Louis, MO, USA) in 0.1 mol l21 solution of ascorbic and sulfamic acids in 4 mol l21 HCl. A 0.5 mol l21 solution of sodium nitrite in water, utilized for Pu oxidation, was prepared fresh weekly. The scintillation cocktail used with the flow-through detector was Ultima Flo-M (Packard, Downers Grove, IL, USA). Ultima Gold (Packard) was used in static LS measurements.Solutions of radionuclide tracers (90Sr,137Cs, 241,243Am, 239, 242Pu, 230Th, 233U) were obtained from in-house standards laboratory. Neptunium-239 source was obtained from “milking” 243Am source fixed on a small TRU-resin column. Nuclear waste samples used in this study included vitrified glass waste, aged irradiated nuclear fuel (K-basin sludge samples, Hanford site, USA), and waste samples derived from the high-level waste storage tanks located at the Hanford site (tank waste supernate).The samples were obtained as dilute processed solutions after the initial sample preparation steps were performed in shielded analytical facilities (hot cells). The initial sample preparation steps and manual analysis procedures have been described previously.11,22,23,26 Apparatus Sequential injection system. An actinide separation instrument, shown schematically in Fig. 1, was configured using a FIALab 3000 (Alitea USA, Medina, WA, USA) sequential injection system equipped with a 24 000 step digital syringe pump (syringe volume 10 ml) and 14 port multiposition valve (Cheminert low-pressure valve, Valco Instruments Co, Inc., Houston, TX, USA).The holding coil was constructed from 1.6 mm id FEP Teflon tubing (Upchurch Scientific, Oak Harbor, WA, USA) of 6 m length (calculated volume 12 ml). All transport and reagent lines were made of 0.76 mm id FEP Teflon tubing (Upchurch Scientific). Sample lines were made of 0.5 mm id FEP Teflon tubing.The size of the columns used in this study was 4 mm id 3 50 mm (calculated bed volume 0.63 ml). The columns were constructed of parts from the Omega- Chrom column system (Upchurch Scientific) and frits from the Quick-Snap column system (IsoLab, Inc., Akron, OH, USA). TRU-resin (Eichrom Industries, Inc., Darien, IL, USA) sorbent extraction material of 20–50 mm particle size was slurried in water and packed into the column under slight pressure using a 10 ml syringe.The free column volume (FCV) of the TRU-resin column was 0.42 ml (gravimetric measurement). Separation experiments were performed using freshly packed columns unless indicated otherwise. Dye dispersion measurements were conducted using a SPD10AV UV-Vis detector (Shimadzu) positioned in place of the sorbent column. The SI system was controlled via a serial cable using FIALab software (Alitea USA) running on a lap-top PC. Radioactivity measurements and fraction collection.The on-line flow-through liquid scintillation detector, a Radiomatic ™ 515A (Packard Instrument Company, Meriden, CT, USA) was configured with a 0.5 ml flow cell and operated as described previously.22 The detector integration time (time to Fig. 1 Schematic diagram of the sequential injection actinide separation instrument. R1, R2, and R3 denote reagents. 1144 Analyst, 1999, 124, 1143–1150accumulate counts for each data point reported) was 6 s. For the off-line radioactivity measurements, the column outlet transport line was connected to a programmable FC205 (Gilson, Middleton WI, USA) fraction collector equipped with a twoway diverter valve. Packard Tri Carb 2550 TR/AB liquid scintillation spectrometer (Packard) was used for off-line liquid scintillation measurements.Alpha spectrometry was performed using 450 mm2 Ortec semiconductor ion implanted detectors (EG&G Ortec, Oak Ridge, TN, USA) equipped with Canberra electronics and data acquisition system.Samples for alpha spectroscopy were prepared using NdF3 microprecipitation procedure.11 Gamma spectroscopy was performed using highpurity germanium detectors (HPGe) detectors (Oxford Instruments, Oak Ridge, TN; and Canberra Industries, Meridien, CT, USA) equipped with Canberra electronics and data acquisition system (Canberra Industries). Procedures SI solution handling. The SI separation instrument, shown schematically in Fig. 1, was set up to deliver various sample and eluent solutions to a TRU-resin column and transport the separated species to either an on-line detector during procedure development or to a fraction collector for off-line sample analysis.A wide-bore holding coil (1.6 mm id) facilitated the rapid loading of large solution volumes into the holding coil at 10 ml min21 without outgassing. Dispersion in the wide bore holding coil was eliminated by introducing a 100 ml air segment between the carrier (0.3 mol l21 nitric acid) and the selected solution.This separation-optimized fluid handling approach has been described in detail in previous papers.19,23,26 Solution delivery for the column conditioning, sample load, column wash, and various actinide elution steps are detailed in Table 1. Aspiration flow rates were 10 ml min21 except for the sample solutions, which were loaded into the holding coil at 1 ml min21. Solutions were delivered to the column at 2 ml min21, except for the column conditioning step, which was performed at 3 ml min21.Sample load matrix for TRU-resin separations. Prior to the separation, 1 mol l21 Fe(NO3)3 solution in 2 mol l21 HNO3, and solid sulfamic and ascorbic acids were added to the sample solution in 2 mol l21 nitric acid to give a resulting solution containing 0.05 mol l21 Fe, 0.1 mol l21sulfamic acid, and 0.1 mol l21 ascorbic acid. This treatment was used to reduce any Fe(iii) species to Fe(ii) and ensure that all Np is present as Np(iv) and fix all the Pu in a single oxidation state of +iii.On-column Pu oxidation. To perform on-column Pu oxidation during the column wash, acidic nitrite wash solution was prepared fresh on-line by first aspirating the required volume of 2 mol l21nitric acid column wash (nominally 8 ml), and then aspirating 150 ml of 0.5 mol l21 NaNO2 in water, followed by an additional 150 ml of 2 mol l21 nitric acid. With this sequence, the 2 mol l21 nitric acid column wash solution is never diluted by the aqueous nitrite more than by a factor of 2 (as established in prior dye injection experiments), thus maintaining the nitric acid concentration in the range required to retain actinides on the column during the wash step.25 Nuclear waste sample analysis.Each analytical sample batch included quality control samples consisting of two blank samples, and a blank sample spiked with the known amounts of 241Am and 239Pu. Known amounts of 243Am and 242Pu were added to each solution in order to trace the overall recovery of the analysis procedure. The recovery tracer approach using these isotopes is valid because the 243Am and 242Pu activity in these samples is negligibly low relative to that of 241Am and 239,240Pu and the activity of the 243Am and 242Pu spikes added.Following the addition of spikes, the samples were evaporated to moist dryness and redissolved in 2 mol l21 HNO3. Just prior to the separation, the samples were subjected to reductive treatment as described in Experimental.Following the TRU-resin column conditioning and sample load steps (injected sample volume 1 ml), the column wash/Pu oxidation sequence was applied (see Table 1). Next trivalent actinides were eluted with 10 ml 4 mol l21 HCl, followed by the selective Pu reduction–elution using 10 ml of the reducing eluent. A 4 mol l21 HCl solution of 0.02 mol l21 Ti(iii) chloride was used as Pu eluent in the analysis of K-basin samples; 4 mol l21 HCl solution of 0.05 mol l21 Fe(ii) 0.1 mol l21 sulfamic–0.1 mol l21 ascorbic acid was used to elute Pu in the analysis of tank waste and vitrified waste samples.The separations were performed using freshly packed columns, which were replaced after the Pu elution step. Results and discussion Removal of the sample matrix All the actinide separations to be described begin by retaining actinides on the TRU-resin column and removing unretained species in the column wash.Nitric acid solutions are used for these steps. Capacity factors, kA, range from ~ 100 for least retained Am(iii) to over 104 for the most strongly retained tetravalent actinides at nitric acid concentrations exceeding 1 mol l21.3,9 The predominant radioactive species in aged nuclear waste that are removed in the column wash are the gamma and beta emitting fission products, 137Cs and 90Sr/90Y. Initial elution of these species in the column wash step is shown in Fig. 2. Sr and Cs ions are unretained on TRU-resin in 2 mol l21 nitric acid. Using fraction collection, we established that a column wash using 4 ml (10 FCV) of 2 mol l21 nitric acid was sufficient to provide decontamination factors exceeding 103 for 137Cs and 104 for 90Sr. Yttrium showed noticeable retention (see Fig. 2) corresponding to capacity factor kA ~ 11 in 2 mol l21 nitric acid. Yttrium removal is virtually complete with 24 FCV (10 ml) of 2 mol l21 nitric acid wash.Fraction collection experiments indicated that 99.9% of 90Y originally present in the sample is removed using 10 ml of the 2 mol l21 nitric acid wash. These fission products account for the majority of the activity of aged high level nuclear waste samples. The removal of the radioactive fission product matrix simplifies subsequent handling and processing of the separated actinide fractions, whose content must be quantified by radiometric or mass spectrometric methods. Table 1 Reagent delivery procedures for the automated actinide separations Condition/Load/Wash Sequencesa— Column conditioning 2 ml 2 mol l21 HNO3 Sample load 0.1–1 ml Column wash–Pu oxidation (1) 2 ml 2 mol l21 HNO3 (2) 0.15 ml 2 mol l21 HNO3– 0.15 ml 0.5 mol l21 NaNO2–8 ml 2 mol l21 HNO3 Column wash without Pu oxidation 10 ml 2 mol l21 HNO3 Actinide Elution Sequences— Group actinide elution 7 ml 0.1 mol l21 NH4HC2O4 Trivalent actinide elutionb 5 (10 ml) 4 mol l21 HCl Pu reduction-elutionc 7 (10 ml) reductant in 4 mol l21 HCl Tetravalent actinide elution 7 ml 1 mol l21 HCl–0.05 mol l21 H2C2O4 Hexavalent actinides (U) elution 7 ml 0.1 mol l21 NH4HC2O4 a 4 3 50 mm sorbent column; separation flow rate 2 ml min21.b 5 ml Used with on-line detection; 10 ml used with fraction collection. c 7 ml Used with on-line detection; 10 ml used with fraction collection. Analyst, 1999, 124, 1143–1150 1145Elution of actinides as a single group Following the sample load and column wash with 2 mol l21 nitric acid, actinides can be eluted as a single group using 0.1 mol l21 ammonium bioxalate complexing eluent.3,9 Rapid separation of actinides from the sample matrix is useful in a number of analytical applications including the determination of the gross actinide content (actinide screen).9,11 The detector trace in Fig. 2A illustrates a SI group actinide separation procedure applied to a vitrified nuclear waste sample spiked with 241Am, 230Th, and 233U.These spikes, representative of actinides in trivalent(Am), tetravalent(Th), and hexavalent(U) oxidation states were added to facilitate the observation of elution profiles via on-line detection. Following the sample load step, the column was washed with 10 ml (24 FCV) of 2 mol l21 nitric acid in order to remove stable and radioactive sample matrix. The actinides were efficiently eluted using 7 ml of 0.1 mol l21 ammonium bioxalate with nominal recoveries of 85–100%.Recoveries were established in separate experiments with fraction collection and off-line detection). An immediate blank run following the spiked nuclear waste sample indicated carryover of 0.3% of the original alpha activity. The carryover in the second and third blank runs using the same column was 0.04% . The carryover was predominantly due to 233U. Provided that this carryover is acceptable for a given application, columns can be reused multiple times as noted by the previous authors.20,21 We found that microprecipitation with NdF3 could be performed directly from 0.1 mol l21 ammonium bioxalate eluent solutions to prepare alpha spectrometry sources for actinide quantification.Actinide precipitation recoveries exceeded 85%. Thus, time consuming digestion of ammonium bioxalate typically done prior to alpha source preparation is unnecessary. Elution of actinides in valence state groups The reductive sample pretreatment described in the Experimental section ensures that any iron present is reduced to Fe(ii), thus eliminating the suppressing effect of Fe(iii) ions on the retention of the trivalent actinides.These conditions reduce all Pu to Pu(iii), and also reduce Np(v) to Np(iv), which is important because Np(v) is unretained on TRU-resin.29–31 Samples loaded on the column after this pretreatment include Pu, Am, and Cm in the trivalent state, Th and Np in the tetravalent state, and U in the hexavalent state. The chemistry for the separation of actinides in groups according to their valence state was described by Horwitz.3 Capacity factors for trivalent actinides are less than one in hydrochloric acid solutions of up to approximately 5 mol l21 concentration, while tetravalent actinides and hexavalent U remain strongly retained with kA > 103 in 4 mol l21 HCl.Therefore, HCl solutions can be used to selectively elute the trivalent actinides. Complexants can then be used to elute the tetra- and hexavalent species.Horwitz used HCl–oxalic acid to elute tetravalent actinides, followed by bioxalate to elute the remaining hexavalent U.3 Subsequent papers by researchers at Argonne National Laboratory described analytical separations of actinides in valence state groups using tetrahydrofuran 2,3,4,5-tetracarboxilic acid (THFTCA) instead of oxalic acid for selective elution of tetravalent actinides.1,14,15,21 We examined the elution of actinides from TRU-resin in valence state groups in an automated SI format.A separation procedure applied to a spiked vitrified glass sample (sample volume 100 ml) is shown in Fig. 2B using on-line detection to monitor the separation process. Sample load and column wash steps were performed just as described above. Trivalent actinides represented by Am were rapidly eluted using 5 ml of 4 mol l21 HCl. Tetravalent actinides represented by Th were eluted using 7 ml of 1 mol l21 HCl–0.05 mol l21 oxalic acid eluent, taking advantage of the preferential complexation of the tetravalent state relative to hexavalent state by the oxalate ion.3,32 Finally, uranium(vi) was eluted with 7 ml of 0.1 mol l21 ammonium bioxalate solution.For triplicate runs performed on the same column, Am and U recoveries exceeded 90%, while Th recoveries exceeded 85%. As determined in separate experiments using 230Th tracer, approximately 1–2% of the Th activity was present in the 0.1 mol l21 ammonium bioxalate fraction.When reusing the column, no detectable carryover was evident in the Am and Th fractions ( < 1% carryover); U carryover was ~ 1% (on-line detection data). After 6 runs on a single column there was no visible degradation in peak shape or change in the onset of the elution time for either Am or U. However, the Th elution peak exhibited slight broadening and shift towards longer retention times by the seventh run on the same column. Therefore, in this separation, extended column reuse (for more than 6 runs) may not be feasible.THFTCA is known to selectively complex tetravalent actinides in preference to hexavalent actinides,33 and solutions of this reagent have been used in the TRU-resin actinide valence state separations in manual and FI formats.1,14,15,21 Using our system with on-line detection, we observed that 0.1 mol l21 Fig. 2 Detector traces showing separation of the sample matrix and actinide elution in a single group (plot A), and group actinide elution sequence according to the valence state (plot B).The separations shown are applied to 100 ml vitrified glass sample spiked with 1.4 MBq 241Am, 2.0 MBq 230Th, and 2.16 MBq 233U. Actinides shown in brackets indicate other species that elute in the same peak(s) as the spiked elements. Eluents: a, 10 ml 2 mol l21 nitric acid; b, 5 ml 4 mol l21 HCl; c, 7 ml 1 mol l21 HCl–0.05 mol l21 oxalic acid; d, 7 ml 0.1 mol l21 ammonium bioxalate. TRU-resin column 4 3 50 mm; separation flow rate 2 ml min21.Time zero corresponds to the beginning of 2 mol l21 nitric acid wash. 1146 Analyst, 1999, 124, 1143–1150THFTCA solutions gave tailed elution peaks for tetravalent actinides (both Th and Pu tracers were tested) with substantial carryover into the subsequent 0.1 mol l21 bioxalate fraction. Aldstadt also reported difficulties with this eluent in a FI format with ICPMS detection and noted the need for further work and better separations.21 Despite the improved selectivity of THFTCA over oxalate for complexing tetravalent actinides, it does not appear to offer any advantage over HCl–oxalic acid for the elution of tetravalent actinide species from TRU-resin.The valence state of Pu can be either Pu(iii), Pu(iv), or both, depending on the sample load and column wash conditions. If loaded as Pu(iii) after reductive sample pretreatment, a reducing agent such as ferrous sulfamate can be included in the column wash solution to maintain Pu in the trivalent state.7 Alternatively, deliberate inclusion of an oxidizing step in conjunction with the column wash (as described in the Experimental section) can be used to assure that Pu elutes in the tetravalent group along with Th and Np.One of the potential uses for the sequential elution of actinides according to their valence states is as a sample pretreatment step prior to actinide determination by ICPMS.1,14,15,21 Using this approach, the actinides are separated from the sample matrix and can be preconcentrated if so desired.The sequential elution of actinides in groups, with Pu in the tetravalent group, addresses a number of potential interferences associated with ICPMS detection, i.e. 241Am/241Pu and 238Pu/ 238U isobaric interferences, 239Pu/238UH and 233U/232ThH polyatomic interferences, and 237Np/ 238U spectral interference. We have successfully developed a SI separation of actinides with on-line ICPMS detection, using a variant of the actinide separation shown in Fig. 2B with HCl–oxalic acid solution for tetravalent actinide elution.19 Experimental details and application towards analysis of the nuclear waste samples will be described in a subsequent report. Selective Pu elution via on-column redox reactions Plutonium can be selectively separated from the other actinide species by manipulation of its oxidation state. In this approach, Pu(iii) is oxidized to Pu(iv) with a reagent included in the column wash, followed by elution of the trivalent actinides with 4 mol l21 HCl.The Pu(iv) retained on the column is then reduced and eluted as Pu(iii). 3,11,24,25 The remaining tetra- and hexavalent actinides can be stripped from the column using bioxalate solution,25 or they can be sequentially eluted using oxalic acid and bioxalate solutions, as described in the valence state separation procedure. The latter elution sequence for actinides in shown in Fig. 3, using a solution of NaNO2 reagent in nitric acid as the oxidant, and Fe(ii) sulfamate–ascorbic acid as the reductant. The sample in this example is a K-basin sample spiked with 241Am, 239Pu, 230Th and 233U. The overall effectiveness of this separation in an automated format is critically dependent on the reliability of the on-column redox reactions. These were investigated previously in a FI system.25 An excess of NaNO2 in acidic solution, prepared freshly on-line in an FI format, was effective for the on-column Pu(iii)–Pu(iv) oxidation step.Fresh preparation of acidic nitrite solutions is necessary because this mixture produces nitrous acid (pKa = 3.35), an unstable species. However, we also found that nitrous acid is retained by the CMPO–TBP stationary phase on the column, and that there is an irreversible interaction between the polymeric support and nitrous acid or one of its decomposition products. Both processes influence the effectiveness of Pu elution in the subsequent on-column reduction step, requiring strong reductants for rapid, quantitative Pu elution. Hydrochloric acid solutions of titanium(iii) chloride and Fe(ii) sulfamate–ascorbic acid provided good Pu elution results with the latter reagent performing less reliably when using larger size columns (1.66 ml bed volume).Hydroquinone was not an effective reductant for converting Pu(iv) to Pu(iii) on columns previously exposed to nitrous acid solutions.25 Issues associated with the redox chemistries were reexamined in the SI format using Pu redox reactions in conjunction with sequential elution of actinide groups by valence state.Because of concerns about nitrite effects arising from our previous studies, we evaluated the IO32 ion as a potential alternative Pu oxidizing reagent. This reagent is known to rapidly oxidize Pu(iii) to Pu(iv) in aqueous solutions. 29,31 Exploratory experiments demonstrated that IO32 ion did oxidize Pu(iii) to Pu(iv) on the TRU-resin column but not with the speed and selectivity required. Using 1 ml of 2 mol l21 HNO3–0.01 mol l21 KIO3 solution as a part of the 2 mol l21 nitric acid column wash resulted in incomplete oxidation to Pu(iv), with detectable amounts of Pu(iii) present in the subsequent 4 mol l21 HCl eluent. Increasing the volume of this solution to 5 ml, and thus increasing the contact time with the oxidizing agent, succeeded in complete oxidation of Pu(iii), with none detectable in the 4 mol l21 HCl elution step.However, a small Pu peak appeared in the bioxalate elution, indicating partial oxidation to Pu(vi). We therefore used NaNO2 for on-column Pu oxidation in all further experiments. It was determined that Pu(iii) could be effectively oxidized on the column to Pu(iv) using 150 ml of 0.5 mol l21 NaNO2 in the column wash as described in the Experimental section. This represents a 2.5 fold smaller amount of injected nitrite relative to our previous work.24,25 In additional experiments, with and without the nitrite oxidation step in place, we observed no effect of the injected nitrite on Np(iv) speciation and elution.Using this revised on-column Pu oxidation procedure, rapid quantitative Pu recovery was possible using 4 mol l21 HCl solutions of either 0.02 mol l21 Ti(iii) chloride or 0.05 mol l21 Fe(ii)–0.1 mol l21 sulfamic acid–0.1 mol l21 ascorbic acid. The effect of the reducing eluent on the speciation of U retained on the column was also examined. When using 4 mol l21 HCl–0.02 mol l21 Ti(iii) chloride as the reductant, approximately 20% of the U activity eluted with 1 mol l21 HCl– 0.05 mol l21 oxalic acid, indicating partial on-column reduction of U(vi) to U(iv).Using 4 mol l21HCl–0.05 mol l21 Fe(ii)–0.1 mol l21 sulfamic–0.1 mol l21 ascorbic acid as the reductant did not result in any detectable U(iv) activity. Based on these results, either reducing agent can be used in the recovery and analysis of Pu, but the 0.05 mol l21 Fe(ii) 0.1 mol l21 sulfamic– Fig. 3 Detector trace showing a variant of the individual actinide separation sequence. The separation applied to a 100 ml nuclear waste sample (K-basin) spiked with 2.1 MBq 241Am, 2.6 MBq 239Pu, 2.5 MBq 230Th and 2.3 MBq 233U. Am is eluted using 5 ml 4 mol l21 HCl (eluent a); Pu is selectively eluted using 7 ml 4 mol l21 HCl solution of 0.05 mol l21 Fe(ii)–0.1 mol l21 sulfamic acid–0.1 mol l21 ascorbic acid (eluent b); Th (Np) are coeluted using 7 ml 1 mol l21 HCl–0.05 mol l21 oxalic acid (eluent c); U is eluted using 7 ml 0.1 mol l21 ammonium bioxalate (eluent d).Analyst, 1999, 124, 1143–1150 11470.1 mol l21 ascorbic acid reagent is preferred if subsequent actinide groups are to be recovered for analysis. For quantification of the separated Pu by alpha spectroscopy, we established that NdF3 microprecipitation can be carried out directly from either of the reducing Pu eluent solutions.The Pu precipitation recoveries were quantitative. Selective Th elution and sequential actinide separations Removal of trivalent actinides and Pu as described in the previous sections leaves Np(iv), Th(iv) and U(vi) species remaining on the TRU-resin column. Elution of these actinides with complexants leads to an actinide elution sequence of Am/ Cm, Pu, Np/Th, and U, as illustrated in Fig. 3. We were interested in the possibility that Th(iv) could be rapidly and selectively eluted before Np(iv), yielding a method for eluting the actinides in the sequence Am/Cm, Pu, Th, Np and U.Horwitz et al. described selective Th elution with a reduced concentration of HCl after eluting trivalent species in 4 mol l21 HCl solutions. Subsequent Np elution was accomplished using 1 mol l21 HCl–0.03 mol l21 oxalic acid and uranium was eluted with 0.1 mol l21 ammonium bioxalate .3 It was noted, however, that Th must be eluted slowly to achieve good reproducibility due to slow kinetics, and that the procedure ‘is probably too tedious for routine analytical use’.3 Using on-line detection, we observed that Th elution using 1.5 mol l21 HCl at a flow rate of 1 ml min21 yielded broad tailed peaks with significant quantities of Th carried over into a subsequent ammonium bioxalate fraction.This approach was not satisfactory for rapid automated separations. Dramatically improved Th elution profiles were observed when HF was incorporated into HCl eluent solutions.We selected this approach on the basis of preferential complexation of Th(iv) by fluoride relative to Np(iv) and U(vi).32 By adjusting the concentrations of HCl and HF, a solution composition was found that rapidly and selectively eluted Th from the TRU-resin column in the presence of Np and U, which remained on the column. Detector traces in Fig. 4 illustrate elution profiles for separate standards of Th and Np using 10 ml 4 mol l21 HCl–0.05 mol l21 HF solution for Th elution and 0.1 mol l21 ammonium bioxalate solution to strip remaining actinides from the column.This selective Th elution method leads to a separation sequence of Am/Cm, Pu, Th, and Np/U. Given selective Th elution with HCL–HF eluent, we then attempted to sequentially elute Np and U using HCl–oxalic acid and bioxalate eluents, respectively, exactly as done in separations based on valence state (see Fig 2B for example). However, Np recovery using 1 mol l21 HCl–0.05 mol l21 oxalic acid after Th elution with 10 ml 4 mol l21 HCl–0.05 mol l21 HF was low and unreproducible, with over 70% of Np activity recovered in a subsequent bioxalate elution.Inclusion of a 4 mol l21 HCl wash step between the Th eluent and the Np eluent offered slight but still unsatisfactory improvement in Np elution. The detailed chemistry underlying these results is unclear, but the results suggest that HF solution either alters the speciation of Np(iv) retained on TRU-resin column, or that the separation material itself is affected when using the HF–HCl eluent.Consequently, the use of an HCl–HF eluent for selective Th elution did not lead to an effective scheme for sequential separation sequence of Am/Cm, Pu, Th, Np, and U. Rapid sequential separation of individual actinides on a single TRU resin column, with fast kinetics, sharp peaks, and minimal carryover of one radionuclide into another, remains a challenge. Nevertheless, the elution sequences described above are potentially useful in the radiochemical analysis of individual actinides.The sequence of Am/Cm, Pu, Th/Np, and U provides a fraction with Th and Np separated from the sample matrix and other actinides. Individual separation of Th and Np may then be accomplished using TEVA-resin extraction chromatographic material, which strongly retains Np(iv) but not Th(iv) from the HCl solutions.5 Alternatively, the new separation sequence of Am/Cm, Pu, Th, and Np/U provides selective recoveries of Pu and Th.Np can then be selectively precipitated from the Np/U fraction using a NdF3 microprecipitation procedure for alpha source preparation. Using this approach, Np(iv) is separated from U(vi) which does not coprecipitate with NdF3. Determination of Am, Cm and Pu in nuclear waste samples Radiometric determination of Am (Cm) and Pu is routinely performed in radiochemical laboratories at the Hanford site.Because of the safety considerations associated with relatively high abundance of these isotopes in aged nuclear wastes, this assay represents one of the most frequently performed radiochemical procedures. Therefore, we were interested in the application of the automated SI separation system towards the analysis of these radionuclides in nuclear waste samples. Three different nuclear waste types (see Experimental) were examined. These included sludge from the aged irradiated nuclear fuel (K-basin samples), tank waste supernate, and vitrified nuclear waste.Nuclear waste and quality control samples were processed as described in the Experimental section, using a procedure consisting of TRU-resin column conditioning, sample load, column wash/Pu oxidation, trivalent actinides elution in 4 mol l21 HCl, and selective Pu reduction–elution using a reducing eluent. The separations were performed using freshly packed columns, which were replaced after the Pu elution step.The separation time including the column conditioning step was 25 min. The Am/Cm and Pu fractions were collected with a fraction collector and analyzed by alpha spectroscopy. Fig. 5 shows the instrumental alpha spectra of the separated Am/Cm (plot A) and Pu (plot B) fractions for the analysis of vitrified waste sample, illustrating excellent separation obtained using automated SI procedure. Counting source preparation using microprecipitation with NdF3 was carried out directly from the eluent matrixes.Using the SI procedure, nominal radiochemical yields were 85% for Pu and 86% for Am. High decontamination factors were typically observed with no detectable crosscontamination of Am/Cm and Pu fractions (individual decon- Fig. 4 Detector traces illustrating the feasibility of Th–Np separation using HCl–HF eluent for Th elution. The elution sequence shown was applied to Th (dashed trace, left y-axis) and Np (solid line, right y-axis) standards.Th trace is offset 30 counts for clarity. Eluents: a, 10 ml 4 mol l21 HCl–0.05 mol l21 HF; b, 7 ml 0.1 mol l21 ammonium bioxalate. TRU-resin column 4 3 50 mm; separation flow rate 2 ml min21. Time zero corresponds to the beginning of 4 mol l21 HCl–0.05 mol l21 HF elution. 1148 Analyst, 1999, 124, 1143–1150tamination factors > 100). The analysis results obtained by the standard manual separation method and our automated SI separation procedure are compared in Table 2 and are in good agreement.Trivalent lanthanides, which may also be present in the fraction containing Am and Cm, do not interfere. They are not alpha emitters and are not present in sufficient quantities in the alpha source to cause energy resolution deterioration when analyzing aged nuclear waste. Conclusions This work demonstrates that a separation-optimized fluid handling technique represents a versatile platform for the automation and detailed elucidation of chemically challenging extraction–chromatographic actinide separations.Chemistries for a variety of actinide separation procedures have now been developed that work reliably, reproducibly, and rapidly for unattended automated procedures. The SI separation system can also serve as the basis for an automated radiochemical separation workstation. With an appropriate set of solutions around the multiposition valve, several actinide separation procedures can be selected through software.All the separations described above were carried out with fixed TRU-resin columns that were periodically replaced. As demonstrated previously for a SI 90Sr separation system, column switching techniques can be used to provide new columns for each sample. Alternatively, we have recently described the ability to pack extraction chromatographic columns on-line, creating a new column for each analysis and disposing of each such column prior to the next analysis.19,27 In this approach, carryover associated with the separation material is eliminated, and it is not necessary to elute the most strongly retained species on the column if they are not going to be analyzed.Either column switching or the renewable column technique could be incorporated as part of an automated actinide separation workstation. Acknowledgements The authors wish to thank professor Jaromir Ruzicka for many helpful discussions on the applications of flow injection techniques.The technical assistance of Matthew J. O’Hara is gratefully acknowledged. This work has been supported with funding from the Office of Biological and Environmental Research of the US Department of Energy. The Pacific Fig. 5 Alpha energy spectra for the analysis of vitrified nuclear waste sample using automated Am/Cm–Pu separation procedure. Plot A corresponds to the trivalent actinide fraction; plot B corresponds to the plutonium fraction. Table 2 Determination of the Am, Cm, and Pu isotopes in nuclear waste samples SI procedure, manual procedure/kBq g21 or kBq ml21a,b,c Sample ID 241Am 243,244Cm 242Cm 239,240Pu 238Pu Vitrified 1270 ± 57 250 ± 13 11.2 ± 1.7 68.5 ± 2.6 159 ± 5.6 waste 1 1190 ± 27 258 ± 7.5 8.51 ± 0.72 68.8 ± 2.6 147 ± 4.4 Vitrified 1270 ± 55 245 ± 13 12.9 ± 1.8 70.7 ± 3.3 165 ± 6.3 waste 2 1190 ± 27 258 ± 7.5 8.51 ± 0.72 68.8 ± 2.6 147 ± 4.4 Tank 25.8 ± 1.3 1.08 ± 0.09 0.107 ± 0.02 2.00 ± 0.08 0.560 ± 0.03 waste 1 22.2 ± 1.1 0.921 ± 0.07 nm 1.99 ± 0.5 0.570 ± 0.03 Tank 23.8 ± 1.2 0.781 ± 0.07 0.112 ± 0.02 2.07 ± 0.07 0.574 ± 0.03 waste 2 22.2 ± 1.1 0.921 ± 0.07 nm 1.99 ± 0.59 0.570 ± 0.03 Tank (5.59 ± 0.34) 3 1022 (3.57 ± 0.23) 3 1022 (3.46 ± 1.5) 3 1024 (3.70 ± 0.16) 3 1022 (4.96 ± 0.20) 3 1022 waste 3 (5.59 ± 0.34) 3 1022 (3.50 ± 0.13) 3 1022 nm (3.63 ± 0.10) 3 1022 (5.18 ± 0.14) 3 1022 K-basin 1 0.770 ± 0.040 nm nm 4.14 ± 0.17 0.770 ± 0.05 0.807 ± 0.052 4.23 ± 0.18 0.855 ± 0.07 K-basin 2 37.0 ± 2.1 nm nm 130 ± 4.9 16.7 ± 0.97 39.6 ± 1.6 139 ± 4 18.4 ± 0.97 K-basin 3 85.5 ± 5.6 nm nm 1200 ± 47 139 ± 8.8 80.3 ± 5.1 1240 ± 35 158 ± 7.3 K-basin 4 122 ± 6.5 nm nm 1390 ± 55 188 ± 11.9 114 ± 9.0 1550 ± 44 193 ± 9.4 a kBq g21 for vitrified glass and K-basin samples; kBq ml21 for tank waste samples.b Error is ± 1s. c nm not measured. Analyst, 1999, 124, 1143–1150 1149Northwest National Laboratory is a multiprogram national laboratory operated for the US Department of Energy by Battelle Memorial Institute.References 1 M. D. Erickson, J. H. Aldstadt, J. S. Alvarado, J. S. Crain, K. A. Orlandin and L. L. Smith, J. Hazard. Mater., 1995, 41, 351. 2 R. L. Murray, Understanding Radioactive Waste, 4th edn., Battelle Press, Richland, Washington, 1994, p. 212. 3 E. P. Horwitz, R. Chiarizia, M. L. Dietz, H. Diamond and D. M. Nelson, Anal. Chim. Acta, 1993, 281, 361. 4 E. P. Horwitz, R. Chiarizia and M. L. Dietz, Solvent Extr. Ion Exch., 1992, 10, 313. 5 E.P. Horwitz, M. L. Dietz, R. Chiarizia, H. Diamond, S. L. Maxwell and M. R. Nelson, Anal. Chim. Acta, 1995, 310, 63. 6 M. L. Dietz and E. P. Horwitz, LC-GC, 1993, 11, 424. 7 S. L. Maxwell, Radioact. Radiochem., 1997, 8, 36. 8 J. L. Cortina and A. Warshawsky, Ion Exch. Solvent Extr., 1997, 13, 195. 9 E. P. Horwitz, M. L. Dietz, H. Diamond, J. J. LaRosa and W. D. Fairman, Anal. Chim. Acta, 1990, 238, 263. 10 R. S. Strebin, Separation of Am and Pu and Actinide Screen by Extraction Chromatography-PNL-ALO-417 in Analytical Chemsitry. Laboratory Procedure Compendium PNL-MA-599, Pacific Northwest National Laboratory, 1993. 11 J. H. Kaye, R. S. Strebin and R. D. Orr, J. Radioanal. Nucl. Chem., 1995, 194, 191. 12 Americium, Plutonium and Uranium in Water. Analytical Procedure, EIChrom Industries, Inc., Darien, IL, ACW03, 1995. 13 S. L. Maxwell and M. R. Nelson, Institute of Nuclear Material Management 35th Annual Meeting, Naples, FL, 1994. 14 L. L. Smith, J. S. Crain, J. S. Yaeger, E. P. Horwitz, H. Diamond and R. Chiarizia, J. Radioanal. Nucl. Chem., 1995, 194, 151. 15 J. S. Crain, L. L. Smith, J. S. Yaeger and J. A. Alvarado, J. Radioanal. Nucl. Chem., 1995, 194, 133. 16 J. Ruzicka and E. H. Hansen, Flow Injection Analysis, 2nd edn.; Wiley, New York, 1988, vol. 62, p. 498. 17 Z. Fang, Flow Injection Separation and Preconcentration, VCH, Weinheim, 1993. 18 J. Ruzicka, Analyst, 1994, 119, 1925. 19 J. W. Grate and O. B. Egorov, Anal. Chem., 1998, 70, 779A. 20 M. Hollenbach, J. Grohs, S. Mamich, M. Kroft and E. R. Denoyer, J. Anal. At. Spectrom., 1994, 9, 927. 21 J. H. Aldstadt, J. M. Kuo, L. L. Smith and M. D. Erickson, Anal. Chim. Acta, 1996, 319, 135. 22 J. W. Grate, R. S. Strebin, J. Janata, O. Egorov and J. Ruzicka, Anal. Chem., 1996, 68, 333. 23 O. Egorov, M. J. O’Hara, J. Ruzicka and J. W. Grate, Anal. Chem., 1998, 70, 977. 24 O. Egorov, J. W. Grate and J. Ruzicka, J. Radioanal. Nucl. Chem., 1998, 234, 231. 25 J. W. Grate and O. Egorov, Anal. Chem., 1998, 70, 3920. 26 J. W. Grate, S. K. Fadeff and O. Egorov, Analyst, 1999, 124, 203. 27 O. Egorov, M. J. O’Hara, J. W. Grate and J. Ruzicka, Anal. Chem., 1998, 71, 345. 28 J. Ruzicka, Anal. Chim. Acta, 1992, 261, 3. 29 J. M. Cleveland, The Chemistry of Plutonium, Gordon and Breach, Science Publishers, New York, 1970. 30 Plutonium Handbook. A Guide to Technology., ed. O. Wick, Gordon and Breach Science Publishers: New York, 1967; vol. 1. 31 T. W. Newton, The Kinetics of the Oxidation-Reduction Reactions of Uranium, Neptunium, Plutonium, and Americium in Aqueous Solutions, ERDA Technical Information Center, Oak Ridge, TN, 1975. 32 The Chemsitry of the Actinide Elements, A. A. Katz, G. T. Seaborg and L. R. Morss, Chapman and Hall, New York, 1986, vol. 1,2. 33 K. L. R. P. G. Nash, E. P. Lessman, M. D. Mendoza, J. F. Feil, J. C. Sullivan, New Water-Soluble Phosphanate and Polycarboxylate Complexants For Enhanced Elements Separations, Plenum Press, New York, 1995. Paper 9/02579K 1150 Analyst, 1999, 124, 1143–1150

ISSN:0003-2654    

DOI:10.1039/a902579k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Development of an automated, simultaneous and continuous measurement system by using a diffusion scrubber coupled to ion chromatography for monitoring trace acidic and basic gases (HCl, HNO3, SO2 and NH3) in the atmosphere Yuichi Komazaki,a Yuichi Hamada,a Shigeru Hashimoto,b Tomomi Fujitac and Shigeru Tanaka*a a Department of Applied Chemistry, Faculty of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama, 223-8522, Japan. E-mail: tanaka@applc.keio.ac.jp b Japan Science and Technology Corporation, 3-12-40 Hiroo, Shibuya-ku, Tokyo, 150-00 12, Japan c HPLC Business Department, Shimadzu Corporation, 380-1 Horiyamashita, Hadano, Kanagawa, 259-1304, Japan Received 5th May 1999, Accepted 23rd June 1999 An automated simultaneous measurement system for monitoring acidic and basic gases such as HCl, HNO3, SO2 and NH3 in the atmosphere was developed by using a diffusion scrubber coupled to an ion chromatograph (DS-IC system).Acidic and basic gases were effectively collected by the diffusion scrubber which consisted of a hydrophobic porous polytetrafluoroethylene (PTFE) tube disposed concentrically within a Pyrex-glass tube.De-ionized water was used as a scrubbing solution for collecting the acidic and basic gases such as HCl, HNO3, SO2 and NH3, which were dissociated to ion species (Cl2, NO32, SO3 22, SO4 22 and NH4 +, respectively). After the collection of gases, the total anions and cations in the scrubbing solution were preconcentrated in anion and cation concentrator columns connected in series.All the analyte ions were separated and determined by ion chromatography. All operations relating to the air sampling, the preconcentration and the ion chromatographic analysis were controlled by a sequencer. Automated simultaneous continuous measurement could be performed at 60 min intervals. The collection efficiencies of the acidic and basic gases were higher than 97% at an air flow rate of 1 l min21. The detection limits (3s of the blank value) of HCl, HNO3, SO2 and NH3 were less than 0.01 ppbv for a 52 l of air sampling volume.The interference from co-existing NO2 for measuring HNO3 was negligible. The concentrations of SO2 measured by this automated measurement (DS-IC) system showed good agreement with those obtained with a pulse-fluorescence meter. Introduction The long-range transport of air pollutants have been increasing, especially in the East Asian Pacific Rim region, by high and rapidly growing anthropogenic emission of nitrogen oxides (NOx), sulfur dioxide (SO2)1 and hydrocarbons.Nitric acid (HNO3) is formed by the reaction of NO2 with hydroxyl radicals (OH) and plays a significant role in the total nitrogen budget and often contributes to the acidification of the atmosphere. Hydrochloric acid (HCl) is emitted directly from municipal incinerators by burning vinyl chloride, etc. HCl also can be formed by a chlorine loss reaction between sodium chloride (NaCl) and acidic gases such as SO2 and HNO3 in the marine atmosphere.On the other hand, ammonia (NH3), which is the dominant basic trace gas in the atmosphere, is derived mainly from biogenic decomposition of organic materials and fertilizers in the land. NH3 reacts with acidic gases and plays an important role in neutralizing the acidic gases and forming aerosols. Most sampling procedures for HCl, HNO3, SO2 and NH3 were based on filter methods in previous studies.In the filter method, a serious disadvantage of positive and negative artifacts from the interaction of gases and aerosols was derived from the configuration that trace gases were collected on filters after removing aerosols.2–4 In order to eliminate the artifacts, diffusion denuders have been used for the collection of trace gases and were reviewed by Appel. 5 The diffusion denuder, where chemicals are coated inside the tube, can selectively collect trace gases of interest from a gas–aerosol mixture. 6 However, the denuder method has some disadvantages. Several kinds of denuders have to be used owing to the simultaneous collection of various gases. The coating procedure and the extraction of collected gases from the denuder were very laborious. Accordingly, it was difficult to automate the measurement of trace gases using the diffusion denuder. While thermal desorption diffusion denuders could eliminate the need for extraction and recoating work and automate the measurement of HNO3, SO2 and NH3,7–9 these systems were too expensive to measure various trace gases simultaneously.Recently, wet denuders and the diffusion scrubbers have been developed for the simultaneous collection of trace gases such as HCl, HNO3, SO2 and NH3 in the atmosphere.10–22 Dasgupta and co-workers11–15 and Frenzel and co-workers16,17 made great contributions to the development of automated continuous measurement systems for monitoring trace gases in the atmosphere by using a wet denuder or a diffusion scrubber coupled to ion chromatography with a preconcentration method using a concentrator column.The wet denuders and the diffusion scrubber have very simple geometries and can collect various trace gases simultaneously in a small volume of scrubbing solution flowing over the surface of the denuder tube or outside the porous polytetrafluoroethylene (PTFE) tube. In addition, automating the collection and on-line analysis give good reproducibility because of lower risks of contamination Analyst, 1999, 124, 1151–1157 1151and less sample manipulation. A fully automated measurement system using a wet denuder or a diffusion scrubber has been developed.18–22 However, trace acidic and basic gases were independently collected with different scrubbing solutions, and anions or cations were preconcentrated and analyzed.In this study, an automated simultaneous measurement system was investigated for the collection of trace acidic and basic gases (HCl, HNO3, SO2 and NH3) in the atmosphere by a diffusion scrubber and the analysis of anions and cations by ion chromatography after the simultaneous preconcentration of the total anions and cations in concentrator columns connected in series.Experimental Diffusion scrubber A schematic diagram of a diffusion scrubber is shown on the left in Fig. 1, and the collection mechanism for trace HCl, HNO3, SO2 and NH3 is illustrated on the right.The specification of the diffusion scrubber is given in Table 1. The diffusion scrubber consists of a hydrophobic porous PTFE tube concentrically disposed within a Pyrex-glass tube. The porous PTFE tube is fixed to polymonochlorotrifluoroethylene (PCTFE) connectors with threaded end fittings at both ends and sealed with tubular silicone rubber spacers and Orings. In order to allow laminar flow of drawn gas into the porous PTFE tube, tapered PTFE tube (3 mm id, 4 mm od) are inserted into each end of the porous PTFE tube.The length of the PTFE tubing (Li) is determined by the equation Li = (0.07R)Re,23 where R is the tube radius and Re the Reynolds number. The diffusion scrubber is always operated in a vertical configuration, to minimize the deposition of particles, such as NH4Cl, NH4NO3 and (NH4)2SO4, on the surface of the porous PTFE tube. Through the inlet, with a threaded male nut and a PTFE tube (2 mm id, 3 mm od), a scrubbing solution of deionized water is fed by gravity into the annular space between the inner surface of the glass tube and the outer surface of the porous PTFE tube in a countercurrent flow to the gas, as illustrated in Fig. 1. The scrubbing solution is static in the diffusion scrubber during the sample collection period. The typical air flow rate is 1.0 l min21. Under laminar flow conditions, trace gases diffuse across the porous PTFE tube and reach the scrubbing solution, in which HCl, HNO3, SO2 and NH3 are collected as Cl2, NO32, SO3 22, SO4 22 and NH4 +, respectively.Standard gas generation Standard gases of HCl, HNO3, SO2 and NH3 were individually generated by passing zero air through a standard gas generator (PD-1B, GASTEC, Japan), in which a permeation tube or a diffusion tube (D-20, GASTEC) was installed and kept at 35 °C. The emission source of HCl or HNO3 was a diffusion tube (D- 20, GASTEC) containing an aqueous solution of HCl or HNO3.For emission of SO2 or NH3, an SO2 permeation tube (P-5-5, GASTEC) or an NH3 permeation tube (P-3, GASTEC) was used. Zero air was supplied by flowing cylindrical dry air (Nippon Sanso, Japan) through a zero air generator (Model 111, Thermo Environmental Instruments, USA). Each concentration of trace gases was controlled by diluting the generated gases with zero air in a multi-gas calibration system (Model 146, Thermo Environmental Instruments). In a preliminary experiment, the concentrations of HCl, HNO3, SO2 and NH3 in the prepared gas downstream of the blending system were individually checked.The prepared gas was collected by an impinger (GL Science, Japan) charged with 30 ml of 10 mmol l21 Na2CO3 solution for HCl, HNO3 and SO2 or deionized water for NH3 as a scrubbing solution at a flow rate of 1.0 l min21 for 60 min. An aliquot of the sample solution was analyzed by ion chromatography under the conditions given in Table 2. Further, the concentrations of HCl, HNO3, SO2 and NH3 in the standard gases measured by the diffusion scrubber method agreed well with those obtained by the impinger method.Therefore, the concentrations of standard gases prepared in this study were highly accurate. Fig. 1 Schematic diagram of a diffusion scrubber with a porous PTFE tube for collecting acidic and basic gases in the atmosphere. Table 1 Specification of the diffusion scrubber Inner tube Porous PTFE, Poreflon TB-54 (Sumitomo Electric, Japan) (4 mm id, 5 mm od, 0.45 mm mean pore size, 79% porosity) Outer tube Pyrex glass (6.6 mm id, 9 mm od) Effective length 50 cm Scrubbing solution De-ionized water Volume of scrubbing solution 4 ml Air flow rate 1.0 l min21 Table 2 Analytical conditions for ion chromatographic analysis Anion species Cation species Separator column Shim-pack IC-A3(s)a (150 3 2 mm id) Shim-pack IC-C3(s)a (150 3 2 mm id) Concentrator column Shim-pack IC-GA3a (10 3 4 mm id) Shim-pack IC-GC3a (10 3 4 mm id) Eluent IC-MA3-1a (0.1 ml min21) 2.5 mM oxalic acid (0.1 ml min21) Detector Conductivity detectora Conductivity detectora Oven temperature 40 °C 40 °C Sample solution volume 12 ml 12 ml a Shimadzu. 1152 Analyst, 1999, 124, 1151–1157Reagents All chemicals were of analytical-reagent grade (Wako, Osaka, Japan), unless indicated otherwise. The eluent for anion analysis was IC-MA3-1 (Shimadzu, Kyoto, Japan). NH4Cl, NH4NO3 and (NH4)2SO4 were used to prepare a mixed standard solution of anions and cations (50 ng ml21 Cl2, 100 ng ml21 NO32, 200 ng ml21 SO4 22 and 122.2 ng ml21 NH4 +).NaHSO3 was independently used for a standard solution of SO3 22. Deionized water was obtained from an ultrapure water generator (CPW-200, Advantec Tokyo, Japan, or Milli-Q Jr., Millipore, Bedford, MA, USA). Ion chromatographic analysis An ion chromatograph system (Model LC-10A, Shimadzu) was used for the determination of Cl2, NO32, SO3 22, SO4 22 and NH4 + in the sample solution.The ion chromatograph system consisted of two liquid delivery pumps (LC-10AD, Shimadzu), two conductivity detectors (CDD-6A, Shimadzu), a column oven (CTO-10A, Shimadzu), two six-port autoinjection valves (FCV-12AH, Shimadzu) and a degasser (DGU-12A, Shimadzu). The system was controlled by a system controller (SCL-10AVP, Shimadzu) and an IBM compatible computer (Dynabook Satellite 300, Toshiba, Japan) with a software package of Class-VP (Shimadzu), which was also used for acquiring and processing the ion chromatograms.The analytical conditions for ion chromatography are summarized in Table 2. In order to conduct automated and continuous analysis during a long-term period, it was necessary to minimize the consumption of the eluents for anion and cation analyses. Therefore, semimicro-type separator columns were selected for both the separation and determination of anions and cations. Consequently, the consumption of the eluent could be less than 5 l during the continuous analysis of anions and cations for 1 month.Simultaneous preconcentration analysis of anions and cations The concentrations of trace gases in the background atmosphere are very low, in the sub-ppbv range. If 1 ppbv of HCl and 1 ppbv of HNO3 are collected by the diffusion scrubber, the concentrations of Cl2 and NO32 in the scrubbing solution are 6.3 and 9.6 ng ml21, respectively, for a 52 l air sampling volume and 12 ml of scrubbing solution.However, a few ng ml21 of Cl2 and NO32 cannot be determined by conventional ion chromatography with a 50 ml sample loop. To achieve adequate sensitivity, anions and cations in the sample solution have to be preconcentrated on anion- and cation-exchange columns, respectively. In this study, anions and cations in the sample solution collected by the diffusion scrubber were preconcentrated using an anion concentrator column and a cation concentrator column (Table 2), respectively, which were individually set on six-port autoinjection valves (FCV-12AH, Shimadzu).By delivering the sample solution from the diffusion scrubber to the concentrator columns in series, anions in the sample solution were first trapped and preconcentrated in the anion concentrator column. Cations that passed through the anion concentrator column were subsequently trapped and preconcentrated in the cation concentrator column. Thus both anions and cations could be loaded simultaneously and separately in these concentrator columns in series. The trapped anions and cations were individually eluted from their concentrator columns to anion and cation separator columns by flowing eluents to their concentrator columns by actuating the autoinjection valves.Very low concentrations (0.01 ng ml21, S/N = 3) of Cl2, NO32, SO4 22 and NH4 + in 12 ml of sample solution could be detected by ion chromatography with the concentrator columns in series.Fig. 2 shows typical ion chromatograms, in which 12 ml of the mixed standard solution containing 50 ng ml21 of Cl2, 100 ng ml21 of NO32, 200 ng ml21 of SO4 22 and 122.2 ng ml21 of NH4 + was preconcentrated and analyzed. The peaks of Cl2, NO32 and SO4 22 were found in the anion chromatogram at about 12, 24 and 29 min, respectively, and the peak of NH4 + was found in the cation chromatogram at about 17 min. As can be seen from these chromatograms, anions and cations were determined within 40 min under the analytical conditions given in Table 2.Therefore, it was found that the simultaneous and continuous determination of Cl2, NO32, SO4 22 and NH4 + could be easily carried out at 60 min intervals by ion chromatography using semimicro separator columns at an eluent flow rate of 0.1 ml min21. In the diffusion scrubber method with de-ionized water as the scrubbing solution, SO2 in the atmosphere was collected as SO3 22 and SO4 22. Therefore, the concentration of SO2 in the atmosphere has to be determined by the total amount of SO3 22 and SO4 22 in the sample solution. A standard solution containing SO3 22 was prepared by dissolving NaHSO3 in deionized water.Actually, the peak of SO3 22 was hardly found in the ion chromatograms of sample solutions in field observations when the concentration of SO2 was very low (a few ppbv) because SO3 22 was almost completely oxidized to SO4 22 in the scrubbing solution. Automated measurement system (DS-IC system) An automated measurement system using a diffusion scrubber coupled to an ion chromatograph (DS-IC system) for measuring trace acidic and basic gases in the atmosphere is illustrated in Fig. 2 Typical anion and cation chromatograms of the standard solution. 12 ml of the standard solution (50 ng ml21 Cl2, 100 ng ml21 NO32, 200 ng ml21 SO4 22 and 122.2 ng ml21 NH4 +) were preconcentrated and analyzed. Analyst, 1999, 124, 1151–1157 1153Fig. 3. De-ionized water as a scrubbing solution was stored in a reservoir, where an ultrapure water generator (Milli-Q Jr.) was attached in order to supply fresh de-ionized water as the scrubbing solution.A cartridge (AF) packed with pelletized Purafil, Puracarbo and Puracoal (Purafil, USA) was placed at the air inlet of the reservoir in order to remove contaminants of trace gases from indoor air into the scrubbing solution. The scrubbing solution was fed by gravity into the diffusion scrubber through PTFE tubing.An air sample was drawn into the diffusion scrubber by an air suction pump (AP; APN-110, Iwaki, Japan) and a mass flow controller (MFC; Model 3650, Kojima Instruments, Japan) at a flow rate of 1.0 l min21 for 52 min. A filter (F; AF2000-01-2, SMC, Japan) and a membrane air dryer (AD; SWF-M06-400, Asahi Glass Engineering, Japan) were placed between the diffusion scrubber and the mass flow controller in order to eliminate problems due to aerosols and moisture.After collecting the air sample, a preconcentration treatment of ions in the sample solution was performed within 8 min as follows. The first autoinjection valve (FCV-12AH, Shimadzu) with an anion concentrator column and a three-port valve V2 (MTV-31, Takasago Electric, Japan) were switched and a plunger pump (NPL-110, Nippon Seimitsu, Japan) was operated at a flow rate of 2.0 ml min21 for 2 min. Fresh scrubbing solution was flowed into the anion concentrator column from the reservoir to purge the eluent in the anion concentrator column.Subsequently, the second autoinjection valve with the cation concentrator column was actuated to purge the eluent in the cation concentrator column for 1 min. Then the scrubbing solution in the diffusion scrubber was introduced into the anion and cation concentrator columns connected in series at a flow rate of 2.0 ml min21 for 6 min by returning the valve V2 to the normal position and opening valve V1 (MTV-21, Takasago Electric).Consequently, the total anions and cations in 12 ml of sample solution were separately loaded on each concentrator column. By actuating both autoinjection valves to the injection position, the analyte ions which were trapped in the concentrator columns were eluted to the anion and cation separator columns in the backflush mode. During the pre-treatment period, the air stream was made to bypass the diffusion scrubber through a three-port valve V4 (MTV-31, Takasago Electric). All operations were fully automated by a sequencer (mFA20, Yokogawa Electric, Japan) which controlled all valves and pumps in a cyclic manner.The measurement of HCl, HNO3, SO2 and NH3 in the atmosphere was performed automatically at 60 min intervals. Hence the DS-IC system developed in this study can eliminate the need for any additional sample preparation and elaborate operational manipulations. Calibration procedure Considering that the collection efficiencies of standard gases by the diffusion scrubber were almost 100% and very stable, the ion chromatographic calibration was sufficient for the overall calibration of the DS-IC system without gas-phase calibration.Therefore, during ambient air measurements in the outside environment such as a remote island, ion chromatographic calibration was carried out, because the generation of mixed standard gas containing HCl, HNO3, SO2 and NH3 is very difficult in field measurements. A 12 ml volume of the mixed standard solution containing 50 ng ml21 of Cl2, 100 ng ml21 of NO32, 200 ng ml21 of SO4 22 and 122.2 ng ml21 of NH4 + was preconcentrated and analyzed by ion chromatography for calibration of this system. Results and discussion Collection efficiencies The collection efficiencies (ƒ) of trace gases by the diffusion scrubber were determined using two diffusion scrubbers in series.18 Concentrations of 3–27 ppbv of trace gases of interest were individually used for this experiment.The sampling duration was 52 min. The collection efficiency can be calculated with the equation ƒ (%) = (1 2 C2/C1) 3 100 (1) where C1 and C2 (ppbv) represent the concentrations of individual trace gases collected by the first and the second diffusion scrubber, respectively. The collection efficiencies of HCl, HNO3, SO2 and NH3 by the diffusion scrubber at a flow rate of 1.0 l min21 are summarized in Table 3. All collection efficiencies for HCl, HNO3, SO2 and NH3 were sufficiently high ( > 97 %).Fig. 3 Automated continuous simultaneous measurement system for monitoring trace acidic and basic gases in the atmosphere by using a diffusion scrubber coupled to an ion chromatograph: AF, air purifier; F, filter; LT, liquid trap; AD, air drier; MFC, mass flow controller; AP, air pump; V1, two-port valve; V2, V3 and V4, three-port valves; STD, standard solution; PP, plunger pump; CC, concentrator column; SC, separator column; CD, conductivity detector. 1154 Analyst, 1999, 124, 1151–1157In general, the ideal collection efficiency (ƒIDEAL) of the diffusion scrubber can be estimated using the Gormley– Kennedy equation as follows:24 ƒIDEAL (%) = [1 2 0.819exp(23.657m) 2 0.0975exp(222.3m) 2 0.033exp(257m)] 3 100 (2) m = pDL/Q (3) where D (cm2 s21) is the diffusion coefficient of the analyte gas, L (cm) is the effective length of the diffusion scrubber, and Q (cm3 s21) is the volumetric gas flow rate.This equation is applicable to ideal conditions such that the surface of the porous PTFE is assumed to be a perfect sink for the trace gases of interest under laminar flow conditions. The diffusion coefficients of HCl, HNO3, SO2 and NH3 used here are 0.16,25 0.107,26 0.13627 and 0.23428 cm2 s21 at 25 °C, respectively. The ideal collection efficiencies of the trace gases are also given in Table 3. The empirical collection efficiencies of trace gases showed good agreement with the ideal collection efficiencies.Nitrogen dioxide interference In urban air, more than 100 ppbv of NO2 is often observed. Although the solubility of NO2 in water is very low, as can be seen from the Henry’s law constant of 1.2 310 mol l21 atm21,29 part of the NO2 can dissolve in water as NO22 and NO32: 2NO2 + H2O ? 2H+ + NO22 + NO32 (4) Therefore, the NO32 from NO2 interferes in the measurement of HNO3 in the atmosphere. In this study, the effect of co-existing NO2 on the collection of HNO3 by the diffusion scrubber using de-ionized water as the scrubbing solution was investigated.A 54 ppbv concentration of NO2 for interference tests was supplied by diluting 10.75 ppmv of NO2 in N2 standard gas (Nippon Sanso) with zero air. The apparatus for zero air generation and the dilution of test gas was almost identical with that used for the collection efficiency test. During all experiments, the NO2 concentration was measured continuously with a chemiluminescent nitrogen oxides analyzer (Model APNA-350E, Horiba, Japan), which had been calibrated with a certified standard of NO (Nippon Sanso).The sample gas was drawn into the diffusion scrubber at a flow rate of 1.0 l min21 for 52 min. As shown in Table 4, less than 0.03% of NO2 was collected by the diffusion scrubber using de-ionized water as the scrubbing solution. The concentration of NO32 in the sample solution was less than the detection limit, which corresponded to < 0.006 ppbv of HNO3 in the case of a 54 ppbv concentration.Therefore, it was found that the interference of NO2 in the HNO3 measurement ( < 0.01%) could be negligible in this diffusion scrubber method. For the above reasons, the diffusion scrubber method is advantageous to the measurement of HNO3 in an atmosphere with a co-existing high concentration of NO2. Stability of ion chromatographic analysis In order to conduct the continuous measurement of trace gases over a long period, stability of the measurement system, including both the air sampling and the sample analysis, is necessary.The instrument was calibrated with the mixed standard solution containing 50 ng ml21 of Cl2, 100 ng ml21 of NO32, 200 ng ml21 of SO4 22 and 122.2 ng ml21 of NH4 +, before and after the continuous measurements. Stability tests of the ion chromatographic analysis were carried out by two different approaches. The repeatability was tested by analyzing the standard solution repeatedly. As shown in Table 5, the uncertainty (RSD) of repeated analyses of the standard solution was < 2.1% (n = 10) for each ion.In addition, the reproducibility of the ion chromatographic analysis during 50 d of continuous operation was checked and the results are also given in Table 5. Each response of the standard solution analysis for Cl2, NO32, SO4 22 and NH4 + was unchanged and the RSD for each ion was < 5% during 50 d. These results indicate that the ion chromatograph system showed good stability for continuous measurement under the analytical conditions in Table 2.Blank analysis of the scrubbing solution and detection limits The limit of detection is governed not only by the instrumental sensitivity and stability but also by the blank concentrations of Cl2, NO32, SO4 22 and NH4 + in the scrubbing solution. The blank concentrations of Cl2, NO32, SO4 22 and NH4 + in the scrubbing solution, that had flowed through the entire liquid delivery lines but not been exposed to the ambient air, were Table 3 Collection efficiencies of trace acidic and basic gases using the diffusion scrubber with de-ionized water as the scrubbing solution Collection efficiency (%) Concentration of Gas gas sample (ppbv) Experimentala (n = 5) Idealb HCl 3.1 ± 0.1 99.5 ± 0.4 99.7 HNO3 4.5 ± 0.3 97.0 ± 0.2 98.0 SO2 27.1 ± 1.3 97.2 ± 0.2 99.2 NH3 25.3 ± 4.0 99.1 ± 0.8 > 99.9 a Effective length, 50 cm; air flow rate, 1.0 l min21; sampling duration, 52 min.b Calculated from the Gormley–Kennedy equation.24 Table 4 Effect of NO2 on the collection of HNO2 and HNO3 by the diffusion scrubber method. Air sampling volume, 52 l; sample solution volume, 12 ml Effect on gas collection (ppbv)a Ratio to NO2 (%) NO2 Collected Run No. (ppbv) (A) HNO2 (B) HNO3 (C) HNO2 (B/A) HNO3 (C/A) NO2 (%) 1 54 0.010 < 0.006b 0.018 < 0.001 < 0.03 2 54 0.009 < 0.006b 0.016 < 0.01 < 0.03 3 54 0.008 < 0.006b 0.015 < 0.01 < 0.03 4 54 0.008 < 0.006b 0.014 < 0.01 < 0.03 5 54 0.007 < 0.006b 0.013 < 0.01 < 0.02 6 54 0.007 < 0.006b 0.013 < 0.01 < 0.02 7 54 0.008 < 0.006b 0.015 < 0.01 < 0.03 8 54 0.007 < 0.006b 0.013 < 0.01 < 0.02 9 54 0.007 < 0.006b 0.013 < 0.01 < 0.02 10 54 0.007 < 0.006b 0.014 < 0.01 < 0.02 Av. 54 0.008 < 0.006b 0.014 < 0.01 < 0.03 s 0.001 0.002 a The concentration of HNO2 or HNO3 (ppbv) was calculated from NO22 or NO32 in the scrubbing solution when 54 ppbv of NO2 were collected by the diffusion scrubber.b Lower detection limit (ppbv). Analyst, 1999, 124, 1151–1157 1155measured. The scrubbing solution was always de-ionized with an ultrapure water generator (Milli-Q Jr.) in this system. However, trace amounts of Cl2, NO32, SO4 22 and NH4 + were measured in the fresh scrubbing solution, as shown in Table 6. The uncertainties (3s) in the blank analyses of Cl2, NO32, SO4 22 and NH4 + gave the detection limits.As shown in Table 6, the detection limits of HCl, HNO3, SO2 and NH3 were less than 0.01 ppbv for a 52 l air sampling volume. The blank concentrations of Cl2, NO32, SO4 22 and NH4 + in the scrubbing solution were also virtually unchanged during 24 d continuous measurement. Ambient air measurement The method described was employed in a field observation as part of the Core Research for Evolutional Science and Technology (CREST) program at Cape Hedo, Okinawa, Japan, from July 13 to August 6, 1997.The sampling site is located at the north end of Okinawa and faces the Pacific Ocean and East China Sea (latitude 26°52AN, longitude 128°16AE). During the field observation, trace gases were measured every hour at an air flow rate of 1.0 l min21. The concentration variations of HCl, HNO3, SO2 and NH3 in the atmosphere are shown in Fig. 4. The average concentrations of HCl, HNO3, SO2 and NH3 were 0.53 ± 0.53 (n = 426), 0.07 ± 0.05 (n = 420), 0.27 ± 0.21 (n = 424) and 1.62 ± 1.16 ppbv (n = 379), respectively.A diurnal dependence was observed for HCl, HNO3 and NH3, reaching maxima during early afternoon and reaching minima at around midnight. Hence the developed DSIC system was found to be very useful and practical for measuring sub-ppbv levels of HCl, HNO3, SO2 and NH3 in the atmosphere. Intercomparison tests In order to evaluate the accuracy of the DS-IC system, an intercomparison of SO2 concentrations in the atmosphere measured by this DS-IC system and a UV pulse fluorescence meter (PFM, Model 43S, Thermo Environmental Instruments) was carried out at the National Acid Rain Monitoring Station in Oki, Japan, during February 25–March 18, 1994.The detection limit of SO2 by the UV pulse fluorescence meter was 0.1 ppbv (S/N = 3). As shown in Fig. 5, the SO2 concentrations obtained by the two methods agreed well. The least-squares fit for the SO2 concentrations measured by the two methods was as follows: [SO2]DS-IC = 1.02[SO2]PFM + 0.07 (r2 = 0.982, n = 195) (5) Conclusions The developed DS-IC system was found to be suitable for the measurement of trace acidic and basic gases such as HCl, Table 5 Repeatability and reproducibility of ion chromatographic analysis of the standard solution. 12 ml of the standard solution (50 ng ml21 Cl2, 100 ng ml21 NO32, 200 ng ml21 SO4 22 and 122.2 ng ml21 NH4 +) were preconcentrated and analyzed Peak area Run No.Cl2 NO32 SO4 22 NH4 + Repeatability— 1 1.44 3 107 1.46 3 107 3.30 3 107 1.12 3 106 2 1.42 3 107 1.41 3 107 3.27 3 107 1.08 3 106 3 1.42 3 107 1.43 3 107 3.34 3 107 1.09 3 106 4 1.46 3 107 1.45 3 107 3.31 3 107 1.12 3 106 5 1.39 3 107 1.42 3 107 3.26 3 107 1.13 3 106 6 1.43 3 107 1.43 3 107 3.27 3 107 1.15 3 106 7 1.47 3 107 1.48 3 107 3.23 3 107 1.09 3 106 8 1.41 3 107 1.41 3 107 3.34 3 107 1.09 3 106 9 1.45 3 107 1.46 3 107 3.29 3 107 1.14 3 106 10 1.39 3 107 1.46 3 107 3.35 3 107 1.09 3 106 Av.(n = 10) 1.43 3 107 1.44 3 107 3.30 3 107 1.11 3 106 s 0.03 3 107 0.02 3 107 0.03 3 107 0.02 3 106 RSD (%) 2.10 1.39 0.91 1.80 Reproducibility— April 19, 1997 1.44 3 107 1.47 3 107 3.21 3 107 1.12 3 106 1.45 3 107 1.47 3 107 3.19 3 107 1.13 3 106 1.44 3 107 1.47 3 107 3.22 3 107 1.12 3 106 May 22, 1997 1.54 3 107 1.40 3 107 3.30 3 107 1.00 3 106 1.53 3 107 1.41 3 107 3.29 3 107 1.02 3 106 1.53 3 107 1.40 3 107 3.23 3 107 0.99 3 106 June 7, 1997 1.43 3 107 1.41 3 107 3.40 3 107 1.08 3 106 1.43 3 107 1.42 3 107 3.41 3 107 1.07 3 106 1.45 3 107 1.42 3 107 3.39 3 107 1.09 3106 Av.(n = 9) 1.47 3 107 1.43 3 107 3.29 3 107 1.07 3 106 s 0.04 3 107 0.03 3 107 0.08 3 107 0.05 3 106 RSD (%) 2.7 2.1 2.4 4.7 Table 6 Blank concentrations of the fresh scrubbing solution and detection limits of trace gases in the atmosphere Concentration/ng ml21 Cl2 NO32 SO4 22 NH4 + 0.040 0.091 0.53 0.32 0.025 0.084 0.57 0.31 0.051 0.079 0.45 0.31 0.022 0.037 0.53 0.32 0.049 0.053 0.42 0.31 0.031 0.059 0.46 0.30 Av.(n = 6) 0.036 0.067 0.49 0.31 s 0.012 0.021 0.06 0.01 Detection limita HCl HNO3 SO2 NH3 (ppbv) 0.006 0.006 0.010 0.006 a Corresponding to 3s of the blank analysis of the scrubbing solution, with a 52 l air sampling volume. Fig. 4 Concentration variations of trace acidic and basic gases (HCl, HNO3, SO2 and NH3) at Cape Hedo, Okinawa, Japan during July 13–August 6, 1997. 1156 Analyst, 1999, 124, 1151–1157HNO3, SO2 and NH3 in the atmosphere. The automated, simultaneous and continuous measurement of these gases at 60 min intervals can be accomplished by collecting analyte gases by the diffusion scrubber, with de-ionized water as the scrubbing solution, preconcentrating the total ions such as Cl2, NO32, SO4 22 and NH4 + in anion and cation concentrator columns connected in series and analyzing by ion chromatography.Using semimicro-type anion and cation separator columns, continuous measurement during 1 month was performed with less than 5 l of eluent.The detection limits of these trace gases were less than 0.01 ppbv for 52 l air sampling volume. The interference from NO2 in the determination of HNO3 was negligible. Acknowledgements This work was supported by funding from the Core Research for Evolutional Science and Technology (CREST) program by the Japan Science and Technology Corporation (JST), the IGAC/ PEACAMPOT program by the National Institute for Environmental Studies (NIES) and the Creative and Fundamental R&D Program for SMES by the Japan Small Business Corporation (JSBC).The authors thank T. Shirai and K. Inoue of Tokyo Dylec Corporation for producing the diffusion scrubber. References 1 H. Akimoto and H. Narita, Atmos. Environ., 1994, 28, 213. 2 B. R. Appel, S. M. Wall, Y. Tokiwa and M. Haik, Atmos. Environ., 1979, 13, 319. 3 D. Klockow, B. Jablonski and R. Niessner, Atmos. Environ., 1979, 13, 1665. 4 J. Forrest, R. L. Tanner, D. J. Spandau, T. D’Ottavio and L. Newman, Atmos. Environ., 1980, 14, 137. 5 B. R. Appel, Adv. Chem. Ser., 1993, 232, 1. 6 M. Ferm, Atmos. Environ., 1979, 13, 1385. 7 W. A. McClenny, P. C. Galley, R. S. Braman and T. J. Shelley, Anal. Chem., 1982, 54, 365. 8 R. S. Braman, T. J. Shelley and W. A. McClenny, Anal. Chem., 1982, 54, 358. 9 D. Klockow, R. Niessner, M. Malejczyk, H. Kiendl, B. vom Berg, M. P. Keuken, A. Wayers-Ypelaan and J. Slanina, Atmos. Environ., 1989, 23, 1131. 10 M. P. Keuken, C. A. M. Schoonebeek, A. Wensveen-Louter and J. Slanina, Atmos. Environ., 1988, 22, 2541. 11 P. K. Dasgupta, Adv. Chem. Ser., 1993, 232, 41. 12 Z. Vecera and P. K. Dasgupta, Environ. Sci. Technol., 1991, 25, 225. 13 Z. Vecera and P. K. Dasgupta, Anal. Chem., 1991, 63, 2210. 14 P. K. Simon, P. K. Dasgupta and Z. Vecera, Anal. Chem., 1991, 63, 1237. 15 P. K. Simon and P. K. Dasgupta, Anal. Chem., 1993, 65, 1134. 16 W. Frenzel, Anal. Chim. Acta, 1994, 291, 305. 17 W. Frenzel, D. Schepers and G. Schulze, Anal. Chim. Acta, 1993, 277, 103. 18 P. K. Simon and P. K. Dasgupta, Environ. Sci. Technol., 1995, 29, 1534. 19 G. P. Wyers, R. P. Otjes and J. Slania, Atmos. Environ., 1993, 27, 2085. 20 M. T. Oms, P. A. C. Jongejan, A. C. Veltkamp, G. P. Wyers and J. Slania, Int. J. Environ. Anal. Chem., 1995, 62, 207. 21 S. M. Buhr, M. P. Buhr, F. C. Fehsenfeld, J. S. Holloway, U. Karst, R. B. Norton, D. D. Parrish and R. E. Sievers, Atmos. Environ., 1995, 29, 2609. 22 Y. Suzuki, Anal. Chim. Acta, 1997, 353, 227. 23 W. MacKays and M. E. Craford, Convective Heat and Mass Transfer, McGraw-Hill, New York, 2nd edn., 1980, pp. 66–68. 24 P. G. Gormley and M. Kennedy, Proc. R. Ir. Acad., 1949, 52, 163. 25 P. Matusca, B. Schwartz and K. Bächmann, Atmos. Environ., 1984, 18, 1667. 26 C. L. Benner, N. L. Eatough, E. A. Lewis and D. J. Eatough, Atmos. Environ., 1988, 22, 1669. 27 B. R. Fish and J. L. Durham, Environ. Lett., 1971, 2, 13. 28 J. H. Jr. Overton, V. P. Aneja and J. L. Duraham, Atmos. Environ., 1979, 13, 355. 29 S. E. Schwartz and W. H. White, in Advances in Environmental Science and Engineering, ed. J. R. Pfafflin and E. N. Ziegler, Gordon and Breach, New York, 1984, vol. 4, pp. 1–45. Paper 9/03565F Fig. 5 Intercomparison of SO2 concentration in the atmosphere measured by the diffusion scrubber method and the pulse fluorescence method. Analyst, 1999, 124, 1151–1157 1157

ISSN:0003-2654    

DOI:10.1039/a903565f

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Multiresidue analysis of pesticides in vegetables and fruits using a high capacity absorbent polymer for water Hirotaka Obana,* Kazuhiko Akutsu, Masahiro Okihashi, Sachiko Kakimoto and Shinjiro Hori Osaka Prefectural Institute of Public Health, 1-3-69, Nakamichi, Higashinari-ku, Osaka, 5370025, Japan. E-mail: obana@iph.pref.osaka.jp Received 26th April 1999, Accepted 8th June 1999 A single extraction and a single clean-up procedure was developed for multi-residue analysis of pesticides in non-fatty vegetables and fruits.The method involves the use of a high capacity absorbent polymer for water as a drying agent in extraction from wet food samples and of a graphitized carbon column for clean-up. A homogeneously chopped food sample (20 g) and polymer (3 g) were mixed to absorb water from the sample and then 10 min later the mixture was vigorously extracted with ethyl acetate (100 ml). The extract (50 ml), separated by filtration, was loaded on a graphitized carbon column without concentration.Additional ethyl acetate (50 ml) was also eluted and both eluates were concentrated to 5 ml for analysis. The procedure for sample preparation was completed within 2 h. In a recovery test, 107 pesticides were spiked and average recoveries were more than 80% from asparagus, orange, potato and strawberry. Most pesticides were recovered in the range 70–120% with usually less than a 10% RSD for six experiments. The results indicated that a single extraction with ethyl acetate in the presence of polymer can be applied to the monitoring of pesticide residues in foods.The control of pesticide residues in food for both regulatory and commercial purposes involves large numbers of samples. Pesticide residue analyses are usually costly and time consuming, hence multi-residue methods are usually applied in regulatory pesticide monitoring.1,2 Most pesticide residue analysis methods involve two key steps: extraction of target analytes from the bulk of the matrix and clean-up of the analyte from matrix co-extractives. For a method to be practical, it is also necessary to consider cost of an analysis, including reagents, equipment, labor and environmental restrictions.Japanese laboratories cannot freely drain waste containing dichloromethane owing to legal restrictions (less than 0.2 mg l21 in Japan).3 Numerous methods have been developed for the analysis of pesticide residues. The first step of a conventional solvent extraction is homogenization of a mixture of wet sample and a water soluble solvent such as acetone or acetonitrile.1,2 Pesticides are transferred to the hydrophobic organic layer by partitioning or salting out and the water layer is discarded.On the other hand, dried samples are extracted in a static mode in an extraction cell with automated extractors such as supercritical fluid extraction (SFE)4 and accelerated solvent extraction (ASE)5,6 and in a glass column with a matrix solid phase extraction.7 We have attempted to develop a new extraction method which is a combination of the above two basic methods in this study.Water in wet food samples is absorbed by a high capacity absorbent polymer in advance, and then the mixture of food and polymer is extracted with ethyl acetate in the dynamic mode with a high speed homogenizer. The new procedure is considered to have advantages with regard to extraction speed and easy operation.Ethyl acetate can extract pesticides effectively since it can easily penetrate into samples without water. A high speed homogenizer reduces the sample particle size and it can also enforce effective extraction by a vigorous spiral stream. High capacity absorbent polymers for water are widely used in industry. A soft polymer is mainly used for sanitation purposes such as in disposable diapers and hard polymer, as used in this study, is applied in the agricultural field to retain water in soil.Experimental Food preparation Samples were obtained at a local market in Osaka and residual pesticides were determined by the method proposed in this study. We confirmed that the samples used in the recovery test were pesticide free. About 500 g of sample were chopped in a conventional food processor for 5 min to obtain thoroughly mixed homogenates. For residue analyses, samples were found to contain pesticide residues in our regular monitoring and were stored in a freezer until this study.Reagents Ethyl acetate, acetone and hexane were of pesticide analysis grade and cyclohexane was of HPLC grade (Wako Pure Chemicals, Osaka, Japan). Pesticide standards were purchased from Wako, Hayashi Pure Chemicals (Osaka, Japan) and Riedel-de Haën (Seelze, Germany). Each compound was dissolved in hexane or acetone to make a 1000 mg ml21 stock standard solution and stored in a refrigerator. The stock standard solutions were mixed and diluted to 2–8 mg ml21 with acetone based on the detector sensitivities in the recovery tests.Polar pesticides, such as organophosphorus and carbamate pesticides, cause matrix effects that might result in over- and/or underestimations of the recovery rates in GC analysis.8,9 To prevent this, working standard solutions were diluted with the cleaned extract from pesticide-free samples in the recovery test. The working standard solutions were also diluted with the mixture Analyst, 1999, 124, 1159–1165 1159of the cleaned extract from pesticide-free samples in the residue analysis.Super-absorbent polymer The super-absorbent polymer used was an industrial product, Aquapearl A3, from Mitsubishi Chemical Industry Ltd. (Tokyo, Japan). Aquapearl A3 is a polymer of acrylic acid with a white color and a 60–300 mm particle size. A 1 g amount of dry Aquapearl A3 occupies about 1 ml and absorbs 200 ml of water, according to the manufacturer. The polymer was used without any clean-up procedure.Graphitized carbon mini-column A graphitized carbon mini-column (Carbograph, 500 mg, 6 ml) (GL Science, Tokyo, Japan) was eluted with 50 ml of ethyl acetate before use for preconditioning. Gel permeation chromatography (GPC) An AUTOVAP AS-2000 (ABC Laboratories, Columbia, MO, USA) was equipped with a CLNpak column (300 3 20 mm id) (Showadenko, Tokyo, Japan). The mobile phase was a mixture of acetone and cyclohexane (3 + 7) at a flow rate of 5 ml min21 as described previously.10 After the extract (5 ml) had been Table 1 Pesticides analyzed and spiked levels in the recovery test Pesticide Detector Detection limita/ ng g21 Monitored (m/z) Ion Spiked level/ mg g21 Echlomezol NCI 3 35 219 0.4 a-BHC NCI 7 71 35 0.2 Dichloran NCI 1 206 204 0.4 b-BHC NCI 11 71 35 0.2 g-BHC NCI 8 71 35 0.2 Quintozene NCI 0.2 249 251 0.2 Tefluthrin NCI 1 241 205 0.4 d-BHC NCI 8 71 35 0.2 Chlorothalonil NCI 0.1 266 264 0.2 Heptachlor NCI 2 266 264 0.2 Dichlorfluanid NCI 0.2 155 199 0.4 Aldrin NCI 1 35 258 0.2 Dicofol NCI 4 250 252 0.4 Fthalide NCI 3 228 226 0.4 Pyrifenox NCI 8 226 228 0.4 HCE NCI 1 35 212 0.2 Captan NCI 1 150 35 0.4 Procymidone NCI 4 35 247 0.4 Triflumizole NCI 1 218 216 0.4 Paclobutrazol NCI 61 166 168 0.4 Endosulfan-a NCI 0.4 242 240 0.2 DDE NCI 1 35 282 0.2 Dieldrin NCI 1 35 237 0.2 Chlorobenzilate NCI 1 278 280 0.4 Endosulfan-b NCI 0.4 242 240 0.2 p,pA-DDD NCI 1 35 249 0.2 o,pA-DDT NCI 1 35 249 0.2 Propiconazole NCI 17 218 220 0.4 p,pA-DDT NCI 3 35 249 0.2 Triphenyl phosphate NCI — 249 325 — Captafol NCI 1 150 35 0.4 Bifenthrin NCI 1 205 241 0.4 Halfenprox NCI 2 79 81 0.4 Tetradifon NCI 0.1 245 247 0.2 Cyhalothrin NCI 7 205 241 0.4 Permethrin NCI 12 207 209 0.4 Cyfluthrin NCI 4 207 209 0.4 Cypermethrin NCI 3 207 209 0.4 Flucythrinate NCI 0.2 199 243 0.4 Fenvalerate NCI 6 167 211 0.4 Fluvalinate NCI 2 294 296 0.4 Deltamethrin NCI 2 79 81 0.4 Isoprocarb EI 11 121 136 0.2 Fenobucarb EI 4 121 150 0.2 Propoxur EI 8 110 152 0.2 Chlorpropham EI 5 127 213 0.2 Bendiocarb EI 5 151 166 0.2 Carbofuran EI 5 164 149 0.2 Pirimicarb EI 4 166 238 0.2 Carbaryl EI 6 144 115 0.2 Metribuzin EI 7 198 215 0.2 Metalaxyl EI 11 206 160 0.2 Methiocarb EI 13 168 153 0.2 Esprocarb EI 13 222 162 0.2 Thiobencarb EI 41 100 257 0.2 Pesticide Detector Detection limita/ ng g21 Monitored (m/z) Ion Spiked level/ mg g21 Diethofencarb EI 16 225 267 0.2 Triadimefon EI 51 208 181 0.2 Pendimethalin EI 5 252 162 0.2 Triadimenol EI 97 112 168 0.4 Quinomethionate EI 10 206 234 0.2 Flutolanil EI 17 173 281 0.2 Pretilachlor EI 29 238 176 0.2 Isoprotiolane EI 41 118 231 0.2 Myclobutanil EI 141 179 150 0.4 Mepronil EI 62 119 269 0.2 Lenacil EI 55 153 136 0.4 Thenylchlor EI 28 127 288 0.2 Triphenyl phosphate EI — 326 215 — Iprodione EI 104 314 316 0.4 Tebufenpyrad EI 18 171 318 0.2 Mefenacet EI 71 192 120 0.4 Fenarimol EI 73 139 107 0.4 Pyridaben EI 18 147 309 0.2 Dichlorvos FPD 5 0.4 Methamidophos FPD 30 0.4 Acephate FPD 30 0.4 Ethoprophos FPD 5 0.4 Thiometon FPD 10 0.4 Dioxabenzofos FPD 10 0.4 Terbufos FPD 10 0.4 Diazinon FPD 5 0.4 Etrimfos FPD 15 0.4 Iprobenfos FPD 10 0.4 Dichlofenthion FPD 10 0.4 Cyanophos FPD 10 0.4 Dimethoate FPD 20 0.4 Tolclofos-methyl FPD 10 0.4 Pyrimiphos-methyl FPD 10 0.4 Chlorpyrifos FPD 10 0.4 Parathion-methyl FPD 15 0.4 Fenthion FPD 10 0.4 Malathion FPD 20 0.4 Fenitrothion FPD 15 0.4 Parathion FPD 15 0.4 Bromophos-methyl FPD 15 0.4 Isofenphos FPD 15 0.4 Phenthoate FPD 15 0.4 Mecarbam FPD 15 0.4 Prothiofos FPD 25 0.4 Methidathion FPD 15 0.4 Butamifos FPD 15 0.4 Ethion FPD 30 0.4 Carbophenothion FPD 10 0.4 Edifenphos FPD 15 0.4 Isoxathion FPD 15 0.4 EPN FPD 15 0.4 Pyridaphenthion FPD 20 0.4 Phosmet FPD 50 0.4 Phosalone FPD 20 0.4 a More than three times the signal-to-noise ratio with sample matrix. 1160 Analyst, 1999, 124, 1159–1165loaded into the GPC system, the first 55 ml of eluate were discarded and the following 95 ml were collected as the pesticide fraction.A further 25 ml were eluted for the GPC wash. Negative chemical ionization (NCI) mode GC-MS An Auto Mas 120M (JEOL, Tokyo, Japan) was connected to an HP5890 gas chromatograph (Hewlett-Packard, Palo Alto, CA, USA). The GC conditions were as follows: column, DB-5 (J&W Scientific, Folsome, CA, USA), 20 m 3 0.25 mm id, 1 mm film thickness; column temperature program, 60 °C (1 min), 60–170 °C at 20 °C min21, 170–300 °C at 6 °C min21, 300 °C (5 min); carrier gas, He; injection temperature, 250 °C; injection volume, 2 ml with HP7673 autosampler (Hewlett-Packard); injection mode, splitless.The MS conditions were as follows: ionization voltage, 150 eV; filament current, 0.3 mA; detector voltage, 20.7 kV; scan time, 100 ms; ion source temperature, 180 °C; transfer line temperature, 250 °C; chemical ionization gas, isobutane (purity 99.99%); vacuum, 0.3 Torr. The ions of pesticides monitored are given in Table 1.Electron ionization (EI) mode GC-MS A Magnum (Thermoquest, CA, USA) was connected to a Model 3300/3400 gas chromatograph (Varian, Palo Alto, CA, USA). The GC conditions were as follows: column, DB-5ms (J&W Scientific), 30 m 3 0.25 mm id, 0.25 mm film thickness; column temperature, 60 °C (1 min), 60–300 °C at 8 °C min21, 300 °C (5 min); carrier gas, He; injection temperature, 250 °C; injection volume, 2 ml with A200S autosampler (Varian); injection mode, splitless.The MS conditions were as follows: filament current, 10 mA; multiplier voltage, 1800 V; scan range, m/z 100–400; scan time, 600 ms; ion trap temperature, 220 °C; transfer line temperature, 250 °C. The ions of pesticides monitored are given in Table 1. GC-FPD An HP5890 gas chromatograph (Hewlett-Packard) equipped with a flame photometric detector (FPD) and DB-1701 capillary column (30 m 3 0.32 mm id, 0.25 mm film thickness) (J&W Scientific) was used. The GC conditions were as follows: column temperature, 80 °C (2 min), 80–180 °C at 20 °C min21, 180 to 260 °C at 4 °C min21, 260–280 °C (at 10 °C min21; carrier gas, He; injection temperature, 250 °C; injection volume, 2 ml with HP7673 autosampler (Hewlett-Packard); injection mode, splitless; detector temperature, 300 °C.Extraction with polymer The sample (20 g) was thoroughly mixed with polymer, using a small spatula, in a 180 ml milk bottle and set aside for 10 min to allow water absorption.Ethyl acetate (100 ml) was added to the mixture and vigorously extracted with a high speed homogenizer (HG-30, Hitachi Koki, Katsuta, Japan) for 2 min. The extract and the mixture were separated using coarse filter paper (No. 5, Advantec, Tokyo, Japan) and 70 ml of solution were collected in a graduated cylinder. The extract was loaded on the graphitized carbon column and eluted by gravitation. Another 50 ml of ethyl acetate were measured into the previously used cylinder and loaded on the column to elute pesticides.The two eluates were collected in the same roundbottomed flask and concentrated nearly to dryness with an evaporator. The residue was dissolved in acetone–cyclohexane (3 + 7) to give a volume of 7 ml. The solution was injected into the 5 ml sample loop of the GPC instrument to make an exact 5 ml injection by overflow. The pesticide fraction was concentrated nearly to dryness with an evaporator and then dissolved in acetone–hexane (1 + 1) with the addition of triphenyl phosphate as an internal standard (IS) to give a volume of 10 ml.When GPC clean-up was not applied, 50 ml of extract were loaded on the graphitized carbon column followed by 50 ml of ethyl acetate for elution and the eluate was concentrated nearly to dryness with an evaporator. The residue was dissolved in acetone–hexane to give a volume of 10 ml after addition of IS solution. The amounts of IS added were 10 mg for GC-FPD and NCI mode GC-MS and 2 mg for EI mode GC-MS.The test solution (1 ml) corresponded to 1 g of sample. Recovery test The recovery tests were performed in duplicate. The first was a preliminary test that was performed to decide the dose of polymer and the necessity for GPC with 12 pesticides. The second test was evaluated with 107 pesticides, which were divided into two groups based on their sensitivities with the corresponding detectors, hence the tests were conducted twice on each sample.One group consisted of 36 organophosphorus pesticides that were determined with GC-FPD and 41 halogenated pesticides that were determined with NCI mode GC-MS. The other group consisted of 30 carbamate and nitrogen containing pesticides that were determined with EI Table 2 Preliminary recovery study with 12 pesticides in apple, green peas and pumpkina Polymer/g Spiked Apple Green bean Pumpkin level/ Pesticide mg g21 0.5 3.0 0.5 3.0 0.5 3.0 g-BHC 0.1 124 128 107 98 64 80 Carbaryl 0.2 127 110 112 92 87 101 Chlorpropham 0.2 119 107 87 89 82 93 Chlorpyrifos 0.2 96 99 68 78 81 87 Cyhalothrin 0.2 87 80 65 70 100 102 DDE 0.1 97 96 92 92 92 96 Dieldrin 0.1 103 109 90 86 84 99 Fenitrothion 0.2 103 104 82 89 83 88 Methamidophos 0.2 56 71 59 71 62 84 Methidathion 0.2 107 107 99 98 103 111 Permethrin 0.2 110 110 68 71 87 84 Pirimicarb 0.2 115 102 96 91 85 95 a Average recoveries (%) of three experiments. Analyst, 1999, 124, 1159–1165 1161mode GC-MS.A pesticide mixture (1 ml) in acetone was added to the sample to give a final concentration of 0.1–0.4 mg g21 on a sample weight basis, and the sample was left for 30 min before extraction.The pesticides spiked are listed in Table 1 together with their spiking levels, appropriate detectors and ions monitored if detected with GC-MS. Acetonitrile extraction Acetonitrile extraction was performed according to the method described by Supelco11 at half the scale of the original method.A sample (25 g) was extracted with acetonitrile (50 ml) for 2 min and further homogenization was performed after addition of 2.5 g of NaCl to salt out water. The extract was left for 30 min to separate into two layers. The upper layer (about 15 ml) was collected and passed through a filter to remove the suspension into a 30 ml glass bottle with a cap. Anhydrous Na2SO4 (5 g) was added to the extract and shaken vigorously for 1 min. An aliquot of 10 ml, equivalent to 5 g of sample, was collected and evaporated nearly to dryness.The residue was dissolved in 2 ml of acetonitrile–toluene (3 + 1) and loaded on the graphitized carbon column, which was eluted with 30 ml of the same mixture solution. The eluate was evaporated and the residue was dissolved in acetone–hexane (1 + 1) to give a volume of 5 ml after addition of IS solution. Results Method development began with the evaluation of the polymer stability in the presence of ethyl acetate because possible fragment(s) from the polymer might interfere in the pesticide determination.The polymer which absorbed water was extracted with ethyl acetate. The polymer swelled and looked transparent after it had absorbed water. It seemed slightly wet to the touch, but water was not transferred to fingers. It was scattered when extracted with ethyl acetate and no breakdown particles were observed after homogenization. The extract was analyzed using ion trap GC-MS and it showed an almost flat chromatogram without interference in pesticide determination.The same chromatograms were also observed in NCI mode GCMS and GC-FPD (data not shown). Aquapearl A3 is classified as a physically hard type among super-absorbent polymers, hence it would be resistant against ethyl acetate even after it absorbed water. We checked several kinds of soft and hard type polymers and found that soft polymers became like a slime or showed interfering peaks in GC-MS determination.A preliminary recovery test with 12 pesticides in three samples was performed to decide the dose of polymer and the necessity for GPC in the clean-up procedure. Most recovery results were within an acceptable range between 70 and 120% regardless of the polymer dose (Table 2). The exception was the recovery of methamidophos with 0.5 g of polymer, which was less than 70% in three samples; however, the rates were 71–84% with a dose of 3 g. Since methamidophos is a water soluble compound, it seemed that a smaller amount of polymer might not be sufficient.The dose of polymer was fixed at 3 g in the subsequent studies. The effect of clean-up of the graphitized carbon mini-column was obvious in removing food color. Green bean and pumpkin extracts changed to yellow from green after column clean-up. A slight yellow extract of apple became colorless after the column treatment (data not shown). Typical total ion chromatograms in EI mode GC-MS of apple extracts are shown in Fig. 2. The major peaks do not differ in the Fig. 1 Total ion chromatogram of apple extract with EI mode GC-MS: (a) crude; (b) after graphitized carbon column clean-up; (c) after graphitized carbon column and GPC clean-up. Fig. 2 Selected ion chromatogram of permethrin residue in romaine lettuce (Table 4) with NCI mode GC-MS. 1162 Analyst, 1999, 124, 1159–1165Table 3 Recovery of 107 pesticides in four vegetables and fruits Asparagus Potato Orange Strawberry Averagea RSD Averagea RSD Averagea RSD Averagea RSD Pesticide (%) (%) (%) (%) (%) (%) (%) (%) Echlomezol 81 10 78 8 80 4 74 8 a-BHC 85 3 88 4 79 5 85 6 Dichloran 93 1 93 5 83 3 85 4 b-BHC 77 1 84 5 85 2 72 4 g-BHC 67 2 80 6 68 2 72 4 Quintozene 70 4 88 5 77 4 75 5 Tefluthrin 81 5 90 7 78 5 75 8 d-BHC 76 3 80 4 96 6 85 5 Chlorothalonil 64 5 81 7 74 4 64 7 Heptachlor 80 3 89 5 86 4 76 4 Dichlorfluanid 15 17 78 5 51 10 34 23 Aldrin 78 5 104 7 84 5 79 5 Dicofol 89 12 82 13 89 14 69 6 Fthalide 83 3 91 3 84 2 87 2 Pyrifenox 67 2 74 5 79 4 75 2 HCE 66 3 88 7 73 4 71 4 Captan 5 26 70 5 31 15 27 30 Procymidone 73 3 82 10 83 3 75 3 Triflumizole 87 3 76 12 77 6 68 3 Paclobutrazol 67 1 78 6 105 9 74 2 Endosulfan-a 76 7 85 3 81 4 93 7 DDE 76 2 95 3 83 4 73 3 Dieldrin 87 2 94 4 77 4 81 4 Chlorobenzilate 79 2 84 3 84 4 77 4 Endosulfan-b 80 8 84 5 76 3 80 6 p,pA-DDD 77 4 91 6 76 2 73 5 o,pA-DDT 83 4 94 4 78 3 74 5 Propiconazole 77 3 80 5 76 3 74 3 p,pA-DDT 82 5 86 3 82 2 78 3 Captafol 8 26 64 3 38 14 38 20 Bifenthrin 94 3 91 2 82 2 82 3 Halfenprox 109 4 116 7 82 3 79 8 Tetradifon 83 4 86 3 78 5 74 6 Cyhalothrin 98 5 88 3 82 4 81 4 Permethrin 100 5 82 5 94 10 82 4 Cyfluthrin 103 7 83 2 82 5 80 6 Cypermethrin 103 8 84 4 81 6 79 6 Flucythrinate 107 9 91 3 84 7 84 7 Fenvalerate 106 9 94 4 87 7 83 5 Fluvalinate 103 8 93 6 88 9 88 7 Deltamethrin 105 9 90 3 82 7 83 6 Isoprocarb 93 4 86 7 89 6 93 4 Fenobucarb 98 3 83 7 89 6 98 3 Propoxur 101 4 84 6 95 5 101 4 Chlorpropham 98 3 76 5 85 6 98 3 Bendiocarb 100 4 85 6 83 5 100 4 Carbofuran 103 2 92 7 85 4 103 2 Pirimicarb 100 3 86 6 85 6 100 3 Carbaryl 100 5 96 5 93 7 100 5 Metribuzin 85 5 89 7 61 9 85 5 Metalaxyl 103 3 91 4 86 4 103 3 Methiocarb 97 7 87 7 91 9 97 7 Esprocarb 78 4 85 9 85 6 78 4 Thiobencarb 81 4 73 13 83 8 81 4 Diethofencarb 93 3 91 5 84 6 93 3 Triadimefon 89 4 99 7 82 8 89 4 Pendimethalin 74 5 86 7 81 5 74 5 Triadimenol 85 5 82 4 76 6 85 5 Quinomethionate 78 5 68 7 74 8 78 5 Flutolanil 81 6 80 8 77 7 81 6 Pretilachlor 81 5 82 6 85 6 81 5 Isoprotiolane 92 7 89 12 89 9 92 7 Myclobutanil 83 7 87 14 82 5 83 7 Mepronil 78 4 85 7 79 8 78 4 Lenacil 80 5 114 10 81 9 80 5 Thenylchlor 85 7 83 10 73 5 85 7 Iprodione 75 4 85 6 75 11 75 4 Tebufenpyrad 70 5 83 4 85 9 70 5 Mefenacet 86 7 85 4 86 6 86 7 Fenarimol 84 7 82 7 87 7 84 7 Pyridaben 71 4 88 4 90 5 71 4 Dichlorvos 82 5 90 3 98 3 95 4 Methamidophos 79 10 79 4 76 6 64 3 Continued on next page Analyst, 1999, 124, 1159–1165 1163three chromatograms.The graphitized carbon column treatment showed some clean-up effect compared with the crude extract, but there seemed no obvious difference before and after GPC clean-up. One reason for using GPC clean-up was to exclude possible large component(s) of the polymer, which might be decomposed in the presence of food matrices. No such polymer component was observed in the three samples after the graphitized carbon clean-up.These results indicated that GPC clean-up was not necessary if samples were not fatty. GPC clean-up was omitted from the recovery tests with 107 pesticides, hence the graphitized carbon eluates were concentrated as a test solution. An actual recovery test was performed on asparagus, potato, orange and strawberry. The average recoveries in six experiments of 107 pesticides are given in Table 3. The best recovery was found in potato at 86% on average for the 107 pesticides and 105 pesticides showed an acceptable recovery range of 70–120%.For the other three samples, the average recoveries of 107 pesticides were more than 80% and an acceptable range was obtained in 99, 98 and 97 pesticides out of 107 in orange, strawberry and asparagus, respectively Among the spiked pesticides, captan, captafol, and dichlorfluanid were commonly less well recovered in the four samples. They have been reported to be unstable owing to reaction with food matrices.12 Other less well recovered pesticides showed more than 55% recovery rates and the low recoveries were observed in at best two samples out of four for in each pesticide.These results indicated that the polymer did not adsorb specific pesticides. Reproducibility was expressed as relative standard devation (RSD). Most RSD values except for the above three unstable pesticides were less than 10%. In strawberry, 104 pesticides showed RSD less than 10% and similarly 100 in asparagus and orange and 99 in potato.Some detection limits of pesticides such as myclobutanil and iprodione were at the 100 mg g21 level with EI mode GC-MS (Table 1), and they were probably affected by co-extracted interferents. The detection limits of the pesticides in the samples were almost the same as those in the standard solution with GCFPD and NCI mode GC-MS. NCI mode GC-MS is effective at excluding interferents in the determination of halogenated compounds. Permethrin in romaine lettuce using NCI mode GC-MS was observed without interference around the permethrin peaks (Fig. 2). The detection limits of pyrethroid pesticides were lower than those obtained by GC with electron capture detection.13 The proposed method with a polymer was compared with the commonly used acetonitrile method in residue analysis. Table 4 gives the levels of six pesticide residues in five samples. Permethrin, chlorpyrifos, triadimefon and triadimenol showed the almost same residue levels with the two methods.Methidathion and methamidophos residues with the polymer method were about 30% lower than those with the acetonitrile method. The lower level of methamidophos might be a reflection of its low recovery in the fortification study. Although the application of residue analysis was limited to six pesticide residues in five sample, the results suggested that the extraction with a polymer is comparable to the acetonitrile method. Table 3 Continued Asparagus Potato Orange Strawberry Averagea RSD Averagea RSD Averagea RSD Averagea RSD Pesticide (%) (%) (%) (%) (%) (%) (%) (%) Acephate NDb — 82 4 72 3 65 3 Ethoprophos 85 8 98 7 97 3 91 3 Thiometon 61 6 72 7 90 3 55 6 Dioxabenzofos 83 5 96 7 100 3 90 3 Terbufos 70 5 88 6 92 3 72 5 Diazinon 74 5 87 7 95 3 81 3 Etrimfos 76 5 88 6 94 3 83 2 Iprobenfos 83 4 89 7 95 2 88 2 Dichlofenthion 78 5 91 7 95 3 79 3 Cyanophos 83 5 91 7 97 2 91 2 Dimethoate 94 5 96 7 98 2 97 1 Tolclofos-methyl 75 4 87 7 94 3 79 3 Pyrimiphos-methyl 76 5 87 6 94 3 79 3 Chlorpyrifos 78 5 88 6 94 3 80 2 Parathion-methyl 80 4 88 6 97 2 88 2 Fenthion 71 4 81 6 93 2 76 3 Malathion 56 9 77 6 92 2 82 2 Fenitrothion 77 4 86 6 94 3 86 1 Parathion 76 5 84 6 95 3 81 2 Bromophos-methyl 84 4 91 7 96 2 80 3 Isofenphos 75 5 85 7 93 3 78 3 Phenthoate 76 4 86 6 93 2 81 2 Mecarbam 77 5 83 7 92 3 82 2 Prothiofos 85 4 90 7 97 3 81 2 Methidathion 87 5 87 6 97 3 91 1 Butamifos 77 5 81 6 92 3 78 3 Ethion 81 3 84 6 94 2 84 3 Carbophenothion 82 4 86 6 93 2 78 3 Edifenphos 83 4 86 8 92 3 79 3 Isoxathion 80 2 77 6 92 3 83 2 EPN 80 4 86 6 92 3 79 4 Pyridaphenthion 81 3 86 6 91 2 85 3 Phosmet 85 3 88 6 94 2 87 4 Phosalone 81 3 86 5 90 1 78 4 Average of 107 pesticides 81 5 86 6 85 5 80 5 a Average of six experiments.b Not determined owing to interference. 1164 Analyst, 1999, 124, 1159–1165Discussion The purpose of this study was to develop a fast, easy and inexpensive multi-residue method for the determination of pesticides in foods.We have previously introduced automated extraction systems in residue analysis such as ASE6 and SFE.14 Both systems need dried food samples to make the extraction fluid penetrated into the samples. Without drying agents the extraction efficiency of those methods was not acceptable. Through experience with those methods, we came to the conclusion that expensive equipment could be replaced by a high speed homogenizer if wet food samples were dried by an appropriate drying agent and extracted vigorously with an organic solvent.The prototype of this study was a method for the determination of acephate and methamidophos, in which samples are extracted with ethyl acetate in the presence of large amounts of anhydrous Na2SO4.15 Bennett et al.16 reported that 59 pesticides in milk could be determined with the same extraction procedure. A problem with this extraction method is the bulky anhydrous Na2SO4, the necessary dose of which is more than four times the sample weight in analysis for acephate.15 Further, the mixture of sample and anhydrous Na2SO4 is likely to become a hard lump, making extraction with a conventional homogenizer difficult.The solution found was to replace anhydrous Na2SO4 with a high capacity absorbent polymer as a drying agent in analyses for acephate and methamidophos in a previous study.17 The extraction efficiency with the polymer was the same as that with anhydrous Na2SO4.We introduced the same extraction system into the multi-residue analysis in this study. One advantage of polymer application was that there was no emulsion after vigorous extraction because it was considered that the emulsifier, water in the sample, was not concerned with extraction. Another benefit of the polymer is its cost, i.e., about US $10 kg21 from the manufacturer. The clean-up procedure, with direct loading of ethyl acetate extracts on the graphitized carbon mini-column, is also a characteristic aspect of this method.Pre-packed mini-columns have been widely used in multi-residue analysis of foods. A typical procedure is to remove interferents by passing a concentrated extract through the column so that the target pesticides are not retained in the column.11 If the pesticides are not retained in the column, it is considered that the crude extracts do not have to be concentrated to be loaded on the column.Hence the extract was directly loaded on the minicolumn without concentration. A graphitized carbon minicolumn has been reported to be applied to more than 200 pesticides in residue analysis when the pesticides were eluted with toluene–acetonitrile (1 + 3).11 However, in this work ethyl acetate (50 ml), also used as the extraction solvent, was used to elute pesticides from the carbon column to simplify the experimental operation. Except for quinomethionate, which required 50 ml for elution, other pesticides were eluted with 20 ml of ethyl acetate (data not shown).The proposed method has the advantages of a short operation time and easy operation. The extraction time including the water absorption time was about 15 min for one sample. The sample preparation time including extraction, clean-up, evaporation and filling was about 2 h in residue analysis. The recovery study showed that most pesticides were more than 70% recovered (Table 3), hence a single extraction with 100 ml of ethyl acetate could be accepted in multi-residue analysis. The reproducibilities were also acceptable in both the recovery test and residues analysis.It is concluded that this new method with a high capacity absorbent polymer can be applied to the monitoring of pesticide residues in foods. References 1 B. M. McMahon and N. F. Hardin, Pesticide Analytical Manual, Food and Drug Administration, Washington, DC, 1994, vol. 1, ch. 3. 2 J. Fillion, R. Hindle, M. Lacroix and J. Selwyn, J. AOAC Int., 1995, 78, 1252. 3 Ordinance of Prime Minister’s Office, Prime Minister’s Office of Japan, Tokyo, 1993, No. 54. 4 S. J. Lehotay, J. Chromatogr., 1997, 785, 289. 5 B. E. Richter, B. A. Jones, J. L. Ezzell, N. L. Porter, N. Avdalovic, and C. Pohl, Anal. Chem., 1996, 68, 1033. 6 H. Obana, K. Kikuchi, M. Okihashi and S. Hori, Analyst, 1997, 122, 217. 7 L. Kadenczki, Z. Arpad, I. Gardi, A. Ambrus, L. Gyorfi, G. Reese, and W. Ebing, J. AOAC Int., 1992, 75, 53. 8 S. J. Lehotay and K. I. Eller, J. AOAC Int., 1995, 78, 821. 9 S. J. Lehotay, N. Aharonson, E. Pfeil and M. A. Ibrahim, J. AOAC Int., 1995, 78, 831. 10 M. Okihashi, H. Obana and S. Hori, J. Food Hyg. Soc. Jpn., 1997, 38, 16. 11 Solid-Phase Extraction of Pesticides from Fruits and Vegetables, for Analysis by GC or HPLC, Bulletin 900A, Supelco, Bellefonte, PA, 1997. 12 Y. Akiyama, N. Yoshioka and M. Tsuji, J. Food Hyg. Soc. Jpn., 1998, 39, 303. 13 Y. Nakamura, Y. Tonogai, Y. Tsumura and Y. Ito, J. AOAC Int., 1993, 76, 1348. 14 H. Obana, M. Okihashi and S. Hori, J. Food Hyg. Soc. Jpn., 1998, 39, 172. 15 Pesticide Analytical Manual, US Food and Drug Administration, Washington, DC, 1991, vol. 2, Sect. 180.108. 16 D. A. Bennett, A. C. Chung and S. M. Lee, J. AOAC Int., 1997, 80, 1065. 17 H. Obana, M. Okihashi, S. Kakimoto and S. Hori, Anal. Commun., 1997, 34, 253. Paper 9/03297E Table 4 Comparison of polymer and acetonitrile extractions in residue analysis Polymer Acetonitrile Average/ Average/ Sample Pesticide mg g21 RSD (%) mg g21 RSD (%) Romaine Lettuce Permethrin 0.061 2.7 0.068 10.7 Orange Chlorpyrifos 0.116 15.1 0.126 1.8 Pineapple Triadimefon 0.498 5.2 0.508 6.5 Triadimenol 0.166 7.8 0.165 3.6 Grapefruit Methidathion 0.086 5.6 0.122 19.1 Cucumber Methamidophos 0.437 2.8 0.630 2.7 Analyst, 1999, 124, 1159–1165 1165

ISSN:0003-2654    

DOI:10.1039/a903297e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Discrimination of diesel fuels with chemical sensors and mass spectrometry based electronic noses Roger Feldhoff,a Claude-Alain Sabyb and Philippe Bernadeta a Elf-Aquitaine, Groupement de Recherches de Lacq, B.P. 34, 64170 Lacq, France b Elf Antar France, Centre de Recherches d’Elf Solaize, B.P. 22, 69360 Saint-Symphorien d’Ozon, France Received 17th March 1999, Accepted 2nd June 1999 The gas phase of about twenty diesel fuels obtained from three different refineries was examined with a FOX 4000 electronic nose (Alpha-M.O.S., France).The data of the eighteen semi-conductor gas sensors were evaluated by principal component analysis (PCA) and linear discriminant analysis (LDA). The sensor responses showed significant differences between the samples corresponding to the fuels’ origin. An LDA model allowed correct identification of the fuels’ production site. The results were compared to measurements with a recently developed electronic nose, which is based on a mass spectrometer (Smart Nose GA 200, LDZ Laboratory, Switzerland).A good correlation was found between the data measured with both types of instrument. The mass spectroscopic data were easier to obtain and more reproducible. Introduction Fuels are mixtures with a strongly varying composition depending on the crude source and refining process. This is accompanied by variations in the composition of the volatiles. A fast and reliable method to determine the odour quality of fuels could be the analysis of their headspace with an electronic nose.Electronic noses and fuel vapour analysis During the last few years the expression ‘electronic nose’ has been used for devices which measure the gas phase of a product by the aid of an assembly of chemical sensors. Recently, LDZ Laboratory, Switzerland and Hewlett Packard independently developed a novel type of ‘electronic nose’, based on mass spectrometry (MS). The difference between these systems and classical mass spectrometers is the following: In classical MS, the different molecules contained in the sample are separated (e.g., by liquid or gas chromatography or even by MS).In a second step these different molecules are analysed by a mass spectrometer. They can then in general be identified by the help of mass spectral libraries. MS-based electronic noses use statistical data treatment to compare and to classify an (unknown) sample with respect to an earlier defined statistical data model.This data model is based on a large number of measurements of a large variety of samples. In this respect, a headspace autosampler guarantees equal and reproducible injection conditions for all samples. Standard test methods for the determination of hydrocarbons in the liquid phase of oil derivatives are based on supercritical fluid chromatography,1 mass spectrometry2 or fluorescent indicator absorption.3 The gas phase is in general analysed by gas chromatography (GC) or gas chromatography coupled with mass spectrometry (GC-MS). Electronic noses are equally good candidates to monitor changes in the composition of the gas phase of chemical products.The advantages of electronic noses over classic GC are their simpler use and higher speed. The greatest disadvantage is that the present compounds can not be analytically identified. Their suitability to the quality control of food, for example meat,4 edible oil5 or blueberries6 has already been studied.Other authors proposed electronic noses to supervise the quality of beer,7 wine,8 cork stoppers9 and alcoholic beverages10 or perfumes.11 Another study describes the application of a (FOX 4000) electronic nose to different foodstuff in the production floor.12 There are two articles describing the application of a (chemical sensor based) electronic nose to fuels.13,14 The first one discusses the qualitative distinction of several automotive fuels (with different octane values) and aviation fuels by aid of an array of Taguchi-sensors.13 The second one proposes electronic noses for forensic and law enforcement applications. 14 These authors investigated accelerant residues in fire debris with a commercial electronic nose consisting of 32 conducting polymer sensors (Aroma Scan Inc., Hollis, NH). Electronic noses and odour perception It is, however, important to keep in mind that an absolute odour description is not possible with electronic noses.Most of these devices have a restricted number of a few dozen sensors responding to a broad variety of chemical compounds. This is very different from the structure of the human nose. The latter has about one thousand different odour receptors, each of them selectively reacting with only a small number of specific molecules.15–18 Further, the molcules producing chemical sensor signals are in general not the same molecules being perceived by the human nose. It even happens that two different samples being well distinguished by the human nose show exactly the same signal pattern on an electronic nose.GC or GC-MS also encounter this problem. Some molecules defining the aroma of a product are present in extremely low concentrations (ppm or ppb) and can sometimes even not be detected by GC-MS. Of course, the reverse case also exists. It is even possible that strong chemical differences between two products can easily be detected by aid of an electronic nose whereas human experts hardly recognise any aroma difference. Additionally, an absolute odour definition does not exist in practice.The high dimensionality of odour components make the definition of a general odour standard complicated. To deal with this situation, sensory (panel) olfaction is in general based Analyst, 1999, 124, 1167–1173 1167on odour standards and odour mapping. A useful method, used by panels dealing with a great variety of products, is based on the so-called odour field.This odour map describes a smell by the aid of 44 odour descriptors.19 However, the main problem hindering an objective odour definition is that every human being perceives the smell of the same product in an individual way.20 The variability of the odour perception from one person to another is enormous. The results of sniffing panels consisting of at most a few dozen experts are, therefore, necessarily encumbered with considerable variations.20 The main advantage of electronic noses over expert panels is, besides their smaller cost and simplicity of use, that their signals are (at least in the short term) very reproducible.The results obtained with such instruments are further highly correlated with the composition of the volatile compounds of the samples investigated. It further happens that typical changes of the gas phase of a certain product are related to specific odour changes. There might even exist a strong correlation between the electronic nose data and the product’s aroma.In these cases, electronic noses could be very useful in the production control of chemicals or foodstuff. The goal of this study is two-fold. First, to find out whether and how far chemical differences of diesel fuels can be monitored by the aid of electronic noses. Second, to compare the performance of chemical sensors and mass spectroscopy with respect to this task. Experimental Chemical sensors The (first) instrument used was a FOX 4000 electronic nose (Alpha-M.O.S., Toulouse, France) equipped with 18 semiconductor sensors.The samples were passed by a HS500 autosampler (CTC Analytics AG, Zwingen, Switzerland) with a capacity of 50 samples. To provide a constant flux of a vector gas (humidified synthetic air) through the electronic nose, a humidifier (air condition unit ACU, model 1997, Alpha- M.O.S.) was used. The air condition unit was supplied with synthetic air (nitrogen–oxygen R80).The vector gas was passed via a 1 l buffer bottle to the electronic nose. A second exit tube of the bottle allowed the superfluous air to escape. The sensor baseline being sensitive to humidity and temperature, the whole set-up was mounted in an air-controlled laboratory with an ambient temperautre of 21 °C. In spite of the use of thermo-stabilised sensor chambers (each chamber contains six of the eighteen gas sensors), the sensor baseline was not stable. It showed up early that instrument maintenance was crucial to obtain reproducible results.Therefore, rigorous care was taken for the operational conditions of the whole equipment: The gas tightness of all tube connections of the ACU and FOX 4000 units was verified. This was done by the aid of an alcohol soaked piece of cleaning paper serving as a probe. During (manual) data acquisition without sample injection the so prepared probe was passed near the tube connections. As most of the sensors are very sensitive to alcohol vapour, they signaled the presence of even extremely small leaks.The vector gas buffer bottle, as well as all the gas tubes, were thermo-isolated with expanded polystyrene. This avoided abrupt baseline shifts of the semiconductor gas sensors due to ambient temperature changes, eventually caused by air draughts. The baseline behaviour was further regularly screened by long-term (e.g., over night) data acquisitions without injection.These measures were crucial to obtain reproducible results. They allowed the reduction of sensor baseline variations by more than a factor of five. At the end, there remained a micro-leak in the first of the three sensor chambers. It was not repaired with respect to cost and time. After optimisation of the signal-to-noise ratio with some of the diesel samples the experimental parameters were fixed as follows: The vector gas flux was set to 300 ml min21, its relative humidity was regulated between 27.2 and 29.5%, its temperature varied between 49.7 and 50.2 °C.By the aid of its builtin vacuum pump the FOX 4000 sucked in 250 ml min21 of this vector gas. The samples were 10 ml gas vials filled with 100 ml of diesel fuel. The sample preparation (five vials per diesel type) was carried out under a fume hood a few hours before the experiment was started. The samples were placed on the tray of the HS500 auto-sampler in arbitrary order. The automatic injection unit heated the samples to 35 °C with an incubation time of 30 min.The temperature of the injection syringe was 40 °C. The syringe was filled two times with a delay of six seconds before injection of 400 ml of the sample headspace into the instrument. This injection volume was chosen to obtain a maximum sensor response between 0.7 and 0.8 units. This represents the range where the signal is considerably high compared to baseline fluctuations but still avoids sensor saturation. It should, thus, provide an optimal signal-to-noiseratio (SNR) and make the application stable against perturbations.The syringe fill speed and injection speed were 100 and 1500 ml s21, respectively. After each injection the syringe was baked out to 50 °C for 4 min and then flushed with nitrogen (N 50) for 6.5 min. The delay time between two injections was 1 h 30 min. This comparatively long interval was necessary because the return of some of the gas sensors to their initial values was extremely slow.The volatile compounds contained in the headspace obviously caused a long desorption process. This may be related to the viscous nature of diesel fuel mainly consisting of long-chained compounds with 10 to 20 C-atoms. The 18 sensors of the FOX electronic nose are listed in Table 1. We only used (doped) metal oxide sensors because they have a higher long-term stability and are less sensitive to humidity fluctuations than their conducting polymer based homologous compounds.The sensor signals were recorded over 600 s with an acquisition point each second. Mass spectrometer The Smart Nose GA 200 system uses a (Balzers instruments, Balzers, Liechtenstein) quadrupole mass spectrometer with 200 channels (1 to 200 u) as the detection unit. The measurements were carried out by LDZ Laboratory, Switzerland. The samples were 20 drops (abot 180 mg) of diesel fuel filled in 10 ml vials. The samples were incubated at 50 °C during 5 min (with agitation).The syringe and injection temperatures were 100 and 120 °C, respectively. After each injection of 2.5 ml of sample headspace, the syringe was purged for 7 min and the injector for 10 min with nitrogen (purity 99.95%) at 0.5 bar. The acquisition time per sample was 10 min. For data examination, three scans (from 10 to 199 u) per sample were averaged. Three samples per diesel type were measured. The most important system parameters are listed in Table 2, which equally gives an overview of both systems’ performance (discussed below).Table 1 Metal oxide sensors of the three chambers (A, B and C) of the FOX 4000 electronic nose Position Chamber A Chamber B Chamber C 1 T30/1 P30/1 SY/LG 2 P10/1 P40/2 SY/G 3 P10/2 P30/2 SY/AA 4 P40/1 P70/1 SY/Gh 5 T70/2 T40/1 SY/gCTl 6 PA2 TA2 SY/gCT 1168 Analyst, 1999, 124, 1167–1173Samples The samples were diesel fuels obtained from three different refineries in France: Feyzin (FZN), Donges (DGS) and Grandpuits (GPS).There are four different fuel types: standard diesel fuel (Di); diesel for hot temperatures, called summer diesel (S); diesel fuel for extremely cold temperatures, called winter diesel (W) and a laboratory synthesised low sulfur derivative. Winter diesels contain less long-chained hydrocarbons to guarantee a reliable motor start also at cold temperatures. The fuels’ origins and designations are listed in Table 3. Most of the fuels were normal diesels. Winter diesels were only provided from the refinery at Feyzin, summer diesels only from Feyzin and Grandpuits (see also Table 4).The 20 diesels are named with a letter indicating the refinery (F, G or D) followed by a number between 01 and 20. The low sulfur derivative (a diesel with a sulfur content of 20 ppm) is named LSD. The latter kind of fuel will play an important role in future because European Union legislation requires a sulfur content of 50 ppm after the year 2005.Discrimination of diesel fuels Sensor signals 1 Signal intensity. Fig. 1 shows the sensor signals of three typical diesel fuels. The sensor responses are given in units of DRmax/R0, which is the maximum change of the sensor’s electrical resistance divided by the initial resistance. Only the absolute values are shown (sensors C2–C6 give negative response). We used DRmax/R0 for data examination because this value gives the most stable result and is more robust against sensor baseline variations.The three fuels F07, D10 and G19 were chosen because they come from three different refineries and exhibit quite different signal patterns. F07 gives the highest and G19 the lowest general intensity. The highest sensor responses are produced in the order C4 > B2 > C3 > B1 > C2 > A6. Sensor B2 has a particular pattern. This is the only sensor giving a nearly equal response for D10 and G19. All other sensors exhibit a pattern of the type F07 > D10 > G19. 2 Signal differences between diesels. We now concentrate on the differences between the signals of the three pairs of samples (I F07–G19, II F07–D10, III D10–G19). The greatest overall differences are observed on the sensors C4 > C3 > B1 > C2 > A6 > B2 (without figure). This corresponds to the same sensors listed before but in another order. These sensors are certainly important for the discrimination of the diesel fuels. 3 Standard deviation.Fig. 2 gives an overview of the standard deviation observed on the 18 sensors. The most important variances are observed on B2 > A6 > B1 > C3 > A1 > C2. This again resembles the same list as before with A1 replacing C4. The lowest variances are observed on B3, C1 and C6, which are also the sensors with the weakest absolute signal. The F07 sample, which is the sample producing the strongest overall signal, exhibits the most severe standard deviation on most of the sensors.Only C3 and C4 behave in the opposite way. This conflicts with the assumption that a higher overall signal should give a better S/N on all sensors. A possible reason for this effect may be that on some of the sensors the chemical desorption process is less reproducible at higher signals (sensor saturation). This may in particular be the case for sensors A1, B1 and B2, which have a considerable relaxation time. These three sensors need about one hour to return to a DRmax/R0 signal of 0.02.The value 0.02 corresponds to about 18, 9 or 5% of the maximum sensor response for A1, B1 and B2, respectively. This effect could (under laboratory conditions) contribute more to the signal variance than the baseline fluctuations. There is Table 2 Response properties of the 18 chemical sensors with respect to the analysed diesel samples. The importance of each sensor is marked with numbers, 6 representing the most important sensor Chamber A Chamber B Chamber C Sensor property 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 Sensor response 1 3 4 2 5 6 2 Response difference 2 4 1 3 5 6 3 Standard deviation 2 1 5 4 6 3 4 Signal-to-noise ratio 1 3 5 4 2 6 5 Separation power 1 4 2 5 6 3 6 Contribution to PC1 1 3 5 2 4 6 7 Contribution to PC2 1 1 3 6 4 5 2 Used for LDA Model X X X X X X Table 3 Origin and designation of the diesel fuels useda Refinery Batch Type Date Name FZN 368 S 18/12/97 F01 FZN 369 W 18/12/97 F02 FZN 369 W 16/12/97 F03 FZN 368 Di 16/12/97 F04 FZN 369 W 04/12/97 F05 FZN 368 Di 08/12/97 F06 FZN 368 Di 03/12/97 F07 FZN 369 W 10/12/97 F08 FZN 368 Di 10/12/97 F09 DGS Di 28/11/97 D10 GPS S 25/11/97 G11 GPS S 02/12/97 G12 DGS Di 25/11/97 D13 DGS Di 04/12/97 D14 GPS Di 10/12/97 G15 GPS Di 18/11/97 G16 GPS Di 04/11/97 G17 GPS Di 23/10/97 G18 GPS Di 29/10/97 G19 DGS Di 12/12/97 D20 Low sulfur derivative LSD a Production sites: FZN = Feyzin, DGS = Donges, GPS = Grandpuits. Types: Di = standard diesel fuel, S = summer diesel, W = winter diesel. Table 4 Number of normal (Di), summer (S) and winter (W) diesels provided from the three refineries Feyzin (FZN), Donges (DGS) and Grandpuits (GPS) FZN DGS GPS Di 4 4 5 S 1 0 0 W 4 0 2 Analyst, 1999, 124, 1167–1173 1169probably still some scope for noise optimisation by reducing the injection volume and, thus, the signal intensity and the relaxation time.On the other hand, such a measure would reduce the system’s robustness against baseline perturbations. 4 Signal-to-noise ratio.A good measure of the signal-tonoise ratio (SNR) provided by each sensor is the mean value of repeated measurements divided by the standard deviation. Sensor C4 gives the best performance, followed by A4 and, after a significant distance, by B1 > A2 > B2 > A1 (without figure). 5 Separation power. Even more important for the distinction of two samples A and B is the ratio of the signal difference (SD) divided by the standard deviation (s), or more precisely the parameter SDA 2 D/(sA + sB).In Fig. 3, the absolute value of this quantity is shown for the three pairs of diesel samples already discussed (I F07–G19, II F07–D10, III D10–G19). This quantity should be directly correlated with the separation power of the sensors. The best performance (for all three pairs of diesels) shows C4, followed, after a significant gap, by C3 > C1 > C5 > C2 > B3. The sensors with the worst performance are B2 and B4.Principal component analysis (PCA) Fig. 4 shows a principal component scores plot of all samples with all sensors used (PCl 86.5%, PC2 6.2%, PC3 2.6%). The strong dominance of PCl shows that the PCA model is nearly one-dimensional. This is in accordance with the observation made on Fig. 1 that, except sensor B2, all sensors behave in a very similar way. The measurement repeatability for each individual diesel is satisfactory. However, the variance is not uniform and often generates inner-group variations (e.g., G11, D14, G15).It is obvious that the fuels show signal preferences characteristic of their origins. The diesels also show an important variation within their (refinery) groups, depending on fuel type and sampling day. This causes an overlap between the diesels of the three production sites. Only the LSD sample, which is of a particular nature, is well distinguished from the other three groups. The measurement of each individual sample is quite reproducible but with a considerable variance in the direction of PC2 (for an explanation see below). The distinction between all the samples is not possible.This is not surprising because some of the samples even come from the same batches only with a sampling delay of a few days. There are, however, some pairs of similar samples. These are from the same batches and have a small sampling delay. Some examples are F02 + F03, F06 + F09, F05 + F08, D10 + D13.Diesels coming from the same batch are in general grouped together. This is, for example, the case for batch FZN-368. The samples F01, F04, F06 and F09, taken between December 8 and 18, are quite close one to the other. Only F07, which also came from this batch but is from December 3, lies far outside. It is probable that either production parameters or the crude source have changed between December 3 and 8. The diesels from batch FZN-369 (F02, F03, F05, F08), which represent another type (winter diesel), are found to be similar to those of batch FZN-368.Thus, the kind of product (Di, S or W) obviously has a less important influence on the sensor’s response than the production site. This is also true for the different products from the refinery at Grandpuits (Di and S), which are all grouped together. Since the data set was too small and very unbalanced with respect to the numbers of the three Fig. 1 Semiconductor sensor responses for three typical diesel samples.The units given are the (maximum of the) electrical resistance change of the chemical sensors divided by the initial resisitance. Fig. 2 Standard deviation of the sensor responses for three typical diesel samples. Fig. 3 Sensor separation power for three pairs of typical diesel fuels. Fig. 4 Principal component scores plot of the diesel data as obtained with the FOX 4000 instrument (five repetitions, all sensors used). F plots = FZN (Di/S/W), G plots = GPS (Di/S), D plots = DGS (Di), LSD as indicated. 1170 Analyst, 1999, 124, 1167–1173kinds of products Di, S and W (see Table 4) we did not investigate further if a distinction between these groups is possible. As already mentioned above, there are important variances in the data, mainly in the direction of PC2. It is, therefore, interesting to have a look at the sensors contributing to the model. Fig. 5 shows the composition of the first two principal components (loadings).PC1 has important contributions from sensors C4 > B2 > C3 > C2 > B1 > A6. PC2 is dominated by sensor B2 with only minor contributions from B1 and C2– C4. As already mentioned in the discussion of Figs. 1 and 2, sensor B2 plays a special role. We reperformed the PCA without it and obtained a nearly one-dimensional scores plot with only PC1 being important (figure not shown). Nothing but noise was found on PC2 under such conditions. Sensor B2 is, thus, indispensable for the distinction of the diesels.A principal component scores plot based on the data of only B1, B2 and C2–C4 (not shown) does not differ much from that with all sensors included. The number of sensors could even be further reduced with only sensors B2 and C4 being essential. Sensors B1, C2, C3 and C4 exhibit a high signal-to-noise level, whereas sensor B2 is among those with the strongest noise contribution. An optimisation of the results would in the first instant aim at noise reduction of this sensor by adjusting the experimental parameters (e.g., the injection volume).Linear discriminant analysis An important question was whether the determination of the production site was possible with these electronic nose data. For that we needed to apply linear discriminant analysis (LDA). To obtain a stable and reliable LDA model, we needed to use only the best performing sensors, so we summarised all the previous observations made concerning the sensors (see Table 2). Listed are the number of the corresponding figure, the sensor property and the intensity of the property.All quantities listed, except the standard deviation (line 3), are positive properties. Some of the sensors occur in nearly every line (B1, B2, C2–C4) while others are not present at all (A3, B4–B6). The most important sensors are obviously C4 (strongest signal, highest S/N, strongest separation power and highest contribution to PC1) and B2 (the worst standard deviation but main contribution to PC2).A1 and A6 may be important because they contribute to the PCA model. The last line indicates the sensors chosen to compute the LDA model. The sensors selected were A6, B1, B2 and C2–C5. To compute the model, we defined four groups corresponding to the three refineries and the LSD sample. The corresponding discriminant function plot is shown in Fig. 6. There is a distinct separation of the four groups now.However, LDA is combined with the risk of being less reproducible than a PCA based model. Therefore, we verified the stability by recomputing the model randomly excluding 30% of the data (30% of each of the four groups). The classification of the omitted data by Mahalanobis distance analysis (MDA) gave 100% correct classification. Further, the landscape of the LDA plot obtained with this reduced data set (not shown) did not visibly differ from that with all data included. Leave-one-out and leave-segment-out cross correlation equally gave 100% correct classification. That is why, under the assumption that the analysed fuels are representative for all the diesel products of these refineries, this model is very reliable.We verified the importance of each sensor by recomputing this LDA model several times, each time leaving out one of the sensors. The result was that no further sensor could be omitted without significantly loosing separation power (except probably sensor B1).The addition of any further sensor did not improve the separation of the four groups defined. We also recomputed the LDA model with all sensors included. The model obtained under these conditions was mainly based on sensors A4, A5, B5, B6 and C6. These sensors are not among those of high reproducibility (see Table 2) and, thus, bring in noise. As a consequence, the corresponding cluster plot (not shown) exhibited larger and less separated clusters.This underlines the importance of an appropriate sensor selection for a reliable LDA model. Total sulfur and total nitrogen content The odour of diesel fuel is said to be mainly related to the contained sulfur or nitrogen compounds. That is why we determined some of the (liquid) samples’ total sulfur and total nitrogen content. The methods used were UV fluorescence and chemiluminescence, respectively. The results are shown in Fig. 7. Nearly all the refinery products show a sulfur content between 400 and 550 ppm.The only exception is the low sulfur derivative LSD (20 ppm). The sulfur content is obviously not characteristic of the production site but the nitrogen content is. The four samples from Feyzin ( > 150 ppm of nitrogen) are well separated from the rest ( < 100 ppm). There are also (less significant) differences between the fuels from the refineries at Donges and at Grandpuits. There is some similarity between the sulfur/nitrogen map and the PCA plot discussed before. We did not investigate further this phenomenon.For a more detailed study it would be interesting to analyse the correlations observed between the sulfur and nitrogen content and the chemical sensor signals. It could also be envisaged to analyse the fuels’ gas phase by GC-MS and to correlate that data with electronic nose data. Fig. 5 Loadings of the first two principal component vectors. Fig. 6 Linear discriminant score plot of the diesel data as obtained with the FOX 4000 instrument (five repetitions, only sensors A6, B1, B2 and C2–C4 used).F plots = FZN (Di/S/W), G plots = GPS (Di/S), D plots = DGS (Di), LSD as indicated. Analyst, 1999, 124, 1167–1173 1171Mass spectrometry results Typical spectra from diesel fuels, as obtained with the Smart Nose instrument are presented in Fig. 8 as a line plot. A logarithmic axis scale was used to provide the full plot in one figure. The mass spectra (10 to 199 u) are normalised on mass 40, which corresponds to argon.Significant differences between the three diesels shown (F07, G19, D10) are observed between 50 and 150 u. The region below 50 u is dominated by atmospheric gases [e.g., masses 18 (H2O), 28 (N2), 32 (O2), 40 (Ar) and 44 (CO2)] and is therefore less discriminant. Beyond 170 u there is hardly any signal response. Before performing a PCA, the whole raw data set was normalised on mass 40. Then, masses 10 to 65 (strong intensity, impurities) and 174 to 199 (less discriminant) were removed.The PCA was thus performed on masses 66 to 173 u without further optimisation by feature selection. The used mass range is illustrated in Fig. 9 (linear axis scale). Significant differences between the three diesel fuels are observed at masses 69–71, 77–79, 82–84, 91, 105, 119 and 120. The corresponding principal components scores plot is presented in Fig. 10. The repeatability (three repetitions per fuel) of the measurement of the individual samples is much better than that found with the chemical sensors.The fuels form separate clusters according to their origins. It should, therefore, be possible to discriminate the three refineries. However, there is still an important variance within each of the clusters (especially in the cluster of Feyzin) being bigger than the intercluster distances. There is obviously a strong correlation between the data obtained with the chemical sensor based FOX 4000 and these mass spectrometric data.The overall landscape of both scores plots is nearly the same. As observed with the FOX electronic nose, LSD and F07 seem to be quite different from the other samples. In both figures LSD, F07 and G12 are found on the extreme edges of the cluster plot. They define a triangle surrounding most of the other samples. The samples found close to each other with the FOX instrument are also close to each other in the mass spectral data. Within these mass spectrometric data it even seems possible to distinguish the different types of diesel fuel (Di, S, W) within the Feyzin cluster.The fuel types Di and S (F1, F4, F6, F7 and F9) and W (F2, F3, F5 and F8) are well grouped together. However, the number of samples from Feyzin were only 4 Di, 1 S and 4 W. Due to this small and not equally distributed data base, we did not try to give a definitive answer to this question. However, the differences between the fuel types are less important than the differences between the refineries.This is even more evident on the data from Grandpuits where the two summer fuels G11 and G12 are found on the opposite sides of the cluster. Summary The measurement parameters used and the results obtained with the two measurement systems are summarised in Table 5. The sampling and injection parameters have already been discussed before. An important parameter for an application in product testing is the delay time between two injections.In the case of diesel fuels (slow and unequal desorption from chemical sensors) there is a significant difference between the chemical sensor technique with 1.5 h delay time and only 10 min with the Fig. 7 Total nitrogen and total sulfur content of some of the diesel fuels. F plots = FZN, G plots = GPS, D plots = DGS, LSD as indicated. Fig. 8 Mass spectral (line) plot (logarithmic scale) of three typical diesel samples. The spectra are normalised on mass 40.The mass range from 66 to 173 (vertical bars) was used for PCA analysis. Fig. 9 Mass spectral range used for PCA analysis (linear scale) of three typical diesel samples. The spectra are normalised on mass 40. Fig. 10 Principal component scores plot of the diesel data as obtained with the Smart Nose instrument (three repetitions, data normalised on mass 40, masses 66 to 173 used). F plots = FZN (Di/S/W), G plots = GPS (Di/S), D plots = DGS (Di), LSD as indicated. 1172 Analyst, 1999, 124, 1167–1173mass spectrometer.The measurement repeatability is much better for the mass spectrometric system. This is mainly due to the observed drift and instability of the chemical sensors. A PCA computed on the data set obtained with the FOX 4000 electronic nose did not allow the three refineries to be distinguished. However, an LDA model, optimised by careful sensor selection (6 of the 18 sensors) allowed the correct discrimination of the diesels according to their different origins.Without knowing which diesels belong to which refinery, it would not have been possible to define this model. A discrimination by PCA between the three refineries was easily obtained with the mass spectrometric data. A reliable distinction of the three refineries should be possible. Further, important requirements for application of an electronic nose in a production site are long term calibration and simplicity of use. The stability of mass spectrometers has been proven and optimised over many years.It is so far not sure whether chemical sensors can be efficiently and reliably recalibrated after a longer working period. Another drawback of chemical sensors are their sensitivity to humidity and temperature. This makes a careful and time consuming maintenance (including thermo-isolation of the system) necessary. An equilibrium time of several hours is necessary after system switch-on or change of the vector gas bottle (every three weeks).Conclusions After a complex hardware optimisation, the eighteen chemical sensors of the FOX 4000 instrument allow the detection of differences between most of the twenty diesel fuels analysed. After sensor selection, linear discriminant analysis (LDA) allowed us to correctly distinguish fuels according to their different origins (three refineries). Strong similarities were found between these results and those obtained from mass spectrometry (MS).The mass spectral measurements were faster and more repeatable than those performed with the chemical sensors. The data from Smart Nose allowed a good distinction between the three refineries even with principal component analysis (PCA). It is more convenient to work with a mass spectrometric electronic nose than with chemical sensor based instruments. Mass spectrometers are more reliable and robust than chemical sensors. There is much less instrument maintenance with MS sensors.Further, long term drift, being a principal problem of chemical sensors, is expected to be less of a problem on mass spectrometer based electronic noses. The advantages of MS over chemical sensors found in this study are certainly not valid for all kinds of chemical products. Other products may exhibit stronger signals on chemical sensors than on mass spectrometers. The chemical desorption process from the sensor surface is in general much faster, bringing the measurement period down to about 15 min.In such cases, and if a long term sensor calibration is possible, chemical sensor based electronic noses can still be an alternative to mass spectrometers. Acknowledgements We wish to thank Dr François Hartmann, Dr Laurent Germanaud, Dr Franck Eydoux (Centre de Recherches d’Elf Solaize) and Prof. P. McLeod (Ecole Pratique des Hautes Etudes, Massy) for support and productive discussions. We further thank Dr Zesiger (LDZ Laboratory, Switzerland) for his kind cooperation.References 1 American Society for Testing & Materials, Philadelphia, 1993a and 1993b (ASTM 5186-91). 2 American Society for Testing & Materials, Philadelphia, 1992b (ASTM D2425-88). 3 American Society for Testing & Materials, Philadelphia, 1992a (ASTM D1319-89). 4 F. Winquist, E. G. H�ornstein, H. Sundgren and I. Lundstr�om, Meas. Sci. Technol., 1993, 4, 1493. 5 S. Bazzo, F. Loubet, T. Tan, J. D. Hewett-Jones, C. E. M. Engelen- Cornax and J. F. A. Quadt, Semin. Food Anal., 1998, 3(1), 15. 6 J. E. Simon, A. Hetzroni, B. Bordelon, G. E. Miles and D. J. Charles, J. Food Sci., 1996, 61(5), 967. 7 T. C. Pearce, J. W. Gardner, S. Friel, P. N. Bartlett and N. Blair, Analyst, 1993, 118, 371. 8 C. Di Natale, F. A. M. Davide, A. Amico, P. Nelli, S. Gropelli and G. Sberveglieri, Sens. Actuators B, 1996, 33(1–3), 83. 9 S. Rocha, I. Delgadillo, A. J. Ferrer Correia, A. Barros and P. Wells, J. Agric. Food Chem., 1998, 46, 145. 10 H. Nanto, K. Kondo, M. Habara, Y. Douguchii, R. I. Waite and H. Nakazumi, Sens. Actuators B, 1996, 35–36, 183–186. 11 T. Carrasco, C. Saby and P. Bernadet, Flavour Fragrance, 1998, accepted. 12 K. J. Strassburger, Semin. Food Anal., 1998, 3, 5. 13 B. S. Hoffheins and R. J. Lauf, Analusis, 1992, 20, 201. 14 S. A. Barshick, W. H. Griest and A. A. Vass, SPIE, 1997, 2941, 63. 15 R. Axel, Sci. Am., 1995, Oct. 95, 130. 16 C. Dulac, Neuron, 1997, 19, 477. 17 K. J. Ressler, S. L. Sullivan and K. B. Buck, Cell, 1993, 73, 597. 18 R. Vassar, J. Ngai and R. Axel, Cell, 1993, 74, 309. 19 J.-N. Jaubert, G. Gordon and J-C. Dor�e, Parfums, Cosm�et., Ar�omes, 1987, 78, 71. 20 N. Neuner-Jehle and F. Etzweiler, in Art, Science and Technology, ed. P. M. M�uller and D. Lamparsky, Elsevier, 1991, 153. Paper 9/02126D Table 5 Comparison of the parameters used and the performance of the two systems with respect to the discrimination of diesel fuels System parameter or property FOX 4000 Smart Nose Detection principle Chemical sensors Mass spectrometer Injection technology Syringe injection Syringe injection Sample volume 0.1 ml ~ 0.2 ml Incubation temperature 35 °C 50 °C Incubation time 30 min 5 min Headspace injection volume 100 ml 2500 ml Time delay between two injections 1.5 h 10 min Analytical information used 18 Sensor signals Mass spectra 10–199 u Measurement repeatability of individual samples (+) + Discrimination of refineries achieved? Yes (with LDA) Yes (with PCA) System long term calibration possible? ? + Simplicity of use and maintenance — + System equilibration time after system start 5–10 h 1–2 h Analyst, 1999, 124, 1167&nda

ISSN:0003-2654    

DOI:10.1039/a902126d

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Experimental demonstration and simulation of electrochemical non-linear responses to glucose and its interferents with an amperometric sensor Satoshi Nakata,*a Hironori Yabuuchi,a Rie Takitani,a Yoko Hirataa and Yoshitaka Masudab a Department of Chemistry, Nara University of Education, Takabatake-cho, Nara 630-8528, Japan. E-mail: nakatas@nara-edu.ac.jp b Department of Chemistry, Faculty of Science, Kobe University, 657, Japan Received 21st April 1999, Accepted 14th June 1999 A novel sensing system based on the multi-dimensional information contained in a dynamic non-linear response is proposed.A sinusoidal potential was applied to an amperometric-type glucose sensor and the resulting current of the sensor was analyzed by fast Fourier transformation (FFT). The amplitudes of the higher harmonics of FFT characterize the non-linear properties of the response. The amplitudes of the higher harmonics of FFT exhibit characteristic changes which depend on the concentration and the kinetics of the reactions of glucose and its interferents at the sensor surface.The essential features of the current–potential curve were reproduced by a computer simulation based on the kinetics of electrochemical reactions. Introduction Amperometric-type glucose sensors, which are used in the medical and scientific fields, have been developed to achieve high selectivity for glucose molecules.1–4 However, such glucose sensors are not very selective for glucose, since they also respond to interferents, such as ascorbate, urate ion and pacetamidophenol, which may be oxidized at the electrode surface, by a diffusion mediator or by enzyme-bound electron relay.3,5,6 Although several attempts have been made to overcome this problem (e.g., modification of the surface to decrease the permeability to organic compounds and lowering the electrode potential by using an electron-transfer mediator), 1–13 it is difficult to reduce the effects of interferents sufficiently solely on the basis of one-dimensional static information (dc voltage and dc current) obtained with a single detector.Hence, it may be useful to develop a strategy other than chemical sensing based on one-dimensional static information. Recently, we described novel sensing systems with an electrode or a semiconductor gas sensor based on the multidimensional information contained in dynamic non-linear responses.14–18 In these systems, a sinusoidal input signal is applied to an experimental system and the resulting output signal is analyzed by fast Fourier transformation (FFT).The dynamic non-linear response is quantitatively evaluated in terms of the amplitude of the higher harmonics of FFT. We found that the amplitude of the higher harmonics changes characteristically, depending on the chemical structure and concentration of the chemical species. That is, the higher harmonics reflect information on the reaction kinetics, e.g., the activation energy, the reaction rate, the diffusion rate and the reaction mechanism.In the measurement of voltage-dependent capacitance, the higher harmonics gives information on dynamic adsorption processes and the adsorption state of chemical species on the electrode surface.14 In the present study, the characteristic responses of the current–potential curve for glucose and its interferents were reproduced by a computer simulation based on the kinetics of the chemical species on the electrode surface.Experimental A sinusoidal potential [f = 0.08 Hz; 0.6 + 0.2cos2pft(V)] was generated by a potentiostat (NPGS-2501, Nikko Keisoku Corp., Kanagawa, Japan) connected to a waveform generator (FG120, Yokogawa Electric Corp., Tokyo, Japan), and applied to an amperometric sensor.17,18 The input sinusoidal potential and the output current were stored in a computer, and the time trace of the conductance was then Fourier transformed to the frequency domain, as shown in Fig. 1. The initial value for FFT was the current at the maximum potential, and the four cycles of the conductance, for which the sample number was 2048, were analyzed by FFT. Mathematica (Wolfram Research, Inc., IL, USA) was used to perform Fourier analysis and calculate the differential equations. The amperometric sensor consisted of a Pt disk electrode (diameter 1.5 mm) as the working electrode, a Ag/AgCl electrode as the reference electrode and a Pt wire (diameter 0.5 mm, length 20 mm) as the counter electrode.Hydrogen peroxide, which was oxidized to proton and oxygen on the hydrogen peroxide detector, was used as the mediator to Fig. 1 Schematic representation of the measurement for evaluating electrochemical non-linearity. Experimental and analytical profiles are as follows: (1) the sinusoidal potential [0.6 + cos 2pft (V), f = 0.08 Hz) is applied to the electrochemical system; (2) the output current is recorded; and (3) the time trace of the current is Fourier transformed to the frequency domain.Higher harmonics on a non-linear system appear with the application of a sinusoidal potential with a single frequency. Note that no higher harmonics appear in a linear system. Analyst, 1999, 124, 1175–1179 1175quantify the concentration of glucose. A filter membrane (diameter 5 mm, pore size 5 mm, thickness 0.16 mm) was immersed in 0.1 M phosphate buffer solution (PBS) (pH 7.0, volume 1 ml) with 5 mg ml21 glucose oxidase for 1 h.The filter membrane with the glucose oxidase (GOD) was attached to the Pt disk surface and then covered with a semipermeable cellulose membrane (pore size 2.5 mm, thickness 2.3 mm) to immobilize the membrane. Glucose and its interferents (sodium ascorbate and sodium urate) were dissolved in 0.1 M phosphate buffer solution. All measurements were performed while stirring the test solution at 313 ± 0.5 K. Glucose oxidase (EC 1.1.3.4, 150 units mg21, from Aspergillus niger) and other reagents were purchased from Sigma (St.Louis, Mo, USA). Results and discussion 1 Characteristic current–potential curves for glucose and its interferents and their quantitative characterization based on higher harmonics Fig. 2 shows the output current (mA) versus the input potential (V) for (a) PBS without an analyte, (b) glucose (0.5, 1, 3 and 5 mM), (c) sodium urate (0.3, 0.5, 1 and 2 mM), and (d) sodium ascorbate (0.5, 1, 3, and 5 mM).The current–potential curves for (b), (c), and (d) were constructed by subtracting that for phosphate buffer solution (a). The capacitance component of the current–potential curve for the PBS is due to the electrical double layer of ions formed around the electrode surface.14 The current–potential curve for glucose is different from those of the interferents. Hence, it is possible to discriminate among glucose and these interferents using only a single electrode based on their dynamic electrochemical responses. To evaluate qualitatively the electrochemical non-linear responses, the relative amplitudes of the higher harmonics in FFT of the output current were examined (Fig. 3). Rn denotes the real part (cosine function) of the nth harmonic in FFT and P1(PBS) is the power spectrum {P1(PBS) = [R1(PBS) 2 + I1(PBS) 2]1/2} of the fundamental harmonic (frequency 0.08 Hz) for PBS [I1(PBS): the imaginary part (sine function) of the fundamental harmonic for the PBS].In Fig. 3, R2 for each sample with PBS subtracted by that for PBS without an analyte, R2 2R 2(PBS) [R2 = real part of second harmonics for each sample with PBS, R2(PBS) = real part of second harmonics for PBS without an analyte], was plotted to detect only the response for each sample by eliminating the characteristic attributable to the current versus potential curve for PBS [Fig. 2(a)]. Rn/P1 (n = 1 or 2) changes characteristically with increase in concentration.These results indicate that it is possible to determine the concentration of glucose in a test solution based on the information found in the higher harmonics of the sensor response. For example, the relative value of R2 for glucose decreased with increase in concentration while those for other substances increased. Therefore, it is possible to distinguish among glucose and the other substances. As for the relative value of R3, the response for urate was enhanced. Hence the current–potential curves for glucose and its interferents can be quantitatively characterized by their higher harmonics.To minimize the experimental error in the determination of concentrations, the amplitudes of the higher harmonics are given as normalized values relative to P1. With the normalization procedure, the errors in the relative amplitudes were found to be less than 5% for six measurements of glucose responses with different GOD membranes and different GOD reagent bottles and to be less than 5% for all of the other measurements. Impedance measurement is generally carried out by assuming a ‘linear system’, i.e., the equivalent circuit for evaluating an electrochemical system consists of linear capacitors and linear resistors.No higher harmonics of the linear system appear for the application of potential with a single frequency (Fig. 1).14,18 For ac voltammetry, the linear sweep potential of a small sinusoidal signal (amplitude 5 mV, frequency 10 Hz–1 kHz) is applied to the electrode cell.19–21 In applying the smallamplitude signal, non-linearity or a higher harmonic is only obtained under quasi-steady state conditions.In contrast, information from dynamic electrochemical non-linearity can be obtained by applying a sinusoidal potential with a large amplitude (200 mV) and a lower frequency (0.08 Hz) (Fig. 1). In order to acquire the characteristic non-linear responses, we examined several experimental conditions.The degree of hysteresis, which corresponds to the imaginary part, on the current–potential curve decreased when f < 0.02 Hz. In contrast, the characteristic response on the current versus potential curve decreased when f > 10 Hz. These results suggest that the surface concentration of the chemical species Fig. 2 Output current (mA) versus the input potential (V) for (a) PBS without an analyte, (b) glucose (0.5, 1, 3 and 5 mM), (c) sodium urate (0.3, 0.5, 1 and 2 mM) and (d) sodium ascorbate (0.5, 1, 3 and 5 mM).The current versus potential curves for (b), (c) and (d) were constructed by subtracting that for PBS (a). The arrow in each figure indicates the direction of time in the current versus potential curve and the numbers beside the curves denote the concentrations (mM) of the samples. Fig. 3 Relative amplitudes of the higher harmonics in FFT of the output current for glucose (2), sodium ascorbate (½) and sodium urate (8). (a) [R2 2 R2(PBS)]/P1(PBS) and (b) [R3 2 R3(PBS)]/P1(PBS). 1176 Analyst, 1999, 124, 1175–1179may approach that in the stationary state at lower frequency and that it is difficult to detect non-linearity in this system at higher frequency. A large amplitude of sinusoidal potential ( > 50 mV) is necessary to obtain non-linearity because the non-equilibrium state can be attained by the large amplitude of perturbation. In addition, the specimen–electrode boundary involving the membrane is a capacitor that will be more or less polarized and charged when a sweep potential is applied.Over this boundary, the sinusoidal wave is deformed, as suggested in the Gouy– Chapman theory.19 Hence, the nature of hysteresis induced at low frequencies is typically associated with the process within a membrane. We have reported the relationship between the electrochemical non-linearity and the differential capacitance. 14 2 Electrochemical significance of non-linear responses on an amperometric sensor These different non-linear responses may be due to the kinetics of the electrochemical reaction.At a high potential, the interferents in biological fluids are electrochemically oxidizable. However, not only is the enzyme reaction with glucose oxidase different from electrochemical oxidation of the interferents, but also the kinetics of electrochemical oxidation vary among the interferents. To clarify the physico-chemical significance of the nonlinear responses, we propose the following reaction-diffusion model for the response of the electrode.1–4,22–25 The kinetics for glucose may be expressed by dCO2(m)/dt = nO2(b) CO2(b) 2 nO2(m) CO2(m) + kb1CH2O2(m) 2 ke2CO2(m)CES(m) (1) dCglu(m)/dt = nglu(b) Cglu(b) + ke1CES(m) 2 ke21Cglu(m)[CE0(m) 2 CES(m)] (2) dCH2O2(m)/dt = ke2 CO2(m) CES(m) 2 kb1CH2O2(m) 2 nH2O2(m) CH2O2(m) (3) The oxidation of glucose includes the Michaelis–Menten kinetics, as indicated in eqns.(1) and (2).For simplification, we assume that the diffusion term of each substance is proportional to the concentration, i.e., DO2(b)”2CO2(b), Dglu(b) ”2Cglu(b), DO2(m) ”2CO2(m) and DH2O2(m)”2CH2O2(m) are replaced with nO2(b)CO2(b), nglu(b)Cglu(b), 2nO2(m)CO2(m) and 2nH2O2(m), CH2O2(m), respectively. The current, i, for glucose is obtained by the anodic current, ia, as expressed by eqn. (4) since the electrochemical oxidation of hydrogen peroxide is an irreversible reaction with the studied potential range, i.e., no cathodic current (ic Å 0):19 i = ia 2 ic Å ia = nFALkb1CH2O2(m) (4) The kinetics for urate may be expressed by dCurate(m)/dt = nurate(b) Curate(b) 2 kb2Curate(m) (5) The current, i, for urate is obtained by eqn.(6) since the electrochemical oxidation of urate is an irreversible reaction within the studied potential range (ic Å 0):19 i = ia 2 ic Å ia = nFALkb2Curate(m) (6) Here, we consider that Durate(b)”2Curate(b) is replaced with nurate(b)Curate(b) and that urate is electrochemically oxidized to allantoin26 at the sensor surface since the membrane attached to the electrode has an effective thickness of 160 mm.The kinetics for ascorbate may be expressed by dCasco(m)/dt = nasco(b) Casco(b) 2 kb3Casco(m) (7) The current, i, for ascorbate is expressed by eqn. (8) since the electrochemical oxidation of ascorbate is an irreversible reaction within the studied potential range (ic Å 0): i = ia 2 ic Å ia = nFALkb3 Casco(m) (8) Here, we consider that Dasco(b)”2Casco(b) is replaced with nasco(b)Casco(b).The rate constant for the electrochemical reactions, kbm (m = 1, 2 or 3), is generally expressed as a function of the electrode potential by kbm = kb 0 mexp[amnF(E2E0 m)/RT] (9) These reaction rates in eqn. (9) suggest that an infinite magnitude of current, |i|, is supplied by the application of a large potential, |E|. However, the current is generally limited owing to the surface area of the electrode, or mass transfer of the reactive compounds.22,23 To consider the limiting current, we assume that kbm (m = 1, 2 or 3) is replaced with kAbm, which is expressed as the logistic function in eqn.(10): ¢ = - - + - k k nF E E RT k nF E E RT k m m m m m m m m m m b b 0 b 0 b 0 0 b b 0 exp [ exp [ G a a G ( ) / ] ( ) / ] ( ) (10) where Gbm (m = 1, 2, or 3) denotes the constant for the logistic function in eqn. (10). When E 2E0 m Å 0 (m = 1, 2 or 3), kAbm = k0b m, i.e., the current obtained using eqn.(10) is similar to that obtained using eqn. (9) around the redox potential, E0. When E 2 E0 m (m = 1, 2 or 3) is highly positive, kAbm = Gbm. When E 2 (m = 1, 2 or 3) is highly negative, kAbm = 0. Fig. 4 shows a numerical simulation of the potential versus current curves based on eqns. (1)–(4) and (10) for (a) eqns. (5), (6) and (10) for (b) and eqns. (7), (8) and (10) for (c) when E is changed in a sinusoidal manner (E = 0.2cos2pft + 0.6 (V), f = 0.08).Here, kb1, kb2 and kb3 in eqns. (1)–(9) are replaced with kAb1,kAb2, and kAb3, in eqn. (10), respectively. In this simulation, the partition constants for O2, H2O2, glucose, urate and ascorbate are involved, but those for the other substances are neglected. We adopted the assumptions of kinetic and diffusionlimited model, i.e., constant enzyme activity and rate limiting concentrations of the substrate. In order to give more realistic parameters, electroactive parameters of H2O2 similar to those of urate were used because the current–potential curves for urate and H2O2 behave similarly.27 The characteristic potential versus current curves in the actual experiments (Fig. 2) can be quantitatively reproduced by a theoretical simulation based on the kinetics of the electrode reaction, i.e., the characteristic responses in Fig. 4(a), (b) and (c) may correspond to those in Fig. 2(b), (c) and (d), respectively. Fig. 5 shows the concentration-dependent nature of the higher harmonics of FFT [(a) R2 and (b) R3)] obtained by the theoretical simulation.The analyzed data correspond to the output current in Fig. 4. In comparison with the concentration dependence in Fig. 4, it is clear that the characteristic concentration-dependent nature of second and third harmonics among different chemical species can also be quantitatively and qualitatively reproduced by a numerical simulation, except R2 for urate and R3 for glucose.These results suggest that the potential versus current curves or the non-linear electrochemical responses change characteristically depending on the kinetics of the electrode reactions and the concentration of the chemical species. Conclusions Although our results and experimental system are only preliminary, it is apparent that the present method can provide abundant information based on the non-linear dynamics of the reactions of glucose and its interferents at the electrode.To improve the limit of detection (in this study, the limit of detection was ca. 0.1–10.0 mM owing to the saturation effect at a high concentration or competition with other chemical species) and to enhance the information available from the sensor, changing the scanning rate, coupling different kinds of sensors and evaluating the saturation effect may be useful for Analyst, 1999, 124, 1175–1179 1177measurements with a periodic change in the potential. Other higher harmonics (second and third harmonics in the imaginary part, fourth and fifth harmonics) also give us abundant information for molecular recognition (data not shown).The characteristic responses for these substances are physicochemically evaluated by considering the mass transfer and kinetics of each substance on the electrode surface, although the order of several parameters may be unrealistic (e.g., the diffusion constant of O2 is generally 1025). Further development of the simulation will not only theoretically clarify the dynamic non-linear responses on the electrode surface but also indicate better experimental conditions for enhancing the differences in the responses to glucose and its interferents.For example, we should consider including the chemical oxidation of ascorbate to dehydroascorbate with O2 and/or H2O2 produced by the biocatalytic cycle, and to include the pH dependence of the reaction rate on ascorbate in the simulation to reproduce the current–potential characteristics in our further study.In this work, we carried out preliminary experiments with a classical glucose sensor, i.e., a non-mediated system for basically characterizing non-linear electrochemical dynamic responses. Most of the mediators, such as ferrocene, potassium ferricyanide and ruthenium and osmium complexes, which have been used were chosen to minimize the effects from interferents. 1–4 The difference in the non-linear dynamic responses to glucose and its interferents should be enhanced with the application of the mediated systems in the present system, which may be promising future work.In addition, we demonstrated the ability to differentiate among individual compounds based on the experimental results and the theoretical simulation. The capability to determine glucose concentration even in the presence of interferences is very important for the practical application of the glucose sensor. We have reported that it is possible to determine the concentration of glucose even in the presence of an interferent at a low concentration, i.e., the dynamic non-linear responses of the mixture consist of the individual responses of the individual samples.18 However, we must consider theoretically and experimentally evaluating the saturation effect on the sensor response and the competition between glucose and the interference in the near future because the sum of the individual responses is not exactly equal to the responses of the mixture of samples at a high concentration.Here, we stress that the present technique should not be judged against traditional approaches but should be applied to enhance the selectivity in the traditional approaches. Fig. 4 Numerical simulation of the potential versus current curves based on eqns. (1)–(4), and (10) for (a) eqns. (5), (6), and (10) for (b) and eqns. (7), (8) and (10) for (c). The arrow in each figure indicates the direction of time in the current versus potential curve and the numbers beside the curves denote the concentrations (mM) of the samples.E = 0.2cos2pft + 0.6, f = 0.08, A = 1.0, n = 1, nO2(b) = 5.0, nO2(m) = 0.75, CO2(b) = 0.01, T = 298. (a) a1 = 0.3, nglu(b) = 0.17, ke1 = 0.2, ke21 = 0.1, ke2 = 12.0, E1 0 = 0.2, Gf1 = 1.0, Gb1 = 1.0, k0 b1 = 0.0005, CE0(m) = 2.0, nH2O2(m) = 0.03, KD(O2) = 0.12, KD(H2O2) = 1.0, KD(glu) = 1.0; (b) a2 = 0.25, nurate(b) = 0.35, E2 0 = 0.25, Gf2 = 1.0, Gb2 = 1.0, ko b2 = 0.0005, KD(urate) = 1.0; (c) a3 = 0.05, nasco(b) = 0.15, k0 b3 = 0.015, E0b 3 = 0, Gf3 = 1.0, Gb3 = 1.0, KD(asco) = 1.0.Fig. 5 Numerical simulation of the relative amplitudes of the higher harmonics in FFT [(a) R2 and (b) R3] of the output current depending on the concentrations obtained by the theoretical simulation. The data for glucose, urate and ascorbate correspond to (a), (b) and (c) in Fig. 4, respectively. 1178 Analyst, 1999, 124, 1175–1179Appendix Symbols A surface area of the electrode [m2] CX(b) concentration of X in the bulk phase [X = O2, glu (glucose), urate or asco (ascorbate)] [molm23] CX(m) concentration of O2 in the membrane [X =O2, glu (glucose), H2O2, urate, or asco (ascorbate)] [molm23] CE0(m) initial concentration of glucose oxidase in the membrane [molm23] CES(m) concentration of the complex ES which is composed of glucose and glucose oxidase in the membrane [molm23] DX(b) diffusion coefficient of X in the bulk phase [X = O2, glu (glucose), urate or asco (ascorbate)] [m2 s21] DX(m) diffusion coefficient of X in the membrane (X =O2 or H2O2) [m2 s21] E potential of the electrode versus the reference electrode [V] E0 m standard electrode potential of H2O2(m = 1), urate (m = 2), or ascorbate (m = 3) [V] F Faraday constant [C mol21] f frequency of the sinusoidal potential [Hz] kbm rate constant of the anodic reaction (‘backward’ reaction) for H2O2 (m = 1), urate (m = 2), or ascorbate (m = 3) on the electrode [s21] ke1 rate constant of the forward reaction on the formation of complex (ES) [s21] ke21 rate constant of the backward reaction on the formation of complex (ES) [m3 mol21 s21] ke2 catalytic rate constant on the breakdown of complex (ES) [m3 mol21 s21] kAbm rate constant of the anodic reaction for H2O2 (m = 1), urate (m = 2) or ascorbate (m = 3) on the electrode in eqn.(10) [s21] k0b m frequency factor of the rate constant of the anodic reaction for H2O2 (m = 1), urate (m = 2) or ascorbate (m = 3) [s21] KD(X) partition constant of X in the bulk/membrane phase = CX(b)/CX(m)][X = O2, H2O2, glu (glucose), urate or asco (ascorbate)] [–] L thickness of the membrane [m] n electrons per molecule oxidized [–] R gas constant [J mol21K21] T absolute temperature [K] am transfer coefficient on the backward reaction of H2O2 (m = 1), urate (m = 2), or ascorbate (m = 3) [–] nX(b) rate constant on the diffusion of X in the bulk phase [X = O2, glu (glucose), urate or asco (ascorbate)] [s21] nX(m) rate constant on the diffusion of X in the membrane (X = O2 or H2O2) [s21] GbX constant of the logistic function on kAbX (X = 1, 2 or 3) [s21] Acknowledgements We thank Professor S.Yamabe (Department of Chemistry, Nara University of Education), Professor Y. Yoshimi (Department of Industrial Chemistry, Shibaura Institute of Technology, Tokyo, Japan) and Professor K. Yoshikawa (Department of Physics, Kyoto University, Kyoto, Japan) for providing useful suggestions.This study was supported in part by a Grant-in-Aid for Scientific Research from the Ministry of Education, Science and Culture of Japan, the Nestlé Science Promotion Committee, the Shimadzu Science Foundation and a President Fellowship from Nara University of Education. References 1 Biosensors: Fundamentals and Applications, ed. A. P. F. Turner, I. Karube and G. S. Wilson, Oxford University Press, Oxford,1987. 2 R. P. Buck, W. E. Hatfield, M. Uma�na and E. F. Bowden, in Biosensor Technology: Fundamentals and Application, Marcel Dekker, New York, 1990. 3 F. Scheller and F. Schbert, in Biosensors, Elsevier, Amsterdam, 1992. 4 D. G. Buerk, Biosensors. Theory and Applications, Technomic Publishing, Lancaster, PA, 1993. 5 R. Maidan and A. Heller, J. Am. Chem. Soc., 1991, 113, 9003. 6 T. Ohsaka, Y. Yamaguchi and N. Oyama, Bull. Chem. Soc. Jpn., 1990, 63, 2646. 7 L. C. Guerente, A. Deronzier, P. Mailley, J.C. Moutet, Anal. Chim. Acta, 1994, 289, 143. 8 N. Oyama, T. Osaka, M. Mizunuma and M. Kobayashi, Anal. Chem., 1988, 60, 2534. 9 H. Liu and J. Deng, Biosens. Bioelectron., 1996, 11, 103. 10 A. E. G. Cass, G. Davis, G. D. Francis, H. A. O. Hill, W. J. Aston, I. J. Higgins, E. V. Plotkin, L. D. L. Scott and A. P. F. Turner, Anal. Chem., 1984, 56, 667. 11 A. L. Crumbliss, H. A. O. Hill and D. J. Page, J. Electroanal. Chem., 1986, 206, 327. 12 I. Willner and A. Riklin, Anal. Chem., 1994, 66, 1535. 13 J. Svorc, S. Miertus, J. Katrlik and M. Stredansky, Anal. Chem., 1997, 69, 2086. 14 S. Nakata, N. Kido, M. Hayashi, M. Hara, H. Sasabe, T. Sugawara and T. Matsuda, Biophys. Chem., 1996, 62, 63. 15 S. Nakata, S. Akakabe, M. Nakasuji and K. Yoshikawa, Anal. Chem., 1996, 68, 2067. 16 S. Nakata, E. Ozaki and N. Ojima, Anal. Chim. Acta, 1998, 361, 93. 17 S. Nakata, Y. Hirata, R. Takitani and K. Yoshikawa, Chem. Lett., 1998, 401. 18 S. Nakata, R. Takitani and Y. Hirata, Anal. Chem., 1998, 70, 4304. 19 A. J. Bard and L. R. Faulkner, Electrochemical Methods, Wiley, New York, 1980. 20 D. E. Smith, Anal. Chem., 1963, 35, 610. 21 D. E. Smith and T. G. McCord, Anal. Chem., 1968, 40, 474. 22 A. V. Sorkirko and F. H. Bark, Electrochim. Acta, 1995, 40, 1983. 23 T. Tatsuma and T. Watanabe, Anal. Chem., 1992, 64, 625 and 630. 24 B. R. Eggins, Biosensors: an Introduction, Wiley and Teubner, 1996. 25 U. E. Sprichiger-Keller, Chemical Sensors and Biosensors for Medical and Biological Applications, Wiley-VCH, Weinheim, 1998. 26 M. Jänchen, G. Walzel, B. Neef, B. Wolf, F. Scheller, M. Kühn, D. Pfeiffer, W. Sojka and W. Jaross, Biomed. Biochim. Acta, 1983, 42, 1055. 27 M. Nanjo and G. G. Guilbault, Anal. Chem., 1974, 46, 1769. Paper 9/03187A Analyst, 1999, 124, 1175–1179 11

ISSN:0003-2654    

DOI:10.1039/a903187a

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Modified gas-permeable silicone rubber membranes for covalent immobilisation of enzymes and their use in biosensor development R. Schüler,* M. Wittkampf† and G.-C. Chemnitius ICB, Institut für Chemo- und Biosensorik, eV, Mendelstrasse 7, D-48149 Münster, Germany Received 13th April 1999, Accepted 18th June 1999 Novel enzyme membranes are introduced. Modified polymeric gas-permeable layers were developed enabling biological components which have available reactive groups (–NH2, –OH, –SH, –COOH) to couple covalently on to their surfaces.Therefore, gas-permeable two component room temperature vulcanising (2K-RTV) silicone rubber was modified using additional cross-linking agents. Triethoxysilanes with functional groups on their side chains such as epoxy or amino groups were used. A special attribute of the resulting gas-permeable membranes is that their formation and modification occur simultaneously during one reaction step. IR spectroscopy was used to observe the changes in the polymeric structure due to the reaction with the additional cross-linking agents.Sensors equipped with these layers are suitable to measure dissolved gases such as O2, CO2 and NH3 consumed or produced by enzymes converting their substrates. Determination of glucose, a well investigated enzymatic detection process, was chosen to demonstrate the applicability of the enzyme immobilisation. Glucose oxidase was immobilised on the membranes and glucose was detected by amperometric measurement of oxygen consumption.It is expected that this immobilisation method will also be useful for miniaturised planar biosensors. Introduction Many biosensors make use of the versatile enzyme groups of oxidases and oxygenases for the specific detection of a variety of different analytes.1 They convert their substrates with oxygen consumption and, in most cases, release of hydrogen peroxide. The decrease in oxygen or the increase in hydrogen peroxide concentration can be determined, e.g., amperometrically.In complex sample matrices the measurement of hydrogen peroxide concentration exhibits problems due to the crossselectivity to electroactive species, e.g., ascorbic acid. Using a Nafion membrane as an anti-interfering layer on the working electrode is a possible approach to minimise this crossselectivity. 2–5 An approach to circumvent the problem of interfering substances is to use membrane-covered dissolved oxygen sensors being developed to separate the internal electrolyte and thus the electrochemical reaction from the electroactive substances in the measurement solution. Only gases can diffuse through the membrane and the consumption of oxygen due to the enzymatic reaction can be measured by the dissolved oxygen sensor without interferences.In 1962, Clark and Lyons6 described the first biosensor for measuring glucose based on this principle. They physically immobilised glucose oxidase in front of a Clark-type oxygen sensor with the help of a dialysis membrane. In subsequent years the sensor design has been modified repeatedly7–11 but the basic components of the sensor have remained the same: a gaspermeable layer and an enzyme layer combined or supplemented with a supporting layer. Enzymes have been immobilised by entrapment in a gel12 or by intermolecular cross-linking,13 but in both cases additional membranes for enzyme fixation in front of the sensor were still necessary owing to the limited adhesion of the immobilised enzyme to the hydrophobic gas-permeable membrane.These sensors are often called sandwich-type sensors. Alternatively, the enzyme was covalently immobilised on a second, hydrophilic membrane14,15 which then was mechanically held in front of the oxygen sensor by an O-ring. Recently, covalent immobilisation of enzymes on commercially available gas-permeable polymer membranes has been described,16–18 but these membranes have to be chemically activated prior to immobilisation during an additional reaction step.The fabrication of the enzyme sensors described above includes several fabrication steps, sometimes an unstable immobilisation and can only be used for the construction of macro electrodes because of the need for mechanical fixation of the membranes. For this reason we describe here novel enzyme membranes suitable both for conventional macro and for planar micro sensors. Casting and modification of the gas-permeable membranes can be accomplished in one reaction step, opening the route to covalent bonding of enzymes directly to the resulting functionalised gas-permeable membranes.As base material we chose two component room temperature vulcanising (2K-RTV) silicone rubber. As additional cross-linking agent 3-aminopropyltriethoxysilane, 3,4-epoxybutyltriethoxysilane or 9,10-epoxydecyltriethoxysilane was introduced to build a multi-component RTV silicone rubber bearing functional groups available for enzyme immobilisation.The functionalised membranes were investigated by infrared spectroscopy and used as gas-permeable membranes in conventional dissolved oxygen sensors. Their properties were compared with those of commercially available membranes. To prove the applicability of the new immobilisation method, glucose sensors were developed by covalent immobilisation of glucose oxidase on differently modified silicone rubber membranes. Experimental Reagents Silopren K 1000, Silopren cross-linking agent K-11, hexane and 3-aminopropyltriethoxysilane were purchased from Fluka (Neu-Ulm, Germany). Dialysis membranes (Enka, Wuppertal, Germany) made from regenerated cellulose (AKZO Cuprophan ®, Typ 80 M) were used as the stabilising support for the silicone rubber membranes.Gas-permeable 13 mm thick PTFE membranes were obtained from BioLytik (Bochum, Germany). † Present address: PhiScience GmbH, Schützenstrasse 41a, D-58239 Schwerte, Germany.Analyst, 1999, 124, 1181–1184 1181The epoxy functionalised alkoxysilanes 3,4-epoxybutyltriethoxysilane and 9,10-epoxydecyltriethoxysilane were kindly donated by G. Sperveslage and Professor Grobe, Inorganic Chemistry Institute, University of Münster, Germany. Glucose oxidase (EC 1.1.3.4, Type VII-S, 115 U mg21, from Aspergillus niger) and d-glucose were obtained from Sigma (Deisenhofen, Germany). Electrochemical instrumentation and software All electrochemical experiments were computer controlled with data acquisition accomplished using an IBM-compatible 486 DX 33 computer to control the potentiostat/galvanostat AUTOLAB with the modules PSTAT10 and ECD and the data acquisition software GPES3 (ECO-Chemie, Utrecht, The Netherlands).The commercially available oxygen sensor Series MS and the corresponding measuring cell (volume 1 ml) were obtained from BioLytik. The measuring cell was thermostated by a Typ M3 thermostat (Lauda, Königshofen, Germany). The gases for the oxygen calibration were mixed using a device from Ludewig + Tillmann (Dortmund, Germany).Sensor measurements The fabrication of the oxygen and the glucose sensor, comprised the making of the gas-permeable membranes, the immobilisation of the enzyme and the assembly of the sensor. In principle, the silicone rubber membranes can be formed in situ on the internal electrolyte of the sensor. For this sensor set-up, dialysis membranes were used as thin supporting material for the cast gas-permeable membranes.Preparation of gas-permeable membranes The silicone rubber membranes were prepared under ambient laboratory conditions. The composition of the pre-polymer for the unmodified membranes contained 100 ml of polydimethylsiloxane (K 1000) in 500 ml of hexane and 10 ml of cross-linking agent (K-11) and for the modified membranes, suitable for enzyme immobilisation, 150 ml of K 1000 in 500 ml of hexane, 20 ml of the functionalised silane and 10 ml of cross-linking agent (K-11).Volumes of 10 ml of the respective polymer mixture were cast on to a piece of dialysis membrane and dried at room temperature. For IR spectroscopy experiments silicone rubber membranes were formed on a PTFE support. Immobilisation For subsequent enzyme immobilisation on the modified membranes, 10 ml of a 5 U ml21 enzyme solution in 0.1 M phosphate buffer (KH2PO4–Na2HPO4, pH 7) was dripped on to and spread over the silicone rubber membrane.The membrane was dried overnight at 4 °C. Unbound enzyme was removed by washing with de-ionised water. When 3-aminopropyltriethoxysilane was used as the additional cross-linking agent, 5 or 10% glutaraldehyde solution was added to the enzyme solution. Assembly Owing to the sensor construction, all membranes were fixed to the BioLytik electrode cell using O-rings. Sodium tetraborate– NaOH buffer (pH 10) (0.1 M KCl) was used as an internal electrolyte.Membrane compositions of dialysis membrane– silicone rubber–enzyme were soaked in 0.1 M phosphate buffer (pH 7) prior to their application as gas-permeable membranes. Oxygen and glucose measurements The measurements of oxygen and glucose were carried out in the thermostated measuring cell at 298 K. Cyclic voltammetry was performed in unstirred solution and for amperometric measurements the solution was constantly stirred. For comparative oxygen measurements conventional PTFE membranes were used for the sensor construction.IR measurements An IFS 48/ASPECT 1000 infrared spectrometer (Bruker, Karlsruhe, Germany) was used to characterise the original and the modified gas-permeable silicone rubber materials. After polymerisation, silicone rubber membrane disks were peeled off the PTFE support and compressed between two NaCl plates. The spectra were recorded as FT-IR transmission spectra after purging the sample chamber with dry nitrogen.All samples were scanned 600 times. The spectra of the silanes were recorded with reference to NaCl. To bring out the important bands more clearly, the spectrum of the unmodified silicone rubber membrane (NaCl reference) was subtracted from the spectra of the modified silicone rubber membranes. Results and discussion IR spectra The IR spectrum of 3-aminopropyltriethoxysilane [Fig. 1(a)] shows two characteristic absorption bands at ~ 3378 and ~ 1629 cm21, indicating the valence and the deformation vibration of a free amino group.Regarding the availability of the amino group for reaction on the polymer with a biological component, the IR spectrum of the amino modified membrane [Fig. 1(b)] was examined. The valence vibration band of the free amino group is reduced to a shoulder in the spectrum of the polymer. At 1629 cm21 the deformation band indicates the free amino group. There are also two new bands at 1574 and 1490 cm21 seen in the spectrum of the amino functionalised silicone rubber.They could be interpreted as the deformation vibration from associated NH3 +.19 The shift of the 1607 cm21 band for free NH deformation vibration to 1574 cm21 in the condensed state was explained by Plueddemann.20 Other models for interand intramolecular reaction were explained by Chiang et al.21 and Moses et al.22 The described models were developed for hydrophilic surfaces but it is predictable that they could also be valid for the reactions in and on the polymer.The IR spectrum of 9,10-epoxydecyltriethoxysilane [Fig. 2(a)] shows an absorption band at 3042 cm21. This band is caused by the CH valence vibration of the C2 atom of the epoxy group. The IR spectrum of the modified silicone rubber with 9,10-epoxydecyltriethoxysilane [Fig. 2(b)] also shows this band. The availability of the epoxy group for the reaction with a biological component could be expected. IR spectra of 3,4-epoxybutyltriethoxysilane and the corresponding modified Fig. 1 IR spectra: (a) 3-aminopropyltriethoxysilane (reference: NaCl); (b) silicone rubber modified with 3-aminopropyltriethoxysilane (the unmodified silicone rubber spectrum was subtracted; reference: NaCl). Grey shadow, range of interest; dotted line, absorption band at 1629 cm21. 1182 Analyst, 1999, 124, 1181–1184silicone rubber show the equivalent absorption band at 3042 cm21. Oxygen calibration The properties of oxygen sensors equipped with PTFE membranes and unmodified and modified silicone rubber membranes were investigated.Cyclic voltammetric experiments showed a plateau for the oxygen reduction current between 2700 and 2900 mV vs. Ag/AgCl. Hence the working potential for all following amperometric measurements was set to 2800 mV vs. Ag/AgCl. For the calibrations of the sensors the measurement solutions were equilibrated using oxygen–nitrogen mixtures of 0, 5, 10, 15, 20, 25, 50 and 100% oxygen saturation which were prepared by the gas mixing device.Fig. 3 depicts the calibration curves of the different oxygen sensors. The sensor equipped with the PTFE membrane showed a sensitivity of 64 nA (% O2)21 and that with the unmodified silicone membrane showed a sensitivity of 69 nA (% O2)21, whereas the sensitivities of the sensors using the modified silicone rubber membranes were higher: for the amino group modified membrane 82 nA (% O2)21, for the 3,4-epoxy group modified membrane 74 nA (% O2)21 and for the 9,10-epoxy group modified membrane 86 nA (% O2)21.This is contrary to expectation as the degree of cross-linking in the modified polymer is higher. It might be that the grade of hydrophobicity was reduced in the modified polymers. The permeability of the membranes was examined by measuring the response time and the stirring dependence of the sensors equipped with the different membranes (Table 1). The stirring dependence was calculated from the equation I I stirrer on stirrer off stirring dependence Æ Æ ¥ Ê Ë Á � � � - = 100 100 (%) Response times were determined after the addition of 100 ml of saturated sodium sulfite solution to 1 ml of air saturated 0.1 M phosphate buffer (pH 7).The stirring dependence of the sensors with the modified silicone rubber membranes was lower than that of the sensors using the PTFE or the unmodified silicone membranes. The response times (t95) of the sensors with the modified membranes were twice as fast as that of the sensor with the PTFE membrane and about 10 seconds slower than that of the sensor with the unmodified silicone membrane. Glucose sensors Glucose sensitive enzyme membranes were obtained by the covalent immobilisation of glucose oxidase on the differently modified silicone rubber membranes under mild ambient conditions.These membranes were used to construct glucose sensors based on the measurement of oxygen consumption. Glucose sensors based on amino group modified silicone rubber membranes The calibration curve for these sensors shown in Fig. 4 is linear over the concentration range from 0.02 to 1.9 mmol l21 glucose. The evaluation of the data led to a sensitivity of 0.36 nA (mM glucose)21 with a correlation coefficient of 0.9995 (n = 18) and a standard deviation of 0.003 nA (mM glucose)21. The oxygen depletion around the enzyme membrane limited the upper Fig. 2 IR spectra: (a) 9,10-epoxydecyltriethoxysilane (reference: NaCl); (b) silicone rubber modified with 9,10-epoxydecyltriethoxysilane (the unmodified silicone rubber spectrum was subtracted; reference, NaCl).Dotted line, absorption band at 3042 cm21. Fig. 3 Calibration for oxygen sensors constructed with different gaspermeable membranes: -, commercially available PTFE membrane; 5, silicone rubber; :, 3-aminopropyltriethoxysilane modified silicone rubber; !, 9,10-epoxydecyltriethoxysilane modified silicone rubber; <, 3,4-epoxybutyltriethoxysilane modified silicone rubber.Lines are regression lines. Conditions: 0.1 M phosphate buffer (pH 7), 298 K, 1013 mbar. Table 1 Comparison of stirring dependence and response times of the different oxygen sensors Oxygen sensor with Stirring dependence (%) (saturated solution) t95/s t90/s PTFE membrane 120 ± 10 60 ± 21 13 ± 1 Silicone rubber membrane 141 ± 12 19 ± 1 11 ± 1 Amino modified silicone rubber membrane 112 ± 9 32 ± 13 18 ± 6 Epoxydecyl modified silicone rubber membrane 111 ± 9 31 ± 5 16 ± 1 Epoxybutyl modified silicone rubber membrane 119 ± 17 28 ± 7 15 ± 3 Fig. 4 Calibration curve for glucose sensor based on the 3-aminopropyltriethoxysilane modified silicone rubber membrane in 0.1 M phosphate buffer (pH 7 298 K. The corresponding steady state current response is shown in the inset. Addition of eight times 5 ml of 0.01 M glucose solution and five times 1 ml and four times 2.5 ml of 0.1 M glucose solution. Analyst, 1999, 124, 1181–1184 1183detection limit to 2.5 mmol l21 glucose.The sensitivities of the calibration curves showed no change over a period of at least 10 d. Glucose sensors based on epoxy group modified silicone rubber membranes No immobilisation could be achieved on the 3,4-epoxybutyltriethoxysilane modified membranes. It is assumed that the butyl spacer between the reactive epoxy group and the polymer backbone is too short for effective immobilisation. Silicone rubber membranes modified by 9,10-epoxydecyltriethoxysilane immobilisation worked fairly well.The resulting sensors showed a higher sensitivity than the glucose sensors with the amino group modified membranes and much lower signal-tonoise ratio. Evaluation of the data led to a linear detection range from 0.01 to 1.1 mmol l21 with a sensitivity of 1.33 nA (mM glucose)21, a correlation coefficient of 0.9999 (n = 19) and a standard deviation of 0.004 nA (mM glucose)21 (Fig. 5). The linear response range is also limited owing to the decreasing oxygen value.The four times higher sensitivity of these sensors might be explained by the thinner enzyme layer allowing faster substrate diffusion and by the milder immobilisation conditions using the epoxy group modified membranes causing less stress on the enzyme conformation. All enzyme membranes can be stored dry before use. Several membranes were tested after storage for 2 weeks and showed similar calibration curves to those measured directly after preparation.The immobilisation method presented here offers several new possibilities for measuring enzyme substrates by detecting gaseous co-substrates or products. For the example of glucose measurement it has been demonstrated that the proposed modification of silicone rubber using various functionalised triethoxysilanes followed by enzyme immobilisation yields powerful enzyme sensors. To obtain sensors with increased linear ranges, the sensors could be used in flow-through devices using automated dilution or by adding a substrate limiting diffusion membrane.23 Using this method, no additional modification of the existing gas-permeable membrane of a gas sensor was necessary because membrane preparation and modification are done in one step.No sandwich set-up of different membranes was necessary for enzyme immobilisation thanks to the direct covalent bonding of the biological component to the gas-permeable membrane.Contrary to the work of von Gentzkow et al.,24 the immobilisation did not result in a porous membrane. Thus compartments of electrochemical detection and sample addition are effectively separated. The modified membranes can be used, as demonstrated in this paper, with support membranes, but it should also be possible to cast these membranes directly on to the internal electrolyte of planar electrodes. For the construction of miniaturised oxygen sensors the use of unmodified silicone rubber as gas-permeable membranes has already been described successfully by Wittkampf et al.25 This opportunity is very significant when using miniaturised planar electrodes made using thin film or thick film technology, as it will not be possible to use O-rings to fit an enzyme membrane mechanically to these sensors.In addition to glucose oxidase, other enzymes could be applied, e.g., other oxidases like lactate oxidase or glutamate oxidase, or oxygenases producing no hydrogen peroxide such as tyrosinase or ascorbic acid oxidase.Also enzymes such as urease or decarboxylases that liberate other gases, in these cases NH3 and CO2, respectively, could be used as biological components. Acknowledgements The authors are grateful to Dr. G. Sperveslage and Professor J. Grobe for supplying the epoxy modified silanes and for helpful discussions. References 1 F. Scheller and F. Schubert, Biosensoren, Birkhäuser Verlag, Basle, 1989. 2 P. Manowitz, P.W. Stoecker and A. M. Yacynych, Biosens. Bioelectron., 1995, 10, 359. 3 R. Vaidya, P. Atanasov and E. Wilkins, Med. Eng. Phys., 1995, 17, 416. 4 H. Frebel, G.-C. Chemnitius, K. Cammann, R. Kakerow, M. Rospert and W. Mokwa, Sens. Actuators B, 1997, 43, 87. 5 S. A. Emr and A. M. Yacynych, Electroanalysis, 1995, 7, 913. 6 L. C. Clark Jr. and C. Lyons, Ann. N.Y. Acad. Sci., 1962, 102, 29. 7 D. P. Newman, US Pat., 4 073 713, 1976. 8 S. L. Xie, E. Wilkins and P. Atanasov, Sens.Actuators B, 1994, 17, 133. 9 S. Yang, P. Atanasov and E. Wilkins, Ann. Biomed. Eng., 1995, 23, 833. 10 S. Gamburzev, P. Atanasov and E. Wilkins, Sens. Actuators B, 1996, 30, 179. 11 L. Stancik, L. Macholan and F. Scheller, Electroanalysis, 1995, 7, 649. 12 C. Tranh-Minh and G. Broun, Anal. Chem., 1975, 47, 1359. 13 C. R. Tillyer and P. T. Gobin, Biosens. Bioelectron., 1991, 6, 569. 14 L. Doretti, D. Ferrara, P. Gattolin and S. Lora, Biosens. Bioelectron., 1996, 11, 365. 15 S. D. Haemmerli, A. A. Suleiman and G. G. Guilbault, Anal. Biochem., 1990, 191, 106. 16 D. J. Tarnowski, E. J. Bekos and C. Korzeniewski, Anal. Chem., 1995, 67, 1546. 17 S. Turmanova, A. Trifonov, O. Kalaijiev and G. Kostov, J. Membr. Sci., 1997, 127, 1. 18 M. Goto, Jpn. Pat., 02061549, 1990. 19 M. Hesse, H. Meier and B. Zeeh, Spektroskopische Methoden in der Organischen Chemie, Georg Thieme, Stuttgart, 1991. 20 E. P. Plueddemann, Compos. Mater., 1974, 6, 6. 21 C. Chiang, H. Ishida and J. L. Koenig, J. Colloid Interface Sci., 1980, 74, 396. 22 P. R. Moses, L. M. Wier, J. C. Lennox, H. O. Tinklea, J. R. Lenhard and R. W. Murray, Anal. Chem., 1977, 50, 576. 23 C. Loechel, G.-C. Chemnitius, M. Borchardt and K. Cammann, Z. Lebensm.-Unters. Forsch., 1998, 207, 381. 24 W. von Gentzkow, H.-D. Feucht, H. Formanek and G. Wanner, Eur. P. Appl., 0562372, 1993. 25 M. Wittkampf, G.-C. Chemnitius, K. Cammann, M. Rospert and W. Mokwa, Sens. Actuators B, 1997, 43, 40. Paper 9/03295I Fig. 5 Calibration curve for glucose sensor based on the 9,10-epoxydecyltriethoxysilane modified silicone rubber membrane in 0.1 M phosphate buffer (pH 7.0) at 298 K. The corresponding steady state current response is shown in the inset. Eighteen times addition of 5 ml of 0.01 M solution of glucose. 1184 Analyst, 1999, 124, 1181–1184

ISSN:0003-2654    

DOI:10.1039/a903295i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Screen-printed zeolite-modified carbon electrodes Alain Walcarius,*a Sandra Rozanska,a Jacques Bessièrea and Joseph Wangb a Laboratoire de Chimie Physique pour l’Environnement, Unité Mixte de Recherche UMR 7564, CNRS - Université H. Poincaré Nancy I, 405, rue de Vandoeuvre, F-54600 Villers-les-Nancy, France. E-mail: walcariu@lcpe.cnrs-nancy.fr b Department of Chemistry and Biochemistry, New Mexico State University, Las Cruces, NM 88003 USA Received 19th May 1999, Accepted 5th July 1999 The evaluation of screen-printed carbon electrodes modified with zeolites for the determination of the herbicides paraquat and diquat is described and compared to the corresponding zeolite-modified carbon paste electrodes.Cyclic voltammetry was used to characterise the electrochemical behaviour of paraquat at both electrodes, indicating better defined peaks for the screen-printed electrodes than for the corresponding carbon pastes and lesser influence of residual oxygen.Square wave voltammetry was then applied to investigate the partitioning of this herbicide into the zeolite particles (Y-type) to the conductive composites. The screen-printed zeolite-modified electrode (SPZME) resulted in faster accumulation and release of the electroactive probe compared to the zeolite-modified carbon paste electrode (ZMCPE). The improved response time, sensitivity, and reproducibility, were attributed to the thin film structure of SPZME compared to the bulky ZMCPE.Substantial enhancement of current signal was observed when operating with disposable SPZMEs that were not soaked in any electrolyte solution prior to the accumulation step. The selective accumulation of paraquat over diquat was demonstrated by using ZSM-5 zeolite particles, while its strong binding to the aluminosilicate prevented efficient voltammetric detection. Introduction Zeolites are attractive electrode modifiers because their presence at an electrode surface allows the combination of selected electrochemical reactions with intrinsic zeolite properties such as size and shape selectivities, ion exchange capacity, or catalytic activity.1 Zeolite-modified electrodes (ZMEs) have generated great interest during the past 15 years, finding numerous applications especially in electroanalysis.2–4 Because zeolites are crystalline in nature and made of solid particles which are not electronically conductive, their use in connection with electrochemistry is not technically easy.This is probably why numerous preparation methods for ZMEs were proposed and assessed previously.3,5,6 They can be classified into four main categories,4 including the dispersion of zeolite particles within a conductive composite matrix, the compression of zeolites onto a conductive substrate, the formation of zeolitic films (embedded in or covered by an inert polymer) on solid electrodes, and the covalent binding of zeolite particles to an electrode surface.Such a diversity illustrates that the ideal preparation procedure has not yet been discovered for getting highly durable ZMEs ensuring fully reproducible measurements. 6 This is especially important with respect to the application of ZMEs for analytical purposes. In particular, zeolite films suffer from poor mechanical stability preventing their use in stirred solution, and zeolite-modified carbon paste electrodes, though being largely exploited in electroanalysis, 7–12 were reported to undergo imbibition of the bulk paste by the surrounding solution,8,13 hindering somewhat the chemical regeneration of the electrode surface (memory effects).A possible way to get round this difficulty could be found in applying the screen-printing technology, which was successfully used for constructing disposable electrochemical sensors and biosensors in the last decade.14–18 The present study aims to prepare screen-printed zeolitemodified carbon electrodes and apply them to the preconcentration/ voltammetric measurements of paraquat and diquat.Zeolite Y was chosen as a representative crystalline aluminosilicate which is known to incorporate these herbicides by ion exchange.19 The main purpose of this paper is to show how this new type of zeolite-modified electrode can overcome some drawbacks encountered with zeolite-modified carbon paste electrodes applied in the accumulation–voltammetric detection scheme. Special attention will be given to the ability to chemically regenerate the electrode surface after the measurement. A particular zeolite ZSM-5 will be tested as a selective sorbent for paraquat over diquat.Experimental Apparatus and reagents Electrochemical experiments were performed in a threeelectrode cell configuration. Working electrodes were laboratory- made screen-printed or carbon paste electrodes modified with zeolite particles. The counter-electrode was made of a platinum wire, and the Ag/AgCl electrode (Metrohm, Herisau, Switzerland) was used as reference.Electrochemical studies were conducted with the EG&G Princeton Applied Research (Princeton, NJ, USA) Model 283 potentiostat/galvanostat monitored by the Model M270 software (EG&G). All reagents and electrolytes were of analytical grade and all solutions were prepared with high purity water (18 M½ cm21) from a Millipore (Watford, Herts., UK) Milli-Q water purification system. Paraquat (methyl viologen dichloride hydrate, N,NA-dimethyl-4,4A-bipyridinium dichloride) was purchased from Aldrich (Bornem, Belgium), and diquat (1,1A-ethylene- Analyst, 1999, 124, 1185–1190 11852,2A-bipyridinium dibromide) was an analytical standard Pestanal® (Riedel-de Haën, Hannover, Germany).Zeolite Y was Linde Molecular sieve Cat. Base L-Y54 powder (UOP, Molecular Sieve Division, Des Plaines, IL, USA) used in its sodium form (formula: Na56Al56Si136O384·250 H2O). Zeolite ZSM-5 was obtained from the Laboratoire des Matériaux Inorganiques (Namur, Belgium) and its chemical composition was Na4Al4Si92O192· ~ 16 H2O.Preparation of zeolite-modified electrodes Screen-printed zeolite-modified electrodes (SPZMEs) were fabricated by using a semi-automatic screen printer (Model TF 100, MPM, Franklin, MA, USA). Samples of zeolite particles were first added to the carbon ink (Ercon ink G-449-I, Ercon, Waltham, MA, USA) in selected ratios and the resulting materials were hand mixed thoroughly (with a spatula) for 30 min, until homogeneous dispersions were obtained (note that the specified zeolite loadings refer to the initial ink composition, prior to the solvent evaporation). These modified inks were printed through a patterned (100 mm thick) stencil onto 10 3 10 cm alumina ceramic plates containing 30 strips (each being sized 3.33 3 1.00 cm, as defined by a laser pre/semi cut).The resulting 0.2 3 3.0 cm printed structures were cured for 60 min at 80 °C.An insulating ink (Ercon R-486(AH)Blue) was then printed on a portion of the plate (and cured for 60 min at 80 °C), to leave 2 3 5 mm sections on both ends for defining the working electrode and the electrical contact. Zeolite-modified carbon paste electrodes (ZMCPEs) were prepared as previously described.9,20 Typically, 0.1 g zeolite and 0.6 g carbon graphite ( < 325 mesh, Johnson Matthey, Royston, Herts., UK) were thoroughly mixed with 0.3 g of the organic binder (Nujol, Aldrich) for ca. 15 min until a uniformly wetted paste was obtained. The paste was then packed into the end of a home-made PTFE cylindrical tube (od 8 mm, id 6 mm) equipped with a screwing stainless steel piston. When necessary, a new surface was obtained by pushing an excess of paste out of the tube (about 200–500 mm thickness) and polishing it on a weighing paper. Unmodified carbon paste was prepared according to the same procedure without adding zeolite to the mixture and was used for comparison purposes.Procedures All experiments were performed at room temperature. Each screen-printed electrode was used either in a single experiment without any preconditioning or in multiple experiments after being immersed in the electrolyte solution (typically 0.1 M NaCl) for 30 min before measurement. The surface of the carbon paste electrodes was mechanically polished prior to each series of experiments. Solutions were purged with pure nitrogen for a selected period of time (see below) before measurements.Cyclic voltammetry and square wave voltammetry were applied to evaluate the electrochemical behaviour of the electrodes. Square wave voltammetry was used to monitor the uptake and release of the analytes from the electrodes because of its better sensitivity and detection limit. Due to the rather high value of the uncompensated resistance when using screenprinted electrodes (1–2 kW), a correction for ohmic drop was applied by positive feedback.Accumulation was performed at open circuit in the analyte solution, the electrode was then removed and rinsed with pure water, and immersed in the detection cell for recording the voltammetric curve. Regeneration was achieved by soaking the electrode loaded with the analyte in a solution containing 1 M NaCl, at open circuit, until no signal for the analyte was obtained. Batch ion exchange experiments involving paraquat and diquat in zeolites Y and ZSM-5 were monitored by UV spectrophotometry (spectrophotometer Beckmann DU 7500; Fullerton, CA, USA) in the supernatant obtained after centrifugation of the zeolite suspension.Paraquat was detected at 258 nm and diquat at 309 nm. Results and discussion Electrochemical behaviour of paraquat at unmodified and zeolite-modified carbon electrodes The electrochemistry of viologen derivatives (such as paraquat, PQ2+; or diquat, DQ2+) has been well-known for a long time on unmodified electrodes,21 and the electrochemical behaviour of paraquat has been well studied on various types of modified electrodes.13,22–25 The electro-reduction of PQ2+ proceeds via the formation of a blue radical cation, PQ· + (eqn. 1), which can reversibly form a dimer.26 The monoelectronic charge transfer is reversible but the presence of oxygen can induce some complications by reacting with PQ· + species (eqn. 2). The further monoelectronic reduction of PQ· + to the fully reduced state, PQ0 (eqn. 3), occurs at more negative potentials and is less reversible because of the conproportionation reaction between PQ0 and PQ2+ to generate 2PQ· + (eqn. 4), which tends to adsorb on the electrode surface. PQ2+ + 1 e2 " PQ· + (1) 2 PQ·+ + 1/2 O2 + H2O ? 2 PQ2+ + 2 OH2 (2) PQ· + + 1 e2 " PQ0 (3) PQ2+ + PQ0 ? 2 PQ· + (4) When incorporated in zeolite Y-modified electrodes, PQ2+ is liable to undergo charge transfer reactions,13,22 which have been demonstrated to proceed after ion exchange for the supporting electrolyte cation, C+ chosen here as monovalent for convenience, according to an extra zeolite electron transfer mechanism [eqns. 5 (a) and (b)].27 PQ2+ (Z) + 2 C+ (S) " PQ2+ (S) + 2 C+ (Z) (5a) PQ2+ (S) + 1 e2 " PQ· + (S) (5b) where subscripts Z and S refer to zeolite and solution phases, respectively. Comparison between zeolite-modified screen-printed and carbon paste electrodes Cyclic voltammetry characterisation. Fig. 1 illustrates typical cyclic voltammograms recorded for PQ2+ in solution (after 15 min nitrogen bubbling) by using carbon paste and screen-printed carbon electrodes, respectively unmodified and modified with 10% zeolite Y.Such electrodes will be noted hereafter as CPE (bare carbon paste electrode), ZMCPE (zeolite-modified carbon paste electrode), SPCE (bare screenprinted carbon electrode), and SPZME (screen-printed zeolitemodified electrode). There are several things to notice on this figure, displaying definite advantages of using screen-printed carbon instead of carbon paste.First, the effect of oxygen was less with use of screen-printed electrodes. Second, the presence of zeolites induced a significant enhancement of peak currents. Third, this enhancement was much more intense with SPZME than with ZMCPE. Purging the solutions with nitrogen during 15 min was sufficient for preventing the effect of oxygen on the electrochemical behaviour of PQ2+ when using SPCE or SPZME, 1186 Analyst, 1999, 124, 1185–1190while the electrocatalytic reduction of oxygen by paraquat was clearly visible when using CPE or ZMCPE (cathodic peak located between 20.50 V and 20.55 V) and required several hours for completing the purge.This electrocatalytic wave was otherwise exploited for sensing oxygen at carbon composite electrodes modified with zeolites in the presence of paraquat.28 After complete deaeration, the two successive reduction steps of PQ2+ were chemically reversible (with cathodic-to-anodic peak currents ratio close to unity) but only the first one was electrochemically reversible, as shown in Table 1.The difference in the cathodic and anodic peak potentials was about 60 mV, as expected for a reversible process, indicating fast electron transfer with both electrodes. When using bare electrodes and scanning potentials up to only 20.85 V (limited to the first electron transfer), a plot of the peak currents versus the square root of the scan rate produced a straight line (r2 = 0.998) for both cathodic and anodic processes, indicating diffusion-controlled charge transfer reactions.While the response of bare electrodes was quite independent of the time afforded to the electrode to contact the solution, prolonged soaking of the zeolite-modified electrodes in the PQ2+ solution resulted in a marked increase in the peak currents. This is fully explained by the accumulation of PQ2+ species at the electrode surface, by ion exchange within zeolite particles.The enhancement was especially outstanding for the second cathodic peak, corresponding to the reduction of PQ· + species (eqn. 3), which was much more intense than the first one (eqn. 1), as compared to the bare electrodes for which the two cathodic peaks are roughly equivalent, giving very different cathodic-to-anodic current ratios (Table 1). This particular voltammetric pattern can be explained by the high reservoir of PQ2+ species (in the bulk zeolites), concomitant to the electrochemical generation of PQ0, inducing the production of large amounts of PQ· + species (eqn. 4) which are reduced at the electrode surface at about 21.0 V (Fig. 1, curves B and D). Similar behaviour was recently observed by Brunetti and Ugo25 at glassy carbon electrodes covered by a poly(estersulfonate) ion exchanger. Interestingly, the enhancement in peak currents observed when passing from bare electrodes to zeolite-modified electrodes was significantly greater with screen-printed electrodes than with carbon paste (look at the scale change on Fig. 1 when passing from SPCE to SPZME). This behaviour was also observed by performing voltammetry in an analyte-free medium after preconcentration at open circuit from a diluted PQ2+ solution. For example, Fig. 2 illustrates the case of 1 3 1026 M PQ2+, analysed by square wave voltammetry after a 30 min preconcentration, indicating a 7.5 enhancement factor for SPZME, compared to 1.7 for the ZMCPE.Such enhancement can be attributed to a higher zeolite loading (in view of the solvent evaporation during the curing step). Trace analysis by square wave voltammetry Release and uptake of paraquat in zeolite Y. ZMCPEs have been applied in the past for electroanalytical determinations based on the preconcentration–voltammetric detection scheme,7,9,29,30 but their chemical regeneration (by back ion exchange) is known to be rather difficult due to the progressive imbibition of the paste interior by the surrounding solution.12,13 This is illustrated in Fig. 3 (curve a) where more than 70% of the initial signal of preconcentrated paraquat was maintained even after prolonged soaking of the electrode in the electrolyte solution. By contrast, the use of SPZMEs resulted in much Fig. 1 Cyclic voltammograms for 2.0 3 1024 M paraquat at bare carbon paste electrode (A), 10% zeolite Y-modified carbon paste electrode (B), bare screen-printed carbon electrode (C), and 10% zeolite Y-modified screen-printed carbon electrode (D), after 15 immersions of the electrodes into the solution (under bubbling with pure nitrogen).Supporting electrolyte: 0.1 M NaCl. Scan rate: 50 mV s21. Table 1 Cyclic voltammetry characteristics observed after soaking the electrodes for 2 h (under nitrogen bubbling) in 2.0 3 1024 M PQ2+ (+0.1 M NaCl); scan rate: 50 mV s21 First electron transfera Second electron transferb Electrode EC1/V EA1/V iC1/iAl EC2/V EC2/V iC2/iA2 c Bare CPE 20.69 20.63 0.99 20.99 20.91 1.1 Bare SPE 20.68 20.62 1.00 21.01 20.94 0.75 Zeolite-modified CPE 20.71 20.62 0.98 20.97 20.89 1.8 Zeolite-modified SPE 20.70 20.61 0.96 21.01 20.87 2.3 a Potential scan ranging from 20.30 V to 20.85 V.b Potential scan ranging from 20.30 V to 21.30 V. c Peak currents were calculated by estimation of background currents after the first peak. Fig. 2 Square wave voltammograms for 1.0 3 1026 M paraquat at carbon paste (A) and screen-printed carbon (B) electrodes, recorded after 30 min soaking in solution, with a 100 Hz frequency, a 5 mV step height and a 50 mV modulation amplitude; (a) bare electrodes, (b) 10% zeolite Y-modified electrodes.Supporting electrolyte: 0.1 M NaCl. Analyst, 1999, 124, 1185–1190 1187faster desorption of paraquat after 5 min accumulation from a 1.0 3 1026 M solution: 30 min soaking in 0.1 M NaCl resulted in fractions ranging from 4 to 11% of the analyte left in the electrode, depending on the zeolite loading (Fig. 3). This could be rationalised by considering the different nature of the electrodes: thin-layer for SPZME which does not allow the solution to diffuse deeply in the bulk electrode, and bulky for ZMCPE which contributes to significant imbibition of the paste interior by the external solution. The release efficiency is concentration dependent, and at 1.0 3 1027 M PQ2+ total desorption was observed with SPZMEs in less than 15 min while maintaining a 55% signal with ZMCPE.The nature of the zeolite-modified electrode was also found to affect the kinetics of the accumulation process (Fig. 4). In all cases, the signal increased with increasing the accumulation period; a faster response time was obtained with using SPZMEs as compared to ZMCPE. SPZMEs yielded 90% of the maximum signal within 5 min, while carbon paste required more than one hour to reach the equilibrium. Once again this behaviour can be attributed to the thin layer structure of SPZMEs preventing them from any significant diffusion of the solution in the bulk of the electrode.Surprisingly, the voltammetric response was not found to rise upon increasing the zeolite content in the screen-printed electrode, contrary to what one would expect from previous work dealing with electrochemical sensors based on chemically modified electrodes. A possible explanation for this could be found in the lower real electrode surface area upon increasing the zeolite content in the carbon ink.Indeed, back-diffusion scanning electron micrographs of SPZMEs containing either 5 or 15% zeolite have revealed a significantly larger occupation of the electrode surface by zeolites in the last case. As a consequence, if a higher amount of zeolite is thought to lead to higher accumulation efficiency, the detection of these species could be somewhat limited by the lower available carbon on the strip surface.Determination of paraquat and diquat at screen-printed electrodes modified with zeolite Y. Square wave voltammograms were recorded at SPZMEs, respectively immersed in the electrolyte solution before carrying out the accumulation step (multiple analysis with the same strip) and untreated (single use electrode), for various concentrations of both PQ2+ and DQ2+. The results indicate linear calibrations between 1 31026 M and 1 3 1025 M for both herbicides, but a higher sensitivity (2.5 times) was observed for single use SPZMEs compared to that obtained after soaking the electrodes in the electrolyte solution before the analyte accumulation. This might be attributed to favoured ion exchange of the analyte in the absence of electrolyte cations.It is noteworthy that this decrease in sensitivity was not observed when soaking the electrode in solutions containing less than 1023 M metal cations. The proposed sensor could be used in case of accidental dissemination of these herbicides in aqueous environments.Selective uptake of paraquat over diquat by zeolite ZSM-5 Paraquat and diquat have very similar electrochemical behaviour so that their voltammetric distinction from each other remains impossible. Moreover the selective accumulation of one of these species in the presence of the other has still to be demonstrated. By considering the different sizes of these herbicides and the possibility for zeolites to accumulate cations by ion exchange and, in the same time, to discriminate between their size, the idea arose to find a zeolite liable to accumulate paraquat while excluding diquat.It is known that entrance to the zeolite framework is usually governed by the size of rings involving 6, 8, 10, or 12 oxygen atoms linked together by either silicon or aluminium atoms.31 In the zeolite A structure, the main aperture is formed by an 8-oxygen ring which does not allow diffusion of species larger than 4-5 Å; indeed, electrodes modified with this zeolite did not display any preconcentration behaviour towards paraquat.20,32 On the other hand, zeolite Y to which the entrance is controlled by a 12-oxygen ring ( ~ 8 Å) is able to accumulate both paraquat and diquat by ion exchange, and this has been exploited for the voltammetric determination of these species (Fig. 5). Therefore, a zeolite structure displaying pore apertures defined by 10-oxygen rings should be promising for the differentiation between PQ2+ and DQ2+.Fig. 3 Relative peak current response for release of 1.0 3 1026 M paraquat from several modified electrodes into 0.1 M NaCl: 10% zeolitemodified carbon paste electrode (a), and screen-printed carbon electrodes modified with 5% (b), 10% (c), and 15% (d) zeolite Y. Other conditions as in Fig. 2. Fig. 4 Peak current response for the uptake of 1.0 3 1026 M paraquat in the modified electrodes (as monitored by square wave voltammetry): 10% zeolite-modified carbon paste electrode (a), and screen-printed carbon electrodes modified with 5% (b), 10% (c), and 15% (d) zeolite Y.Other conditions as in Fig. 2. Fig. 5 Structures of paraquat (PQ2+) and diquat (DQ2+) compared to the 12-oxygen ring entrance to zeolite Y (face 111) and the 10-oxygen ring entrances (faces 010 and 100) to zeolite ZSM-5. 1188 Analyst, 1999, 124, 1185–1190Zeolite ZSM-5 offers this specification.33–35 In ZSM-5, the 10-oxygen rings are distorted two different degrees, defining a tridimensional lattice made of right and zigzag channels sizing respectively 5.3 3 5.6 Å and 5.1 3 5.5 Å.Fig. 5 illustrates the possibility for selective accumulation, showing that both PQ2+ and DQ2+ can readily enter the 12-oxygen ring of zeolite Y while only PQ2+ species could be accommodated within the channels of ZSM-5, DQ2+ being excluded because of its larger size. The hypothesis has been checked and results are presented in Table 2, indicating clearly the selective accumulation of PQ2+ over DQ2+ in ZSM-5. Whether alone or in the presence of DQ2+, PQ2+ is incorporated within the ZSM-5 structure at the same level (up to about 0.37 mmol g21) while DQ2+ is totally excluded from the internal ZSM-5 channels.By comparison, both herbicides are readily incorporated to a large extent (0.98 for PQ2+, and 0.81 for DQ2+) in the supercages of zeolite Y; when used in mixture, a slight preference for DQ2+ was observed, most probably because of favourable charge distribution in the zeolite framework.Another feature of the results of Table 2 is the high affinity of zeolites for PQ2+ and DQ2+ species (Y for both these herbicides, and ZSM-5 for PQ2+): when the initial concentration of the analytes was less than that corresponding to the saturation of the accessible ion exchange sites, all of them (macroscopically) were incorporated into the aluminosilicate with a concomitant leaching of equivalent concentrations of sodium ions in the solution. While the accumulation of PQ2+ and DQ2+ in zeolite Y can be exploited for their voltammetric quantification after back ion exchange with the electrolyte cation (Na+), by using zeolitemodified electrodes, the selective accumulation of PQ2+ in ZSM-5 was not appropriate to the sensor field because desorption was not possible.Indeed, the interactions between PQ2+ and ZSM-5 were so strong that no desorption occurred even under severe experimental conditions. No desorption was observed in high ionic strength medium (6 M NaCl), in concentrated acid (1 M HNO3 or HCl) or basic (1 M NaOH) solutions, in organic solvents (acetonitrile or dimethylsulfoxide), or even in the presence of an ion pairing agent for PQ2+ species (0.1 M sodium dodecylsulfate).Accordingly, no voltammetric signal was observed in these media for PQ2+ after accumulation at ZSM-5-modified (either carbon paste or screen-printed) electrodes.Sodium dithionite, known to reduce PQ2+ into the corresponding blue radical cation, was also tested and resulted in the blue coloration of zeolite particles, but no significant leaching of paraquat species in solution was observed. This indicates a very strong binding of this toxic herbicide to ZSM-5 which is most probably due to confinement effects in the sorption process36 because of the very similar size of PQ2+ compared to that of the zeolite channels.This makes zeolite ZSM-5 a promising sorbent for the treatment of solutions infected by this herbicide. Conclusions We have demonstrated that screen-printing technology allowed the preparation of new types of zeolite-modified electrodes, which are promising for the mass production of low-cost and single-use zeolite-based sensors, with significant advantages (faster response time and easier chemical regeneration) compared to the corresponding zeolite-modified carbon paste electrodes.Using the herbicides paraquat and diquat as representative electroactive probes, it was shown that the screen-printed carbon electrodes modified with zeolite Y were characterised by restricted diffusion of the solution in the bulk of the electrode, which is known to be a serious limitation in the use of zeolite-modified carbon paste electrodes for electroanalysis. In addition, the use of zeolite ZSM-5 resulted in the selective accumulation of paraquat over diquat. The interactions of this herbicide with the aluminosilicate framework were so strong that desorption was prevented in a wide range of experimental conditions, making ZSM-5 a good substrate for the irreversible uptake of this toxic species. Future applications of screen-printed zeolite-modified electrodes, particularly onsite field screening, would require dual-electrode strips, with a printed reference electrode along the working one.Acknowledgement Professor Bao-Lian Su from the Laboratoire des Matériaux Inorganiques at the Facultés Universitaires Notre-Dame de la Paix (Namur, Belgium) is gratefully acknowledged for providing us with a sample of zeolite ZSM-5.References 1 Recent Advances and New Horizons in Zeolite Science and Technology, ed. H. Chon, S. I. Woo and S.-E. Park, Stud. Surf. Sci. Catal., 1996, 102. 2 D. R. Rolison, R. J. Nowak, T. Welsh and C. G. Murray, Talanta, 1991, 38, 27. 3 A. Walcarius, Electroanalysis, 1996, 8, 971. 4 A. Walcarius, Anal. Chim. Acta, 1999, 384, 1. 5 D. R. Rolison, Chem. Rev., 1990, 90, 867. 6 D. R. Rolison, Stud. Surf. Sci. Catal., 1994, 85, 543. 7 J. Wang and T. Martinez, Anal. Chim. Acta, 1988, 207, 95. 8 J. Wang and A. Walcarius, J. Electroanal. Chem., 1996, 404, 237. 9 J. Wang and A. Walcarius, J. Electroanal. Chem., 1996, 407, 183. 10 G. Marko-Varga, E. Burestedt, C. J. Svensson, J. Emnéus, L. Gorton, T. Ruzgas, M. Lutz and K. Unger, Electroanalysis, 1996, 8, 1121. 11 S.V. Guerra, C. R. Xavier, S. Nakagaki and L. T. Kubota, Electroanalysis, 1998, 10, 462. Table 2 Distribution of paraquat (PQ2+) and diquat (DQ2+) between solution and zeolite phases as a function of the zeolite type Initial conditions in solutiona Composition of zeolites at the equilibriumb PQ2+ added/ mM DQ2+ added/ mM PQ2+ in Y/ mmol g21 DQ2+ inY/ mmol g21 PQ2+ in ZSM-5/ mmol g21 DQ2+ in ZSM-5/ mmol g21 0.10 — 0.33 — 0.32 — 0.20 — 0.67 — 0.36 — 0.30 — 0.98 — 0.37 — — 0.10 — 0.32 — < 0.005c — 0.20 — 0.57 — < 0.005c — 0.30 — 0.81 — < 0.005c 0.05 0.05 0.16 0.17 0.15 < 0.005c 0.10 0.10 0.29 0.33 0.29 < 0.005c 0.15 0.15 0.43 0.50 0.35 < 0.005c a Slurry: 10 mg zeolite in 30 ml solution.b Calculated from the measured residual concentration of PQ2+ and DQ2+ in solution after equilibration. c Lowest measurable value. Analyst, 1999, 124, 1185–1190 118912 A. Walcarius, Anal. Chim. Acta, 1999, 388, 79. 13 A. Walcarius, L. Lamberts and E.G. Derouane, Electrochim. Acta, 1993, 38, 2257. 14 S. A. Wring and J. P. Hart, Analyst, 1992, 117, 1281. 15 J. P. Hart and S. A. Wring, Electroanalysis, 1994, 6, 617. 16 J. Wang, Analyst, 1994, 119, 763. 17 C. A. Galan-Vidal, J. Moñoz, C. Dominguez and S. Alegret, Trends Anal. Chem., 1995, 14, 225. 18 J. P. Hart and S. A. Wring, Trends Anal. Chem., 1997, 16, 89. 19 K. B. Yoon and J. K. Kochi, J. Am. Chem. Soc., 1989, 111, 1128. 20 A. Walcarius, T. Barbaise and J. Bessiere, Anal. Chim. Acta, 1997, 340, 61. 21 L. Bird and A. T. Kuhn, Chem. Soc. Rev., 1981, 10, 49. 22 H. A. Gemborys and B. R. Shaw, J. Electroanal. Chem., 1986, 208, 95. 23 J.-M. Zen, S.-H. Jeng and H.-J. Chen, Anal. Chem., 1996, 68, 498. 24 B. Barroso-Fernandez,, M. T. Lee-Alvarez,, C. J. Seliskar and W. R. Heineman, Anal. Chim. Acta, 1998, 370, 221. 25 B. Brunetti and P. Ugo, J. Electroanal. Chem., 1999, 460, 38. 26 J. A. Alden, J. A. Cooper, F. Hutchinson, F. Prieto and R. G. Compton, J. Electroanal. Chem., 1997, 432, 63. 27 A. Walcarius,, L. Lamberts and E. G. Derouane, Electrochim. Acta, 1993, 38, 2267. 28 B. R. Shaw and K. E. Creasy, J. Electroanal. Chem., 1988, 243, 209. 29 C. Bing and L. Kryger, Talanta, 1996, 43, 153. 30 B. Chen, N.-K. Goh and L.-S. Chia, Electrochim. Acta, 1997, 42, 595. 31 D. W. Breck, Zeolites Molecular Sieves, Structure, Chemistry and Use, Wiley, New York, 1974, (original edition); R. E. Krieger, Malabar, Florida 1984 (new edition). 32 B. R. Shaw, K. E. Creasy, C. J. Lanczycki, J. A. Sargeant and M. Tirhado, J. Electrochem. Soc., 1988, 135, 869. 33 E. M. Flanigen, J. M. Bennett, R. W. Grose, J. P. Cohen, R. L. Patton, R. M. Kirchner and J. V. Smith, Nature (London), 1978, 271, 512. 34 G. T. Kokotailo, S. L. Lawton, D. H. Olson and W. M. Meier, Nature (London), 1978, 272, 437. 35 D. H. Olson, G. T. Kokotailo, S. L. Lawton and W. M. Meier, J. Phys. Chem., 1981, 85, 2238. 36 E. G. Derouane, J. Mol. Catal. A, 1998, 134, 29. Paper 9/04025K 1190 Analyst, 1999, 124, 1185–1190

ISSN:0003-2654    

DOI:10.1039/a904025k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

A program for the weighted linear least-squares regression of unbalanced response arrays Elio Desimoni* DIFCA, Università degli Studi di Milano, Via Celoria 2, 20133 Milan, Italy. E-mail: desimoni@unimi.it Received 22nd March 1999, Accepted 28th June 1999 A program is described to establish calibration diagrams by weighted, linear, least-squares regression of unbalanced response arrays. Whatever the confidence level, the number of analysed standard solutions and of their available replicates, the program allows (i) testing for scedasticity, linearity, outliers and normality, (ii) evaluation of slope and intercept of the calibration function and their confidence interval and (iii) evaluation of an unknown concentration and its confidence interval by interpolation/extrapolation. For negative results of the linearity test, the program structure allows a rapid evaluation of data to be discarded to attempt entering the linear range.The program was validated by analysing response arrays obtained by adding a Gaussian noise to known response/concentration functional relationships. The results of validation tests led to the implementation of an empirical but efficient way to correct regression results when certain experimental situations lead to unjustified over-weighting of some responses.The analysis of some real calibration data sets allows the versatility of the software to be evaluated. Introduction Calibration graphs in instrumental analysis can be obtained by ordinary, linear, least-squares regression (LLSR) if (i) errors occur only in the y-direction, (ii) errors are normally distributed, (iii) variance does not change with concentration and, of course, (iv) responses are linearly related to concentration.1–5 The first assumption is usually valid, unless the insufficient quality of standard solutions used to establish the calibration graph suggests the application of more suitable regression techniques. 1,5,6 The second assumption is also usually valid, unless when working, for example, close to physical limits, as with measurements at the lowest detectable concentrations. The third assumption has always to be tested because, when data are heteroscedastic, weighted LLSR (wLLSR) is mandatory. The last assumption has always to be tested also, because deviations from linearity are also frequent. Of course, exponentials, logarithms, parabolas, etc., are perfectly valid (and often theoretically appropriate) calibration relationships but, when choosing linear least-squares regression, testing for linearity is mandatory.Recommended tests of normality, scedasticity and linearity require repeated observations of the standard solutions used in establishing the calibration graph. Depending on the verdict of these tests, the analysis of regression is a creative and interactive process, which needs to be quality-controlled.7 Surprisingly, as underlined in a recent IUPAC Technical Report,8 the analytical literature rarely describes important statistical details related to the calibration design and to the processing of data in the regression analysis.Also, many popular statistical packages do not always warn the user to perform scedasticity and linearity tests before LLSR and/or rarely offer wLLSR. Moreover, they do not easily allow processing of unbalanced data arrays, e.g. those containing a different number of observations of some tested solutions.Similar arrays of responses can result, for example, after elimination of outliers from balanced data arrays acquired by modern processor-driven instruments. In these cases, adequate weighting is necessary.8 Because of this situation, some workers developed in-house programs. An example is the program CALWER.9 Developed in Excel from Microsoft® and proposed to perform weighted regression, CALWER allows the use of several types of calibration functions and variance models and, moreover, it permits a traceable link between raw data and reported values.More recently,10 a program developed in Mathcad 7 Professional from Mathsoft®, LLSRR, was specifically aimed to test automatically scedasticity and linearity of unbalanced response arrays by specific F-tests, but when performing ordinary LLSR only. This paper describes a new version of LLSRR, namely uwLLSR, suitable for performing interactively the weighted LLSR of unbalanced, heteroscedastic data in the light of internal routines for testing linearity, scedasticity, outliers and normality.The approach is as pragmatic as possible, avoiding the use of sophisticated or robust techniques, to allow an easy understanding by analysts interested in improving the quality of their data processing without a full immersion in high level statistics. As was previously done,10 the proposed worksheet is validated by analysing noised data whose true calibration function and signal-to-noise ratio are a priori known.Experimental uwLLSR was prepared as a template in Mathcad 7.02a Professional (MathSoft Inc.®, Cambridge, MA, USA). Since the software always maintains 15 digits of precision internally, even if displayed data can be rounded as convenient or logical, people processing the same response arrays by using a different precision might obtain results slightly different from those reported in this paper. Unbalanced arrays of noised responses used in validation tests were obtained by: (i) choosing a response/concentration functional relationship S(C) = b.C + a (1) and a standard deviation/concentration functional relationship s(C) = p.C + q (2) s(C) = Ap2·C2 + q2 (3) s(C) = p.Ck + q (4) Analyst, 1999, 124, 1191–1196 1191[even though eqn. (3) is the most correct,11,12 in this work all relationships were used in turn to add a Gaussian noise to responses obtained from eqn.(1)]; (ii) generating a balanced array of noised responses, Si,j, by the Mathcad function Si,j = rnorm(jj,S(Ci), s(Ci)) (5) which, for 1@i@ii, returns a vector of jj random responses from normal distributions with means S(Ci) and standard deviations s(Ci), respectively (ii and jj are the maximum values of row and column range variables, i and j, respectively); (iii) cutting-off the 15-digit noised responses obtained from eqn.(5) at the second decimal place by the Mathcad function Si j , ) ¡¿ oor(102 102 (6) (iv) and, finally, arbitrarily dropping some of the noised responses to obtain an unbalanced, noised data set: for example, after discarding S2,3, S2,4 and S4,4 the original balanced 4 3 4 array S S S S S S S S S S S S S S S S S S S S S S S S m m S S : : , , , , , , , , , , , , , , , , , , , , , , , = E I IIIII ¢� ¡Æ ¢«¢«¢«¢«¢« �¡ = 11 1 2 1 3 1 4 2 1 2 2 2 3 2 4 3 1 3 2 3 3 3 4 4 1 4 2 4 3 4 4 11 1 2 1 3 1 4 2 1 2 2 3 1 3, , , , , , 2 3 3 3 4 4 1 4 2 4 3 S S S S S m E I IIIII ¢� ¡Æ ¢«¢«¢«¢«¢« turns into an unbalanced array characterised by the 4,2,4,3 repetition design.Once an experimental situation has been defined by choosing proper values of a, b, p, q, k [if necessary, as in eqn. (4)] and a repetition design, one can generate and analyse as many data sets as necessary by simply clicking eqn. (5) in the Mathcad template. Results and discussion Description of the worksheet The program runs nine subsequent sections. Section 1: data input. Here one enters the selected level of significance, a, (from which any critical value necessary to tand F-tests is obtained through the Mathcad software probability density functions), the number of analysed solutions, ii, and the maximum available number of replicates, jj.Next, one enters concentration and response data in the default 9 3 1 and, respectively, 9 3 8 arrays. This is because uwLLSR is developed to analyse repeated observations of at least some of the standard solutions used to evaluate the calibration graph.The above-mentioned dimensions of the default arrays were chosen since, according to literature suggestions,13,14, the calibration is often unstable when the number of calibration points is less than six, while additional points beyond 20 do not give urther help. Moreover, the minimum number of observations necessary to obtain a useful estimate of standard deviation is about six.15 On considering that in most routine operations these numbers can barely be considered realistic, default arrays should suffice in almost all cases (if necessary, any other array dimension can be set). By using repeated observations, uwLLSR differs from CALWER.9 Eventual missing data in the arrays are entered as ¡®m¡�, so that the program can ignore them in the following calculations and in the final plots.Finally, one has also to set the values of Force and K parameters. The use of Force is explained in the section dealing with validation tests (see below).K is used to select the calibration graph or the standard additions mode (K = 1 or 0, respectively). When choosing the calibration graph mode, one has also to enter the number of observations of the unknown, M, and, in a specific array, the unknown response(s). Section 2: Bartlett test for homogeneity of variances. Among those available,1,3,16 the Bartlett test is the only test suitable for comparing more than two variances whose degrees of freedom (dof) are not equal,1 i.e.for managing unbalanced data arrays. The program estimates the standard deviations of all sets of repeated observations, and uses them to calculate and compare the experimental F value with the critical (1-tail) value relevant to the selected confidence level and to proper dof. If the experimental value of F is larger than the critical value, in other words if the VAR parameter displayed by the worksheet is higher than unity (VAR = test value/critical value), variances are not homogeneous.Even though not essential, since the worksheet always performs a weighted regression, the Bartlett test helps in understanding the actual degree of heteroscedasticity. Section 3: linear regression of standard deviation. The functional relationship between standard deviation and concentration is necessary to calculate weights, and consequently to evaluate confidence intervals (CIs), over the explored concentration range. Since usually only a few repeated measurements are available, standard deviation estimates are likely to be very poor.As suggested by the Analytical Methods Committee of the Royal Society of Chemistry,13 under the assumption that standard deviation can be approximated to a linear function of concentration, uwLLSR fits raw, experimental standard deviations by ordinary LLSR and plots the standard deviation functional relationship dv(C) = p¡�C + q (7) On knowing p and q, the program can calculate weights, wi, at any concentration by the equation4 w ii C C i i i i = ¡¿ - - A dv( dv( ) ) 2 2 (8) Per cent.weights are also calculated and displayed. By using experimental standard deviation values, uwLLSR differs from CALWER,9, which applies all the available variance models to single response arrays and selects the best one by comparing method performances via the logarithmic likelihood. Weights obtained by eqn. (8) cannot be used after data transformation, such as when attempting to obtain linear data from non-linear data.In these cases transformation-dependent weights should replace those defined above.4,17,18 Section 4: the line of regression of y on x. Since the lack-offit test13,19 needs the functional relationship between response and concentration, the weighted regression is performed by a priori hypothesising a linear relationship. Weighted correlation (r) and determination (r2) coefficients are evaluated by a properly modified equation, suitable for working with unbalanced data arrays. Standard equations of slope (b), intercept (a) and of their standard deviations (sb and sa, respectively) were also modified to manage unbalanced data arrays. Finally, the program returns the CI of slope and intercept, t.sb and t.sa, respectively (t is the two-tail, critical value at the desired confidence level and N 2 2 dof; N is the number of responses used in establishing the diagram). 1192 Analyst, 1999, 124, 1191¡©1196Section 5: lack-of-fit test.The program performs the lackof- fit test by analysing the residual variance.13,19. Also, in this case the usual equations are conveniently modified to manage unbalanced arrays. If the experimental value of F is lower than the critical, one-tail value, that is, if the LIN parameter (LIN = test value/critical value) is lower than 1.0, the hypothesis of linearity is not disproved: linearity can never be proved.19,20 Should linearity be disproved, transformations or curvilinear, spline functions or other non-linear regression methods would be mandatory. A simpler approach,2 involving the LLSR of data arrays in which the highest, next-highest, etc.point is omitted, can be easily attempted by simply changing ii to ii 2 1, ii 2 2, etc. As soon as this is done, uwLLSR returns updated results (see under Validation of the program). Section 6: evaluation of unknown concentrations. If linearity is not disproved, the regression line can be used to interpolate (calibration graph) or extrapolate (standard additions method) the unknown concentration, Cuk.uwLLSR estimates the standard deviation, s(Cuk), by using an equation which holds even when the slope is not highly significant,21 and the relevant CI, t.s(Cuk). When using the standard additions method, whatever the value of Suk eventually entered by the operator, Cuk is set to b/a and the equation of s(Cuk) is automatically corrected to take into account that no experimental response is used.2,4 When using the calibration graph method and M is greater than 1, the average unknown signal is used to interpolate the unknown concentration.Section 7: plotting regression line and confidence intervals of interpolated unknown. In this section, uwLLSR plots the calibration graph, with original data points, weighted regression line and CIs of interpolated unknowns. Underneath, it shows the plot of weighted residuals. Horizontal lines in Fig. 1 are drawn at ±sy/x and ±z(1-a/2).sy/x (sy/x is the residual standard deviation). Since weighted residuals can be considered a random sample from a normal distribution with mean zero and standard deviation sy/x, data outside the ±z(1-a/2).sy/x band in the plot of weighted residuals are suspect outliers, having only an a probability to pertain to the distribution of weighted residuals. Section 8: outliers test. In this section, eventual outliers can be checked by an F-test.3 To perform the test, one has to enter the actual values of the dof and of sy/x relevant to the response/ concentration functional relationship, and to go back to Section 1, where he/she can drop the suspect outlier from the response matrix by substituting it with m (false missing value).For the program to be running in the right way, the remaining data must occupy the leftmost places of the row. As soon as he/she returns to Section 8 after dropping the suspect response, the program recalculates all data and performs the F-test by using old and new sy/x and dof values:3 if the OUT parameter (OUT = test value/critical 1-tail value) is larger than 1, the suspect response can be permanently discarded.It should be noted that discarding outliers is controversial, because of the a risk of stating that the point is an outlier when, in fact, it is not. However, one should also consider that the lower is the selected a value, the lower the risk.Section 9: Shapiro–Wilk test. The Shapiro–Wilk test implemented in uwLLSR allows for testing normality of not less than three replicates.22 After choosing the specific significance level (0.5, 0.1, 0.05 or 0.01) and the row index, i, of the concentration whose data have to be tested, the worksheet calculates the NOR parameter (NOR = critical value/test value): if NOR is larger than unity, normality is not verified. However, because of the limited number of replicates usually available, testing normality for calibration graphs obtained by instrumental analyses is a challenging task: if normality is to be tested, a conveniently large number of replicates is necessary.If normality is not verified under real anacal conditions (e.g. when processing data sets containing few responses) one can also choose to use Gaussian statistics as a fairly good approximation, because alternatives usually require much effort. Below Section 9, the program summarises some regression results (LIN, VAR, correlation and determination coefficients, sy/x, b, a, Cuk and relevant CIs, etc.).As in CALWER,9 under this section the operator can save, as a text region, all information necessary to ensure a full traceability of the calibration procedure. Validation of the program Four experimental situations (A?D) were tested. Each was characterised by different functional relationship, RSD%, slopes, intercepts, calibration mode, dof, repetition designs, etc.These parameters, together with the theoretical unknown concentration value, Cuk, relevant to an arbitrarily selected response, Suk, simulating the results of the analysis of an unknown sample, are reported in Table 1. By choosing K equal to 0 or 1 (see Section 1), the standard additions or, respectively, the calibration graph mode was selected. Twenty-five data sets were repeatedly generated and analysed in each of the four situations. The results were averaged over the 25 runs.Average coefficient of determination, rm 2, slope, bm, and intercept, am, of the response functional relationship [eqn. (1)] are summarised in Table 2. Cukm in Table 2 is the average value of the interpolated, or extrapolated, unknown concentration. Table 2 also shows the average CIs of bm, am and Cukm (in all cases, the confidence level was chosen equal to 0.95). According to suggestions,23225 data are rounded to the second uncertain figure of their uncertainty (that is of their CIs).One can compare these results with the true data of Table 1. Outb, Outa and Outc in Table 2 indicate the number of times the true value of a parameter, (slope, intercept and Cuk values of Table 1) was found outside the relevant CI. During run D, Cuk Fig. 1 Example of calibration diagram obtained during run B; the upper plot displays original responses, weighted regression line and confidence band of interpolated values. The lower plot displays weighted residuals and ±sy/x and ±z(1-a/2).sy/x bands.Analyst, 1999, 124, 1191–1196 1193was found outside its experimental CI seven times. The average distance of these seven Cuk values from the closest limit of their experimental CI was 0.12 arbitrary units (a.u.). Nevertheless, the average CI over the 25 runs, 4.02 ± 0.22 a.u., embraced the true CI (4.0 a.u.). Sometimes the addition of a Gaussian noise to responses obtained from theoretically linear relationships led to non-linear data sets (as revealed by the lack-of-fit test).During validation runs A–D, no attempt was made to enter the linear range by data reduction or other methods, so that average rm 2, bm, am and Cukm values are those obtained whatever the response of the lack-of-fit test. The efficiency of the lack-of-fit test was checked by run E (see Table 1), simulating the establishment of a precision calibration graph by analysing in triplicate eight standard solutions: in this case the true response of the solutions having the two highest concentrations (70 and 80 a.u.in Table 1) was decreased by 4.25 and 8.68%, respectively, before adding the Gaussian noise. This allowed the simulation of a net deviation from linearity at the upper end of the explored concentration range. After the addition of Gaussian noise according to the selected standard deviation functional relationship (see Table 1), 25 data sets were generated and analysed. In nine cases the response/concentration functional relationship was found to be linear, in 13 cases non-linearity was avoided by dropping data relevant to C = 80 a.u., and in three cases non-linearity was avoided by dropping data relevant to C = 80 a.u.and C = 70 a.u. The smoothing of raw standard deviations by ordinary LLSR13 is not problem-free. Because of the unfavourable experimental situation C, the intercept of the standard deviation regression function [q in eqn. (7)] was frequently found to be much too low.The closer to zero it is, the higher the weight assigned to responses of the relevant concentration. If weights at the lowest concentration are overestimated, uwLLSR returns an abnormally wide and unjustifiable CI of interpolated values (see Fig. 2). Should this happen in real cases, additional measurements should be attempted to by-pass the problem. Alternatively, uwLLSR allows the standard deviation regression line to be forced through the first point, s1/C1, by simply changing the Force parameter in Section 1 from 0 (default) to 1: this forcing (empirical, but standard deviation cannot be zero) usually suffices to increase slightly the standard deviation intercept and to obtain reasonably narrow CIs (see Fig. 3). In the course of all the 100 simulations in runs A–D, the above situation was met only six times, always when analysing data sets of situation C. In these cases, the results of forced standard deviation regression were retained for calculating average values displayed in Table 2.As evident in Fig. 2 and 3, two points lie outside the ±2sy/x band. After their elimination, in agreement with the results of the outlier test, the results displayed in Fig. 4 were obtained. Table 3 shows weights, determination coefficients and standard deviations of interpolated concentration obtained in Fig. 2–4: the best sy/x and sCuk values were obtained after forcing and eliminating the two outliers.Fig. 2–4 represent a good example of creative and interactive analysis.7 Examples of real calibration data sets The program is being routinely used in the author’s laboratory. Table 4 presents a real calibration data set, relevant to the quantification of sulfides by flow injection analysis with amperometric detection at a palladium–vitreous carbon chemically modified electrode.26 The program outputs when choos- Table 1 Parameters selected for the validation tests (all parameters are in arbitrary units) Run A Run B Run C Run D Run E S(C) = b.C + a 15.0.C + 20.0 -2.5.C-1.0 5.0.C + 10.0 4.0.C + 16 5.0.C+3 C values 0.0–10.0–20.0– 30.0–40.0–50.0 0.0–50–100–150 0.0–2.0–6.0– 8.0–10.0–12.0 0.0–4.0–8.0–12 10,20,30,40,50, 60,70,80 Repetition design (ri) 4.2.1.1.3.4 3.3.1.3 4.3.1.4.3.4 5.3.1.5 3.3.3.3.3.3.3.3 N 15 10 19 14 24 s = f(C) 0.8.C + 0.5 0.05.C1.1 + 0.03 0.5.C + 0.3 A0.07·C2 + 0.8 A0.04·C2 + 0.05 s range 0.50?40.50 0.03?12.41 0.30?6.30 0.89?3.30 0.22?12.60 RSD% range 2.50?5.26 –2.96?–3.30 3.00?9.00 4.33?5.15 3.96?7.35 Suk/Cuk 0.0/1.33 –200/79.6 45.0/7.0 0.0/4.0 52.3/9.86 Experimental method Standard additions Calibration graph Calibration graph Standard additions Calibration graph Table 2 Average results of validation tests (all parameters are in arbitrary units) Run A Run B Run C Run D rm 2 0.9952 0.9991 0.9798 0.9921 bm 14.98 ± 0.55 –2.500 ± 0.056 4.98 ± 0.32 3.99 ± 0.21 am 20.1 ± 2.4 –1.1 ± 1.4 10.08 ± 0.41 16.02 ± 0.77 Cukm 1.34 ± 0.18 79.6 ± 4.7 7.0 ± 1.8 4.02 ± 0.22 Outb 5 1 3 3 Outa 0 0 3 4 OutC 1 1 – 7 Fig. 2 Example of over-weighting in analysing situation C. 1194 Analyst, 1999, 124, 1191–1196ing different a risks are summarised in Table 5. It can be observed that the raw data set results are non-linear, even when choosing a very low a risk (0.005). The curvature of the calibration plot can be eliminated by discarding data relevant to the two most concentrated standard solutions. After this, a suspect outlier can be identified when choosing a = 0.05.However, no outlier is evident when choosing a = 0.01 or lower. The use of a = 0.01 and of responses of only the first six standard solutions allows the coefficient of determination to be increased from 0.9990 to 0.9997 and sy/x to be lowered from 0.0496 to 0.0267. The final result is displayed in Fig. 5. Two additional examples of real calibration data sets are supplied as Electronic Supplementary Information.† Conclusions The results of validation tests confirmed that uwLLSR allows one to evaluate slope and intercept and their CIs, to interpolate/ extrapolate the unknown concentration and to evaluate its CI whatever the actual experimental conditions (scedasticity, number of standards, number of replicate observations of each standard and of the unknown).The available (facultative) outliers and normality tests, the possibility to reduce the explored concentration range easily (by changing ii to ii 2 1, ii 2 2, and so on) and to adjust weights by forcing the standard deviation regression line, permit fully interactive operations.The program can be used to process homoscedastic data arrays also, since weighted regression always produces superior results than ordinary regression.2,4,9 Of course, uwLLSR can be used to process balanced data arrays also, since they are particular cases of unbalanced data arrays. When using a Pentium II IBM-compatible PC, results are returned about 10–20 s after the necessary inputs.† Available as Electronic Supplementary Material; see http://www.rsc.org/ suppdata/an/1999/1191. Fig. 3 Same example as in Fig. 1 but after forcing the standard deviation/ concentration functional relationship through the first data point. Fig. 4 Same example as in Fig. 3 but after elimination of the lowest and highest responses of the lowest concentration. Table 3 Per cent. weights, coefficients of determination and sCuk values relevant to Fig. 2 (before forcing), 3 (after forcing) and 4 (after forcing and elimination of two outliers) Fig. 2 Fig. 3 Fig. 4 w1 99.88% 96.61% 98.94% w2 0.095% 2.58% 0.83% w3 0.011% 0.36% 0.10% w4 0.0062% 0.21% 0.060% w5 0.0040% 0.14% 0.039% w6 0.0028% 0.096% 0.027% r2 0.8928 0.9836 0.9860 sCuk 0.1774 0.0654 0.0635 Table 4 Calibration data set of sulfides in flow injection analysis at a palladium–vitreous carbon chemically modified electrode Concentration/mM Response/mA 0.88 0.170 0.211 0.179 1.77 0.341 0.349 0.353 3.55 0.663 0.684 0.656 7.10 1.37 1.34 1.39 14.2 2.60 2.58 2.64 28.4 5.29 5.24 5.26 56.8 10.83 11.10 11.05 81.2 15.85 15.75 15.94 Table 5 Effect of the selected a level on Bartlett, lack–of–fit and outlier tests No.of standard solutions a = 0.05 a = 0.01 a = 0.005 n = 8 Heteroscedastic, non–linear Heteroscedastic, non–linear Heteroscedastic, non–linear n = 7 Heteroscedastic, non–linear Heteroscedastic, non-linear Heteroscedastic, non-linear n = 6 Homoscedastic, non–linear 1 suspect outlier Homoscedastic, linear Homoscedastic, linear Analyst, 1999, 124, 1191–1196 1195The described worksheet can be obtained free from the author by E-mail.Acknowledgements This work was carried out with the financial support of the Italian National Research Council (C.N.R., Rome) and of Ministero dell’Università e della Ricerca Scientifica (MURST, Rome). References 1 R. L. Anderson, Practical Statistics for Analytical Chemists, Van Nostrand Reinhold, New York, 1987. 2 J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Ellis Horwood, Prentice Hall, New York, 1993. 3 W. Funk, V. Damman and G. Donnevert, Quality Assurance in Analytical Chemistry, VCH, Weinheim, 1995. 4 J. N. Miller, Analyst, 1991, 116, 3. 5 J. W. Einax, H. W. Zwanziger and S. Geiß, Chemometrics in Environmental Analysis, VCH, Weinheim, 1997. 6 W. Bremser and W. Hässelbarth, Anal. Chim. Acta, 1997, 348, 61. 7 D. C. Crocker, How to use Regression Analysis in Quality Control, ASQC Quality Press, Milwaukee, WI, 1990. 8 J. Mocak, A. M. Bond, S. Mitchell and G. Scollary, Pure Appl. Chem., 1997, 69, 297. 9 S. M. Gort and R. Hoogerbrugge, Chemom. Intell. Lab. Syst., 1995, 28, 193. 10 E. Desimoni, A. Daghetta and S. Valsecchi, Ann. Chim. (Rome), 1998, 88, 601. 11 M. Thompson, Analyst, 1988, 113, 1579. 12 D. M. Rocke and S. Lorenzato, Technometrics, 1995, 37, 176. 13 Analytical Methods Committee, Analyst, 1994, 119, 2363. 14 G. T. Wernimont, Use of Statistics to Develop and Evaluate Analytical Methods, AOAC International, Arlington, VA, 1987. 15 L. A. Currie and G. Svehla, Pure Appl. Chem., 1994, 66, 595. 16 K. Baumann and H. Wätzig, J. Chromatogr., 1995, 700, 9. 17 R. de Levie, J. Chem. Educ., 1986, 63, 10. 18 B. R. Ramachandran, J. N. Allen and A. M. Halpern, Anal. Chem., 1996, 68, 281. 19 M. G. Natrella, Experimental Statistics Handbook 1991, National Bureau of Standards, Gaithersburg, MD, 1991. 20 Analytical Methods Committee, Analyst, 1988, 113, 1469. 21 I. L. Larsen, N. A. Hartmann and J. J. Wagner, Anal. Chem., 1973, 45, 1511. 22 S. S. Shapiro, How to Test Normality and Other Distributional Assumptions, ASQC Quality Press, Milwaukee, WI, 1990. 23 W. Horwitz, Pure Appl. Chem., 1995, 67, 331. 24 Analytical Methods Committee, Analyst, 1987, 112, 679. 25 Guide to the Expression of Uncertainty in Measurement, ISO, Geneva, 1993. 26 I. G. Casella, M. R. Guascito and E. Desimoni, unpublished work. Paper 9/02251A Fig. 5 Calibration graph relevant to the quantification of sulfides by flow injection analysis with amperometric detection at a palladium–vitreous carbon chemically modified electrode. Calibration function: S(C) = (0.1841 ± 0.0025).C+(0.024 ± 0.024); a = 0.01; n = 16. 1196 Analyst, 1999, 124, 1191–1196

ISSN:0003-2654    

DOI:10.1039/a902251a

出版商:RSC    

年代:1999

数据来源: RSC