摘要:

Tutorial Review Electrofunctional polymers: their role in the development of new analytical systems T. W. Lewis,a G. G. Wallace*a and M. R. Smythb a Intelligent Polymer Research Institute, University of Wollongong, Northfields Avenue, Wollongong, NSW 2522, Australia b School of Chemical Sciences, Dublin City University, Dublin 9, Ireland Received 15th October 1998, Accepted 22nd December 1998 1 Introduction 2 Electrofunctional polymers 2.1 Ferrocene containing polymers 2.2 Quinone containing polymers 2.3 Polymers containing Os, Ru complexes 2.4 Electronically conducting polymers 3 On-line derivatisation, controlled release, preconcentration, separation 4 Sensors 5 Conclusions and future developments 6 Acknowledgements 7 References 1 Introduction The overall analytical process involves a number of steps: sample collection, transportation and storage, analysis, data collection, processing and evaluation.Most of these essential steps commonly involve the use of polymeric materials in one form or another. Inert polymers are the usual materials of choice for sample collection, transportation and storage, and are indirectly involved in the form of engineering polymers in the other steps.However, it is the so-called ‘functional polymers’ that have most to contribute to the actual analytical step, including separation of chemical components, specific interactions and signal generation. Functional polymers are designed to undergo predetermined molecular interactions enabling their use in analyte preconcentration, separation (membrane or chromatographic) or derivatisation.Polymers containing ligand groups, for example, are used to preconcentrate metals while others with biological receptors attached are used in protein purification. More recently, the design and use of electrofunctional polymers has added a new dimension to these capabilities. With these polymers, the molecular interaction parameters can be adjusted in situ by the use of electrical stimulation to induce redox reactions within the polymer. The injection/removal of charge and the subsequent changes in chemical and/or physical properties of the polymer markedly alter the ability of the material to interact with other chemical species.For such an affect to be of direct benefit, the polymer should be an electronic conductor. This then gives rise to the ability to use such electrofunctional polymers as sensors since electrical signals (DC, Di, DR) can be readily transmitted back through the material to the electronic measuring device. 2 Electrofunctional polymers As described above, the application of an electrical potential to an electrofunctional polymer will alter the physical and/or chemical properties of the material. This in turn alters its ability to undergo particular molecular interactions or to trigger subsequent chemical events. These changes/events can be used to advantage in the development of analytical systems.The most common application of electrofunctional polymers in the analytical sciences is in the development of new sensors. Professor Gordon Wallace is Director of the Intelligent Polymer Research Institute (IPRI) at the University of Wollongong. He received his PhD degree in 1984 from Deakin University. Since then he has pioneered research into the development of Intelligent Polymer Systems. His first lecturing appointment was at University College Cork.He took up a position as a Lecturer in the Chemistry Department (University of Wollongong) in September 1985 and was promoted to Professorial Research Fellow in 1990. He has published more than 200 refereed scientific articles and presented more than 200 conference lectures at international symposia. Along with colleagues in IPRI he has developed several new sensing, membrane separations, controlled release and biomaterial technologies in recent years. Trevor W.Lewis was a parttime Lecturer in Analytical/Environmental Chemistry at the University of Wollongong, NSW, Australia, from 1985 to 1990, and has occupied his current university position as a full-time Lecturer since 1994. From 1990 to 1993 he was Director of a contract chemical and microbiological analysis laboratory within the University. He has authored over 30 journal articles and conference presentations and presented 12 invited lectures, largely in the areas of functional polymers and their applications. Since 1990 he has also supervised 20 research students in various aspects of environmental chemistry and functional polymers.For Professor Malcolm Smyth’s biography, see Analyst, 1993, 118, 315. Analyst, 1999, 124, 213–219 213However, other emerging areas of novel application are poised to have a significant impact. 2.1 Ferrocene containing polymers It is well known that ferrocene (Fc) undergoes a reversible one electron transfer process in solution: Fc " Fc+ + e (1) Ferrocene is often used as an electroactive component of gels and also other soluble polymers.For example, previous workers1 have incorporated Fc as an integral part of a crosslinked polyacrylamide gel (Scheme 1). The gel like material provides an excellent media for encapsulation of active enzymes while the Fc provides a non-diffusional electron transfer mediator. As such, the gel material has been used in the development of biosensors for detection of glucose.Others2 have attached ferrocene to an acrylamide backbone (Scheme 2) and produced useful electrofunctional coatings by forming interpolymer complexes with polyacrylic acid. Again, glucose oxidase was used to demonstrate the usefulness of this material in providing electron transfer mediation in the development of a biosensor. Another interesting example involves copolymerisation of vinylferrocene with other vinyl monomers containing complexing quaternary groups.Such polymers have been used to ion pair with useful ligands such as shown in Scheme 3. These can be placed on electrode surfaces and used to bind metal ions and thus form the basis of a sensor. Interestingly, the potential applied to the polymer influences the oxidation state of the ferrocene and the subsequent metal ion binding capabilities.3 The change in polymer charge density upon oxidation of the ferrocene groups indirectly influences the ability of the incorporated ligands to bind metal ions.Similar effects can be obtained by producing copolymers containing Fc pendant groups and other complexing groups, as shown in Scheme 4. Again, it has been demonstrated that the applied potential influences the complexing ability of the ligand (L). 2.2 Quinone containing polymers Another common redox group attached to polymer backbones is quinones. Quinones are known to undergo reversible electron transfer processes in solution that can be described according to C6H4O2 + 2e2 + 2H+ ? C6H6O2 (2) The simplest of these are polymers or copolymers containing the benzoquinones4 (Scheme 5).Scheme 1 A cross-linked polyacrylamide gel containing ferrocene (from ref. 1). Scheme 2 Ferrocene attached to an acrylamide backbone (from ref. 2). Scheme 3 Ferrocene–methylpyridinium vinyl copolymers with a representative range of useful ligands (from ref. 3). Scheme 4 Copolymerisation of vinylferrocene and other vinyl monomers (from ref. 3). 214 Analyst, 1999, 124, 213–219The use of electroinactive monomers (e.g., styrene) as spacers enhances the rate observed for the electrochemical reduction and oxidation of the quinone groups. These workers highlight the importance of the nature of the solvent when dealing with electrofunctional polymers. In solvents that readily solvate the polymer (e.g., DMSO), chains will be extended giving greater permeability to ion and solvent with enhanced electroactivity. In a non-solvating medium (e.g., water) the hydrophobic polymer will become more tightly packed and the reverse will be true.Other workers5 have used quinone-containing polymers (Scheme 6) to produce coatings that induce an electrocatalytic response towards oxygen. Another interesting example of quinone containing polymers acting as electron transfer mediators has been their use in development of biosensors,6 a glucose biosensor in this case. The polymers shown in Scheme 7 were employed. Both polymers proved effective as electron transfer mediators.However, the cross-linked polymers obtained by heating at 45–50 °C were more stable. Initial electroactivity decreased to about 60% of the non-crosslinked material, presumably owing to decreased flexibility of the polymer chain. Polyvinylferrocene has also been utilised as a redox active polymer for electron transfer mediation in the development of biosensors.7 2.3 Polymers containing Os, Ru complexes Another important group of polymers with applications to sensing are those containing Os or Ru centres coordinated within the polymers.The electroactivity of these polymers is associated with the reversible redox behaviour of the metal centres. These polymers have been utilised as electron transfer mediators for the detection of inorganic species such as nitrite,8–12 Fe(ii)/Fe(iii)13 or as biosensors. The employment of redox polymers in the construction of enzyme-containing sensitive films on electrode surfaces has led to the appearance of a new generation of amperometric biosensors. 14,15 The enzyme on the sensor surface is molecularly wired with non-diffusionally mediating species bound in long polymer chains, which results in a direct electrical communication between the enzyme and the electrode surface. There are several advantages to the use of non-diffusional mediators; most notably, the mediator is prevented from leaching out of the sensor membrane, thereby eliminating the need for a containment membrane, while simultaneously improving the lifetime of the sensor.16 Many redox polymers, such as ferrocene-modified polysiloxane, 17 polypyrrole,18 polyquinone19 and polyvinylpyridine, 12 have been investigated as electron transfer mediators.Hale et al.15,17 used a ferrocene derivative bound to an insoluble siloxane polymer to incorporate the enzyme and facilitate rapid electron transfer. Heller and co-workers14,20–24 have reported attempts with modified enzymes based on covalent bonding of the enzyme with polymers containing rapidly responding redox couples.In this so-called ‘molecular wiring’ on the electrode surface, hydrogels of long and flexible molecules of [Os(bpy)2(PVP)10Cl]Cl (Os-polymer; bpy = 2,2A-bipyridyl, PVP = poly-4-vinylpyridine) are formed, which result in the redox centres in the enzyme being electrically connected to the electrode surface via molecular wires containing the redox active species. This approach has been successfully employed to develop a compact and self-contained microelectrode for subcutaneous in vivo amperometric monitoring of glucose.25 In addition, the construction of other osmium redox polymercontaining sensors has been investigated by Smyth and coworkers26 –28 to increase both the sensitivity and stability of such sensors and, at the same time, to eliminate the influence of interferents on their response.The performance of such sensors has also been investigated in the organic phase, where the maintenance of the enzyme activity in organic solvents with minimal water content suggests the high stability of the sensing and electron transferring element.29 Recently, Lu et al.30 reported the development of an amperometric immunosensor with a non-diffusional redox polymer to transfer electrons between the electrode surface and the antigen horseradish peroxidase (HRP) bound to the anti- HRP antibody on the sensing surface.The redox polymer [Os(bpy)2(PVP)10Cl]Cl was co-immobilised with the antibody on the electrode surface by cross-linking between the antibody with glutaraldehyde to form a combined sensing and electron transfer system. When the antigen (HRP) was bound to the antibody on the surface of the sensing film, ‘electrical wiring’ occurs between the electrode and the redox centres in the bound HRP. The immunosensors showed greatly enhanced performance in terms of the magnitude of the response and the detection limit compared with that observed in a similar sensor employing a diffusional mediator. 2.4 Electronically conducting polymers Perhaps the most fascinating group of electrofunctional polymers as far as use in analytical devices is concerned is the Scheme 5 Quinone electron transfer mediators (from ref. 4). Scheme 6 Poly(naphthoquinone) (from ref. 5). Scheme 7 Cross-linked quinone-containing polymers obtained by heating (from ref. 6). Analyst, 1999, 124, 213–219 215electronically conducting polymers such as polypyrrole (PPy), polyaniline (PAn) and polythiophene (PTh).With these materials, redox changes are not localised at a specific centre but rather delocalised over a number of conducting polymer groups. It is then the polymer backbone that is amenable to oxidation/ reduction and, for PPy, this can be represented by the equation. Polythiophenes undergo a similar oxidation reduction response, whereas polyanilines undergo two distinct redox processes in acidic media, as illustrated in eqns.(4) and (5). These conducting polymers are capable of functioning as electron transfer mediators,31–33 as has been demonstrated in the oxidation of ascorbic and other organic acids32 or cytochrome c33 at polypyrrole coated electrodes. Polyanilines have been shown to provide electrocatalytic effects towards organic acids.34 This role can be amplified by incorporation of metallic sites, e.g., Pt, Pd, to function as electrocatalysts.35,36 The ability of conducting polymers to act as electron transfer mediators for proteins such as cytochrome c has been shown to be facilitated by the attachment of appropriate functional groups, e.g., 3-methyl-4-pyrrolecarboxylic acid33 (Scheme 8).Alternatively, electron transfer mediating counter ions can be incorporated. For example, Fe(CN)6 42 can be incorporated into conducting polymers at the time of synthesis and then act as an efficient electron transfer mediator for cytochrome c.Interestingly, Fe(CN)6 42 in solution does not function as an electron transfer mediator for cytochrome c and does so only when incorporated into the polymer. The polymer itself without Fe(CN)6 42 incorporated does not act as an electron mediator.37 Others38 have utilised the monomer shown in Scheme 9. After polymerisation, Fe(CN)6 42 was incorporated to function as the electrocatalytic site. Other electron transfer mediators incorporated into inherently conducting polymers include ferrocenes39 that have been incorporated as counter ions after sulfonation (Scheme 10).Others40,41 have attached porphyrins to act as electron transfer mediators for chemical sensors and biosensors. In some cases, sulfonated porphyrins (Scheme 11) have been incorporated as counter anions. Others have covalently attached the porphyrin to suitable monomers prior to polymerisation (Scheme 12). Polyanilines, with no additional electron transfer mediator,42 have been used to provide electron transfer mediation in the amperometric detection of the copper protein ceruloplasmin.In the case of conducting polymers, a range of other functional moieties have been incorporated to introduce specific (3) Oxidation/reduction of polypyrrole (4) (5) Scheme 8 3-Methyl-4-pyrrolecarboxylic acid repeat unit. Scheme 9 Protonated or quaternised 3-(pyrrol-1-yl)pyridine (from ref. 38). Scheme 10 Poly(1-vinylferrocene-2-sulfonate) (PVFS).Scheme 11 Tetrasulfophthalocyanine metal chelate (M = Co, Cu, Ni). Scheme 12 An example of an electropolymerisable porphyrin. 216 Analyst, 1999, 124, 213–219molecular interaction capabilities. These include metal complexing groups43 and enzymes44–51 which have been incorporated into Ppy or Pan. Bilirubin oxidase has been incorporated into poly(1,5-dihydroxynaphthalene).49 In fact, in this case the enzyme was utilised to initiate the polymerisation reaction and was subsequently incorporated into the conducting polymer.Antibodies52,53 and even whole living red blood cells (RBCs)54,55 have been incorporated. In the case of the RBCs, the ability of antibodies, retained as part of the integrated cell membrane structure, to provide molecular recognition capabilities has been demonstrated. 3 On-line derivatisation, controlled release, preconcentration, separation The dramatic physical and chemical changes that accompany oxidation/reduction of electrofunctional polymers can be used in the development of novel analytical systems.For simple conducting electroactive polymer (CEP) systems, mobile anions are released from the electrofunctional polymer structure upon reduction, as shown in eqn. (3). This has been used to advantage in the development of controlled release devices,56,57 which form the basis of on-line derivatisation58 prior to spectrophotometric detection. For other conducting polymer systems, e.g., when high molecular mass polyelectrolytes (PEs) are incorporated as counter ions, the electrochemical controlled release of cations is possible.57,59,60 Incorporation of PEs also results in the production of more open, porous conducting polymer structures, allowing the incorporation and release of larger molecules such as proteins.61 On-line reactors designed to encapsulate (but not release) the derivatising component have also been described. For example, when enzymes or antibodies are incorporated into the conducting polymer structure, the transitions that accompany oxidation/ reduction control the bioactivity of incorporated species.62,63 This can be used in on-line, real time control of bioderivatisation/ reactor systems.Electrofunctional polymers have also been used in the development of novel separation processes for analytical purposes. This again is based on the fact that the oxidation state of the polymer determines the molecular interaction capabilities.As such, construction of the columns or membrane systems that enable potentials to be applied to the polymer during operation are required (Fig. 1). Electrofunctional polymers can be prepared as stand-alone membranes or particles or may be coated on to other solid supports or particulates such as poly(vinylene difluoride) (PVDF) or silica, respectively. Electrochemically controlled chromatographic systems using polypyrroles for separations of simple ions64 or amino acids65 have been described.Similarly, ion exchange chromatographic systems based on polyaniline have been developed.66,67 Others68 have coated polyaniline stationary phases with polymers containing complexing groups (crown ethers) and still utilised electrochemical control over the separation. Another interesting example of the use of electrofunctional polymers for analytical separation was the development of a novel form of affinity chromatography. In this case the antigen was incorporated into the CEP and it was demonstrated that the potential applied to the column had a dramatic effect on the binding capacity.69 In order to obtain high capacity (high surface area/volume ratio), electrode substrates such as carbon fibres or reticulated vitreous carbon were used.An alternative to using electrochemical control of polymers during chromatography is to use redox reagents to control the oxidation state of the polymer and hence the retention characteristics.Such systems have been developed for both polypyrrole and polyaniline.70 Using simple concentration gradients of sodium sulfite as a reductant, the effect on retention of amino acids has been investigated. Electrofunctional polymers have also been used in the development of membrane separation systems. Membrane separations usually require large surface area systems for efficient separation. Stand-alone conducting polymer membranes can be produced, but the need for adequate mechanical and electrical properties limits the range of counter ions (and hence molecular functionality) that can be incorporated.The effect of even minor structural changes to dopants on the mechanical and electrical properties of stand-alone CEP membranes is well known.71 However, these have been successfully used for the separation/transport of simple ions such as K+,71 Cu2+,72 and small organic molecules.73 A more suitable approach is to coat the layers of the electrofunctional material on to conventional membranes as supports.If the membrane is first (sputter) coated with a layer of metal, e.g., Pt, then excellent electrical control can be achieved, allowing efficient electrocoating of the polymer and electrochemical control of separations (Fig. 2). The underlying polysulfone substrate also provides a template to control the final porosity of the membrane. The platinum coating has a minimal effect on the porosity and the thickness of the final CEP coating. This is particularly important for design of separation systems that must deal with transport of larger macro-molecules such as proteins.74,75 Interesting membrane structures have also been produced by electrochemical growth of CEP through a substrate membrane (Fig. 3). The highly asymmetric (with respect to surface area) structure obtained is attractive for membrane separations.76,77 A similar approach, but using photopolymerisation, has been employed to prepare membranes containing electroactive ferrocene groups.78 Fig. 1 (a) An electrochromatography column. 1 = Stainless steel column casing–auxiliary electrode; 2 = membrane; and 3 = porous conducting polymer–stationary phase–electrode. (b) An electromembrane cell. 1 = Conducting polymer membrane. Fig. 2 Schematic diagram outlining the production of a polysulfone–Pt– conducting polymer membrane. Analyst, 1999, 124, 213–219 217Conventional membranes can also be coated using chemical oxidation–interfacial polymerisation, where the monomer and oxidant solutions are separated by the membrane to be coated.Both polypyrrole79,80 and polyaniline81–84 membranes have also been used for gas separations. In both systems the dramatic influence of the membrane oxidation state on gas transport properties has been demonstrated. For example, the selectivity factors for CO2/CH4 using a poly-N-methylpyrrole membrane increased by a factor of 21 when the polymer was reduced.The same group80 demonstrated that extraordinarily large selectivity factors for O2/N2 (up to 18) could be obtained using the partially oxidised form of polypyrrole. These selectivity factors appear to have been enhanced using polyaniline81 in that factors of 30 for O2/N2 and 333 for CO2/CH4 have been obtained. Once again, doping, dedoping and redoping have a dramatic effect on the selectivity factors attainable. A further example of the possible applications of these conducting polymer membranes involves their use as an interface to a mass spectrometer.85 4 Sensors A logical extension of the ability to control the partitioning of selected analytes into/out of electrofunctional polymers is to monitor the electronic changes that accompany these processes and employ this to develop sensors.The simplest form of such devices is the electrochemical sensor. This involves coating the electrofunctional polymer on to a conventional electrode substrate such as gold, platinum or glassy carbon.The polymer may be formed on such supports by chemical polymerisation or electrodeposited. Both conducting and non-conducting polymers can be deposited electrochemically. Spin coating, airbrush spraying or dip coating may be used to deposit soluble conducting polymers. In all cases it is important that a thin, coherent and reproducible layer is obtained if reliable analytical performance is to be achieved. Coating of small (micro dimension) structures presents some additional challenges since electrodeposition of polymers on structures of these dimensions is not always efficient.86 The use of inert precoatings (e.g., Nafion) may be required to assist in fabrication.Micro arrays are used commercially for resistance measurements (as required for the electronic nose application discussed below) based on CEPs. In this case the polymer must be localised and must also coat the conductor in addition to an insulating component of the substrate (Fig. 4). When electrodeposition is used, lateral growth must be encouraged. This is usually achieved by chemical modification of the insulating surface to make it more hydrophobic.87 The most common example involving resistance measurements is in the development of various gas sensors. Much work has been carried out on the development of conducting polymer sensors for volatiles. Both simple conducting polymers88,89 and composites, e.g., polypyrrole–polyamide90,91 containing them have been developed for use in ‘electronic noses’.The basis of operation of these sensing devices for volatile components is the changes in resistance in the polymer material in the presence of the volatile component. Selectivity can be altered by varying the nature of substituents on the CEP backbone or the use of different counter ions. Arrays of polymer coated electrodes give excellent quantitative information when combined with computer processing methodologies such as principle component analysis.This approach has been used for applications in the beer industry92 and even for the detection of microorganisms93 via the detection of volatiles, and recently similar approaches have been used for the detection and characterisation of proteins in solution.94 5 Conclusions and future developments The advent of electrofunctional polymers has presented the analytical scientist with a solid phase reagent with chemical and physical properties that can be purposefully manipulated during the analysis procedure.As such they provide an extra dimension in the development of novel purification and separation procedures, on-line derivatisation and sensing technologies. The current focus of research in this area is on the development of processing and device fabrication technologies that allow stable, reproducible systems containing electrofunctional polymers to be produced. 6 Acknowledgements Professor Gordon Wallace acknowledges the continued support of the Australian Research Council.The preparation of this manuscript was made possible by the award of a Royal Society of Chemistry Travel Grant. Professor Gordon Wallace acknowledges the gracious provision of facilities by Dublin City University during this trip. 7 References 1 H. Z. Bu, S. R. Mikkelson and A. M. English, Anal. Chem., 1995, 67, 4071. 2 E. J. Calvo, C. Danilowicz and L. Diaz, J. Chem.Soc., Faraday. Trans., 1993, 89, 377. 3 A. R. Guadalupe and H. D. Abruna, Anal. Chem., 1985, 57, 142. 4 B. L. Funt and P. M. Hoang, J. Electroanal. Chem., 1983, 154, 229. 5 M. C. Pham and J. E. Dubois, J. Electroanal. Chem., 1986, 199, 153. 6 T. Kaku, H. I. Kiran and Y. Okamoto, Anal. Chem., 1994, 66, 1231. 7 C. J. Chen, C. C. Liu and F. Savinell, J. Electroanal. Chem., 1993, 348, 317. 8 N. Barisci, G. G. Wallace, E. Wilke, M. Meaney, M. R. Smyth and J. G. Vos, Electroanalysis, 1989, 1, 245. 9 G.G. Wallace, M. Meaney, M. R. Smyth and J. G. Vos, Electroanalysis, 1989, 1, 357. 10 A. P. Doherty, R. J. Forster, M. R. Smyth and J. G. Vos, Anal. Chim. Acta, 1991, 225, 45. 11 T. J. O’Shea, D. Leech, M. R. Smyth and J. G. Vos, Talanta, 1992, 39, 443. 12 M. M. Malone, A. P. Doherty, M. R. Smyth and J. G. Vos, Analyst, 1992, 117, 1259. Fig. 3 Schematic diagram of the production of a ‘keyed’ Ppy membrane. Fig. 4 Interdigitated micro array components coated with PPy. 218 Analyst, 1999, 124, 213–21913 A. P. Doherty, R. J. Forster, M. R. Smyth and J. G. Vos, Anal. Chem., 1992, 64, 572. 14 A. Heller, J. Phys. Chem., 1992, 96, 258. 15 P. D. Hale, T. Inagaki, H. J. Karan, Y. Okamoto and T. A. Skotheim, J. Am. Chem. Soc., 1989, 111, 3482. 16 M. G. Garguilo, N. Huynh, A. Proctor and A. C. Michael, Anal. Chem., 1993, 65, 523. 17 P. D. Hale, L. J. Boyuslavsky, T. Inagaki, H. J. Karan, H. S. Lee, T. A. Skotheim and Y. Okamoto, Anal. Chem., 1991, 63, 677. 18 N. C. Foulds and C. R. Lowe, Anal. Chem., 1988, 60, 2473. 19 T. Kaku, H. I. Karan and Y. Okamoto, Anal. Chem., 1994, 66, 1231. 20 B. A. Gregg and A. Heller, Anal. Chem., 1990, 62, 258. 21 B. A. Gregg and A. Heller, J. Phys. Chem., 1991, 95, 5976. 22 T. J. Ohara, R. Rajagoplan and A. Heller, Anal. Chem., 1993, 65, 3512. 23 T. de Lumley-Woodyear, P. Rocca, J. Lindsay, Y. Dror, A. Freeman and A. Heller, Anal. Chem., 1995, 67, 1332. 24 B. Linke, W. Kerner, M. Kiwit, M.Pishko and A. Heller, Biosens. Bioelectron., 1994, 9, 151. 25 E. Cs·regi, C. P. Quinn, D. W. Schmidtke, S. Linquist, M. V. Pishko, L. Ye, I. Katakis, J. A. Hubbell and A. Heller, Anal. Chem., 1994, 66, 3131. 26 M. Pravda, C. M. Jungar, E. I. Iwuoha, M. R. Smyth, K. Vytras and A. Ivaska, Anal. Chim. Acta, 1995, 304, 127. 27 M. Pravda, O. Adeyoju, E. I. Iwuoha, J. G. Vos, M. R. Smyth and K. Vytras, Electroanalysis, 1995, 7, 619. 28 E. Rohde, E. Dempsey, M. R. Smyth, J. G. Vos and H.Emons, Anal. Chim. Acta, 1993, 278, 5. 29 E. I. Iwuoha, M. R. Smyth and J. G. Vos, Electroanalysis, 1994, 6, 982. 30 B. Lu, E. I. Iwuoha, M. R. Smyth and R. O’Kennedy, Anal. Commun., 1997, 34, 21. 31 M. E. G. Lyons, W. Breen and J. Cassidy, J. Chem. Soc., Faraday Trans., 1991, 87, 115. 32 R. A. Saruceno, J. G. Pack and A. G. Ewing, J. Electroanal. Chem., 1986, 197, 265. 33 J. M. Cooper, D. G. Morris and K. S. Ryder, J. Chem. Soc., Chem. Commun., 1995, 697. 34 M. Gholamian, J.Sundaram and A. Q. Contractor, Langmuir, 1987, 3, 741. 35 H. Yoneyama, Y. Shoji and K. Kawai, Chem. Lett., 1989, 1067. 36 K. M. Kost, D. E. Bartah, B. Kazee and T. Kuwana, Anal. Chem., 1988, 60, 2379. 37 W. Lu, H. Zhao and G. G. Wallace, Electroanalysis, 1996, 8, 248. 38 H. Mao and P. G. Pickup, J. Electroanal. Chem., 1989, 265, 127. 39 M. H. Lee, Y. T. Hang and S. B. Rhee, Synth. Met., 1995, 69, 515. 40 F. Bedioui, J. Devynck and C. Bied-Charreton, Acc. Chem. Res., 1995, 28, 30, and references cited therein. 41 B. R. Saunders, R. J. Fleming and K. S. Murray, Chem. Mater., 1995, 7, 1082, and references cited therein. 42 J. Ye, R. P. Baldwin and J. W. Schlager, Electroanalysis, 1989, 1, 133. 43 A. Deronzier and J. C. Moulet, Coord. Chem. Rev., 1996, 147, 339, and references cited therein. 44 L. Wang, E. Kobatake, Y. Ikiriyama and M. Aizawa, Denki Kagaku, 1992, 60, 1050. 45 D. Compagnone, G. Federici and J. V. Bannister, Electroanalysis, 1995, 7, 1151. 46 H. Shinohara, T. Chiba and M. Aizawa, Sens. Actuators, 1988, 13, 79. 47 M. Umana and J. Waller, Anal. Chem., 1986, 58, 2979. 48 S. B. Adeloju, S. J. Shaw and G. G. Wallace, Electroanalysis, 1994, 6, 865. 49 S. B. Adeolju, S. J. Shaw and G. G. Wallace, Anal. Chim. Acta, 1993, 281, 611. 50 S. B. Adeloju, S. J. Shaw and G. G. Wallace, Anal. Chim. Acta, 1993, 281, 621. 51 W. Schuhmann, C. Kranz, J. Huber and H. Wohlschlager, Synth. Met., 1993, 61, 31. 52 O. A. Sadik and G. G. Wallace, Anal.Chim. Acta, 1993, 279, 209. 53 D. Barnett, O. A. Sadik, M. John and G. G. Wallace, Analyst, 1994, 119, 1997. 54 M. V. Deshpande and E. A. H. Hall, Biosens. Bioelectron., 1990, 5, 431. 55 A. J. Hodgson, M. J. John, T. Campbell, A. Georgevich, S. Woodhouse and G. G. Wallace, Proc. SPIE, 1996, 2716, 164. 56 B. Zinger and L. L. Miller, J. Am. Chem. Soc., 1984, 106, 6861. 57 Q. X. Zhou, L. L. Miller and J. R. Valentine, J. Electroanal. Chem., 1989, 261, 147. 58 Y. Lin, PhD Thesis, University of Wollongong, 1991. 59 A. Wojda and K. Matsymiuk, J. Electroanal. Chem., 1997, 424, 93. 60 X. Ren and P. G. Pickup, J. Phys. Chem., 1993, 97, 5356. 61 A. J. Hodgson, K. J. Gilmore, C. Small, G. G. Wallace, I. L. McKenzie, T. Aoki and N. Ogata, Supramol. Sci., 1994, 1, 77. 62 M. Aizawa, S. Yabuki and H. Shinohara, Stud. Org. Chem., 1987, 30, 353. 63 M. Aizawa, T. Haruyama, G. I. Khan, E. Kobatake and Y. Kariyama, Biosens. Bioelectron., 1994, 9, 601. 64 H. Ge, P.R. Teasdale and G. G. Wallace, J. Chromatogr., 1991, 544, 305. 65 R. S. Deinhammer, M. D. Parker and K. Shimazu, J. Electroanal. Chem., 1995, 387, 35. 66 A. A. Syed and M. K. Dinesan, React. Polym., 1992, 17, 145. 67 A. A. Syed and M. K. Dinesan, Analyst, 1992, 117, 61. 68 T. Nagoka, M. Fujimoto, H. Nakao, K. Kakun, J. Yano and K. Ogura, J. Electroanal. Chem., 1993, 350, 337. 69 G. G. Wallace, K. E. Maxwell, T. W. Lewis, A. J. Hodgson and M. J. Spencer, J. Liq. Chromatogr., 1990, 13, 3091. 70 H. Chriswanto and G. G. Wallace, J. Liq. Chromatogr., 1996, 19, 2457. 71 W. E. Price, G. G. Wallace and H. Zhao, J. Membr. Sci., 1994, 87, 47. 72 H. Zhao, W. E. Price and G. G. Wallace, Polymer, 1993, 34, 16. 73 H. Zhao, W. E. Price and G. G. Wallace, J. Membr. Sci., 1995, 100, 239. 74 C. O. Too, G. G. Wallace and D. Zhou, Proc. SPIE, 1997, 3241, 106. 75 D. Zhou, C. O. Too and G. G. Wallace, J. Int. Mater. Syst. Struct., 1997, 8, 1052. 76 Z. Cai and C. R. Martin, J. Am. Chem. Soc., 1989, 11, 4138. 77 J. Mansouri and R. P. Burford, J. Membr. Sci., 1994, 87, 23. 78 C. Liu and C. R. Martin, Nature (London), 1991, 353, 50. 79 W. Liang and C. R. Martin, Chem. Mater., 1991, 3, 390. 80 R. J. Parthasorathy, V. P. Manon and C. R. Martin, Chem. Mater., 1997, 9, 560. 81 M. R. Anderson, B. R. Mattes, H. Reiss and R. B. Kanor, Science, 1991, 252, 1412. 82 M. R. Anderson, B. R. Mattes, H. Reiss and R. B. Kanor, Synth. Met., 1991, 41, 1151. 83 S. Kuwabata and C. R. Martin, J. Membr. Sci., 1994, 91, 1. 84 J. Yang, Q. Sun, X. Hou and M. Wan, Chin. J. Polym. Sci., 1993, 11, 121. 85 V. M. Schmidt and J. Heitbaum, Synth. Met., 1991, 41, 425. 86 J. N. Barisci, P. M. Murray, C. J. Small and G. G. Wallace, Electroanalysis, 1996, 8, 330. 87 M. Nishizawa, M. Shibuya, J. Sawaguchi, T. Matsue and I. Uchida, J. Phys. Chem., 1991, 95, 1042. 88 J. N. Barisci, M. K. Andrews, P. Harris, A. C. Partridge and G. G. Wallace, Proc. SPIE, 1997, 3242, 164. 89 J. J. Miasik, A. Hooper and B. C. Tofield, J. Chem. Soc., Faraday. Trans., 1986, 82, 1117. 90 P. N. Bartlett and S. K. Ling-Chung, Sens. Actuators, 1989, 20, 287. 91 F. Selampinar, L. Toppare, U. Akbulut, T. Yalain and S. Suzer, Synth. Met., 1995, 68, 109. 92 J. W. Gardner, T. C. Pearce, S. Friel, P. N. Bartlett and N. Blair, Sens. Actuators, 1994, 18–19, 240. 93 P. K. Namdev, Y. Alroy and V. Singh, Biotechnol. Prog., 1998, 14, 75, and references cited therein. 94 W. Lu, T. Nguyen and G. G. Wallace, Electroanalysis, 1998, 10, 1101. Paper 8/08015A Analyst, 1999, 124, 213–219 219

ISSN:0003-2654    

DOI:10.1039/a808015a

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Universal detection in capillary electrophoresis with a micro-interferometric backscatter detector Kelly Swinney, Jana Pennington and Darryl J. Bornhop* Department of Chemistry and Biochemistry, Texas Tech University, Box 41061, Lubbock, TX 79409-1061, USA. E-mail: djbornhop@ttu.edu Received 11th December 1998, Accepted 24th January 1999 An optically simple, inexpensive, micro-volume refractive index detector was applied to capillary electrophoresis (CE), allowing universal solute detection at the sub-picogram level.The micro-interferometric backscatter detector (MIBD) employs direct, side illumination of an unmodified capillary by an He–Ne laser, producing a 360° fan of scattered light that contains a set of high contrast interference fringes. These light and dark spots are viewed on a flat plane in the direct backscatter configuration. A slit–photodetector assembly accomplishes signal interrogation of the time-dependent fringe shifts, produced or imparted by refractive index (RI) changes.Using an unfocused laser beam to prove the unmodified separation capillary produces a detector volume of 4.7 3 1029 L. The separation and quantification of a mixture of organic dyes and simple sugars demonstrate the system’s utility. Submicromolar concentration detection limits of 0.46, 1.1 and 0.72 mm for Bromothymol Blue, Thymol Blue and Bromocresol Green, respectively, are achievable with CE-MIBD in the simplest configuration. The 3s RI concentration detection limits are 2.5 times superior to those obtained by UV/VIS detection performed under the same conditions.Several carbohydrates (maltose, lactose and d-ribose) are separable and detectable at the ppm level, using no active thermal stabilization. Further demonstrating the utility of MIBD for universal detection with CE. Introduction The refractive index (RI) detector is a workhorse detector in conventional high performance liquid chromatographic (HPLC) separations because it is a concentration sensitive, bulk property, non-destructive sensor.These detectors are reasonably sensitive, providing a signal for essentially all analytes. However, except for use in basic research,1–10 RI measurement schemes have found little acceptance in either capillary HPLC or in capillary electrophoresis (CE).1 The lack of a commercial device and the limited use of RI detection in CE stem not from opulence, as universal detection for micro-volumes is surely needed, but from the inherent difficulty in reducing detection volumes to 1 nL or smaller without a loss of sensitivity.The properties of the laser, such as high spatial coherence and monochromaticity have led to spectacular improvements in detection limits for spectroscopic measurements.11,12 Using a laser’s high spatial coherence, the beam can be focused to a very small spot without loss of power13–15 or launched into a fiber optic efficiently.12,16 Because of these and other optical properties, lasers have become the ‘source of choice’ in the effort to construct micro-volume RI detectors.2–10,17–23 Most state-of-the-art RI measurements involve some form of interferometry, a measurement technique that is critically dependent on the characteristics of laser light. In general, very small phase changes caused by optical path length differences, in response to the analyte, allow for high sensitivity.12,24 While the use of interferometric techniques has produced some impressive results toward measuring RI changes in small volumes,2–5,16 sensitivity dependence on pathlength has persisted.For example, although Woodruff and Yeung5 demonstrated a Fabry– Perot interferometer with excellent sensitivity and detection volumes in the low microliter region, flow cell volume constraints limit this method to schemes larger than capillarybased techniques. Other RI detection methods have been investigated, including a concentration gradient method, which probes the on-axis optical perturbation produced by a transient solute band,20,22 and the use of a tapered fiber optic,25 which probes the reduction in transmitted beam intensity due to refractive index interfacial beam intensity coupling.Additionally, refractometric interference spectrometry shows promise for ‘difficult’ analytes and can be used to measure organic pollutants in water, as demonstrated by Gaugliz and co-workers.17 These alternative approaches to RI sensing have limitations.For example, the concentration gradient detector20,22 is insensitive to thermal noise, but most suitable for detection schemes in capillary isoelectric focusing or for separation systems where postcolumn detection is acceptable. Bornhop and Dovichi2 used a laser-based, off-axis, oncapillary technique to probe nanoliter volumes and detect nanogram amounts of sugars in micro-LC separations,2 but this technique was found to be tedious to align and required the removal of the coating from the capillary.Bruno and coworkers3 further developed the forward scatter, off-axis technique. They showed that improved performance is possible by using an RI-matching fluid to surround the capillary tube, a flow cell assembly with active thermal control and position sensitive detection. These improvements facilitated the use of the RI detector for capillary electrophoresis,3 yet the device still requires removal of the polymer coating to aid in index matching and off-axis alignment.Another technique for detecting changes in RI within small volumes uses a holographic grating to produce two-beam interference in a forward scatter configuration.4 Krattinger and co-workers4 were able to separate and detect metal ions by CE using the holographic grating, a capillary that is encapsulated in an index matching glue and a photodiode array wired to produce position sensitive detection. Bruggraf et al.23 reported the application of the holographic forward scatter RI detector for CE on a chip showing that the holographic technique is a meritorious advance in universal detection in small volumes, eliminating the need for the capillary to serve as the optic.Yet as another arrangement of the grazing angle forward-scattering Analyst, 1999, 124, 221–225 221Plexiglas Box HV Power Supply 4 mW He/Ne Computer DAQ Photodetector Buffer Reservoir Aluminium Block Capillary Buffer Reservoir refractive index technique it has an inherent pathlength dependence which ultimately hinders detection limits in ultrasmall volumes.2–4,21 Deng and Li,10 acknowledging the difficulties with forward scattering techniques, have recently shown that a retro-reflected beam interference technique similar to the use of a microinterferometric backscatter detector (MIBD)6–9 can be used for RI detection in CE.In their scheme, a focused laser beam impinges on a bare capillary surface causing interference of two retro-reflected beams originating from the outer surface of the capillary.By observing the retro-reflected interference pattern, RI measurements are possible. The main drawbacks of the retroreflected beam interference RI detector are poorer detection limits than with the previously reported RI detectors2–9 and enhanced complexity of the optical train.6–9 Work in our laboratory has shown that on-capillary RI analysis and capillary HPLC detection can be performed using interferometric backscatter.6–9 The MIBD, previously termed laser interferometric backscatter (LIB) detector, employs a simple optical train, produces high sensitivity RI measurements in small diameter capillaries with minimal pathlength sensitivity8 and requires no modification of the capillary tube.Success of the MIBD with flowing streams was shown with a separation of aromatic compounds using capillary HPLC.9 The MIBD response proved to be linear over a dynamic range of about 2.5 decades.Picogram mass detection limits were achieved in nanoliter probe volumes using a massive aluminum flow-cell block on which the capillary was mounted for passive thermal control. Here we describe the use of the MIBD for universal detection in CE. In the optical configuration for the MIBD,8 an unfocused He–Ne laser beam impinges on a capillary tube producing a 360° fan of radiation, spatially contained and perpendicular to the long axis of the tube.When viewed on a flat surface placed coincident with but above or below the incident laser beam, the scattered light is seen as a series of high contrast interference fringes (light and dark spots). The central fringe pattern at a nearly 0° backscattering angle is similar in appearance to that produced by single slit diffraction. The positions of these maxima and minima can be employed in the sensitive measurement of fluid bulk properties. We show that the MIBDCE system can be used to detect solutes at the ppm level, with mass detection limits in the picogram range.Such performance is illustrated with the separation of a mixture of organic dyes and a three component carbohydrate mixture. Experimental The basic optical configuration for the MIBD has been described in detail elsewhere6–9 and is depicted in the generalized block diagram presented in Fig. 1. A low power, linearly polarized beam at 632.8 nm was provided by a 4 mW helium–neon laser (Melles Griot, Irvine, CA, USA).The laser output was directed on to the capillary, located 55 cm from the laser’s aperture, creating an array of backscattered light or interference pattern. The capillary was mounted on a massive aluminum block acting as a flow cell assembly with no active thermal control. The complete flow cell assembly was tilted at an angle convenient for placement of the fringe on the photodetector and normally not exceeding 7°. The backscatter radiation was allowed to propagate ca. 30 cm to a small area (1.0 cm2) photodiode (UDT Sensors, Hawthorn, NV, USA) housed in a fixture containing a 632.8 nm interference filter (Optical Coating Technology, Southampton, MA, USA) and a 150 mm precision air slit (Melles Griot). The photodetector assembly was mounted on a high precision translation stage. Translation of the detector allowed for easy selection of the central fringes. These fringes have been found to move significantly with an RI change caused by the analyte.6 The output of the photodiode is conditioned with a current-tovoltage converter, consisting of a JFET operational amplifier, wired with a 10 MW feedback resistor in parallel with a 0.01 pF capacitor.No additional electronic filtering was applied to the voltage signal. The analog voltage signal from the RI detector was digitized with a DAQ board (National Instruments, Austin, TX, USA) and displayed on a PC running a digital strip-chart recorder (Labtech for Windows, Wilmington, MA, USA).All optical components and detectors were rigidly mounted on a 4 3 3 ft optical bench (Newport Research, Irvine, CA, USA). Manual micrometer driven translation stages were used to provide reproducible translation of the photodetector, capillary tube and optical components. Excluding the laser, the entire experiment is enclosed in a Plexiglas box. High sensitivity is easily achieved in MIBD by simply insuring that the laser beam illuminates the entire capillary in a region where the coating has not been scratched or marred.Upon obtaining a high contrast interference pattern, the flow cell assembly is tilted to direct the beam on to the photodetector assembly. Once a signal is obtained, the detector assembly is translated across two full fringes, allowing the identification of the minima. The detector is then placed so that a small voltage output is observed. This position corresponds to the edge of the sloping intensity gradient of the working fringe and is located at I = 1/e2 of the essentially Gaussian intensity distribution.A 100 mm id, 350 mm od capillary was used in all separations. The 17 mm polyimide coating was not modified or removed in the RI detection zone. For the dye separation (a) the RI detection window was located 34.5 cm from the injection end of a 42 cm long capillary and (b) injection was made by pressure driven (commercial instrument controlled) hydrodynamic injection at 1.5 psi for 1.0 s.For the carbohydrate separation (a) the detection window was 83.5 cm from the injection end of a 95 cm column and (b) injection was made by hydrodynamic siphoning at a height of 35 cm for a period of 5 s. The working buffer for the dye separation consisted of 10 mm anthraquinone-2-sulfonic acid, 4 mm sodium borate and 10% ethanol. A buffer consisting of 33 mm borate and 13.3 mm CAPS (3-cyclohexylanino- 1-propane-sulfonic acid) was used for the carbohydrate separation.An applied voltage of 20 kV (56.7 mA) was administered to the capillary in order to elute the dyes (Bromothymol Blue, Thymol Blue and Bromocresol Green), while 12 kV (36 mA) were applied to the capillary to induce component separation in the carbohydrate analysis (maltose, lactose and d-ribose). All separations were conducted at room temperature (22–25 °C). The dye mixture and individual dye standard solutions were prepared from analytical-reagent grade or better chemicals (Aldrich, Milwaukee, WI, USA) and dissolved in a solution of 4 mm sodium borate–10% ethanol.The carbohydrate standards and mixture were also prepared from analytical-reagent grade chemicals (Aldrich), but dissolved in de-ionized water. No other sample preparation or work-up was performed before injection. Calibration curves were generated in duplicate for each solute Fig. 1 Block diagram for the MIBD (RI) detector. 222 Analyst, 1999, 124, 221–225using peak height and peak area, and the separation samples were analyzed in triplicate for each separation system.Results and discussion The principle of operation for the RI measurement using the MIBD has been described in detail elsewhere6–9 and involves the detection of optical pathlength changes within the tube, as RI changes for materials contained in the tube. Upon solute introduction, the position of the backscattered fringes shift in response to a change in RI within the probe volume.We employ a simple approach to measure this fringe shift by placing the slit–photodetector assembly on the edge of a fringe. The positional movement is found to be proportional to analyte concentration and is generally linear over three decades.14,15 This simple intensity-based detection method is inexpensive and easy to implement. However, such a configuration will ultimately limit the dynamic operation range of the MIBD to a refractive index shift equal to a distance corresponding to 2s of the Gaussian-like profile of the backscattered fringe (e.g., once the fringe maxima shifts to a position near or past the slit, nonlinear operation ensues).Previous investigations that have demonstrated CE with RI detection showed promise, but were constrained by pathlength sensitivity.4,19 Here we performed analyses of several test mixtures to evaluate the performance of the MIBD (which has a many pass optical configuration) with CE.Fig. 2 shows a typical electropherogram for a three component mixture of dyes (in order of elution), Bromothymol Blue, Thymol Blue and Bromocresol Green. These solutes where chosen because they are known to be separable quickly and to provide a detection limit comparison with standard UV/VIS absorption (619 nm). Using the conventional configuration for CE shown in Fig. 1 (without active thermal control of either the capillary or the detection region), a high voltage was applied to a 42 cm 3 100 mm diameter fused capillary, producing a current of 30 mA.A solution containing micromolar amounts of the dyes was injected by hydrodynamic introduction, producing an estimated sample volume of 22.9 nL.26 The detection limit, three standard deviations above the background, was estimated from the peak-to-peak noise in the electropherogram.27 Calibration curves were generated for each solute using both peak height and peak area. The response was linear over three decades of concentration with correlation coefficients ranging from 0.993 to 0.999.As shown in Table 1, MIBD-CE can facilitate mass detection limits for the above three dyes, at the 3s level, of 6.7, 11.9 and 11.2 pg or 10.7, 25.6 and 16.0 fmol, respectively. The concentration detection limit for Bromothymol Blue is 4.66 3 1027 mol L21 or about 0.29 ppm, and is in a region that facilitates analytical utility. In order to compare the performance of a standard UV detector with the MIBD (RI) detector, a Linear Instruments (Reno, NV, USA) UV/VIS detector was affixed to the CE system.A typical electropherogram for the three component dye mixtures is shown in Fig. 3. Calibration was performed and the detection limits in CE-UV were determined (Table 2). It is noteworthy that for the dye solutes which have strong molar absorptivities at 619 nm (23 200 L mol21 cm21 for Thymol Fig. 2 Electropherogram of the dye mixture using the MIBD (RI) detector. Peaks: 1 = 30 mm Bromothymol Blue; 2 = 60 mm Thymol Blue, 3 = 30 mm Bromocresol Green.Table1 Comparison of detection limits (DL) for the dye samples Solute Mr/g mol21 Concentration DLa/m (RI detector) Injected mass DLa (RI detector) Concentration DLa/m (UV detector) Injected mass DLa (UV detector) Bromothymol Blue 624.4 4.66 3 1027 6.7 pg or 10.7 fmol 1.16 3 1026 16.5 pg or 26.5 fmol Thymol Blue 466.6 1.12 3 1026 11.9 pg or 25.6 fmol 2.23 3 1026 23.8 pg or 50.1 fmol Bromocresol Green 698.0 7.23 3 1027 11.2 pg or 16.0 fmol 1.16 3 1026 18.5 pg or 26.5 fmol a Calculated at 3s.Table 2 Detection limits (DL) and separation efficiency for the carbohydrate separation Solute Identity Mr/g mol21 Concentration DLa/m (RI detector) Injected mass DLa (RI detector) Number of theoretical plates (RI detector) HETP (RI detector) Maltose 360.32 1.28 3 1025 0.22 ng or 0.60 pmol 8.35 3 104 1.73 3 1024 Lactose 360.31 3.33 3 1025 0.57 ng or 1.57 pmol 1.21 3 105 1.22 3 1024 d-Ribose 150.13 6.53 3 1025 0.46 ng or 3.09 pmol 3.27 3 104 5.52 3 1024 a Calculated at 3s.Fig. 3 Electropherogram of the dye mixture using UV/VIS detection (619 nm). Peaks: 1 = 30 mm Bromothymol Blue; 2 = 60 mm Thymol Blue, 3 = 30 mM Bromocresol Green. Analyst, 1999, 124, 221–225 223Blue, 30 400 L mol21 cm21 for Bromocresol Green and 37 475 L mol21 cm21 for Bromothymol Blue),28 the functional detection limits of the MIBD-RI detector are a factor of 1.6–2.5 times better than those reported for the standard UV/VIS absorption detector.Hence MIBD-RI can provide detection limits comparable to or even better than those with the conventional UV absorbance monitor under CE conditions using a 100 mm diameter capillary. If the capillary diameter was further reduced, as desired for many CE separations, the detection limit for the UV detector would be poorer, as predicted by Beer’s law. However, no such performance reduction is expected for the relatively pathlength insensitive RI detector (MIBD has been applied to shorter pathlength sample cells without a decrease in sensitivity).7 Consequently, for CE performed in capillaries of < 50 mm id, the detection limit for the RI detector is expected be at least five times better than that for the standard UV detector.It is well known that applying a voltage to the CE capillary produces joule heat throughout the tube, and it is also known that dn/dT for H2O corresponds to 1.7 3 1023 RIU °C21 (RIU = refractive index units).8 This heating can perturb the RI detector and interfere with detection of solutes.We have developed a simple procedure to effectively eliminate this RI response to the rapid joule heating produced by applying a high voltage to the capillary on injection. Briefly, a voltage is applied for 15 s to a capillary filled with fresh buffer immediately before hydrodynamic injection of the sample, heating the capillary by about 10 °C.Then the injection is rapidly performed while the system is near the separation temperature and before the temperature can decay. We then quickly begin the separation. By implementing this procedure, the large temperature shock to the system which corresponds to Dn = 1.7 3 1022 RIU for DT = 10 °C are avoided, and therefore not observed by the RI detector upon reapplying the voltage to the capillary in order to drive the separation. Furthermore, it is noteworthy that the MIBD is relatively insensitive to short term thermal perturbations during the separation.Fig. 2 and 4 show typical baseline noise levels that are stable for the duration of the separation. Glycoproteins, an important class of molecules, have been studied extensively, yet the rapid separation and detection of these analytes is still difficult because this class of molecules possess relatively low absorption coefficients. To demonstrate further the utility of MIBD as a ‘universal’ detector for CE, carbohydrate samples were studied. Fig. 4 shows the separation of (in order of elution) (1) maltose, (2) lactose and (4) d-ribose (the third peak is a solvent anomaly). Again, the simplest MIBD-CE system configuration was used, employing an 83.5 cm 3 100 mm id capillary and no active thermal control for either the separation or detection system. Borate–CAPS buffer was used and 12 kV were applied to the capillary, producing 37 mA of current. Hydrodynamic injection for 5 s produces a volume of 47.3 nL26 for a 5.0 mm solution of the sugars.Solute detection limits are in the range 196–38 mm (Table 2), which is an improvement over those previously reported29 for RI-CE of underivatized species. These detection limits are 1–2 decades better than those obtained by standard UV absorbance at 195 nm28 and are near those needed for glycoprotein analysis.30 Ignoring dilution from the separation process, the RI change at the detection limit for maltose corresponds to Dn = 1.39 3 1026.The system produces linear calibration plots (r2 > 0.995) over three decades of concentration. As noted before, dn/dT (thermal) perturbations ultimately limit the use of dc RI measurements in CE. MIBD is no exception, with the detection limits approaching those dictated by dT changes. Surprisingly, baseline stability for CE with MIBD is fairly good (particularly since no active control was employed) and the detection limits are in the region of analytical interest.The separation also shows that MIBD has the potential to be used for the determination of biological solutes, lacking good absorption, under CE conditions. We predict that with active thermal control and electronic filtering of the photodetector signal, nanomolar detection limits should be possible. Band broadening induced by the detector dead volume can be estimated by calculating the number of theoretical plates and HETP from the electropherograms, as shown in Tables 2 and 3.As expected using an unfocused beam that produces a probe volume of approximately 4.7 nL, a slight reduction in separation efficiency is observed for MIBD compared with UV detection, which has a probe volume in the region of the 0.5 nL. This observation further illustrates the importance of the detector probe volume for separation performance in CE. In any event, the separation efficiency produced in the MIBD-CE system is comparable to that with conventional, on-column, UV detection.Although the electrophoretic efficiency could be improved, the performance is good, with about 6.80 3 105 theoretical plates for the dye separation and 1.21 3 105 plates for the carbohydrate separation. Even though the MIBD detection volume can be reduced by spatially masking the excitation source (currently under investigation), thus producing a smaller diameter probe beam, in these experiments the separation efficiency was primarily limited by other factors, including thermal instability in the column, during injection and separation.Reducing the diameter of the probe beam and applying active thermal control should improve the separation efficiency and the detection limits for MIBD significantly. In summary, it has been shown that the MIBD can be used for universal detection in CE. The optical configuration is simple, facilitates the use of unmodified capillaries and can be constructed inexpensively (less the optical bench, the total cost is about US$ 1000).The MIBD is sensitive to small changes in refractive index (picogram amounts can be detected currently), allowing ppm quantities of poorly absorbing solutes to be quantified. Considering that the current trend toward ‘system’ miniaturization23,31–33 is limited by the availability of an ultrasmall universal detection system, a sensitive, low volume detector such as the MIBD represents a major advancement. Current investigations in our laboratories include the further application of the MIBD for RI detection in CE, for polarimetry measurements in capillary dimensions33 and for RI detection with CEC. Acknowledgements Funding for this work from the Robert A.Welch Foundation (Grant No. D-1312) and from the Texas Tech University Research Enhancement Fund is gratefully acknowledged. Fig. 4 Electropherogram of the carbohydrate mixture using the MIBD (RI) detector. Peaks: 1 = 5 mm maltose; 2 = 5 mm lactose; 3 = solvent anomaly; 4 = 5 mm d-ribose. 224 Analyst, 1999, 124, 221–225References 1 J. W. Jorgenson and J. D. Wit, in Detectors for Capillary Chromatography, ed. H.H. Hill and D.G. McMinn, Wiley, New York, 1992, ch. 15. 2 D. J. Bornhop and N. J. Dovichi, Anal. Chem., 1986, 58, 504. 3 A. E. Bruno, B. Krattinger, F. Maystre and H. M. Widmer, Anal. Chem., 1991, 63, 2689. 4 B. Krattinger, A. E. Bruno and G. J. Bruin, Anal. Chem., 1994, 66, 1. 5 S. D. Woodruff and E.S. Yeung, Anal. Chem., 1982, 54, 1174. 6 D. J. Bornhop, US Pat., 53 25 170, 1994. 7 D. J. Bornhop, Appl. Opt., 1995, 34, 3234. 8 H. J. Tarigan, P. Neill, C. K. Kenmore and D. J. Bornhop, Anal. Chem., 1996, 15, 1763. 9 C. K. Kenmore, S. R. Erskine and D. J. Bornhop, J. Chromatogr. A, 1997, 762, 219. 10 Y. Deng and B. Li, Appl. Opt., 1998, 37, 998. 11 D. L. Andrews, Lasers in Chemistry, Springer, Berlin, 1990. 12 B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley- Interscience, New York, 1991. 13 T. Herishfeld, Appl. Opt., 1976, 15, 2965. 14 C. Radzewicz,P. Glowezewski and J. Kramimske, Appl. Phys., 1978, 17, 423. 15 E. K. Gustafson and R. L. Byer, Opt. Lett., 1984, 220. 16 J. D. Ingle and S. R. Crouch, Spectrochemical Analysis, Prentice- Hall, Englewood Cliffs, NJ, 1988, pp. 59–60 and 98–106. 17 G. Gauglitz, J. Krause-Bonte, H. Schlemmer and A. Matthes, Anal. Chem., 1988, 60, 2609. 18 H. M. Yan, G. Kraus and G. Gauglitz, Anal. Chim. Acta, 1995, 312, 1. 19 D. J. Bornhop, T. G. Nolan and N. J. Dovichi, J. Chromatogr., 1987, 384, 181. 20 J. Wu and I. Pawliszyn, Anal. Chem., 1992, 64, 224. 21 R. E. Synovec, Anal. Chem., 1987, 59, 2877. 22 J. Wu and J. Pawliszyn, Anal. Chem., 1992, 64, 2934. 23 N. Bruggraf, B. Krattiger, A. de Mello, N. de Rooij and A. Manz, Analyst, 1988, 123, 1443. 24 M. Born and E. Wolf, Principles of Optics, Pergamon Press, New York, 1975. 25 D. A. Buttry, T. C. Vogelmann, G. Chen and R. Goodwin, US Pat., 5 600 433, 1997. 26 R. Weinberger, Practical Capillary Electrophoresis, Academic Press, Boston, 1993. 27 J. E. Knoll, J. Chromatogr. Sci., 1985, 23, 422. 28 The molar absorptivities were determined by standard methods using a UV/VIS spectrometer and Beer’s law. 29 A. E. Bruno, F. Maystre, B. Krattiger, P. Nussbaum and E. Gassmann, Trends Anal. Chem., 1994, 13, 190. 30 A. Paulus and A. Klockow, J. Chromatogr. A, 1996, 720, 353. 31 S. V. Ermakov, S. C. Jacobson and J. M. Ramsey, Anal. Chem., 1998, 70, 4494. 32 C. L. Colyer, S. D. Mangru and D. J. Harrison, J. Chromatogr. A, 1997, 781, 271. 33 D. J. Bornhop and J. Hankins, Anal. Chem., 1996, 68, 1677. Paper 8/09691K Analyst, 1999, 124, 221–225 225

ISSN:0003-2654    

DOI:10.1039/a809691k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Procrustes analysis for the comparison of test methods in reversed-phase high performance liquid chromatography of basic compounds Richard G. Brereton*a and David V. McCalleyb a School of Chemistry, University of Bristol, Cantock’s Close, Bristol, UK BS8 1TS b Faculty of Applied Sciences, University of the West of England, Frenchay, Bristol, UK BS16 1QY Received 3rd November 1998, Accepted 6th January 1999 A dataset consisting of three column performance parameters obtained by testing eight reversed-phase HPLC columns with nine basic compounds and three mobile phases (methanol, acetonitrile and tetrahydrofuran) was assessed using chemometrics.The main aim was to investigate loss of information as the number of tests is reduced. Principal components analysis was performed on a full dataset using methanol as a mobile phase and procrustes analysis was used to compare numerically the scores plots as test compounds and individual test parameters were removed from the full dataset.The scores plots using different mobile phases were compared to determine whether the performance under one condition could predict that under another. 1. Introduction There is continuing interest in test methods for the identification of reversed-phase (RP) columns which give good performance in the analysis of basic substances by high performance liquid chromatography (HPLC). This large group of substances, which includes important pharmaceuticals, compounds of biomedical significance and environmental pollutants, can give poor chromatographic peak shapes, apparently due to detrimental interactions with silanol groups on such columns.1–4 Whereas the selection of test compounds and mobile phase conditions is mostly unproblematic for neutral and acidic compounds, establishing a simple test procedure for assessing column activity towards bases appears to be fraught with difficulties.For example, we have shown that the pH of the mobile phase and the organic modifier utilised [methanol, acetonitrile or tetrahydrofuran (THF)] can influence the relative performance of columns even when the same test compounds are utilised.5–7 Furthermore, there are considerable variations in the relative performance for a given mobile phase when different basic substances are chosen as test compounds.For example, we noted a remarkable lack of correlation between peak asymmetry for nortriptyline and pyridine when obtaining data on eight different RP columns in a methanolic mobile phase buffered at pH 3.This means that a column that gives good results for one of these commonly used probes may not give good results for the other. It is hardly surprising, therefore, that different test procedures using a single mobile phase and one or two basic test substances fail to produce any agreement.4,5 Instead, we have used much more comprehensive test procedures involving three different modifiers buffered at low and high pH (six mobile phases per column) and up to 10 basic probe substances to generate a better overall picture of the activity of a given column towards bases.6,7 However, these procedures are laborious, and we have sought to use chemometric procedures to try to identify a reduced number of test compounds and mobile phases which still can generate a valid column assessment8 (i.e., one giving an assessment similar to that from the full dataset). In this paper, we have concentrated on data obtained at the high pH end of the stability of most RP columns (pH 7).We have shown that the performance of columns is generally much better at acid pH (pH 3).6,7 This is presumably due to suppression of the ionisation of silanol groups and reduced ionexchange interactions with bases. However, the selectivity of a separation may differ at pH 3 and 7; moreover, the retention of some bases may be too low at pH 3 for a separation to be obtained.Thus testing at pH 7 is of practical interest, and represents for strong bases a more critical condition where the limitations of the column may be more apparent. It should be noted that for strong bases (pKa > 8) the solute is still likely to be ionised at pH 7, leading to greater silanophilic interactions than at pH 3. However, for weak bases (pKa = 4–6), pH 7 can reduce silanophilic effects because the base will be nonionised. Chemometric methods can be employed to determine which tests are significant. By manual inspection of PC (principal components) plots it is possible to visualise the influence of different tests.For example, is it possible to obtain equivalent information using fewer than the nine test compounds used at pH 7? If so, new columns can be evaluated with less work. However, a difficulty relates in part to the large number of possible subgroups of tests. For example, there are 84 possible ways of selecting six out of nine test compounds.It is clearly impracticable to perform such experiments independently, and even using simple computation approaches, viewing 84 scores plots is prohibitively time consuming. We have shown that procrustes analysis9,10 can be used to compare directly the information obtained in two scores plots. A numerical indicator of how similar the plots are can be obtained and is outlined below. Hence it is possible to obtain information on the similarity between scores plots using a subset of the overall dataset.This helps answer the question of whether a particular subset of compounds or chromatographic tests provides similar information to the full set of compounds. It is therefore possible to rank each subset of compounds, and so examine the effect of reducing the number of compounds in the dataset, and determining whether more efficient approaches are possible and, if so, which ones. Previously at pH 7 we described a dataset consisting of 27 individual tests as discussed in Section 2.1.Another question that could be answered is whether reducing the number of tests produces acceptable information. However, there are Analyst, 1999, 124, 227–238 22727!/[N!(27 2 N)!] ways of doing this, which equals over 80 million ways of reducing the number of tests by 10. It is possible, however, to explore the changes when a small number of tests are removed, to see which tests are the key tests. The greater the change, the more significant is the test. 2. Methods 2.1. Chromatography Full details of the HPLC equipment and columns used have been given previously.6,7 Eight chromatographic columns were tested, at pH 7, as summarised in Table 1. Nine test compounds were used. Single letter mnemonics will be used below, for brevity, as indicated in Table 1(b). Three isoeluotropic mobile phases were used [Table 1(c)], and in all cases isocratic conditions using a single solvent were employed. Finally, three column evaluation parameters, kA, N and As values, were recorded for the probes at pH 7. 2.2. Principal components analysis and finding optimum fits For each mobile phase a full dataset consisting of eight rows ( = chromatographic columns) and 27 columns ( = evaluation parameters) is formed as a matrix. Because all parameters are measured on different physical scales, the first step is to standardise the data as follows: x c c s ij ij j j = - ( ) where �c j is the mean value of parameter j over all eight columns and sj is the corresponding standard deviation, giving a new matrix X, and cij is the corresponding raw column evaluation parameter for column i and parameter j.Principal components analysis (PCA)11,12 can be performed on the full dataset of 27 evaluation parameters as follows for each individual mobile phase: X = T·P + E A similar procedure involves performing PCA on a reduced dataset, with certain parameters removed. For example, all measurements on two of the nine compounds could be omitted, to give a new scores matrix: rX = rT·rP + rE where rX is the reduced standardised data matrix.If two compounds are omitted there will be 21 measurements left. The method in this paper compares the scores plot from the first two principal components, which as discussed earlier appear adequate to describe the data.8 It could be extended to more components. The aim is to obtain a scores plot from e reduced dataset that resembles as closely as possible the overall dataset scores plot, by expansion/contraction of the scale and by rotation.If tik is the score of the ith column and kth principal component ( = 1 or 2) for the full dataset, and rtik for the reduced dataset, then the mean distance of the scores from the origin for the full dataset is D t t i t i = + ( ) = Â 1 2 2 2 1 8 8 / with a corresponding equation for Dr for the reduced dataset, then the scores of the reduced dataset are scaled so that s,rtik = rtik (D/Dr) meaning that the average distance from the origin is identical for both the full and reduced dataset.The next step is to rotate the scores plot from the reduced dataset, so that it fits the scores plot from the overall dataset as closely as possible. As stated in previous papers,8–10 a difficulty that sometimes occurs is that the sign of a PC can change unpredictably according to algorithm. This is because PC calculations involve a square root step.However, it is only necessary to consider one sign change (a reflection) of the second PC. Changing the signs of both PCs is the same as a 180° rotation. The optimum rotation angle, q, is calculated so that d q = - ( ) - = Â t p t ik k s r ik k 1 2 1 2 , , is a minimum where q,s,rtik is the score of the reduced and scaled dataset rotated through q and p is a parity parameter, equalling +1 if there is no reflection and 21 if there is reflection. This value of d is obtained using a simple simplex method, involving starting with two angles of 0 and 10°, determining which results in the lowest error, and moving in that direction, then reducing the step-size until convergence. In most cases the functions are monotonic and, since there is only one variable (q), the method for finding the optimum is not too crucial, as many standard approaches would provide the same solution.It is important to recognise that, although theoretically multiple optima are possible, this only occurs if the two PC scores plots differ considerably, in which case a large value of d is obtained no matter what method of optimisation is employed.Table 1 Datasets used in this paper (a) Columns— Inertsil ODS Inertsil ODS-2 Inertsil ODS-3 Kromasil C-18 Kromasil C-8 Symmetry C-18 Supelco ABZ+ Purospher (b) Test compounds— Compound Letter Pyridine P Nicotine N Amphetamine A Codeine C Diphenhydramine D Nortriptyline R Procainamide M Quinine Q 2-[N-methyl-N-(2-pyridyl)-amino]ethanol E (c) Test conditions— Mobile phase a Methanol–phosphate buffer (pH 7) b THF–phosphate buffer (pH 7) c Acetonitrile–phosphate buffer (pH 7) 228 Analyst, 1999, 124, 227–2382.3.Graphical output and removal of individual compounds Parameters are systematically removed from the dataset. There are nine possible ways in which a single compound can be removed, in each case resulting in 24 remaining parameters. The three chromatographic test measurements from each compound are left out of the dataset.There are 9!/[(9 2 N)!N!] ways of removing N compounds from the dataset, as can be demonstrated by straightforward binomial formulae, i.e., 36 ways of removing two compounds, 84 for three compounds and 126 for four compounds. Removal of more compounds was not studied in this case, for brevity. The best rotation and scaling are computed as above for each reduced dataset, and then the residual distance d calculated for each combination of compounds, as an indicator of how much difference is made by omitting certain compounds.Finally, the solutions are ranked in order of significance. Those with lowest values of d have the least effect on the scores plots and so represent good sets of compounds to remove. Finally, it is possible to obtain the scores and loadings plots for the reduced datasets. Superimposing the optimal plots for both the full and reduced datasets provides a graphical illustration of how similar the information is.This method can be extended to removing any single or set of parameters, for example one of the 27 individual measured column evaluation parameters can be removed. In practice, of course, it is not necessarily a good strategy to omit a few random tests, especially if they correspond to different compounds and peak parameters. However, if groups of compounds and parameters are seen to have a significant influence on the scores plots, then this provides excellent evidence for either retaining or omitting such groups.Finally, an entire set of peak-shape parameters such as kA or asymmetry can be removed, reducing the dataset by one third. 2.4. Comparison of two full datasets Another aspect of interest is to compare two full datasets of 27 measurements under two different conditions, for example when methanol and THF are used as mobile phases. The stage of removing parameters is not necessary. The procedure simply involves obtaining the scores and loadings for each condition independently.One condition (using methanol as a mobile phase) is set as the reference and the aim is to ask how similar are the patterns obtained using methanol as a mobile phase to the patterns obtained using THF or acetonitrile. The second scores are then scaled, reflected (if necessary) and rotated to fit the scores of the data obtained using the methods described above. The value of d can be obtained as a measure of similarity. 3.Results 3.1. Comparison of scores plots A first step is to compare two PC scores plots, one with all the performance parameters used and another when only a subset of the parameters is employed. Fig. 1(a) is the plot at pH 7 using methanol as a mobile phase with all 27 chromatographic column performance parameters, after the data have been standardised. Fig. 1(b) represents the same data but with only 18 parameters, leaving out those for compounds P, C and D, and Fig. 1(c) omits C, M and Q. Fig. 2(a)–(c) are the corresponding loadings plots. An aim is to ensure that the two scores plots fit as closely as possible, by rotation and scaling as discussed above. Fig. 3 shows the error (d) between the scores map produced using all nine compounds and that produced by excluding P, C and D [Fig. 3(a)] and that produced by excluding C, M and Q [Fig. 3(b)] as a function of rotation angle. The optimum rotation angle for the former group is 0° (within 1°) and for the latter it is 18°.Note that in these cases no reflection is needed, although this is by no means always the case. The optimum (minimum) error (d) is 0.236 for Fig. 3(a) and 1.172 for Fig. 3(b). The corresponding scores plots are given in Fig. 1, and it can be seen that 1.172 is a substantial error. Fig. 1(a) and (b) are almost identical with Purospher in the top right hand corner, Kromasil C-18 at the bottom left, with Inertsil ODS and ODS-3 almost superimposed, Symmetry C-18 and Kromasil C-8 having low scores on PC1 and in corresponding positions along PC2.However, the relative positions of the chromatographic columns have changed considerably in Fig. 1(c). For example, Purospher is very close to Symmetry C-18, suggesting that by using a reduced subset of six compounds, the performance of these latter two columns is very similar, which is shown not to be the case when the full set of nine compounds is employed. Furthermore, the behaviour of Inertsil ODS and ODS-3 is now significantly different.In order to illustrate this further, the correlation coefficients between columns i and j over all K parameters given by Fig. 1 Scores plots at pH 7 using methanol mobile phase. From top to bottom: (a) including all compounds; (b) excluding P, C and D; and (c) excluding C, M and Q. Analyst, 1999, 124, 227–238 229r x x x x x x x x ij s ik s ik s jk s jk k K s ik s ik s jk s jk k K k K = - ( ) - ( ) - ( ) - ( ) = = = Â Â Â 1 2 2 1 1 is calculated, where sxik is the value of the standardised parameter k for column i.These matrices are calculated for the datasets consising of all nine compounds ( = 27 parameters) and those for subsets of six compounds ( = 18 parameters). Then the root mean square difference between these sets of correlation coefficients is calculated as follows: p all ij p ij j i i r r D = - ( ) = = Â Â 2 8 1 8 36 / where allrij is the correlation coefficient defined as above between the columns with all the compounds included and prij is that for subset p.In the example above, these values are 0.0546 and 0.1566 for the datasets of Fig. 1(b) and (c), respectively, demonstrating that removing the compounds of Fig. 1(c) has a three times larger influence on relative correlations. It is important to recognise the need for rotation and scaling to reduce this error, as demonstrated in Fig. 3. The use of simplex (or any sensible alternative optimisation method) to find the lowest value of d is important. 3.2.Removal of compounds The next step is to study the influence of different compounds on the overall scores map. This is done by removing a number of compounds and calculating an optimal error as described above. Table 2 lists the values of d as an increasing number of compounds are removed from the dataset. There are a number of important conclusions to be drawn from these results. The first is that it is not sensible to simply rank compounds in order of usefulness. One compound is not ‘better’ than another compound.What is important is that groups of compounds can be left out of the analysis, and, on the whole, these groups are mutually exclusive. A set of compounds can be omitted from the dataset and still provide a PC map similar to that for the full dataset. Removing a very different set of compounds may also achieve similar quality results. An analogy is if the compounds are regarded as distributed on the circumference of a circle.In order to span the circle evenly, it is necessary to remove compounds evenly from around the Fig. 2 (a) Loadings plot using methanol mobile phase at pH 7 and all compounds included. (b) Loadings plot using methanol mobile phase at pH 7 excluding P, C and D. (c) Loadings plot using methanol mobile phase at pH 7 excluding C, M and Q. (d) Superimposed loadings plots using methanol mobile phase at pH 7 including all (solid symbols) and excluding P, C and D (open symbols).(e) Superimposed loadings plots using methanol mobile phase at pH 7 including all (solid symbols) and excluding C, M and Q (open symbols). Note that there is some overlap between symbols in (d) and (e), e.g. ANa7 in (d). 230 Analyst, 1999, 124, 227–238circumference, and there are several mutually exclusive ways of doing this. Table 3(a) lists the best 11 solutions when four compounds are removed [as obtained from Table 2(d)].It is fairly clear that there are mutually exclusive groups. For example, P and N never co-occur, but one occurs in all cases. This implies that although it is possible to omit one of these compounds from the analysis and obtain results similar to those of the full dataset, removing both at the same time is not a good approach. There appear several important relationships: (i) P and N are mutually exclusive, but one occurs in all the best solutions. (ii) D and R are mutually exclusive but one occurs in all the best solutions.(iii) E and M are mutually exclusive apart from the eleventh solution [Table 3(a)] and on average one occurs in all the best solutions. (iv) C and Q are mutually exclusive and one occurs in all solutions apart from 8 and 11. (v) A hardly features in this table. (vi) The pairs R and E and M and D show fairly similar trends. Hence it is possible to obtain the following three groups of contrasts: (i) P versus N; (ii) D and E versus R and M; (iii) C versus Q.Contrasting compounds are expected to show very similar behaviour. If one of a pair is removed from the test set, little difference is made to the column assessment. If both are removed, this can have a major influence on the results from the column assessment. Eight possible subsets of compounds can be predicted simply by the groupings above, either including or excluding each contrasting compound as in Table 3(b). It can be seen that there are many similarities in the patterns between Table 3(a) and (b).The opposite effect is seen with the least favourable solutions. For example, P and N occur in most of the worst solutions in Table 2(d). Hence removing both compounds from the dataset is a bad decision. Removing only one is reasonable. Removal of A occurs four times in the worst five solutions in Table 2(d), indicating that it may be making important contributions to the overall assessment. The absence of the omission of A from the best solutions also adds weight to this Fig. 3 Influence of rotation angle on errors for (a) P, C and D and (b) C, Q and M. Table 2 Optimised errors (a) Optimised errors after one compound removed from the dataset— A 0.30 C 0.33 N 0.34 M 0.35 Q 0.37 D 0.37 E 0.40 R 0.41 P 0.41 (b) Optimised errors after two compounds removed from the dataset— N D 0.34 A M 0.37 N R 0.37 A C 0.38 A Q 0.40 D M 0.41 C D 0.42 P C 0.43 R M 0.44 P M 0.46 Q E 0.46 A E 0.47 P D 0.48 C R 0.49 R E 0.50 C E 0.51 P Q 0.51 N C 0.51 N Q 0.52 N M 0.54 D Q 0.56 P R 0.57 R Q 0.58 D E 0.58 A D 0.60 N A 0.62 M E 0.63 M Q 0.63 N E 0.65 P E 0.67 A R 0.68 C Q 0.70 P A 0.74 C M 0.76 D R 0.77 P N 0.78 (c) Optimised errors after three compounds removed from the dataset— P C D 0.24 P D M 0.29 N C D 0.31 N D M 0.36 N R M 0.36 N C R 0.39 P C R 0.43 A Q E 0.44 P R M 0.45 N R E 0.47 N R Q 0.51 A C E 0.51 A D M 0.51 N D Q 0.53 A C D 0.54 C R E 0.54 R Q E 0.54 P R E 0.55 P D Q 0.57 A M Q 0.61 C D E 0.62 A R M 0.62 P R Q 0.63 A M E 0.63 P D E 0.64 P M Q 0.65 N D E 0.65 R M E 0.65 D Q E 0.66 Analyst, 1999, 124, 227–238 231Table 2 Continued (c) Continued— A R E 0.66 A C R 0.68 A D Q 0.68 C Q E 0.68 N A Q 0.69 N Q E 0.69 D M Q 0.70 P C Q 0.70 P Q E 0.70 N A M 0.71 A C Q 0.71 P A C 0.71 N A C 0.72 A D E 0.72 M Q E 0.72 P A M 0.72 R M Q 0.72 D M E 0.73 P C E 0.73 N A D 0.74 C D M 0.75 A R Q 0.75 N D R 0.75 P N D 0.75 D R M 0.76 P A Q 0.76 N C E 0.77 P C M 0.78 A C M 0.78 P N R 0.79 C D Q 0.80 C R M 0.80 P M E 0.81 C D R 0.82 N M Q 0.83 P N C 0.84 C R Q 0.84 N A R 0.86 P N M 0.87 P N Q 0.87 N C Q 0.88 D R E 0.88 N M E 0.89 N A E 0.90 P D R 0.91 D R Q 0.93 C M E 0.97 N C M 0.99 P A D 0.99 P A E 0.99 P A R 1.05 A D R 1.07 P N E 1.13 P N A 1.17 C M Q 1.17 (d) Optimised errors after four compounds removed from the dataset— N R Q E 0.42 P C R E 0.44 P R Q E 0.48 N C R E 0.49 P C D M 0.50 P D M Q 0.50 N A C D 0.53 N A D M 0.55 P C D E 0.58 P R M Q 0.60 P R M E 0.60 P N C D 0.61 P C D Q 0.61 A R Q E 0.61 A M Q E 0.62 A C R E 0.62 A C Q E 0.64 P C R M 0.65 P D Q E 0.65 N D R M 0.65 P N C R 0.66 P N D M 0.66 Table 2 Continued (d) Continued— A C D E 0.68 N C D E 0.69 N R M Q 0.69 P C R Q 0.69 N D Q E 0.70 N R M E 0.70 P N R M 0.70 N C D R 0.70 R M Q E 0.71 N D M Q 0.72 A D Q E 0.72 P D M E 0.72 N A D Q 0.72 A D M Q 0.73 P A C D 0.73 C R Q E 0.74 P D R M 0.74 A R M E 0.74 N A R E 0.75 N A R M 0.76 P C D R 0.76 N A C R 0.77 P A D M 0.78 P N R E 0.80 A C D M 0.81 C D Q E 0.81 D M Q E 0.81 A R M Q 0.81 A D M E 0.82 N C D M 0.82 P M Q E 0.82 N C D Q 0.83 P A M Q 0.83 P N R Q 0.83 N A R Q 0.84 P C Q E 0.84 P N D Q 0.84 N C R Q 0.84 N C R M 0.85 N A D E 0.86 N D M E 0.87 N D R E 0.87 A C D Q 0.88 P A C Q 0.90 N D R Q 0.90 P A C R 0.90 P D R E 0.91 N C Q E 0.92 P A D E 0.92 C D R E 0.93 N A Q E 0.93 A C R M 0.93 P A R E 0.93 P A R M 0.94 C R M E 0.96 N A M Q 0.96 A C M E 0.96 A C R Q 0.97 P A C M 0.97 D R Q E 0.97 D R M E 0.99 N M Q E 0.99 C D M E 0.99 N A C E 1.00 P D R Q 1.01 P A D Q 1.01 A D R M 1.02 D R M Q 1.03 C D R M 1.04 P N D E 1.04 P A Q E 1.05 P N M Q 1.05 P A R Q 1.06 N A C Q 1.06 N A M E 1.08 P C M E 1.08 A C D R 1.09 P A C E 1.09 232 Analyst, 1999, 124, 227–238argument.Similar types of conclusions can be obtained when examining Table 2(b) and (c). In Table 2(b) the worst pairs to leave out are P and N and D and R, for example.The best solutions in Table 2(c) exclude either P or N but not both. Some chemical deductions can be made. For example, P and N are both hydrophilic bases containing a pyridine ring, whereas D and R are larger hydrophobic bases. Omitting one of two similar structures from the analysis does not significantly influence the nature of the PC plots, whereas omitting both structures has a major influence as the chromatography of such a type of compound is no longer represented.In reality, this conclusion is to be expected. For example, if a column is to be tested for a variety of different types of compounds such as bases, acids and zwitterions, it is necessary to obtain compounds representative of each group. If the number of compounds is limited, and it is still desired to assess the quality of chromatography over the entire range, it is better to take a selection of compounds from each group than simply concentrate on one group. In contrast, however, if only one class of compound is of interest, this constraint is not so important. In the work described in this paper, we define the type of chromatographic assessment required by the overall scores map obtained using nine target compounds. The only information we can then provide is which subset of compounds results in similar patterns.We cannot provide completely general recommendations for all conceivable compound classes and groups. 3.3. Removal of individual tests There are an immense number of ways in which the 27 tests can be reduced in size.For example, there are 888 030 possible selections of 20 tests, and 8 436 285 possible selections of 17 tests. To check all these computationally is difficult. Using a 350 MHz Pentium II requires 3 days to scan all combinations of 17 tests. However, it is possible to examine the quality of solution as the number of tests is reduced. Table 4 lists the best and worst solutions when up to six independent tests have been removed together with the optimised value of d.A value of around 0.1 indicates an excellent fit whereas that of 1 a much poorer fit. Fig. 4(a) is a graph of d against the solution number ranked in order of quality when four tests are omitted from the dataset. It can be seen that there are a small number of very poor and very good solutions. It is constructive to examine which parameters are most significant in terms of quality of solution. A comparison of kA, N and As can be performed by calculating the average frequency of occurrence in a window of 41 ranked solutions.For example, if four tests are omitted from the dataset, then there will be 164 possible tests corresponding to 41 solutions ( = 4 3 41). If the best 41 solutions contain 80 kA parameters, 60 N parameters and 24 As parameters (totalling 164 tests), then these three values are proportional to the average frequency of occurrence centred around the solution number.This procedure is continued for solutions 2–43, 3–44 and so on. Graphs of relative occurrence against rank of solution when four tests are removed are given in Fig. 4(b) and (c) for the best and worse solutions, respectively. It is very clear that very few of the poor solutions involve removing kA, which is clearly less represented in the poor solution than either N or As. This suggests that kA has relatively little impact on the quality of solution. There is less difference between the three types of parameter for the best solutions.Table 4 suggests a number of features. The most obvious reinforce the observations about kA described above, hardly being represented in the worst solutions [only appearing in Table 4(a) and (f)]. Removing one out of 27 tests makes only a small difference to the scores plots. This contrasts to the case where six tests are removed. The worst solution has a value of d of 1.211 whereas the best leaves the scores plot almost unchanged at 0.129.It is interesting that the worst solution of all corresponds to removal of both peak-shape parameters for three separate compounds. In many of the worst solutions, pairs of peak-shape parameters for two or three compounds are omitted. Fig. 5(a) and (b) are the optimised rotated, reflected and scaled scores plots when the least and most significant six independent column parameters [see Table 4(f)] are removed. It can be shown that there is extremely little change in Fig. 5(a), suggesting that an appropriate reduction in column assessment parameters hardly changes the pattern, but Fig. 5(b) suggests that an inappropriate reduction in column assessment parameters has a substantial effect on the scores. Table 2 Continued (d) Continued— P N C Q 1.10 P A M E 1.13 A D R E 1.14 N A C M 1.14 C D R Q 1.16 P N D R 1.17 A D R Q 1.19 P N Q E 1.19 P N C M 1.20 P C M Q 1.20 A C M Q 1.24 P N A Q 1.25 P N A M 1.25 P N A C 1.26 C D M Q 1.26 P N C E 1.27 C M Q E 1.27 N C M E 1.27 C R M Q 1.30 N A D R 1.34 P N M E 1.35 P N A R 1.44 P N A D 1.44 P A D R 1.48 N C M Q 1.49 P N A E 1.60 Table 3 Best solutions (a) Rank of the best solutions when four compounds are left out with the compounds indicated— Rank P N A C D R M Q E 1 ] ] ] ] 2 ] ] ] ] 3 ] ] ] ] 4 ] ] ] ] 5 ] ] ] ] 6 ] ] ] ] 7 ] ] ] ] 8 ] ] ] ] 9 ] ] ] ] 10 ] ] ] ] 11 ] ] ] ] (b) Eight possible best solutions as described in the text— P/N C/Q D&M/ R&E P N A C D R M Q E 0 0 0 ] ] ] ] 0 0 1 ] ] ] ] 0 1 0 ] ] ] ] 0 1 1 ] ] ] ] 1 0 0 ] ] ] ] 1 0 1 ] ] ] ] 1 1 0 ] ] ] ] 1 1 1 ] ] ] ] Analyst, 1999, 124, 227–238 233Table 4 Solutions (a) Solutions when one test is removed— Aka7 0.110 Cka7 0.128 Nka7 0.130 QAsa7 0.152 Rka7 0.154 QNa7 0.155 Mka7 0.157 CAsa7 0.164 Dka7 0.164 MAsa7 0.167 NNa7 0.170 AAsa7 0.172 NAsa7 0.172 Qka7 0.173 DNa7 0.177 CNa7 0.177 MNa7 0.181 ANa7 0.186 RNa7 0.186 Eka7 0.188 Pka7 0.192 ENa7 0.195 DAsa7 0.204 PAsa7 0.209 PNa7 0.226 RAsa7 0.228 EAsa7 0.258 (b) Best and worst solutions when two tests are removed— AAsa Nka 0.159 PAsa NNa 0.351 Cka Aka 0.162 NNa NAsa 0.353 Nka NAsa 0.166 NAsa EAsa 0.355 NAsa Mka 0.167 DNa RAsa 0.355 RNa MAsa 0.170 ANa NAsa 0.357 Cka NAsa 0.172 DAsa EAsa 0.358 Aka QAsa 0.174 MAsa EAsa 0.359 ANa Nka 0.174 CNa MNa 0.362 Dka MAsa 0.176 Eka EAsa 0.366 Cka NNa 0.177 DAsa RNa 0.366 Cka AAsa 0.179 PNa NNa 0.375 CAsa Dka 0.180 PAsa NAsa 0.379 Aka Nka 0.180 RNa RAsa 0.382 PAsa Nka 0.180 PNa NAsa 0.383 Dka AAsa 0.183 DAsa AAsa 0.385 Aka QNa 0.183 ANa PAsa 0.391 AAsa Rka 0.183 AAsa RAsa 0.402 CNa Eka 0.184 ANa PNa 0.410 Aka Mka 0.185 DAsa RAsa 0.429 Cka MAsa 0.186 PNa PAsa 0.435 (c) Best and worst solutions when three tests are removed— AAsa CAsa Dka 0.141 RAsa DAsa EAsa 0.525 Cka DNa NNa 0.147 PAsa RAsa AAsa 0.531 DAsa Nka NAsa 0.147 PAsa ANa AAsa 0.541 Cka DAsa NNa 0.148 RNa AAsa DAsa 0.543 ANa DNa Nka 0.149 PNa ANa AAsa 0.547 AAsa CAsa Nka 0.150 RNa RAsa AAsa 0.549 AAsa Dka MAsa 0.150 ANa NNa NAsa 0.550 Cka DAsa NAsa 0.153 PAsa NNa NAsa 0.558 ANa Cka DNa 0.155 RAsa DNa DAsa 0.558 RNa Cka MAsa 0.159 PNa NNa NAsa 0.560 DAsa Nka NNa 0.161 PAsa ANa NNa 0.561 PNa DNa Nka 0.163 PNa ANa NNa 0.571 RNa Nka MAsa 0.163 RNa RAsa DAsa 0.572 RNa NAsa Mka 0.164 PNa PAsa AAsa 0.585 AAsa Nka MAsa 0.164 PNa PAsa NNa 0.585 RNa Nka NAsa 0.165 PNa ANa NAsa 0.588 ANa DNa Mka 0.166 PAsa ANa NAsa 0.593 RNa Cka NNa 0.167 PNa PAsa NAsa 0.617 RNa Cka NAsa 0.169 RAsa AAsa DAsa 0.620 DNa NNa Mka 0.169 PNa PAsa ANa 0.636 (d) Best and worst solutions when four tests are removed— PAsa RNa Nka MAsa 0.126 RAsa AAsa DAsa EAsa 0.716 PNa DNa Nka MAsa 0.146 PNa RAsa AAsa DAsa 0.717 PNa RNa Nka MAsa 0.148 PNa PAsa AAsa DAsa 0.721 RNa Cka NNa MAsa 0.148 PNa PAsa NAsa EAsa 0.731 RNa Cka NAsa MAsa 0.152 PNa ANa AAsa NAsa 0.731 Dka DAsa NAsa MNa 0.152 PNa PAsa AAsa NNa 0.735 234 Analyst, 1999, 124, 227–2383.4.Removal of groups of test parameters As discussed above, it is also desirable to compare the scores plots when a group of parameters is removed.Fig. 6(a) compares the scores plot using methanol as a mobile phase and keeping all the 27 column assessment parameters, with that retaining the two peak-shape parameters, N and As and omitting the kA values, resulting in 18 column assessment parameters, which corresponds to a d value of 1.62. In contrast, retaining only the nine retention parameters results in a d value of 3.49 as illustrated in Fig. 6(b). This is primarily because the retention parameters are similar in behaviour and are closely clustered in the loadings plot. In contrast, the peak-shape parameters span the principal component loadings space much more evenly, and so are better at characterising the columns. Fig. 6(c) and (d) involve removing one of the two peak-shape parameters resulting in relatively small d values (of 0.68 and 1.10, respectively), suggesting that these two have a correlation close to ±1, meaning that they both convey fairly similar information (typical of correlated parameters), so removing one set does not significantly change the overall correlation structure. 3.5. Influence of mobile phase modifiers It is also constructive to compare the performance of columns using different mobile phase modifiers. This has important practical implications, because it guides the experimenter as to Table 4 Continued (d) Continued— AAsa CAsa Dka MNa 0.152 PNa PAsa ANa EAsa 0.738 RNa Nka NAsa MAsa 0.153 PNa PAsa RAsa AAsa 0.741 RNa NAsa Mka MAsa 0.154 PAsa ANa NAsa EAsa 0.742 PNa RNa CAsa Nka 0.156 RAsa AAsa DNa DAsa 0.750 PNa CNa DNa Eka 0.157 PAsa ANa AAsa NAsa 0.756 RNa ANa Nka MAsa 0.158 RNa RAsa AAsa DAsa 0.757 Pka Rka AAsa CAsa 0.158 PAsa RAsa AAsa DAsa 0.759 PNa DNa Mka MAsa 0.160 PNa ANa NNa NAsa 0.772 RNa Cka CAsa NNa 0.161 PNa PAsa AAsa NAsa 0.776 PAsa RNa CAsa Nka 0.161 PAsa ANa NNa NAsa 0.787 AAsa CNa QNa Eka 0.162 PNa PAsa NNa NAsa 0.790 PAsa RNa Mka MAsa 0.163 PNa PAsa ANa NNa 0.799 Pka AAsa CAsa Qka 0.163 PNa PAsa ANa AAsa 0.810 PAsa CAsa DAsa Nka 0.163 PNa PAsa ANa NAsa 0.839 (e) Best and worst solutions when five tests are removed— PAsa RAsa Dka Nka MAsa 0.127 PNa PAsa AAsa DAsa NAsa 0.904 CNa DAsa NAsa Qka Eka 0.130 PAsa ANa AAsa NAsa EAsa 0.905 PAsa CAsa Dka DAsa Nka 0.130 PNa PAsa ANa NAsa MAsa 0.908 PAsa Dka DAsa Nka MAsa 0.131 PNa PAsa ANa AAsa EAsa 0.909 Pka PAsa RNa MAsa Qka 0.133 PNa ANa NNa NAsa EAsa 0.912 PAsa RAsa CAsa Dka Nka 0.133 PNa PAsa RAsa AAsa NAsa 0.913 PAsa RNa MAsa Qka Eka 0.140 PNa PAsa NNa NAsa EAsa 0.922 Pka PAsa RNa Cka MAsa 0.143 PNa ANa AAsa NNa NAsa 0.922 ANa CNa DNa Qka Eka 0.145 PNa PAsa ANa NNa EAsa 0.923 Pka PAsa CAsa DAsa Qka 0.148 PAsa RAsa ANa AAsa DAsa 0.939 PNa CNa DNa Nka EAsa 0.149 PAsa ANa NNa NAsa EAsa 0.952 Pka PAsa Rka RAsa CAsa 0.150 PNa PAsa ANa AAsa DAsa 0.957 PNa RNa CAsa MNa Eka 0.151 PNa PAsa AAsa NNa NAsa 0.957 Pka PAsa Rka CAsa DAsa 0.152 PNa PAsa RAsa AAsa DAsa 0.960 RNa CNa NAsa Qka Eka 0.152 PNa PAsa RAsa ANa AAsa 0.963 RAsa CAsa Dka Nka NAsa 0.153 PAsa ANa AAsa NNa NAsa 0.964 Pka PAsa Cka CAsa DAsa 0.154 PNa PAsa ANa AAsa NNa 0.977 RNa Cka Nka NAsa MAsa 0.155 PNa PAsa ANa NAsa EAsa 0.986 PAsa RNa Cka Nka MAsa 0.156 PNa PAsa ANa NNa NAsa 1.017 CNa DAsa NNa Qka Eka 0.157 PNa PAsa ANa AAsa NAsa 1.030 (f) Best and worst solutions when six tests are removed— PNa Rka CAsa DAsa MNa Eka 0.129 PNa PAsa RNa RAsa AAsa DAsa 1.097 Pka PAsa Aka CAsa DAsa Qka 0.129 PNa PAsa ANa NAsa MAsa EAsa 1.098 Rka AAsa NAsa MNa QNa Eka 0.130 PNa PAsa RNa RAsa ANa AAsa 1.098 PNa Rka DAsa MNa QNa Eka 0.131 PNa PAsa RAsa AAsa DNa DAsa 1.101 Rka AAsa NAsa MNa QAsa Eka 0.133 Cka CNa Nka Mka MNa Qka 1.105 Rka RAsa NAsa MNa QNa Eka 0.135 PNa PAsa ANa AAsa NNa EAsa 1.107 AAsa CNa Dka NAsa Mka QAsa 0.135 PNa PAsa ANa NNa NAsa MAsa 1.109 AAsa CNa DNa NNa Qka Eka 0.136 PNa PAsa RNa ANa AAsa DAsa 1.111 PNa RNa CNa Nka QNa EAsa 0.138 PNa PAsa RNa ANa AAsa NAsa 1.113 PAsa RAsa Aka CAsa Qka Eka 0.140 PNa PAsa RAsa ANa AAsa NNa 1.119 Rka ANa DAsa MNa QNa Eka 0.143 PNa PAsa ANa AAsa DAsa NNa 1.122 PNa RNa Aka MNa QNa EAsa 0.145 PAsa ANa AAsa NNa NAsa EAsa 1.142 AAsa Cka CNa NAsa QAsa Eka 0.146 PNa PAsa RAsa AAsa DAsa NAsa 1.158 Rka DAsa NAsa MNa QNa Eka 0.146 PAsa RAsa ANa AAsa DAsa NAsa 1.161 Pka PAsa CNa CAsa Dka DAsa 0.147 PNa PAsa ANa NNa NAsa EAsa 1.175 Rka RAsa CNa NAsa QNa Eka 0.147 PNa PAsa RAsa ANa AAsa NAsa 1.181 PAsa Rka RAsa CAsa MNa Eka 0.148 PNa PAsa ANa AAsa NAsa EAsa 1.185 Pka PAsa Aka Cka CAsa DAsa 0.148 PNa PAsa ANa AAsa DAsa NAsa 1.189 CNa DAsa NAsa Qka QAsa Eka 0.148 PNa PAsa RAsa ANa AAsa DAsa 1.202 AAsa CNa Dka Nka NAsa QAsa 0.149 PNa PAsa ANa AAsa NNa NAsa 1.211 Analyst, 1999, 124, 227–238 235whether it is necessary to use more than one solvent for assessment, or whether the behaviour under a different solvent regime can be predicted. Fig. 6(e) shows the scores when methanol and acetonitrile are used, with a d value of 1.55. This suggests some small changes, as the behaviour of the Inertsil ODS-3 column is somewhat different, but nevertheless is sufficiently small to indicate reasonably similar behaviour. For a preliminary assessment, it suggests that only one of these solvents is necessary to provide exploratory information on column performance, but for a detailed assessment of individual columns, as a second step, it is advisable to use both mobile phase modifiers. (A similar assessment at pH 3 of these two solvent systems gives a d value of 0.98, suggesting that at pH 3 there is less difference.) THF, in contrast, provides a very different picture of the columns [Fig. 6(f)], corresponding to a d value of 3.71. There is no easy way to superimpose the two patterns. Since only eight columns are used in this test study, and the geometric transformations of rotation, reflection and scaling are employed (leaving five degrees of freedom), the pattern of Fig. 6(f) is really significant. Note that the picture is the best that can be done to try to overlap the two scores plots. Hence the behaviour in THF is very different to that in methanol (and acetonitrile). (A similar d value of 3.61 exists at pH 3.) 4. Conclusions An important function of chemometrics is to be able to reduce the number of analytical tests to obtain a given amount of information.In this paper, we propose approaches for chromatographic column assessment, but the methodology is applicable more widely in other areas where multiple analytical tests are performed. A very important aspect, however, relates to the target PC scores plot. In the cases studied in this paper, the target is of a full set of 27 tests at pH 7 in methanol. Furthermore, only eight columns are explored. It is crucial to understand that chemometric methods are not pure statistical approaches, that is, they depend critically on prior embedded chemical assumptions, as opposed to straight statistical methods which often deal with supposedly infinite and readily sampled populations.The methods proposed here can only converge on to a chemically defined target. What is essential to recognise is that this pattern is the reference for all the computations. If the number and range of compounds and columns in this reference dataset is too limited, of course, the methods described will aim only to reproduce these limited data.In the work described above, the test compounds are restricted to basic compounds. A question might arise as to whether even nine solutes and eight columns are sufficient to generate a satisfactory initial picture for all potential basic solutes and all columns. However, most of the nine solutes were not chosen at random and on the contrary have been used individually to illustrate differences between columns: the solutes have a fairly wide range of stereochemical features and pKa values. It would have been possible to use a larger number of columns but the number of experiments becomes prohibitive; again there is an optimisation between increasing the experimentation unacceptably or not performing sufficient experiments. For good analytical chemical reasons, it is felt that the target data are reasonably sized for acceptable conclusions.Other compounds such as zwitterions (e.g., amino acids), proteins, acidic compounds or hydrocarbons may exhibit a Fig. 4 (a) Value of d against solution number as four tests are removed. (b) Relative occurrences of the parameters (from top left to bottom left) As, N and kA for the best solution when four parameters are removed. (c) Relative occurrences of the parameters (from top to bottom) As, N and kA for the worst solution when four parameters are removed. Fig. 5 (a) Optimised comparison of scores plots when the least significant six column assessment parameters have been removed.(b) Optimised comparison of scores plots when the most significant six column assessment parameters have been removed. 236 Analyst, 1999, 124, 227–238different pattern. The procrustes analysis does not guarantee that the subset of tests will be adequate to check column performance for these other groups of compounds. This is under the control of the experimenter, who must first have found a sufficiently representative group of tests for the particular application in hand.In a previous paper, we recommended compounds and tests that might be removed. This paper refines these recommendations. (i) There are mutually exclusive groups of compounds that can be removed. (ii) The kA values have less influence on the overall assessment than the peak-shape parameters. (iii) Owing to the negative correlation coefficients between N and As, as a preliminary step, it may be sufficient just to measure one of these.(iv) At pH 7 there are considerable differences between using methanol and THF as mobile phases, but less between methanol and acetonitrile. (v) An efficient approach to assessment could be to record a single peak-shape parameter (e.g., N or As) using methanol and around five compounds, with a second step which increases the number of assessment parameters in cases where detailed information is required. It is important to recognise that there is no single best solution for reducing the number of test compounds.Obviously, in contrast, the most certain approach is to perform as many tests on as many compounds as possible. This will always give the most information but in reality is not practicable, as in many complex optimisation problems there are a huge number of local maxima. In certain types of study such as calibration or curve resolution one solution stands out above the others. In the type of problem studied in this paper, this is not the case.For example, it is possible to change the relative ranks of solutions by such simple approaches as data preprocessing and weighting, with little reference to the physical needs of the chromatographer. What can be obtained are general conclusions about possible strategies for reducing the number of tests and a fairly certain statement that equivalent information is possible using a well selected (but not random) subset of the original parameters: this can often be rationalised in terms of chemistry and molecular structure. Important trends, such as the requirement that one of P and N are included in the test set to avoid disastrous loss of information or that kA is not a particularly informative parameter, are evident from a variety of different calculations. As in most chemometrics, there are often a huge number of variations on a generic method which, if sufficiently robust, result in relatively similar conclusions. The conclusions from the work in this paper are sufficiently robust that they suggest a satisfactory target dataset and chemometric methodology. 5. References 1 D. Chan Leach, M. A. Stadalius, J. S. Berus and L. R. Snyder, LC–GC Int. 1988, 1, 22. 2 J. Nawrocki, J. Chromatogr., 1997, 779, 29. 3 G. B. Cox, J. Chromatogr. A, 1993, 656, 353. Fig. 6 (a) Optimised comparison of scores plots when kA values have been removed, using methanol as mobile phase. (b) Optimised comparison of scores plots when only kA values have been retained, using methanol as mobile phase. (c) Optimised comparison of scores plots when N values have been removed, using methanol as mobile phase. (d) Optimised comparison of scores plots when As values have been removed, using methanol as mobile phase. (e) Comparison of scores plots at pH 7 when methanol and acetonitrile are employed as mobile phases. (f) Comparison of scores plots at pH 7 when methanol and THF are employed as mobile phases. Analyst, 1999, 124, 227–238 2374 H. A. Claessens, M. A. van Straten, C. A. Cramers, M. Jezierska and B. Buszewski, J. Chromatogr. A, 1998, 826, 135. 5 D. V. McCalley and R. G. Brereton, J. Chromatogr. A, 1998, 828, 407. 6 D. V. McCalley, J. Chromatogr. A, 1997, 769, 169. 7 D. V. McCalley, J. Chromatogr. A, 1996, 738, 169. 8 R. G. Brereton and D. V. McCalley, Analyst, 1998, 123, 1175. 9 C. Demir, P. Hindmarch and R. G. Brereton, Analyst, 1996, 121, 1443. 10 S. Dunkerley, J. Crosby, R. G. Brereton, K. D. Zissis and R. E. A. Escott, Analyst, 1998, 123, 2021. 11 S. Wold, K. Esbensen and P. Geladi, Chemom. Intell. Lab. Syst., 1987, 2, 37 12 I. T. Jolliffe, Principal Components Analysis, Springer, New York, 1986. Paper 8/08537D 238 Analyst, 1999, 124, 227–238

ISSN:0003-2654    

DOI:10.1039/a808537d

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Automated trace enrichment for screening and/or determination of primary, secondary and tertiary amphetamines in biological samples by liquid chromatography R. Herráez-Hernández and P. Campíns-Falcó* Departamento de Química Analítica, Facultad de Química, Universidad de Valencia, Doctor Moliner 50, 46100-Burjassot, Valencia, Spain Received 17th December 1998, Accepted 28th January 1999 A rapid and simple liquid chromatographic method for the automated determination of amphetamines in biological fluids was developed.The proposed procedure is based on the injection of 250 mL of sample into a 20 3 2.1 mm id precolumn (packed with a 30 mm Hypersil C18 stationary phase) for enrichment and purification of the analytes. Next, the analytes are transferred to a 5 mm LiChrospher 100 RP18, 125 3 4 mm id analytical column for their separation under reversed-phase conditions. Water was used to eliminate the matrix components from the precolumn and a 0.2 m phosphate buffer (pH 3) containing 2% triethylamine was the mobile phase for the resolution of the amphetamines.The UV detector was set at 210 nm. The method was applied to the determination of different primary, secondary and tertiary amphetamines in plasma and urine: b-phenylethylamine, norephedrine, ephedrine, N-methylpseudoephedrine, pseudoephedrine, N-methylephedrine, amphetamine, 3-phenylpropylamine and methamphetamine. The method provides satisfactory linearity and reproducibility within the tested concentration range (1.0–10.0 mg mL21) and limits of detection of 50–500 ng/mL21.Introduction The rapid and sensitive determination of amphetamine and amphetamine analogues is an area of growing interest because these compounds have become popular drugs of abuse. Different analytical techniques have been used for this purpose, but gas chromatography (GC) with mass spectrometry (MS) seems to be the most reliable method for the identification and determination of amphetamines in biological fluids.GC-MS, however, has not been widely available in clinical, forensic and toxicological laboratories. Liquid chromatography (LC) is also well suited for the determination of amphetamines in biological fluids, but its application is limited by the poor sensitivity achieved with most LC detectors. In addition, amphetamines often show poor chromatographic performance in comparison with GC. Nevertheless, a great number of LC methods using a chemical reaction have been proposed to improve both sensitivity and resolution.Most of these methods are based on analyte isolation, followed by chemical derivatization and subsequent injection of the analyte derivative into the chromatographic system. In spite of their inherent difficulties, liquid– liquid extraction and solid-phase extraction are commonly used to isolate the analytes from the matrix .1,2 Since many reagents are reactive towards matrix components, re-extraction of the derivatized analytes is often necessary.As a result, the procedures are very time consuming and prone to errors. For these reasons, and since there is an increasing demand for the analysis of large number of samples, special attention has been devoted in recent years to the development of procedures which permit the automation of the entire analysis. One of the most elegant proposals for the on-line determination of amphetamines in biological fluids is the employment of solid-phase extraction reagents, which was introduced by Krull and co-workers.3,4 This approach is based on coupling a resinbased derivatization reagent to the analytical column.Controlled pore resins allow the direct injection of the samples and the exclusion of macromolecular matrix components. At the same time or somewhat later, the analytes are made to react with the reactive part of the solid-phase reagent. Finally, the derivatives formed are transferred to the separation column by means of a switching arrangement. Successful results have been reported for different tags such as 9-fluoreneacetyl, 3,5-dinitrobenzoyl, p-nitrobenzoyl and 9-fluorenylmethyl chloroformate (FMOC), among others.3,4 However, this methodology has not gained wide acceptance, probably because of the difficulties in synthesizing the reagents, and also because of their limited stability.As an alternative, we have demonstrated the possibility of performing the derivatizations in an on-line mode by combining a (pre)column, packed with a conventional support (e.g., octadecylsilica), with the separation column.The precolumn effects purification of the analytes and retains the derivatives formed when an aliquot of the reagent is flushed through it. Finally, the retained derivatives are transferred to the analytical column for their separation. The reliability of this methodology has been demonstrated for a variety of reagents such as FMOC, o-phthaldialdehyde (OPA) and 1,2-naphthoquinone- 4-sulfonate.5–7 The on-line determination of amines in biological fluids by on-column and post-column derivatization is also possible.8,9 In the latter approaches, however, the requirement for a non-responsive reagent has limited the applicability to derivatization with OPA and analogous reagents.Therefore, these modalities can only be applied to the determination of primary amphetamines. Whereas chemical derivatization allows the determination of primary and secondary amphetamines at mg mL21 levels, no satisfactory reagents are known for tertiary amphetamines and only a few procedures for the derivatization of these compounds have been reported.In such methods, derivatization typically entails dealkylation of the amine with 2-naphthyl chloroformate. 10,11 However, the conditions required to obtain satisfactory conversion yields (heating for 1 h at 100 °C) make this reaction very difficult to integrate into an on-line system.For these reasons, the determination of tertiary amphetamines by LC is usually based on the (off-line) extraction of the analytes from very large sample volumes and subsequent evaporation of the extraction solvent.2 Since the analytes are usually reconstituted in a volume much lower than that of the sample, enrichment factors in the 5–50 range can easily be Analyst, 1999, 124, 239–244 239reached. There are, however, two major drawbacks.One is the risk of losing the analytes during the evaporation step because of their volatility. The other is the limited detection capability because only a relatively small fraction of the extract is finally loaded into the LC column. Both limitations can be overcome by applying a column-switching approach. In recent years, column-switching chromatography has gained popularity in the context of analyte purification and enrichment, especially in the biomedical field.12,13 The employment of a small primary column (often called a precolumn) permits the direct injection of relatively large samples volumes. The diluted analytes are then concentrated at the head of the precolumn and then clean-up is carried out on the same column.Finally, the analytes are switched to a secondary (separation) column. In this work, we evaluated the possibility of determining primary, secondary and tertiary amphetamines in a fully automated way, without any chemical derivatization.For this purpose, a column-switching system was employed to enrich and purify the amphetamines. Conditions were optimized to obtain the maximum sensitivity in the analysis of untreated plasma and urine samples. The reliability of the method was evaluated in terms of linearity, reproducibility, accuracy and sensitivity. Experimental Apparatus The chromatographic system used consisted of two quaternary pumps (1050 Series, Hewlett-Packard, Palo Alto, CA, USA) and an automatic sample injector (Hewlett-Packard 1050 Series) with a sample loop injector of 100 mL.The UV detector (Hewlett-Packard 1100 Series) was linked to a data system (Hewlett-Packard HPLC ChemStation) for data acquisition and storage. The chromatographic signal was monitored at 210 nm. All the assays were carried out at ambient temperature. Reagents All the reagents were of analytical-reagent grade. Amphetamine sulfate, ephedrine hydrochloride, N-methylephedrine, N-methylpseudoephedrine, pseudoephedrine hydrochloride, methamphetamine hydrochloride and b-phenylethylamine hydrochloride were obtained from Sigma (St.Louis, MO, USA) and norephedrine hydrochloride, 3-phenylpropylamine and triethylamine from Aldrich (Steinheim, Germany). Acetonitrile of HPLC grade (Scharlau, Barcelona, Spain) was used for conditioning both the precolumn and the analytical column. Phosphoric acid (Probus, Badalona, Spain) and sodium dihydrogenphosphate monohydrate (Merck, Darmstadt, Germany) were also used.Preparation of solutions Stock standard solutions of the amphetamines (1000 mg mL21) were prepared in water. Working standard solutions of the amines were prepared by dilution of the stock standard solutions with water. Water was distilled, de-ionized and filtered using 0.45 mm nylon membranes (Teknokroma, Barcelona, Spain). All solutions were stored in the dark at 2°C. Columns and mobile phases For sample clean-up, a 20 3 2.1 mm id precolumn packed with 30 mm Hypersil ODS-C18 stationary phase (Merck) was used.A 5 mm, LiChrospher 100 RP18, 125 3 4 mm id column (Merck) was used as the analytical column. The precolumn and the analytical column were combined by means of a high pressure six-port valve (Hewlett Packard 1100 Series) in a straight-flush configuration. Distilled water, delivered at a flow rate of 1 mL min21 by one of the pumps, was used as washing solvent to eliminate matrix components from the precolumn.A 0.1 m phosphate buffer of pH 3 containing 2% (v/v) triethylamine was used as the eluent for the analytical column (delivered by the other pump at a flow rate of 1.3 mL min21). The mobile phase was prepared by dissolving sodium dihydrogenphosphate monohydrate in water. Next, triethylamine was added and the pH was adjusted to 3.0 with phosphoric acid. All solvents were filtered with 0.45 mm nylon membranes (Teknokroma) and degassed with helium before use. Preconcentration and purification studies The potential of the described chromatographic system for the enrichment of the amphetamines was evaluated by processing different volumes of aqueous solutions of the analytes.The volumes assayed were in the 20–350 mL range (sample volumes greater than 100 mL were injected by multiple draw). The samples were injected into the precolumn, which was then flushed with various volumes of water. After the flushing stage, the switching valve was rotated, so the precolumn was inserted into the analytical column flow scheme.At the end of each assay, the switching valve was turned to the original position, then the precolumn was re-equilibrated with 2.5 mL of water before the next sample injection. Each sample was assayed in triplicate. Analysis of urine and plasma samples Untreated urine or plasma samples were spiked with the amphetamines, producing concentrations in the 1.0–10.0 mg mL21 range. Volumes of 1 mL of these samples were placed in 2 mL injection glass vials and 250 mL were processed on-line as described above. For analyte purification, the precolumn was flushed with 1 mL of water.Each concentration was assayed in triplicate. Results and discussion Column-switching operation Chromatographic conditions. Initially, we optimized the chromatographic conditions for the separation of the amphetamines under investigation. The resolution was found to be mainly dependent on the percentage of triethylamine in the mobile phase and on the pH.The best resolution in the minimum time of analysis was obtained with a mobile phase containing 2% triethylamine at pH 3. Fig. 1(a) depicts a chromatogram obtained for an aqueous mixture of the amphetamines assayed (25 mL, 100 mg mL21 each compound), which was directly injected into the analytical column. Fig. 1(b) shows the chromatogram obtained when processing a standard aqueous solution of the amphetamines with the described columnswitching system (for an equivalent amount of drug injected).In this instance, the clean-up stage was omitted, so the precolumn was linked to the analytical column immediately after sample loading. Fig. 1 indicates that the inclusion of the precolumn in the chromatographic system lead to two drops in the chromatographic baseline. These drops are due to the insertion of water into the flow scheme of the analytical column. Negligible baseline distortions were observed at wavelengths higher than 240 Analyst, 1999, 124, 239–244230 nm.However, the UV absorbance of the amphetamines increases with decreasing wavelength. For this reason, detection wavelengths as low as 198–205 nm have been proposed for the determination of amphetamines.14 However, considerable baseline fluctuations were observed at such wavelengths with the described system. As a compromise between sensitivity and baseline stability, we selected 210 nm as the best option for further work.As can be deduced by comparing Fig. 1(a) and (b), the presence of the precolumn also introduces peak broadening. Many studies have demonstrated that band broadening can be minimized by connecting the precolumn and the analytical column in the back-flush mode.12 However, in the present study a straight-flush configuration was preferred because in such a way the analytical column is protected from contamination by matrix compounds strongly retained at the head of the precolumn.This is particularly important when large sample volumes are to be injected. Nevertheless, the resolution can be considered satisfactory provided that not all drugs would appear in the same sample. Enrichment of the amphetamines. In column-switching systems, the degree of concentration possible depends mainly on the volume of sample injected and on the stationary phase used for the precolumn. For a given precolumn packing, the sample volume also determines the number of samples that can be analyzed before replacing the precolumn.As regards the biomedical field, the most problematic fluids are those containing a large fraction of proteins, such as plasma, blood and serum. In these cases, sample volumes are typically of few hundred microlitres in order to ensure adequate stability and performance of the chromatographic system. Larger sample volumes can be processed when analyzing fluids with low protein contents, such as urine. Nevertheless, previous treatment of the samples (filtration, centrifugation or protein precipitation) may be necessary when large volumes of biological fluids are processed. Otherwise, rapid development of back-pressure occurs and the precolumn must be replaced after every few injections.On the other hand, the enrichment process also concentrates some matrix components in the precolumn. Therefore, flushing of the precolumn with a proper solvent before linking it to the analytical column is required in order to improve the selectivity and to extend the lifetime of the analytical column.The duration of the flushing stage (and hence the volume of eluent) necessary for cleaning the precolumn depends on the sample type and also on the volume of sample injected. It should be keep in mind that the washing volume should be as low as possible in order to prevent losses of the analytes by breakthrough. Although other packings (such as restricted access materials) have been claimed to tolerate larger sample volumes,15 based on previous studies we used a 30 mm Hypersil C18, stationary phase for the precolumn.5 C18 packed precolumns retain the amphetamines satisfactorily, while flushing of the precolumn simply with water provides the required selectivity in most instances.We investigated the recoveries of the amphetamines when flushing the precolumn with different volumes of water (1–5 mL). Smaller volumes were not investigated because, a significant amount of matrix components cannot be washed out, which results in poor selectivity.16 The results of this studies are shown in Fig. 2. As can be deduced from Fig. 2, effective retention of the amphetamines is achieved with the precolumn for most analytes, within the 1–2.5 mL range. Recoveries higher than 80% were observed for most analytes, the only exception being b-phenylethylamine, which was significantly less retained. Although flushing the precolumn with 5 mL of water also provided satisfactory recoveries for most analytes, unsuitable values were observed for some amphetamines such as bphenylethylamine and methamphetamine.On the basis of these results, 1 mL was selected as the optimum volume for flushing the precolumn for the simultaneous determination of the amphetamines. Next, we investigated the maximum volume of real samples that can be processed when flushing the precolumn with 1 mL of water. For this purpose, different sample volumes (50–350 mL) of untreated urine or untreated plasma were investigated.A sample volume of 250 mL was finally selected as a compromise between sensitivity, selectivity and stability of the system. In Fig. 3 are depicted the chromatograms obtained under such conditions for blank urine and blank plasma. These chromatograms demonstrate that suitable selectivity is achieved under the proposed conditions. Fig. 3 also shows the chromatograms obtained for urine and plasma spiked with a mixture of the amphetamines assayed (at a concentration of 5 mg mL21 of each compound).The recoveries obtained for urine and plasma under the selected conditions are given in Table 1. As can be seen, satisfactory recoveries were obtained for all the amphetamines. Fig. 1 Chromatograms obtained for an standard mixture of the amphetamines (a) directly injected into the analytical column and (b) injected into the described column-switching system. The amount of the analytes injected was 2.5 mg (each compound).For other experimental details, see text. Compounds: b-phenylethylamine (PEA), norephedrine (NOREPE), ephedrine (EPE), N-methylpseudoephedrine (N-MPSEPE), pseudoephedrine (PSEPE), Nmethylephedrine (N-MEPE), amphetamine (AMP), 3-phenylpropylamine (PPA) and methamphetamine (MET). Analyst, 1999, 124, 239–244 241The lowest retention was achieved for methamphetamine, with 81 and 76% of drug recovered for urine and plasma, respectively.The values obtained are similar to those obtained for aqueous solutions of the amphetamines assayed (see Fig. 2). This indicates that retention of the amphetamines in the precolumn is not significantly affected by the matrix. Linearity, reproducibility and accuracy studies The practicability of the assay was tested with urine and plasma samples spiked with amphetamines at therapeutic concentrations. 2,17 The linearity was evaluated by plotting the peak area against the concentration of amphetamine.The results are summarized in Table 2. As can be seen, the method provides good linearity over the working concentration range of 1.0–10.0 mg mL21, and no significant differences between urine and plasma were found. It should be noted that the quantification of b-phenylethylamine at the lowest concentration assayed (1.0 mg mL21) was not satisfactory, because as mentioned above, the presence of the drop in the baseline just before the peak of bphenylethylamine made the allocation of the baseline for integration very difficult.Therefore, the lowest concentration assayed for this compound was 2.0 mg mL21. The reproducibility was tested at different concentrations through the consecutive injection of five aliquots of the samples. The results obtained are also given in Table 2. The data demonstrate that the proposed procedure provides satisfactory reproducibility for all the analytes. For the highest concentrations assayed (5.0 and 10.0 mg mL21), the RSDs were < 9%.Higher RSDs were observed for the lowest concentration assayed (4–12%), but even in this case, they are clearly below the value recommended for the analysis of biological samples.2 Again, the worst reproducibilities were found for those amphetamines that eluted close to the baseline drops. The accuracy of the method was tested by calculating the relative errors obtained in the quantification of urine and plasma samples spiked with different concentrations of the analytes in the range 1.0–10.0 mg mL21.The results are given in Table 3. The method provides concentrations close to the real values, with relative errors ranging from 210 to +8% for urine and from 28 to +9% for plasma samples. In all cases tested, the concentrations calculated for b-phenylethylamine were lower than the real concentrations, which can be explained by difficulties in measuring the peak areas for this compound. Sensitivity studies In Table 4 are listed the limits of detection, for a signal-to-noise ratio of 3, obtained for the amphetamines under investigation in biological samples (no significant differences in the sensitivity were found between urine and plasma samples).As can be seen, the limit of detection most commonly found was 50 ng mL21, which was observed for four of the compounds assayed. This value is comparable to the limits of detection reported for other LC methods for primary and secondary amphetamines based on chemical derivatization and UV detection, which are typically in the 25–100 ng mL21 range; chemical derivatization and fluorescence detection usually permit the detection of ampheta- Fig. 2 Effect of the volume of water used for cleaning the precolumn after injection of the sample (standard solution of amphetamines containing 10 mg mL21 of each compound) on the analyte recoveries.For other experimental details, see text. Compounds: b-phenylethylamine (+), norephedrine (5), ephedrine (/), N-methylpseudoephedrine (8), pseudoephedrine (X), N-methylephedrine (2), amphetamine (½), 3-phenylpropylamine (:) and methamphetamine (.).Fig. 3 Chromatograms obtained for (a) blank urine and urine spiked with the amphetamines, and (b) blank plasma and plasma spiked with the amphetamines; concentration of each compound, 5 mg mL21. For other experimental details, see text. Compounds: b-phenylethylamine (PEA), norephedrine (NOREPE), ephedrine (EPE), N-methylpseudoephedrine (N-MPSEPE), pseudoephedrine (PSEPE), N-methylephedrine (N-MEPE), amphetamine (AMP), 3-phenylpropylamine (PPA) and methamphetamine (MET). 242 Analyst, 1999, 124, 239–244mines at low ng mL21 levels.2,7,18 The relatively poor sensitivity found for b-phenylethylamine, pseudoephedrine, Nmethylephedrine and methamphetamine is comparable to that reported by LC methods without chemical derivatization; it should be realized that in such methods, sample volumes of 5–10 mL are typically used.19,20 The highest limit of detection was found for methamphetamine.This is probably due to the poor peak shape obtained for this compound, as it is eluted at the longest retention time, and also due to the presence of the drop in the baseline close to it. The presence of the baseline distortion at 5.6 min would also explain the high limit of detection found for b-phenylethylamine (in principle, similar sensitivity could be expected for b-phenylethylamine and norephedrine, if only peak height and analyte recoveries are considered).Minor modification of the mobile phase composition would probably overcome the presence of system peaks near the peaks of the analytes, which can be of interest when only the determination of b-phenylethylamine or methamphetamine is required. According to the literature, the sensitivity achieved with the proposed method can be considered satisfactory for many applications concerning the determination of amphetamines at therapeutic levels2,17 and in doping control.19 However, more sensitive procedures, such as LC with chemical derivatization or GC, would be neccesary for the quantification of amphetamines at sub-ppm levels (in pharmacokinetic studies, for instance).Stability of the system The stability of the system was evaluated by analyzing consecutively 25 aliquots of plasma according to the proposed procedure. Next, the precolumn and the analytical column were flushed with acetonitrile (approximately 15 mL).The precolumn and the analytical column were then re-equilibrated with water and the phosphate buffer, respectively. Finally, another five samples were processed. As expected, the successive injection of 250 mL of untreated plasma resulted in a significant development of back-pressure in the precolumn. The pressure was approximately duplicated after 25 injections (equivalent to 4.5 mL of plasma). The backpressure observed on linking the precolumn and the analytical Table 1 Mean recoveries ± standard deviations obtained for the amphetamines in urine and plasma (n = 3).Concentration of the analytes, 10 mg mL21 Compound Recovery in urine (%) Recovery in plasma (%) b-Phenylethylamine 86 ± 3 90 ± 5 Norephedrine 95 ± 5 101 ± 2 Ephedrine 91 ± 1 92 ± 2 N-Methylpseudoephedrine 93 ± 1 95 ± 6 Pseudoephedrine 102 ± 1 98.3 ± 0.7 N-Methylephedrine 98 ± 3 95 ± 5 Amphetamine 93 ± 5 96 ± 1 3-Phenylpropylamine 103 ± 8 101± 6 Methamphetamine 81 ± 2 76 ± 4 Table 2 Reproducibility and linearity of the method for urine and plasma Urine Plasma Reproducibilityb (%) Reproducibilityb (%) Linearitya: Linearitya: Compound y = a + bx 1 mg mL21 5 mg mL21 10 mg mL21 y = a + bx 1 mg mL21 5 mg mL21 10 mg mL21 b-Phenylethylamine y = 120 + 69x — 7 3 y = 147 + 71x — 9 4 r = 0.9990 r = 0.998 Norephedrine y = 237 + 103x 11 7 5 y = 259 + 100x 12 8 3 r = 0.990 r = 0.990 Ephedrine y = 7 + 77x 9 4 1 y = 217 + 74x 9 9 4 r = 0.995 r = 0.993 N-Methylpseudoephedrine y = 227 + 80x 11 5 3 y = 217 + 72x 12 7 5 r = 0.993 r = 0.990 Pseudoephedrine y = 6 + 59x 5 7 3 y = 216 + 68x 7 7 4 r = 0.990 r = 0.994 N-Methylephedrine y = 233 + 68x 4 5 0.7 y = 235 + 62 x 12 5 1 r = 0.996 r = 0.99992 Amphetamine y = 88 + 201x 6 6 1 y = 110 + 183x 6 6 4 r = 0.9990 r = 0.997 3-Phenylpropylamine y = 7 + 101x 6 0.2 2 y = 29 + 109x 4 0.2 4 r = 0.998 r = 0.992 Methamphetamine y = 264 + 87x 11 5 2 y = 285 + 92x 10 5 8 r = 0.998 r = 0.992 a Five data points in triplicate in the 1.0–10.0 mg mL21 concentration range (2.0–10.0 mg mL21 for b-phenylethylamine). b Relative standard deviation obtained from five consecutive injections Table 3 Accuracy of the method (n = 5) Relative errora (%) Compound Urine Plasma b-Phenylethylamine 210 to 22 27.0 to 20.1 Norephedrine 24 to +6 23 to +8 Ephedrine 20.7 to +6 24 to +7 N-Methylpseudoephedrine 28 to +3 27 to +8 Pseudoephedrine 22 to +5 23 to +3 N-methylephedrine 22 to +8 24 to +9 Amphetamine 25 to +1 27 to +5 3-Phenylpropylamine 26 to +2 28 to +0.5 Methamphetamine 21 to +3 24 to +2 a Lower and upper values obtained from spiked samples containing 1.0–10.0 mg mL21 of the amphetamines (2.5–10.0 mg mL21 for bphenylethylamine).Table 4 Limits of detection for biological samples Compound Concentration/ ng mL21 Compound Concentration/ mg mL21 b-Phenylethylamine 250 N-Methylephedrine 250 Norephedrine 50 Amphetamine 100 Ephedrine 50 3-Phenylpropylamine 50 N-Methylpseudoephe- Methamphetamine 500 drine 50 Pseudoephedrine 250 Analyst, 1999, 124, 239–244 243column increased by a factor of 1.1.However, the pressure at the top of the analytical column remained constant. This indicates that the described straight-flush configuration provided effective protection of the analytical column, as possibly solid particles or matrix components retained at the head of the precolumn are not transferred to the analytical column.Conditioning of the precolumn with acetonitrile after processing 25 samples led to a significant decrease in the back-pressure in the precolumn. The same effect was also observed for the analytical column. In the latter instance, no variation in the back-pressure was observed with respect to the first injections. The retention properties of the precolumn do not depend significantly on the number of injections. This is illustrated in Table 5, which shows the recoveries obtained at different times during the study.The results obtained suggest that cleaning of the precolumn with 1 mL of water after sample injection eliminates most sample proteins. Although the precolumn can probably be used for the analysis of many more samples, as a preventive measure the packing was changed after every 50–60 injections. Satisfactory stability was also observed when processing urine instead of plasma samples. Similarly, the repetitive injection of real samples did not modify the performance of the analytical column (no significant differences in resolution were observed in the course of this investigation, in which several hundred samples were processed).Conclusions Column switching is a reliable alternative for the automated determination of amphetamines in biological fluids. Trace enrichment via a precolumn packed with a C18 stationary phase provides suitable linearity, reproducibility and selectivity. In spite of the injection of a sample volume as large as 250 mL, satisfactory stability and performance of the system are also achieved, with the only precaution of occasionally cleaning the precolumn and the analytical column with acetonitrile (for example, when the daily work is finished).Therefore, the system can be used repeatedly for several analyses. The proposed method, although less sensitive than methods using chemical derivatization, provides the required sensitivity for many applications concerning the determination of amphetamines at therapeutic levels and for rapid screening tests.The main advantage of the described procedure over methods involving chemical reactions is that it can also be applied to the determination of tertiary amphetamines. Compared with assays which do not employ chemical reactions, the described assay is much simpler and faster because no off-line manipulations of the samples are involved. References 1 P. Campíns-Falcó, A.Sevillano-Cabeza and C. Molíns-Legua, J. Liq. Chromatogr., 1994, 17, 731. 2 T. Kraemer and H. H. Maurer, J. Chromatogr., 1998, 713, 163. 3 I. S. Krull, Z. Deyl and H. Lingeman, J. Chromatogr., 1994, 659, 1. 4 I. S. Krull, M. E. Szulc, A. J. Bourque, F.-X. Zhou, J. Yu and R. Strong, J. Chromatogr., 1994, 659, 19. 5 R. Herráez-Hernández, P. Campíns-Falcó and A. Sevillano-Cabeza, Anal. Chem., 1996, 68, 734. 6 R. Herráez-Hernández, P. Campíns-Falcó and A. Sevillano-Cabeza, J. Chromatogr., 1996, 679, 69. 7 M. D. Pastor-Navarro, R. Porras-Serrano, R. Herráez-Hernández and P. Campíns-Falcó, Analyst, 1998, 123, 319. 8 R. Herráez-Hernández, P. Campíns-Falcó and S. Díaz-Oltra, Chromatographia, in the press. 9 R. Herráez-Hernández and P. Campíns-Falcó, Anal. Chim. Acta, submitted for publication. 10 G. Gübitz, R. Wintersteiger and A. Hartinger, J. Chromatogr., 1981, 218, 51. 11 T. D. Doyle, W. M. Adams, S. F. Fry and I. W. Wainer, J. Liq. Chromatogr., 1986, 9, 455. 12 P. Campíns-Falcó, R. Herráez-Hernández and A. Sevillano-Cabeza, J. Chromatogr., 1993, 619, 177. 13 U. A. Th. Brinkman, J. Chromatogr., 1994, 665, 217. 14 H. J. Helmlin and R. Brenneisen, J. Chromatogr., 1992, 593, 87. 15 Z. Yu and D. Wasterlund, J. Chromatogr., 1996, 725, 137. 16 P. Campíns-Falcó, R. Herráez-Hernández and A. Sevillano-Cabeza, Anal. Chem., 1994, 66, 244. 17 A. H. Beckket and M. Rowland, Nature (London), 1964, 204, 1023. 18 Y. Ohkura, M. Kai and H. Nohta, H., J. Chromatogr., 1994, 659, 85. 19 C. Imaz, D. Carreras, C. Rodríguez, A. F. Rodríguez, J. Maynar and R. Cortés, J. Chromatogr., 1993, 631, 201. 20 M. Katagi, H. Nishioka, K. Nakajima, H. Tsuchihashi, H. Fujima, H. Wada, K. Nakamura and K. Makino, J. Chromatogr., 1996, 676, 35. Paper 8/09825E Table 5 Influence of multiple plasma injections on recoveries. Concentration of the amphetamines, 10 mg mL21 Recoverya (%) Compound Injections 1–3 Injections 18–20 Injections 26, 27 b-Phenylethylamine 91 ± 4 88 ± 2 93 ± 2 Norephedrine 92 ± 3 99 ± 3 86.5 ± 0.7 Ephedrine 87 ± 6 81 ± 2 81 ± 1 N-Methylpseudoephedrine 82 ± 4 78 ± 5 84 ± 1 Pseudoephedrine 87 ± 6 89.0 ± 0.6 84 ± 1 N-Methylephedrine 92.1 ± 0.7 97 ± 3 86 Amphetamine 95.0 ± 1.0 89 ± 2 89 ± 4 3-Phenylpropylamine 90 ± 4 86 ± 4 83 Methamphetamine 80.1 ± 0.7 76 ± 2 79 ± 2 a Mean value ± standard deviation. 244 Analyst, 1999, 124, 239–244

ISSN:0003-2654    

DOI:10.1039/a809825e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of copper and zinc in blood plasma by ion chromatography using a cobalt internal standard Edmund Lane,a Alexis J. Holden*a and Robert A. Cowardb a Centre for Toxicology, University of Central Lancashire, Preston, UK PR1 2HE. E-mail: a.j.holden@uclan.ac.uk b Dialysis Unit, Royal Preston Hospital, Sharoe Green Lane North, Fulwood, Preston, UK PR2 9HT Received 12th November 1998, Accepted 13th January 1999 Ion chromatography was used to detect levels of copper and zinc in blood plasma from renal dialysis patients on continuous ambulatory peritoneal dialysis (CAPD) and haemodialysis (HD).The developed method used cobalt as an internal standard and when combined with the standard additions method improved the overall precision of the results by between 20.3 and 6.0% and 20.8 and 5.7% for copper and zinc, respectively. The method was compared with inductively coupled plasma optical emission spectrometry (ICP-OES) and the results indicated no significant difference between the two methods with or without an internal standard.Without an internal standard, tcalc was 0.869 with a tcrit of 2.201 (n = 12, P = 0.05) and with an internal standard, tcalc was 0.189 compared to a tcrit of 2.201 (n = 12, P = 0.05). The copper and zinc levels in blood plasma in both dialysis groups were not significantly different to the copper and zinc levels in blood plasma of the control patients. A convenient method of analysis of trace elements in blood such as ion chromatography with UV/VIS detection is useful for determining whether inorganic elements may be disrupted in the body due to changes in the state of health.Introduction The trace elements copper and zinc serve chiefly as key components of proteins and of enzyme systems in the human body and are vital for their proper functions.1 Zinc is involved in the synthesis of nucleic acids and proteins and is necessary for growth.1 The majority of these enzymes act as antioxidants in biological systems [e.g., Cu, Zn in superoxide dismutase (SOD) and Cu in caeruloplasmin].Deficiencies or excess of the trace elements may have a detrimental effect on the function of these enzymes.2,3 Levels of many trace elements are disturbed in various diseases, e.g., abnormal levels of copper and zinc have been found in patients on renal dialysis,4–8 lowered copper levels are associated with Wilson’s disease and hypocupremia is observed with leukaemia and with a number of acute diseases.The detection of trace elements in blood is an important factor for determining any disruption in trace element status of the body. The method used to determine elemental levels needs to be sensitive enough to detect the amounts present in blood (levels of Cu and Zn ~1 mg dm23) in an all too often limited sample volume.3 Analysis of trace elements in biological samples requires: precise and accurate separation of the analyte from the bulk matrix for some measurements, a suitable means of detection which is of the appropriate sensitivity and to be free of contamination.Careful consideration of factors such as diet and disease state are critical in the assessment of the results before any conclusions can be made about cause–effect relationships for the disruption of metal status.3 Atomic absorption spectrometry (AAS) is widely used for the detection of many elements. Techniques such as flame atomic absorption spectrometry (FAAS) and electrothermal atomic absorption spectrometry (ETAAS) are suitable for the detection of Zn and Cu in blood serum and plasma.With the development of such routine methods often samples only require dilution.9 Detection limits for flame AAS have been reported to be as low as 9.4 mg dm23 for zinc 10 and 0.2 mg dm23 for copper when preconcentration techniques have been employed.11 Detection limits for ETAAS are typically 0.4 mg dm23 and 0.1 mg dm23 for Cu12 and Zn,13 respectively.However, for most of the AAS techniques, only one element can be determined at a time in many routine laboratories, although the relatively new Perkin- Elmer SIMAA 6000 method of simultaneous multi-element ETAAS is able to detect up to six elements with the same detection limits as conventional ETAAS analysis due to the very low noise of the solid-state detector.14 Inductively coupled plasma optical emission spectrometry (ICP-OES) and mass spectrometry (ICP-MS) can be used in the direct analysis of biological fluids.The advantages include simultaneous detection of several elements in the one sample and its ability to overcome matrix interferences.15 The sample is efficiently desolvated, vaporised, excited and ionised. Many chemical interferences are greatly reduced by the high temperature of 7000–8000 °C. Typical levels of detection range from < 0.5 to 100 mg dm23 for most metals. One of the major advantages of ICP-MS over ICP-OES is a three orders of magnitude higher sensitivity for metal detection although there may be problems due to the matrix.15 Ion chromatography has been increasingly used for the determination of metals and is becoming an alternative to conventional spectrometric methods.16,17 The main advantage is that multiple elements can be simultaneously determined in one sample using similar sample pre-treatment methods.Ion chromatography can separate a mixture of ions by their net charge according to the principles of ion exchange.Several detection methods are available including conductance and detection by UV by the use of a post column derivatisation reagent. Ion chromatography is suitable for small sample volumes, with injection volumes of 25 ml per analysis. Detection limits of less than 13 mg dm23 have been reported by a number of workers.18,19 The current study involved the development of an ion chromatography (IC) method which would enable the determination of copper and zinc in the same aliquot of a sample. The final method made use of a cobalt internal standard which was used to minimise errors due to sample loading and fluctuations, Analyst, 1999, 124, 245–249 245which occurred during the running of the column.The peak area results for the copper and zinc peaks were expressed as a ratio of the peak area result from the cobalt peak. Hence, if a varied volume of sample was injected or there was a variation in the flow rate of the mobile phase or the detector did not respond in a uniform manner throughout all the chromatography runs, then the results could still be used.The cobalt peak would be affected in the same manner as the copper and zinc peaks and so the ratio of the peak areas would remain the same although the raw peak area data values would change according to the inconsistencies within the instrument. Hence the use of an internal standard in analyses allows the use of data which may otherwise be deemed worthless. The method of standard additions was also used.To test the developed method a number of blood plasma samples from patients on haemodialysis (HD) and continuous ambulatory peritoneal dialysis (CAPD) as well as control samples were analysed. Materials and methods Specimens Blood samples from eight renal dialysis patients, four on HD and four on CAPD, plus four control samples were obtained from the Royal Preston Hospital.The blood samples were collected from patients on renal dialysis in S-Monovette LHMetall- Analytik tubes (Sarstedt Ltd., Leicester, UK) which contained lithium–heparin as an anticoagulant. These tubes are recommended by the suppliers for the collection of blood plasma which is to be analysed for analytes such as copper, zinc, lead, manganese, cadmium, iron, aluminium, nickel, selenium, chromium and mercury. The plasma was removed from the blood sample by centrifugation and the samples stored in the refrigerator at 4 °C before use.The time between sample collection and centrifugation was between 1–2 h because of the staggered collection time on the hospital ward. The dialysis subjects ranged in age (35–60 years) and in the number of years that they had received dialysis (2–25 years) (Table 1). The control samples were age matched where possible. Reagents All chemicals were of analytical-reagent grade and where possible all analyses were carried out in plasticware to reduce the risk of contamination during storage, transport and analysis. All containers were acid washed by soaking in 10% nitric acid (BDH-Merck, Poole, Dorset, UK) overnight and then in deionised water (Barnstead E-Pure, Fisons, Loughborough, Leicestershire, UK, 17.6 mW cm21) overnight.Stock standard solutions containing 1000 mg dm23 of copper, zinc and cobalt were diluted to 5 mg dm23 working solutions. All these solutions were prepared daily from copper, zinc and cobalt atomic absorption spectrometric standards (Fisher, Loughborough, Leicestershire, UK).Instrumentation A 1 cm3 sample of the blood plasma sample was digested with 70% nitric acid in a ratio of 1:3 in pressure sealed Teflon bombs (150 cm3 CEM, Buckingham, UK). An MDS 8D microwave (CEM) was used at 70% power for a total of 60 min in 3 3 20 min periods allowing the samples to cool for 10 min between each 20 min period. The digested material was diluted to 10 cm3 with deionised water.For the standard additions method, four 1 cm3 aliquots of the plasma digest were spiked with increasing amounts of copper (0, 0.2, 0.4, 0.8 cm3 from the working solution) and zinc (0, 0.2, 0.4, 0.8 cm3 from the working solution). A 0.6 mg dm23 cobalt internal standard (0.6 cm3 from the working solution) was also added to each sample. The solutions were made up to 5 cm3 with deionised water. Measurements were performed in triplicate on a Dionex DX500 ion chromatography system.A 25 cm HPIC-CS5 cation exchange column (Dionex, Camberley, Surrey, UK) with a 5 cm HPIC-CG5 guard column (Dionex) and a 25 ml sample loop was used. The column packing consisted of a cross-linked styrene and vinylbenzene polymer with sulfonic functional groups. The eluent was 50 mM oxalic acid and 95 mM lithium hydroxide at pH 4.8 (BDH-Merck) at a flow rate of 1 cm3 min21. The postcolumn reagent used was 0.3 mM 4-(2-pyridylazo) resorcinol (PAR) (Dionex) in 1 M acetic acid and 3 M ammonium hydroxide solution at a flow rate of 0.7 cm3 min21.The detection was at 520 nm using a Dionex A20 absorbance detector (Dionex). An inductively coupled plasma optical emission spectrometer (ICP-OES) (Spectro, Halesowen, Worcestershire, UK, Analytical Model P) was used to detect copper and zinc. Copper was detected at 324.8 nm and zinc at 213.9 nm. The sample was introduced into the plasma at a flow rate of 1 cm3 min21 through silicone tubing via a peristaltic pump. Five replicate measurements were made and approximately 0.5 cm3 of sample was required per sample.The system was flushed with deionised water between individual solutions. Results and discussion Copper, zinc and cobalt were easily separated by ion chromatography with elution times of 2.9, 4.8 and 6.6 min, respectively, as shown in Fig 1. With the oxalate eluent used it is reported that only cations lead, copper, zinc, manganese, cobalt, cadmium and nickel are detected. It was hypothesised that either iron(ii) or iron(iii) may interfere with the analysis of zinc or copper but preliminary experiments suggested that this was not the case.The cobalt internal standard was subjected to the same conditions as the sample, hence any fluctuations in the running conditions would be eliminated by plotting the absorbance ratio between copper and cobalt against the concentration of spiked copper. Similarly for zinc, the absorbance ratio between zinc Table 1 Details of the patient sample used in each dialysis and control group Dialysis Years on Years on type Male dialysis Female dialysisysis CAPD 38 1 50 2 73 10 52 11 HD 54 1 33 3 70 4 52 1 Control 34 N/A 69 N/A 50 N/A 61 N/A Fig. 1 Chromatograms of spiked blood plasma. Copper, cobalt and zinc eluted at 2.9, 4.8 and 6.6 min, respectively. Each of the three samples was spiked with cobalt (0.6 mg dm23) and with increasing concentrations of copper and zinc (0, 0.4, 0.8 mg dm23). 246 Analyst, 1999, 124, 245–249and cobalt against the concentration of spiked zinc was plotted. By using cobalt as an internal standard it is suggested that the ratio of the analyte to the internal standard provides a more accurate determination than the use of the analyte response alone. The precision was further improved by combining the internal standard method with the method of standard additions. This produced an overall improvement in the correlation coefficient of between 20.3 and 6% for copper and between 20.8 and 5.7% for zinc (Fig. 2 and Table 2). For this method to be reliable, the chosen internal standard must be well separated from the components but must appear close to the peaks of interest. Cobalt can be used as an internal standard as it is found in serum at very low levels (< 0.05 mg dm23),1 well below the detection limit of IC with VIS detection and ICP-OES. The detection limit was calculated as three times the signal to noise ratio.The signal to noise ratio was calculated by measuring the concentration of the baseline noise in comparison with a peak of a 0.5 mg dm23 sample of the analyte. The detection limits were calculated as 0.09 mg dm23 for Cu and 0.03 mg dm23 for Zn. The results obtained from the ion chromatography analysis were validated by comparing the results with those obtained from ICP-OES. The t-test (at 95% confidence limit) revealed no significant differences between the values obtained by the two methods.There was reasonable agreement between both methods, although more samples would indicate if a better correlation was possible. The correlation coefficient for copper was 0.88 with an equation of y = 1.04 x 2 0.019 (n = 12) (Fig. 3) and the correlation coefficient for zinc was 0.89 with an equation of y = 1.08 x 2 0.022 (n = 12). A Bland–Altman plot is where the difference between the results from the two methods is expressed as a percentage of the mean of the two results for each sample, plotted against the mean concentration of the two measurements.If there is no systematic difference in the results from the two techniques then the data points would oscillate about the zero line on the y-axis. The Bland–Altman Fig. 2 Comparison of calibration and standard addition graphs of plasma from one haemodialysis patient (* in Table 2) for Cu [(a) and (b)] and Zn [(c) and (d)] with [(b) and (d)] and without [(a) and (c)] the use of an internal standard.Table 2 Comparison of the correlation coefficient for each patient group when samples were analysed with and without an internal standard (IS). Results marked (*) correspond to data presented in Fig. 2 Copper Zinc Dialysis Improve- Improvetype No IS With IS ment (%) No IS With IS ment (%) CAPD 0.917 0.976 6.05 0.951 0.943 20.83 0.922 0.954 3.35 0.907 0.951 4.64 0.944 0.991 4.76 0.950 0.977 2.70 0.954 0.991 3.75 0.921 0.976 5.64 HD 0.923 0.943 2.14 0.945 0.954 0.93 0.943 0.951 0.82 0.928 0.932 0.43 0.951 0.977 2.61 0.897 0.942 4.78 * 0.925 0.976 5.23 0.922 0.951 3.05 Control 0.931 0.954 2.41 0.944 0.977 3.33 0.921 0.932 1.18 0.954 0.976 2.25 0.945 0.942 20.33 0.932 0.954 2.31 0.914 0.951 3.89 0.934 0.991 5.77 Average 2.98 ± 1.89 2.92 ± 2.07 Analyst, 1999, 124, 245–249 247plot is used as Pollock et al. 20 suggested it to be more sensitive to potential bias within techniques than the standard regression techniques which can be used.In Fig. 4 if the difference between the results is 10%, this represents that one of the results is 1.1 times the value of the other, if the difference is 66%, then one result is twice the other and if the difference is 100%, one result is three times the other. Using the results from this work the Bland–Altman plot (Fig. 4) demonstrates that there is an overall positive bias for both copper and zinc of 0.006 mg dm23 (0.35%) and 0.128 mg dm23 (7.07%), respectively.This suggests that the results for zinc obtained from the ICP-OES method are higher than those obtained with IC-VIS analysis. This could result from matrix effects in either the IC-VIS or ICP-OES methods. Teixeira et al. 21 reported that the use of ICP-OES for the determination of zinc at low concentrations in matrices containing high concentrations of copper was difficult because of the interference by copper in the main emission wavelength of zinc at 213.856 nm. There were no significant differences in both the copper and zinc levels between CAPD patients and HD patients and there was no significant difference when dialysis groups were compared to the control patients (Tables 3 and 4).It must be stressed that these individual values represent the levels of copper and zinc in different patients and thus the concentrations of copper and zinc may vary from patient to patient due to factors such as sex, age, and time on dialysis. The normal range for copper in serum is 0.8–1.5 mg dm23 and for zinc it is 0.6 –1.3 mg dm23.Overall, the average control copper levels were lower than the normal levels expected in the blood. The measured copper concentration for this study was 0.501 ± 0.05 mg dm23 for copper and was 1.350 ± 0.05 mg dm23 for zinc in the control samples. This paper is concerned with the development of an alternative analytical method to conventional spectrometric methods for the determination of copper and zinc in biological samples and as such the levels of copper and zinc reported in this study should not be used exclusively as the measured concentration of the zinc in the samples was probably affected by the lapse time between collection and centrifugation as suggested by English and Hambridge.22 They reported that Fig. 3 The correlation between samples run on both ICP-OES and IC-VIS: (a) the correlation coefficient for copper values was r = 0.88 and the equation of the line is y = 1.04 x 2 0.019 (n = 12); (b) the correlation coefficient for zinc was r = 0.89 with an equation of the line y = 1.08 x 2 0.022 (n = 12).Table 3 Concentrations of copper and zinc in acid digested blood plasma as detected by ion chromatography with VIS detection using the standard additions method with a cobalt internal standard and with ICP-OES Copper/mg dm23 Zinc/mg dm23 Dialysis Method type (n = 4) Average Range Average Range IC-VIS Control 0.501 ± 0.05 0.45–0.67 1.350 ± 0.05 0.73–1.83 CAPD 0.549 ± 0.05 0.33–0.68 1.576 ± 0.06 0.89–1.95 HD 0.594 ± 0.05 0.46–0.78 1.348 ± 0.06 0.98–1.79 ICP-OES Control 0.534 ± 0.03 0.31–0.62 1.417 ± 0.02 0.79–1.95 CAPD 0.597 ± 0.02 0.51–0.70 1.789 ± 0.03 0.91–2.28 HD 0.575 ± 0.02 0.43–0.73 1.481 ± 0.02 0.93–2.02 Table 4 One tailed t-test for unequal variances was used to compare between the means of the dialysis group with the control groups and between the two methods (P < 0.05) Copper Zinc Comparison n tcalc tcrit tcalc tcrit IC-VIS (with IS) vs.ICP-OES 12 0.189 2.201 0.265 2.201 IC-VIS (no IS) vs. ICP-OES 12 0.869 2.201 1.170 2.201 CAPD vs. control 4 0.448 3.182 0.747 3.182 HD vs. control 4 0.938 3.182 0.007 3.182 CAPD vs. HD 4 0.427 3.182 0.747 3.182 Fig. 4 The Bland–Altman plots for copper and zinc which compares the two methods IC-VIS and ICP-OES. The plots show a small positive bias indicating that the results obtained with ICP-OES are higher than those obtained using IC-VIS. 248 Analyst, 1999, 124, 245–249the level of Zn determined in plasma and serum increased by 6% for the first 2 h after collection which was thought to be attributed to the release of some erythrocyte membrane zinc.A similar effect was not found for copper. Many factors are involved in determining blood and tissue levels of individual elements in patients with renal failure. The processes involved in renal disease itself can result in either excretion or retention of individual elements whilst dialysis treatment can cause either removal or exposure to these elements.Trace metal disruption causes chronic renal insufficiency in dialysis patients that have been studied but many of these studies are contradictory due to different techniques that have been used. For instance, copper levels have shown conflicting results with high, low and normal levels with some evidence that the levels increase with age and the female sex but not with dialysis duration.Hypozincaemia is also common in patients with end stage renal disease and those on dialysis (haemodialysis and CAPD). In haemodialysis the blood is pumped at 200–400 cm3 min21 into an artificial kidney where it is filtered by a semi-permeable dialysis membrane against the dialysate solution at a flow of 500 cm3 min21. Transfer occurs across the membrane of waste products into the dialysate which is drained away. The dialysate solution is made from combining a concentrate with a filtered and reverse osmosis treated tap water.In CAPD the dialysate (typically 2 dm3) is introduced into the abdominal cavity via a catheter and remains there for 4–6 hours. During this time, excess fluid and waste products are transferred into the dialysate solution from the blood across the peritoneal membrane and subsequently drained out to be replaced by fluid dialysate. Dialysis fluid (CAPD, haemodialysis concentrate and reverse osmosis treated tap water) has a role in contributing to trace element abnormalities in renal patients.Metal contaminants in dialysis fluid such as copper and zinc should be eliminated by the manufacturing process. Other sources of contamination such as dialysate tubing could be a possible source of metal contamination.23 Further factors may be important in determining blood and tissue levels, in particular the ongoing ageing process with poor dietary intake and decreased gastrointestinal absorption of zinc, whilst copper absorption in the small intestine is facilitated by zinc deficiency owing to the loss of absorption competition.Ion chromatography is an attractive alternative to the usually more conventional spectrometric methods for the determination of metals. Ion chromatography offers a more cost effective choice as the running expenses of ion chromatography often are cheaper than those faced by ICP-OES users. The most obvious advantage of this technique is that multiple elements can be determined in one sample of 25 ml volume and complete analysis can be performed when coupled with a suitable detection system or systems. Generally, the sample requires minimal sample pre-treatment.16 Although it is accepted that microwave digestion seems laborious when compared to the simple dilution methods used with some atomic spectrometric methods, the use of acid digestion to totally breakdown the matrix of a sample in order to ensure all the analyte is recovered offers its own advantages.The selectivity and peak sharpness in IC can be enhanced by the use of complexing agents such as PAR for the detection of cations followed by spectrophotometric detection giving a highly sensitive method. The use of the standard additions method and of an internal standard can improve the precision of the results. Acknowledgements Thanks to Dr. Philip H. E. Gardiner and Heather Birtwistle at Sheffield Hallam University for the use of the ICP-OES system.This work was supported by a studentship from the Lancashire Centre for Medical Studies. References 1 W. Kaim and B. Chwederski, Bioinorganic Chemistry: Inorganic Elements in the Chemistry of Life, An Introduction and Guide, Wiley, New York, 1994, pp. 151–161. 2 J. M. Campistol and A. Argiles, Nephrol. Dial. Transplant, 1996, 11, 142. 3 M. Gallieni, D. Brancaccio, M. Cozzolio and E. Sabioni, Nephrol. Dial. Transplant, 1996, 11, 1232. 4 N. J. Emenaker, R. A. Disilvestro and N. S. Nahman, Am. J. Clin. Nutr., 1996, 64, 757. 5 R. J. Cousins, Physiol. Rev, 1985, 65, 309. 6 P. L. Kimmel, T. M. Philips and S. Q. Lew, Kidney Int. 1996, 49, 1412. 7 L. Yver and D. Blanchier, Nephrol. Dial. Transplant, 1987, 2, 451. 8 C. Canavesse, Nephron, 1990, 56, 455. 9 K. Nomiyama and H. Nomiyama, Nephrol. Dial. Transplant, 1989, 4, (suppl) 114. 10 K. Cundeva and T. Stafilov, Talanta, 1997, 44(3), 451. 11 H. W. Chen, S. K. Xu and Z. L. Fang, Anal. Chim. Acta, 1994, 298(2), 167. 12 P. B. Barrera, R. D. Gonzalez and A. B. Barrera, Fresenius’ J. Anal. Chem., 1997, 357(4), 457. 13 S. D. Huang and K. Y. Shih, Spectrochim. Acta, Part B, 1995, 50(8), 837. 14 M. Hoenig and A. Cilissen, Spectrochim. Acta, Part B, 1997, 52, 1443. 15 J. Schoppenthau and L. Dunemann, Fresenius’ J. Anal. Chem., 1994, 349, 794. 16 C. N. Ong, H. Y. Ong and L. H. Chua, Anal. Biochem., 1988, 173, 64. 17 M. En-ling and J. Zhu-ming, Chin. Med. J., 1993, 106(2), 118. 18 H. T. Lu, S. F. Mou, Y. Yan, S. Y. Tong and J. M. Riviello, J. Chromatogr. A, 1998, 800(2), 247. 19 S. Okawa and T. Ishikawa, Bunseki Kagaku, 1998, 47(1), 9. 20 M. A. Pollock, S. G. Jefferson, J. W. Kane, K. Lomax, G. MacKinnon and C. B. Winnard, Ann. Clin. Biochem., 1992, 29, 556. 21 L. S. G. Teixeira, J. O. N. Reis, A. C. S. Costa, S. L. C. Ferreira, M. D. S. G. Korn and J. B. deAndrade, Talanta, 1998, 46(6), 1279. 22 J. L. English and K. M. Hambridge, Clin. Chim. Acta, 1988, 175, 211. 23 R. Milacic, M. Benedik and S. Knezevic, Clin. Chim. Acta, 1997, 265, 169. Paper 8/08852G Analyst, 1999, 124, 245–249 249

ISSN:0003-2654    

DOI:10.1039/a808852g

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Investigations by HPLC-electrospray mass spectrometry and NMR spectroscopy into the isomerisation of salinomycin Adrienne L. Davis, James A. Harris,* Charlotte A. L. Russell and John P. G. Wilkins Unilever Research, Colworth House, Sharnbrook, Bedfordshire, UK MK44 1LQ Received 8th December 1998, Accepted 14th January 1999 HPLC-MS studies have indicated that certain polyether ionophore veterinary drugs are prone to degradation when stored as water–methanol solutions at ambient temperature.Salinomycin and narasin were particularly susceptible, disappearing completely within weeks to produce more polar species, which were identified as isomers of the original compounds. Lasalocid appeared to be stable under such conditions. Structural elucidation of the principal ultimate salinomycin isomerisation product was achieved by 2D NMR spectroscopy. This indicated that the isomerisation process consists of the opening of the spiro rings in the salinomycin structure with the concomitant formation of a furan moiety.The MS data indicated that the isomers retain the ability to complex alkali metal ions and may therefore retain their pharmacological activity. These discoveries may have implications both for the development of legislation covering acceptable levels of polyether ionophore residues in foodstuffs and also for analytical protocols designed to detect them. Introduction The polyether ionophores are used intensively throughout the world in poultry production.They are coccidiostats, and may be incorporated in feedstuffs. They include lasalocid, monensin, narasin and salinomycin. Currently, there are no European Union (EU) maximum residue limits (MRLs) controlling their residues in food. However, their status is under review because of concerns over their toxicity and potential drug resistance problems. We have been developing mass spectrometric methods for the determination of residues of these compounds in foodstuffs. 1 During the course of this work, we have observed that certain polyether ionophores isomerise in aqueous solution (such solutions are used, for example, during the preparation of spiked materials for recovery determination).The current work reports an investigation of this process. The rates of isomerisation of lasalocid, monensin, salinomycin and narasin were studied by HPLC-MS, and the structure of the salinomycin isomer was determined by 2D NMR spectroscopy. Experimental aim To investigate the isomerisation of polyether ionophores in aqueous solutions and to determine the structure of the principal ultimate isomer of salinomycin.Experimental Materials Lasalocid (97% purity), monensin (90–95%), salinomycin (96%) and narasin (97%), all as sodium salts, were obtained from the Sigma Chemical Corporation (Poole, Dorset). Individual 1 mg ml21 stock solutions were prepared in methanol in glass volumetric flasks and stored at 4 °C for up to 3 months.Dilutions of these stock solutions were prepared (see below). Solvents and reagents were of high purity. Deionised water (pH 6.3) was obtained from a Milli-Q Plus water purification system (Millipore, Watford, Hertfordshire). HPLC-electrospray MS HPLC was performed using a Hewlett Packard (Bracknell, Berkshire) HP1050 system. The mobile phase was 0.02 M aqueous ammonium acetate (adjusted to pH 7.0 by the addition of 3–4 drops of 5% aqueous ammonia solution)–acetonitrile– methanol (20 + 60 + 20 v/v/v).Solvent delivery was at 1 ml min21. Automated injections of 50 ml were made. The HPLC column was a Phenomenex (Macclesfield, Cheshire) Nucleosil 15 cm 3 4.6 mm C18 5 mm column. Approximately 0.15 ml min21 of the eluent flow was delivered to the Micromass (Altrincham, Cheshire) Quattro I triple quadrupole MS (the rest being diverted to waste via a pneumatic splitter). The MS was equipped with a megaflow electrospray source operated in positive ion mode.The source temperature was 150 °C, and the nitrogen bath and sheath gas flows were 300 and 40 l h21 respectively. The cone voltage (CV) potential settings were 10 to 100 V. Various acquisition regimes were used; scanning (typically m/z 200–800) and/or selected ion recording (SIR) to monitor characteristic sodiated and ammoniated molecular ions plus diagnostic fragment ions1 (mass window, 1.0 Da, dwell time, 0.1 s; interscan delay, 0.02 s). Data were acquired for 25 min.MS analysis of salinomycin and its ultimate isomerisation product A freshly diluted aqueous 10 mg ml21 salinomycin solution, a similar solution that had been prepared 2 months earlier and stored at ambient temperature (20–25 °C) and a mixed solution made from equal volumes of these two solutions were analysed by HPLC electrospray MS (acquiring data from m/z 100–1000, with CV of 20, 30, 50 and 70 V). Analyst, 1999, 124, 251–256 251MS analysis of time course experiment Duplicate dilutions of the stock solutions of the four polyether ionophores were prepared in water, to give 20 mg ml21 solutions (i.e., 2% methanol).One of the duplicate solutions was stored at ambient temperature (20–25 °C) and the other at 4 °C. The solutions were analysed by HPLC-MS using SIR, within hours of preparation (day 0) and after 2, 4, 7 and 11 days. Quantification was performed versus standard solutions in 0.02 M aqueous ammonium acetate solution (pH 7.0)–acetonitrile (80 + 20 v/v), stored at 4 °C.NMR spectroscopy To elucidate the structure of the isomer of salinomycin (compared to salinomycin), NMR spectra were measured on a Bruker AMX400 spectrometer (Rheinstetten/Karlsruhe, Germany) operating at a probe temperature of 294 K or 300 K using either a dual 1H/13C 5 mm probe or a multinuclear 5 mm inverse probe as appropriate. The solvent used was D2O-exchanged acetonitrile-d3 and spectra were referenced relative to internal tetramethylsilane (TMS).Aqueous dilution of the stock solution was performed to give 50 ml of 40 mg ml21 salinomycin solution. This was stored at ambient temperature for 17 days. A similar aqueous dilution was then freshly prepared. The two solutions were analysed by HPLC-MS, then both solutions were evaporated to dryness using a freeze dryer. Residues were redissolved in 0.5 ml acetonitrile-d3. A 50 ml aliquot was diluted to 5 ml with acetonitrile and analysed by HPLC-MS.The remaining extracts (approximate concentration, 4 mg ml21) were then analysed by NMR. 2D TOCSY experiments2 were acquired using a multiple of the MLEV-17 sequence for spin locking flanked by two 2.5 ms trim pulses. The spin-lock field strength was 10 kHz, the mixing time was 10 ms and a relaxation delay of 1 s was employed. The HMQC experiment3 was acquired with a 400 ms delay after the BIRD pulse, a dephasing/refocusing delay of 3.5 ms and a relaxation delay of 0.5 s between scans.GARP decoupling was applied during acquisition. The 2D HMBC experiment3 was recorded with a 3.5 ms delay for the low-pass J filter, a 60 ms delay for evolution of long-range couplings and a relaxation delay of 0.5 s. All 2D experiments were recorded phase sensitive in both dimensions (using TPPI for quadrature detection,4 with the exception of the HMBC experiment which was recorded in mixed mode format.5 Results and discussion It should be noted that, when referring to pseudomolecular ions, etc., M is based on the polyether ionophore free acid (not the sodium salt).Thus the sodiated pseudomolecular ion of salinomycin (free acid C42H70O11; nominal monoisotopic molecular weight, 750), described as [M + Na]+, represents C42H70O11Na+, observed at m/z 773. HPLC–electrospray MS was used to characterise the freshly prepared, the 2 month old and the mixed fresh and aged 10 mg ml21 largely aqueous (water–methanol, 98 + 2 v/v) salinomycin solutions.(It was not convenient to prepare purely aqueous solutions because of solubility.) The main response from the fresh solution, observed at 15.3 min, was ascribed to unchanged salinomycin. (Some other minor responses were also observed. These have been identified as technical contaminants. 1 That from the aged solution, at ca. 2.8 min, exhibited a rather similar mass spectrum and was tentatively identified as an isomer of salinomycin (there was negligible response at the retention time of intact salinomycin).The electrospray, positive ion mass spectra of the unchanged and the isomerised salinomycin, obtained with a cone voltage potential of 20 V, are both dominated by ammoniated [M + NH4]+ and sodiated [M + Na]+ molecular ions at m/z 768 and 773 (both with ca. 100% relative abundance). Additional responses in the spectrum of the isomer were observed at m/z 733, with ca. 70% relative abundance, probably due to [M + H 2 H2O]+, and m/z 809, with ca. 30% relative abundance, possibly due to the solvent adduct ion [M + CH3CN + NH4]+. The shorter retention time indicates that the apparent isomer is more polar than its parent. The overall MS response of the isomer peak was approximately 30% of that of the unchanged salinomycin peak. HPLC-electrospray MS analysis of the 10 mg ml21 water– methanol (98 + 2 v/v) solutions of lasalocid, monensin, narasin and salinomycin, stored either at 20–25 °C or at 4 °C, over a period of 11 days, was undertaken to investigate the rates of conversion of the polyether ionophores (Fig. 1). The responses for salinomycin and narasin in the water– methanol solutions stored at 20–25 °C had disappeared by day 11, and that for monensin had dropped by > 30%. The reduction in the lasalocid response from the water–methanol solution was less (ca. 10%). For the water–methanol solution of salinomycin stored at 20–25 °C for 2 days, three LC peaks in addition to that of salinomycin were observed.Their retention times (2.8, 4.3 and 5.3 min, with the 2.8 and 5.3 min peaks being most abundant) were shorter than that of salinomycin (15.3 min). The additional peaks all exhibited ions at m/z 768 and 773, indicating that they were isomers of salinomycin. By day 11, the shortest retention time peak was the only significant peak observed. For the water–methanol solution of narasin stored at 20–25 °C for 2 days, additional LC peaks were observed at 3.1, 4.7, 5.2, 6.5 and 7.4 min, with the 3.1 and the 6.5 min peaks being the most abundant (narasin eluted at 17.1 min). They all exhibited the m/z 782 and 787 pseudomolecular ions of narasin.Again, by day 11, the shortest retention time peak was the only significant peak observed. For the water–methanol solution of monensin stored at 20–25 °C, the rate of change was slower. By day 11, three minor additional LC peaks were observed at 2.5, 4.7 and 8.7 min, the major one being at 2.5 min (monensin eluted at 10.4 min).As expected, the water–methanol solutions of polyether ionophores stored at 4 °C exhibited a similar behaviour, but a slower rate of change, with an approximately 60% drop in salinomycin and narasin levels and a 20% drop in monensin levels over the 11 days (lasalocid levels were again unchanged). These experiments indicate that some polyether ionophores, particularly salinomycin and narasin, are unstable in water– methanol solution. No significant degradation was observed in the solvent containing acetonitrile and ammonium acetate at pH 7.0.The mechanism of degradation for salinomycin appears to Fig. 1 Effect of storage at 20–25 °C on the concentration of various polyether ionophores in water–methanol (98 + 2 v/v), as determined by HPLC-electrospray MS (initial concentration ca. 10 mg ml21 = 100 %). 252 Analyst, 1999, 124, 251–256involve several relatively transient species, but results in one principal ultimate product (under the conditions and timescales described), which appears to be an isomer of salinomycin.In order to investigate the mechanism of degradation, the ultimate isomerisation product of salinomycin was characterised by NMR. The structure and numbering scheme of salinomycin and its isomer are shown in Fig. 2. Salinomycin itself has been previously studied by NMR spectroscopy6–8 and 13C NMR assignments have been published. 6,7 However, these assignments were made by comparison with shift data in smaller molecules,6 or by comparison with data from the spectra of the alkali metal complexes,7 not by 2D methods giving unambiguous results.Therefore, in order to aid interpretation of the NMR data for the isomer, the NMR spectra of salinomycin were acquired in the same solvent (acetonitrile-d3) and unequivocal assignments were obtained by 2D NMR spectroscopy, primarily the 1H–13C correlation techniques HMQC and HMBC. 1H and 13C NMR assignments for salinomycin are given in Tables 1 and 2 respectively.Long-range 1H–13C correlations are listed in Table 3; all data given in Tables 1 and 2 are consistent with the correlations shown. In those cases where overlap in the 13C dimension prevented identification of 1H resonances in the HMQC spectrum [e.g., C(8) and C(22)], uncertainties were resolved by observation of 1H–1H correlations in a 2D TOCSY spectrum. The 13C assignments given in Table 2 are in agreement (allowing for solvent differences) with those given in the literature,7 with the exception that C(17) and C(21) are interchanged. NMR data for the isomer were obtained by the same methods and are also shown in Tables 1–3.Comparison of the spectra of salinomycin and the isomer immediately indicated that the isomerisation process had caused rearrangement of the spiro rings in the central part of the molecule. The carbon resonances C(17), C(21) and C(24) are all shifted very significantly, as are the two proton resonances at low field [H(18) and H(19)].The coupling from H(18) to H(19) is reduced from 10.6 to 3.4 Hz and no coupling from H(18) and H(19) to H(20) is observed. In the HMBC spectrum of the isomer, H(18) and H(19) are observed to correlate to C(19) and C(18) respectively, and also to C(17) and C(20). The shifts of these carbons are 166.6, 108.8, 120.0 and 152.1 ppm [C(17)–C(20)]. These data, and the magnitude of the H(18)–H(19) coupling constant (3.4 Hz) strongly suggest the formation of a (substituted) furan moiety in the isomer.9,10 Further structural elucidation was achieved mainly by observing correlations from the Me protons in the HMBC spectrum as these were generally well resolved and had good S/N.The most upfield Me proton resonance (at 0.759 ppm) was observed to have long-range correlations to three carbons, two of which were protonated carbons with shifts indicative of the Fig. 2 The structure and numbering scheme of salinomycin (top) and its isomer (bottom).Table 1 1H NMR data for salinomycin and its isomera Atom Salinomycin Isomer Atom Salinomycin Isomer 1 — — 22 2.280, 1.959 2.979, 2.789 2 —b —b 23 2.113, 1.852 1.86, 1.83 3 3.885 3.87c 24 — — 4 1.921, 1.459 1.915, 1.467 25 3.618 3.459 5 1.919, 1.491 1.910, 1.475 26 1.45, 1.45 1.732, 1.545 6 1.842 1.832 27 1.63, 1.63 1.700, 1.613 7 3.667 3.671 28 — — 8 1.490 1.510 29 4.002c 3.778 9 4.119 4.11c 30 1.252 d 1.238 d 10 2.911 3.051 31 1.326, 1.345 1.348, 1.348 11 — — 32 0.902 t 0.901 t 12 2.804 2.715 33 1.579 s 1.146 s 13 3.689 3.821 34 0.735 d 1.315 d 14 1.742 1.412 35 0.931 d 0.906 d 15 1.684, 1.179 2.317, 1.304 36 1.923, 1.410 1.786, 1.161 16 1.677 3.142 37 0.817 t 0.774 t 17 — — 38 0.847 d 0.838 d 18 6.115 ddd 6.364 de 39 0.759 d 0.759 d 19 5.891 ddf 7.266 de 40 0.962 d 0.958 d 20 4.041 — 41 1.35–1.44 1.35–1.45 21 — — 42 0.929 t 0.923 t a Chemical shifts are in units of ppm.Spectra were measured at 294 K and are referenced relative to internal TMS. b Not observed.c Broad resonance. d Coupling constants: 10.6, 2.4 Hz. e Coupling constant: 3.4 Hz. f Coupling constants: 10.6, 1.5 Hz. Analyst, 1999, 124, 251–256 253attachment of oxygen, whilst the other had a shift very close to that observed for C(8) in salinomycin. The carbon directly attached to the upfield Me proton had a shift coincident with C(39) in salinomycin. Therefore, these isomer resonances could be identified as H(39), C(39), C(7), C(8) and C(9), with C(7) and C(9) distinguished by the correlation of H(7) to a further Me carbon, C(40).Correlations from C(40) enabled the methine carbon C(6) and the methylene carbon C(5) to be identified. C(7), C(6), C(5) and their attached protons have shifts nearly identical to the analogous resonances in salinomycin, strongly suggesting that this part of the molecule is unaltered in the isomer. This observation allowed C(2), C(3), C(4), C(41), C(42) and their attached protons [except H(2)] to be identified, all having shifts invariant from the analogous salinomycin resonances, as expected.The C(3) and C(41) resonances are both broad, as is C(2). H(2) and C(1) are not observable. Clearly, a dynamic process is broadening some of the signals relating to this part of the molecule. Proceeding from a further Me resonance at 0.838 ppm, correlations to C(9), C(10) and C(11) are observed, identifying the Me group as H(38). C(8), C(9), C(10) and C(11) and their attached protons, where appropriate, have chemical shifts comparable with those in salinomycin, once again suggesting that this part of the molecule is unchanged.The ethyl sidechain consisting of C(36) and C(37) can be linked to this fragment by the correlation from one of the methylene protons on C(36) to the ketone carbonyl C(11). Correlations from H(37) to C(36) and C(12) allow C(37) and C(12) to be identified. No further linkages relating to this part of the molecule are observable in the HMBC spectrum. Correlations observed in the TOCSY spectrum were consistent with the above assignments, giving fragment 1 as shown in Fig. 3. At the other end of the molecule [C(25) to C(29) and substituents], Me group resonances H(32) and C(32) can be identified by comparison of the shifts with those observed in salinomycin. Correlations from H(32) identify the methylene carbon C(31) and a quaternary carbon C(28). Correlations from the Me resonance H(30) to C(28) and C(29) allow C(30) and C(29) to be identified, whilst a methine carbon C(25) and a methylene carbon C(27) are identified by their correlations to H(29).Once again all of the 13C shifts are very similar to those observed in salinomycin, indicating that this part of the molecule is unchanged. The methylene carbon C(26) can therefore be assigned by analogy to salinomycin. H(25) correlates to a quaternary carbon C(24) with a shift (73.8 ppm) indicative of an electronegative substituent, -OR.H(25) also correlates to an Me group C(33). The Me proton H(33) correlates to C(25), C(24) and a methylene carbon C(23). Consequently, given these correlations and the structure of salinomycin, it can be deduced that C(25) is attached to a group 2C(OR)(CH2-RA)Me. The methylene protons H(23) and H(23A) correlate to another Table 2 13C NMR data for salinomycin and its isomera Atom Salinomycin Isomer Atom Salinomycin Isomer 1 —b —b 22 37.06d 33.75 2 50.76c 50.36c 23 32.46cf 33.41 3 76.42 76.66c 24 89.15 73.77 4 20.67 20.62 25 75.15 74.70 5 27.04 27.00 26 21.26cg 21.31 6 28.97 28.95 27 29.73g 30.25 7 72.54 72.28 28 71.66 71.67 8 37.00d 37.11 29 77.80 77.87 9 69.31 70.56 30 15.21 14.99 10 49.85 47.98 31 32.33f 31.91 11 218.9e 219.26 32 6.80 6.85 12 56.63 58.70 33 27.46c 23.27 13 77.39 74.32 34 16.07 21.40 14 33.62 35.16 35 17.89 16.10 15 39.20 40.57 36 17.46c 15.76 16 41.65 32.41 37 13.38 12.97 17 100.27 166.60 38 13.46 13.82 18 123.44 108.78 39 7.43 7.44 19 132.04c 120.03 40 11.34 11.32 20 67.54c 152.15 41 23.90 23.74c 21 107.65 190.84 42 12.62 12.67 a Chemical shifts are in units of ppm.Spectra were measured at 294 K and are referenced relative to internal TMS. b Not observed. c Broad resonance. d Assignment Interchangeable. e Observable in 2D spectrum only. f Assignment Interchangeable. g Distinguished by comparison with data for the isomer. Table 3 Long-range proton–carbon correlations for salinomycin and its isomera Salinomycin Isomer Proton Carbon Proton Carbon 3 2, 4, 5, 7 7 5, 9, (8, 22), 40 7 5, 8, 9, 40 (5,8) 7, 39 9 7, 10, 11, (8, 22), 39 10 9, 38 10 9, 38 12 11, 13, 36, 37 13 11, 12, 15, 36 13 14, 15, (35, 36) 15 14, 16 15 13, 14, 16, 17, 35 16 15, 17, 18, (26, 34) 18 17, 19 ,20 18 17, 19, 20 19 17, 18 ,21 19 17, 18, 20 20 18, 19, (8, 22) 22 (23, 31) 22 21, 23, 24 (6, 23) 33 23 22, 24, 33 25 24, 27, 33 29 25, 27, 28, 30 30 28, 29 30 28, 29 31 28, 29 32 28, 31 32 28, 31 33 (23. 31), 24, 25 33 23, 24, 25 34 15, 16, 17 34 15, 16, 17 (35, 42) 2, 13, 14, 15, 41 35 13, 14, 15 36 11 36 11 37 12, 36 37 12, 36 38 9, 10, 11 38 9, 10, 11 39 7, (8, 22), 9 39 7, 8, 9 40 5, 6, 7 40 5, 6, 7 a Measured in HMBC spectra at 400 MHz. Parentheses indicate ambiguous data. 254 Analyst, 1999, 124, 251–256methylene carbon, C(22), which in turn correlates to C(23) and C(24), as expected, and a carbonyl group C(21). This gives fragment 2 (see Fig. 3). The final molecular fragment is constructed starting from H(13), which correlates to a methine carbon H(14) and a methylene carbon H(15). The chemical shifts of H(13) and C(13) are indicative that C(13) contains an electronegative substituent, -OR. Two Me signals, H(34) and H(35), correlate to C(15), one of which, H(34), also correlates to a further methine carbon, C(16). H(16) correlates to the previously identified furan carbon, C(17), giving fragment 3, as shown in Fig. 3 (other correlations listed in Table 3 are consistent with this structure). In order to obtain the structure of the isomer, the three fragments must be linked together. The chemical shift of the carbonyl group C(21) indicates that it is probably either an abunsaturated ketone or a ketone attached to an aromatic ring. Therefore, fragments 2 and 3 are linked by the C(20) to C(21) bond. In the TOCSY spectrum, a correlation is observed from H(12) to H(13), linking fragments 1 and 2.Two protons must be added to the structure in order to give the correct molecular mass. Consequently, C(13) and C(24) must each possess a hydroxy group, leading to the structure shown in Fig. 2. No attempts were made to establish the stereochemistry of the molecule; it is assumed that the stereochemistry of salinomycin has been maintained in the isomer. All 13C chemical shifts are fully consistent with the structure. In particular, C(18) to C(21) are in excellent agreement with previous data,9 as indicated in Fig. 4 (shifts from the isomer outside parentheses). A proposed mechanism for the formation of the isomer from salinomycin is shown in Fig. 5. Having elucidated the structure of the ultimate isomerisation product of salinomycin, its electrospray mass spectra obtained at various CV potentials and those of unchanged salinomycin were re-evaluated. The greater abundance of the ion at m/z 733 in the isomer spectrum indicates that the pseudomolecular ions of the salinomycin isomer are more prone to fragmentation/ dehydration reactions than those of salinomycin (the stability of the sodiated pseudomolecular ion of salinomycin, which reflects the biological activity of salinomycin, has been reported.1 The elevated abundance of the apparent solvent adduct ion at m/z 809 may also indicate the more open structure of the ionised salinomycin isomer.It was not readily possible to rationalise any of the minor differences in the abundances of the lower mass ions (below ca.m/z 650). Conclusions This work has demonstrated that certain polyether ionophore compounds are prone to isomerisation when stored in aqueous solutions at ambient temperature. Salinomycin and narasin isomerise more rapidly than monensin, whilst no significant isomerisation was observed for lasalocid. The effect of changing the pH of the aqueous solution was not studied, but it is very likely that this will accelerate the rate of isomerisation (and further hydrolytic degradation).NMR studies indicated that the principal ultimate isomerisation product of salinomycin differs from salinomycin itself by the opening up of the spiro rings and the formation of a substituted furan moiety. Narasin, which is a methylated homologue of salinomycin and which displayed a similar rate of degradation in aqueous solution, presumably undergoes a similar conversion. It would be interesting to compare the observed degradation process with that pertaining in veterinary use. Given that the electrospray mass spectra of the isomers are dominated by sodiated molecular ions, it is likely that they retain at least some of their propensity to complex alkali metal ions, and thus preserve some of their biological activity. If they do retain Fig. 3 Salinomycin fragments, as determined by 2D NMR spectroscopy. Fig. 4 Reported 13C chemical shift data9 (shifts of analogous resonances from salinomycin isomer shown in parentheses). Fig. 5 Mechanism of formation of salinomycin isomer (C42H70O11: menoisotopic molecular weight, 750; average molecular weight, 750.01. Analyst, 1999, 124, 251–256 255biological activity, it may be necessary to include them in residue monitoring studies. References 1 J. A. Harris, C. A. L. Russell and J. P. G. Wilkins, Analyst, 1998, 123, 2625. 2 A. Bax and D. G. Davis, J. Magn. Reson., 1985, 65, 355. 3 L. Lerner and A. Bax, Carbohydr. Res., 1987, 166, 35. 4 D. Marion and K. Wüthrich, Biochem. Biophys. Res. Commun., 1983, 113, 967. 5 A. Bax and D. Marion, J. Magn. Reson., 1988, 78, 186. 6 H. Seto, Y. Miyazaki, K. Fujita and N. Otake, Tetrahedron Lett., 1977, 28, 2417. 7 F. J. Riddell and S. J. Tompsett, Tetrahedron Lett., 1991, 47, 10 109. 8 S. Mronga, G. M�uller, J. Fischer and F. Riddell, J. Am. Chem. Soc, 1993, 115, 8414. 9 SADTLER CSEARCH database: entry SADT-536 for methyl (5-methyl-2-furyl) ketone, 1998, Bio Rad Laboratories, Sadtler Division, Grand Junction, CO, USA. 10 J. A. Elvidge in Nuclear magnetic Resonance for Chemists, ed. D. W. Mathieson, Academic Press, London, 1967, p. 188. Paper 8/09594I 256 Analyst, 1999, 124, 251–2

ISSN:0003-2654    

DOI:10.1039/a809594i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of total tin in sediment reference materials by isotope dilution inductively coupled plasma mass spectrometry after alkali fusion Jun Yoshinaga, Atsuko Nakama and Kyoko Takata National Institute for Environmental Studies, Onogawa 16-2, Tsukuba, Ibaraki 305-0053, Japan Received 28th September 1998, Accepted 18th January 1999 The total tin concentration in a candidate sediment reference material (NIES CRM No.12 Marine Sediment) was determined by inductively coupled plasma mass spectrometry (ICP-MS) with a combination of different sample decomposition methods (acid digestion and alkali fusion) and modes of quantification [standard addition and isotope dilution (ID)] during a collaborative analysis for certification.Good agreement between the methods was obtained (10.2–11.0 mg kg21) and the data were consistent with those obtained from collaboration laboratories (9.38–11.6 mg kg21). Among the analytical methods used in this study, the most precise method, ID-ICP-MS, was examined further for its accuracy by analyzing other sediment CRMs.The analytical values obtained after acid digestion and alkali fusion differed significantly for all seven sediment CRMs analyzed; the total tin value was consistently higher and the relative standard deviation of the analysis was larger when the sample was decomposed by alkali fusion than by acid digestion. The difference varied from CRM to CRM (3–80%). This result indicated that there was a fraction of tin that was resistant to acid attack and the distribution of the fraction in the sediment was inhomogeneous.For an accurate determination of total tin in sediment, alkali fusion-ID-ICP-MS is the most suitable method. Introduction Tin (Sn) in geological materials has been determined in the context of ore exploitation and understanding of the geochemistry of this element. A number of methods are available for the determination of total Sn in this matrix, such as atomic absorption spectrometry (AAS),1–6 inductively coupled plasma atomic emission spectrometry (ICP-AES)7 and X-ray fluorescence spectrometry (XRF),7 although the sensitivity of the last two methods is not high enough for the trace determination of Sn.Total Sn determination in geological materials has tended to be concentrated on rocks and ores and relatively little attention has been paid to Sn in sediments. The National Institute for Environmental Studies (NIES) recently issued a marine sediment certified reference material, NIES CRM No.12 Marine Sediment, prepared for the quality assurance of organic Sn (tributyltin and triphenyltin).8 Organic Sn compounds have been used as antifouling agents on ships’ hulls and are now known to have a toxic effect on non-target marine organisms. During the certification of this CRM, collaborative analysis for total Sn content was also carried out.Four out of five laboratories that participated in the collaborative analysis for total Sn used mixed acid digestion for the decomposition of this CRM and the other used non-destructive neutron activation analysis.Since some geological materials contain acid-insoluble Sn(iv) oxide, data derived from sample decomposition by alkali fusion should also be included in the certification. In certifying an element content in a candidate CRM, inclusion of a definitive analytical method is desirable to obtain a reliable certified value.Our laboratory has been using isotope dilution (ID) inductively coupled plasma mass spectrometry (ICP-MS) for the certification of a variety of elements in environmental and biological CRMs. Establishment of an alkali fusion-ID-ICP-MS procedure was therefore necessary for the accurate determination of total Sn in the candidate sediment CRM. This study consisted of two parts. First, the total Sn content of NIES CRM No.12 Marine Sediment was measured by AAS, ICP-MS and ID-ICP-MS preceded by alkali fusion for the certification. Second, based on the ID-ICP-MS developed for the certification, determinations of total Sn in other sediment CRMs were carried out to examine the accuracy of the method further.Experimental Samples NIES CRM No.12 Marine Sediment was analyzed for total Sn content for certification. It was prepared from a surface sediment from Tokyo Bay sampled in 1989. The detailed preparation procedure, homogeneity assessment and certification of the tributyltin content of the CRM have been described in a separate paper.8 Sediment CRMs with certified or information values for total Sn were used for the validation of the proposed alkali fusion-ID-ICP-MS procedure.They included two CRMs from the Geological Survey of Japan (GSJ) with preferable values for total Sn:9 Lake Sediment JLk-1 (preferable value for Sn 5.7 mg kg21) and Stream Sediment JSd-1 (2.77 mg kg21). There were also three CRMs from the National Research Council of Canada (NRC) with certified values for total Sn: MESS-1 (certified value 3.98 ± 0.44 mg kg21), BCSS-1 (1.85 ± 2.5 mg kg21) and PACS-2 (19.8 ± Analyst, 1999, 124, 257–261 2572.5 mg kg21).Another CRM from NIES, CRM No.16 River Sediment for PAHs, was analyzed to determine information values for total Sn. Reagents ACS grade LiBO2 used for fusion was purchased from Aldrich (Tokyo, Japan) and ultrapure HCl (Kanto Chemicals, Tokyo, Japan) was used for the dissolution of fusion cake.For the solvent extraction of Sn from the fused sample, trioctylphosphine oxide (TOPO) and ascorbic acid (both from Merck, Darmstadt, Germany) were used. 4-Methylpentan-2-one (IBMK) (Kanto Chemicals) and 2,6-dimethylheptan-4-one (DIBK) (Wako, Osaka, Japan) used as an extraction solvents were of AAS grade. Ultrapure HNO3, HClO4 and HF used for acid decomposition were purchased from Kanto Chemicals. A tin stock standard solution was prepared by dissolving 0.5 g of high purity Sn metal (99.999%) (Wako) in 180 g of 30% ultrapure HCl and diluting to 500 g.Working standard solutions were prepared by diluting this stock standard solution with 1 m HCl. A 118Sn spike solution was prepared as described by Okamoto10 and used for ID analysis. The spike concentration was accurately determined by a reverse isotope dilution technique using a standard solution prepared from the high purity metal. Water used throughout the study was purified with a Millipore (Bedford, MA, USA) water purification system without further in-house distillation.Instrumentation The ICP-MS instrument used was a Hewlett-Packard (Avondale, PA, USA) HP-4500. The operating conditions were as follows: rf power 1.3 kW; reflected power, < 1 W; plasma gas flow rate, 16 L min21; auxiliary gas flow rate, 1.10 L min21; nebulizer gas flow rate, 1.15 L min21; and sample uptake rate, 0.2 mL min21. A flow injection (FI) system with a 100 mL sample loop was made from a six-way valve and Teflon tubing and was operated manually.Sample was delivered to the ICP by means of the peristaltic pump of the ICP-MS instrument. Scannning conditions for standard addition (SA)- and IDICP- MS analyses were as follows: ions monitored, m/z 120 for SA and 118 and 120 for ID, integration, 1 s for SA and 3 s for ID; data acquisition, three points per mass (peak center ± 0.05 u) by peak jumping mode; and number of measurements per sample, three for SA and five for ID.The software of the HP- 4500 fixes the scan number at 100; the dwell time was automatically set at 3.3 ms per point or 10 ms per mass for SA and 10 ms per point or 30 ms per mass for ID. In the ID analysis, the within-run precision of the 120Sn/118Sn measurement, i.e., the relative standard deviation (RSD) of the five replicates, was typically 0.3% for 20 ppb Sn standard solution (105 cps at m/z 118). Other major Sn isotopes, i.e., 116, 117, 119, 122 and 124, were also monitored in the preliminary experiments to check if spectroscopic interferences were present.Mass discrimination was corrected by periodic measurements of a standard solution (20 ppb) prepared from pure Sn metal. A Z-5100 Zeeman-effect background correction AAS instrument (Perkin-Elmer, Norwalk, CT, USA) fitted with an Sn hollow cathode lamp and AS-60 autosampler was also used for the determination of solvent extracted Sn. Alkali fusion–FI-ICP-MS The method of alkali fusion employed in this study was adopted from Elsheimer and Fries.5 A 0.15–0.2 g amount of the sample was mixed with 0.75 g of LiBO2 in a high-purity glassy carbon (Toho Carbon, Tokyo, Japan) or platinum crucible and heated in a muffle furnace at 1000 °C for 15 min.Sediment samples were precombusted at 500 °C for 4 h in a Pyrex glass beaker to remove organic carbon when a platinum crucible was used for fusion. After cooling overnight, the crucible was placed in a weighed Teflon beaker which was then filled with 1 m HCl.The fusion cake was dissolved by gentle heating and magnetic stirring. The solution was made up to 100 g with 1 m HCl and stored in a Teflon or polypropylene bottle as a stock sample solution. An aliquot of the stock sample solution was further diluted 10-fold with 1 m HCl and an appropriate amount of Sn was added. Three point standard addition (including 0 ng addition) was employed. The sample was analyzed by FI-ICP-MS, monitoring the ion at m/z 120 in the transient signal acquisition mode.Quantification was effected by peak area integration. Alkali fusion–solvent extraction-AAS The procedure for alkali fusion to prepare the stock sample solution was the same as above. The subsequent solvent extraction procedure was based on that of Elsheimer and Fries.5 An aliquot of the stock sample solution (typically 25 g) was placed in a capped 50 mL Teflon centrifuge tube and 0.45 g of ascorbic acid was added.Then 10 mL of 4% TOPO in IBMK was added and the mixture was shaken for 10 min by using a mechanical shaker. Phase separation was performed by centrifuging the tube at 3000 rpm for 10 min. The organic layer was transferred into another Teflon centrifuge tube and stored at 4 °C until measurement. The extracted Sn concentration was determined by AAS using standard solutions similarly extracted. A tungsten-impregnated graphite tube was used without any chemical modifier.2 Alkali fusion–solvent extraction-ID-ICP-MS An appropriate amount of 118Sn spike was added to the sample and fused as described above.After TOPO–IBMK or TOPO– DIBK extraction, 3-4 mL of the organic layer were transferred into another Teflon centrifuge tube and the Sn was backextracted into 1 mL of 5 m HNO3 by using a vortex mixer for 1 min. The HNO3 layer was diluted fivefold with water and subjected to ICP-MS measurement. Total Sn concentrations in the samples were calculated from the measured 120Sn/118Sn ratios according to the ID equation (e.g., ref. 11). Acid digestion-ICP-MS and -ID-ICP-MS Approximately 0.3 g of the sample (and 118Sn spike solution in ID analysis) was weighed into a Teflon beaker and decomposed with HNO3–HClO4–HF. The sample solution was made up to 50 g with 0.5 m HNO3 after heating to dryness. Standard addition-ICP-MS determination carried out using a similar procedure to the LiBO2 fused sample, although FI introduction was not used.Ion counts at m/z 118 and 120 were measured directly from the 118Sn spiked, decomposed and diluted solution for ID-ICP-MS quantification. Results and discussion Contamination It was found, in an earlier phase of the present study, that significant contamination was present in the LiBO2 fusion– TOPO–IBMK extraction procedure. The sources of the con- 258 Analyst, 1999, 124, 257–261tamination were identified: glassware, reagents and the atmosphere. Leaching of Sn from glass was apparent when 1 m HCl was stored in a glass calibrated flask, although all of the glassware was vigorously washed with acid prior to use.Therefore, all of the glassware was replaced with Teflon or polyethylene vessels. The tin contents of all the reagents used were checked by ETAAS and ICP-MS. It was found that TOPO contained a significant amount of Sn as an impurity. TOPO samples from various manufacturers were surveyed for their Sn contents and the purest one, in terms of Sn impurity content, was selected and used throughout.Other reagents used did not contain detectable Sn. Some Sn contamination was found to occur on decomposing the sample in a muffle furnace. Two different muffle furnaces installed in different laboratories were tested but the results were similar. Therefore, a procedural blank was prepared for each series of sample fusions. The mean Sn contamination level in the subsequent ID-ICP-MS analysis was 11 ± 9 ng, but typically 5–7 ng when samples with 300–2000 ng Sn contents were treated.The blank level was typically 1 ng for the acid digested samples with 500–6000 ng Sn contents. Total Sn contents of NIES CRM No.12 Marine Sediment General. The total Sn contents of NIES CRM No. 12 Marine Sediment and NRC PACS-2, which was analyzed concurrently for quality control of the analysis, determined by a variety of methods, are given in Table 1. There were no significant differences in the mean values between ETAAS and FI-ICPMS, both preceded by alkali fusion, and acid digestion-ICP-MS for NIES CRM No.12. A small difference was found between the values derived from acid digestion-and alkali fusion-IDICP- MS.All of the analytical values for NIES CRM No.12 in Table 1 were in agreement with the data from collaborating laboratories (9.38–11.6 mg kg21).8 The variation between the data from collaborating laboratories was much larger than the small difference observed between acid digestion- and alkali fusion-ID-ICP-MS.Therefore, all of the data in Table 1 were included in the certification of total Sn in this CRM.8 The LiBO2 fusions for these samples were performed exclusively using a glassy carbon crucible. Alkali fusion-FI-ICP-MS and alkali fusion-ETAAS. Analytical results obtained by ICP-MS and ETAAS after LiBO2 fusion in a glassy carbon crucible showed a slightly poorer precision compared with those obtained after acid digestion. This is attributed to the fact that LiBO2 fusion results in high, matrix (Li and B) content in the sample solution, which necessitates solvent extraction in the case of ETAAS.Involvement of solvent extraction lengthens the analytical procedure and it becomes more liable to procedural contamination. This contamination problem, together with the possible variability of the extraction efficiency, might have led to poorer precision in solvent extraction AAS analysis. ICP-MS also suffers from the high matrix content of the sample solution prepared by alkali fusion.Although the higher sensitivity to Sn of ICP-MS did not require solvent extraction, the presence of a high matrix content necessitates the introduction of the sample by FI, which is an established sample introduction method for ICP-MS but the small sample volume (consequently small absolute amount of analyte) involved together with the high matrix concentration affects the precision of the determination.Although the number of analyses was only one for each analytical method, the analytical results for PACS-2 by ETAAS and FI-ICP-MS were out of the uncertainty range of the certified Sn value of this CRM. This might have resulted from the poorer precision of the methods as mentioned above. Another possibility is the inhomogeneity of this CRM in terms of total Sn content, and this will be discussed later. Alkali fusion-ID-ICP-MS. Theoretically, ID-ICP-MS provides better precision and accuracy than do other ICP-MS quantification modes.The present ID-ICP-MS result was in line with this expectation for its better precision; the RSD of the analyses of NIES CRM No.12 was 3%, which was smaller than those with ETAAS and FI-ICP-MS. However, this RSD is larger than the within-run RSD (0.3%) of the isotope ratio measurement. The poorer than expected precision might be explained by a combination of atmospheric contamination variability, as mentioned earlier, dissolved solvent in the backextracted sample solution and sample inhomogeneity.A preliminary experiment on the efficiency of extraction of Sn by the present procedure using a standard solution demonstrated a 99% recovery. However, the ion counts at m/z 118 and 120 for the back-extracted samples were sometimes much smaller than those for the aqueous standard solutions of a similar Sn concentration. This should have been caused by the matrix effect arising from the IBMK dissolved in the backextracted solution.It was also possible that the presence of organic solvent in the sample might destabilize plasma. The within-run precision of isotope ratio measurement of the backextracted sample solution sometimes exceeded 1%, which contrasted with the 0.3% RSD expected in the aqueous standard solution. This indicated that a less water soluble solvent (e.g., DIBK) should be used for the extraction of Sn for more precise ID-ICP-MS determination. The possibility of sample inhomogeneity of NIES CRM No. 12 Marine Sediment will be discussed later. ID-ICP-MS can be an accurate analytical method only if no spectroscopic interference is present. Fig. 1 shows the ICP mass spectrum of m/z 85–130 for an unspiked NIES CRM No. 12 sample prepared by LiBO2 fusion, TOPO–IBMK extraction and back-extraction. Detailed measurement of the isotope ratios of Sn in the unspiked samples revealed consistent values with IUPAC values12 for 119/118, 120/118 and 124/118.A slight bias was found for 116/118, probably owing to molecular interference from 100Mo16O as this extraction procedure was not Sn-specific. The spectrum demonstrates extraction of Mo, Table 1 Results of total Sn determination of marine sediment CRMs by various methods (mg kg21 dry mass) Method NIES CRM No.12 NRC PACS-2a Fusion–extraction-ETAASb 10.5 ± 0.7 (n = 9) 23.0 Fusion-FI-ICP-MSb 11.0 ± 1.0 (n = 5) 23.9 Fusion–extraction-ID-ICP-MSb 10.5 ± 0.3 (n = 4) 21.0; 21.6 (n = 2) Acid digestion-ICP-MS 10.5 ± 0.3 (n = 6) 20.0 ± 1.6 (n = 3) Acid digestion-ID-ICP-MS 10.2 ± 0.2 (n = 6) 21.5 ± 0.4 (n = 3) a Certified value: 19.8 ± 2.5 mg kg21 dry mass.b Alkali fusion was performed using a glassy carbon crucible. Fig. 1 An ICP mass spectrum of the m/z range 85–130 for a NIES CRM No.12 Marine Sediment sample prepared by LiBO2 fusion, TOPO–IBMK extraction and HNO3 back-extraction. Analyst, 1999, 124, 257–261 259Ag and In in addition to Sn.An experiment using a standard solution showed a > 80% recovery of Mo and In by the present extraction-back extraction procedure (data not shown). Coextraction of Mo by a TOPO extraction procedure was also reported by Terashima.3 It should also be noted that dissolved solvent in the sample solution did not bias the mass calibration, excluding the possibility of systemtic error in the present results. Acid digestion ICP-MS and ID-ICP-MS. The analytical values obtained by acid digestion showed a slightly better precision than those obtained by alkali fusion.This may be related to the absence of matrix arising from flux (ICP-MS) or solvent (ID-ICP-MS). The lower blank level found in the acid digestion procedure than that in the alkali fusion might also contribute to the better precision. The results for PACS-2 were in good agreement with the certified value. The co-existing elements in an acid-digested, unseparated sample did not give any spectral interference in ID-ICP-MS when a 120/118 isotope ratio was used for the quantification. Isotope ratio measurement of unspiked, acid-digested and unseparated NIES CRM No. 12 showed that some spectral interference was observed for isotope pairs other than 120/118. Interference on m/z 112, 114, 115, 116 and 119 was seen; these should be from the isobaric interference from Cd and In, but the interfering molecules responsible for the 119/118 isotope ratio deviation was not identified. Since 119/118 was consistent with the IUPAC value in the solvent extracted sample, as described earlier, a molecule containing an element unextracted by TOPO should be involved in this spectral interference on m/z 119.ID-ICP-MS of Sn in other sediment CRMs The proposed ID-ICP-MS method preceded by either acid digestion or alkali fusion was applied to the accurate determination of total Sn in other CRMs with a sediment matrix to examine the applicability of the method.For the fusion of sediments, a platinum crucible was used; precombustion of the sediment CRMs was essential to prevent insoluble platinum–tin alloy formation during the fusion process of an organic matter rich sediment.3 The extraction solvent was switched from IBMK to the less water soluble DIBK. Comparison of alkali fusion and acid digestion. Table 2 shows a comparison of the results obtained by alkali fusion and by acid digestion for seven CRMs. The analytical results for the CRMs were in good agreement with the certified/information values when acid digestion was employed for sample decomposition.The certified Sn contents of NRC MESS-1 and PACS- 2 were exclusively based on the analytical results derived from acid digestion.13 The recent hydride generation AAS results on NIES CRM No.16, JLk-1 and JSd-1 from GSJ were also preceded by acid digestion.14 However, a significant difference was found between the values obtained by acid digestion and alkali fusion for all of the CRMs analyzed, the values obtained by alkali fusion being consistently higher.The absolute difference was particularly large for NIES CRM No.16 and NRC MESS-1. In addition, the RSD of the mean obtained by the alkali fusion of these CRMs was much larger ( > 10%) than the expected precision of the ID-ICP-MS analysis (0.3%). The slightly higher blank level encountered in alkali fusion than in acid digestion might only partly contribute to this variability, but it is unlikely to be the main factor.Another possibility that might be responsible for the bias and the variability was loss of Sn during fusion. Loss of Sn in the sample should apparently lead to a lower result. However, when it was combined with selective loss of an enriched isotope spike during the fusion process, it might have led to a biased result. Alkali fusion-ETAAS analysis of MESS-1, the sample with the largest difference from the acid digestion value, gave a similar result (7.5 ± 1.3 mg kg21, n = 3) to the alkali fusion-ID-ICPMS value.This result shows that loss of Sn in sediment samples during fusion can be ruled out. The calculation revealed that selective loss of the enriched isotope spike would have resulted in a lower Sn value if no loss of Sn in the sample occurred. Hence loss of Sn cannot be the cause of the difference and the variability of the present alkali fusion-ID-ICP-MS results. The results, therefore, should rather be interpreted as indicating that all of the sediment CRMs analyzed in this study contain a fraction of acid-insoluble Sn and the distribution of the fraction is inhomogeneous.The latter was unexpected in this case because the sediments analyzed in this study were all reference materials which should have been prepared to have a homogeneous element composition as a CRM. Re-evaluation of the certified value for total Sn content of NIES CRM No.12 Marine Sediment. The small difference between the values derived from alkali fusion and acid digestion in NIES CRM No.12 Marine Sediment shown in Table 1 was reproduced. The observed small difference may indicate that this CRM also contains a small fraction of the acid-insoluble Sn.Fortunately, the difference is small enough (3%) to be covered by the uncertainty range. In addition, the RSD of the analytical result was, again, 3%, which was worse than the analytical precision of the present ID analysis (0.3%).The use of DIBK improved the within-run precision of the isotope ratio measurement of the sample solution to the level of aqueous standard solution, indicating that the deterioration of the precision due to dissolved solvent could be negligible. However, the replacement of the solvent did not improve the overall precision of analysis. It indicated that NIES CRM No.12 had sample inhomogeneity of this level (3%) in terms of Sn content at a 150–200 mg sample intake, which was also covered by the uncertainty range.It should, therefore, be stressed that the uncertainty range of the certified total Sn content of NIES CRM No.12 Marine Table 2 Total Sn content of sediment CRMs determined by ID-ICP-MS after acid digestion or precombustion–alkali fusion using a platinum crucible (mg kg21 dry mass) Sample Acid digestion Precombustion– LiBO2 fusion Certified/reference value NIES CRM No.12 Marine Sediment 10.3 ± 0.05 (n = 3) 10.6 ± 0.3 (n = 4) 10.7 ± 1.4a NIES CRM No.16 River Sediment 7.72 ± 0.18 (n = 6) 9.2 ± 0.9 (n = 6) 7.7 ± 0.1b GSJ JLk-1 Lake Sediment 4.95 ± 0.04 (n = 3) 5.49 ± 0.07 (n = 9) 5.1 ± 0.2b 5.7c GSJ JSd-1 Stream Sediment 1.82 ± 0.03 (n = 6) 2.14 ± 0.08 (n = 3) 2.1 ± 0.1b 2.77c NRC MESS-1 3.88 ± 0.04 (n = 3) 7.0 ± 2.7 (n = 11) 3.98 ± 0.44a NRC BCSS-1 1.95 ± 0.18 (n = 6) — 1.85 ± 0.20a NRC PACS-2 21.5 ± 0.4 (n = 3) 22.4 ± 1.7 (n = 7) 19.8 ± 2.5a a Certified value.b Acid digestion–hydride generation AAS value from.Terashima.14 c Preferable value from GSJ.8 260 Analyst, 1999, 124, 257–261Sediment includes a bias resulting from the sample decomposition method and some inhomogeneity. The geochemical implication of the presence of acidinsoluble Sn in sediment and its inhomogeneous distribution is another issue and it will be a subject of further investigations. ID-ICP-MS will be an important tool for such investigations. Conclusion ID-ICP-MS was found to be a precise method for the determination of Sn in sediments. It was precise enough to detect the presence of acid-insoluble Sn in sediments at low levels (e.g., 3%) when acid digestion and alkali fusion were employed for the decomposition of the same sediment sample. To obtain an accurate total Sn content of a sediment, such as in the certification of a CRM, alkali fusion-ID-ICP-MS was essential. Acknowledgements The authors thank Dr. N. Imai, Geological Survey of Japan, for supplying sediment CRMs Jlk-1 and JSd-1. Thanks are also due to Drs. A. Tanaka, NIES, and S. Terashima, GSJ, for their valuable comments on the manuscript. References 1 M. Tominaga and Y. Umezaki, Anal. Chim. Acta, 1979, 110, 55. 2 L. Zhou, T. T. Chao and A. L. Meier, Talanta, 1984, 31, 73. 3 S. Terashima, Bull. Geol. Surv. Jpn, 1985, 36, 375. 4 E. Lundberg and B. Bergmark, Anal. Chim. Acta, 1986, 188, 111. 5 H. N. Elsheimer and T.L. Fries, Anal. Chim. Acta, 1990, 239, 145. 6 K. Ide, S. Hashimoto and H. Ohkochi, Bunseki Kagaku, 1995, 44, 617. 7 J. S. Kane, J. R. Evans and J. C. Jackson, Chem. Geol., 1989, 78, 1. 8 J. Yoshinaga, H. Kon, T. Horiguchi, M. Morita and K. Okamoto, Anal. Sci., 1998, 14, 1121. 9 N. Imai, S. Terashima, S. Itoh and A. Ando, Geostand. Newsl., 1996, 20, 165. 10 K. Okamoto, Spectrochim. Acta, Part B, 1991, 46, 1615. 11 J. W. McLaren, D. Beauchemin and S. S. Berman, Anal. Chem., 1987, 59, 610. 12 IUPAC, Pure Appl. Chem., 1991, 63, 991. 13 J. W. McLaren, National Research Council of Canada, personal communication, 1998. 14 S. Terashima, Geological Survey of Japan, personal communication, 1998. Paper 8/07515H Analyst, 1999, 124, 257–261 261

ISSN:0003-2654    

DOI:10.1039/a807515h

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Rapid separation of uranium and plutonium by extraction chromatography for determination by thermal ionisation mass spectrometry P. Goodall* and C. Lythgoe BNFL, B229, Sellafield, Seascale, Cumbria, UK CA20 1PG Received 25th November 1998, Accepted 2nd February 1999 A rapid method based upon extraction chromatography was developed for the separation of U and Pu from solutions of spent nuclear fuel for thermal ionization mass spectrometry. The method involves retention of Pu(iv) and U(vi) on a UTEVA column, thereby accomplishing a separation of these actinides from fission products.A separation of Pu from U was then achieved by selective elution of Pu. Two approaches for accomplishing this selective elution were investigated: (a) competitive complexation using oxalic acid and (b) reduction of Pu(iv) to Pu(iii) using ascorbic acid. The latter method yielded a purer isolate as the former method yielded a Pu isolate that was contaminated with an unacceptable amount of U.The UTEVA separation was 3–5 times more rapid than the existing separation and was considered to be more suitable for automation. Introduction Materials control and accountancy (MC&A) is a vital task within the nuclear industry and is demanded by both international treaty and the requirements of plant safety. As part of this task, the precise and accurate determination of uranium and plutonium is essential and thermal ionization mass spectrometry (TIMS) is accepted widely as the benchmark. The determination of U and Pu by TIMS requires pure U and Pu isolates.The current method of chemical separation used in these laboratories is based upon two separation columns. An initial separation of Pu from uranium/fission products is accomplished on a Dowex AG-1X4 column. A second separation of U from fission products utilizes a di(2-ethylhexyl) phosphoric acid extractant supported on an inert polymeric substrate. The effluent from the first separation, containing U and accompanying fission products, is the feed for the second stage of the separation.The separation procedure is slow, costly and, owing to the use of concentrated mineral acids, not suited to automation. Extraction chromatography, using a number of highly selective extractants, has been developed to permit the rapid separation of the actinides on an analytical scale. The characteristics and behaviour of these materials were discussed fully by Horwitz et al. 1 Analytical separations of the actinides have generally utilized a multi-column approach1–6 and often utilize a mixture of classical ion exchange and extraction chromatography. The use or adaptation of one of the known multi-column extraction chromatographic procedures would require little development but would offer little advantage over the separation schemes employed currently within these laboratories. However, careful examination of the properties of the commercially available materials suggested that a single solid phase extraction (SPE) cartridge could form the basis for a rapid separation of U and Pu from solutions of spent fuel.The advantages of a single column method were perceived to be (1) fewer manipulations and increased throughput; (2) reduction in solid waste generated per analytical cycle; (3) reduction in liquid waste generated per analytical cycle; and (4) ease of automation as the most aggressive reagent, 3 m nitric acid, was compatible with stainless steel.The chemical separation has the following requirements: (1) removal of fission products from the analytes of interest; (2) effect a mutual separation of U from Pu; (3) separation of Pu from Am due to the m/z = 241 isobar; and (4) analytes of interest in the nitrate form. It was considered that a UTEVA column (EiChroM Industries, Darien, IL, USA), consisting of an alkyl phosphonate extractant supported on an inert polymer, was the basis for a single column separation of U from Pu and of U/Pu from fission products.The initial approach relied upon competitive complexation using an oxalate ligand to elute Pu selectively with respect to U. An alternative route was based upon manipulation of the redox states of Pu to allow the selective elution of Pu with respect to U. Other commercial extraction chromatographic materials such as TEVA or TRU.Spec were considered not to have the required properties. TEVA does not retain U from nitric acid and would therefore not be capable of providing a single column solution.TRU.Spec, although retaining U and Pu strongly, had four failings: (i) the extractant had some affinity for selected fission products; (ii) the Pu complex with the extractant was very strong and would probably not allow selective elution via a competitive complexation; (iii) the extractant retains +3 actinides, thus preventing selective elution of Pu via manipulation of redox states; and (iv) retention of +3 actinides would not allow the separation of Pu from Am(iii).Experimental Sample pre-treatment Two procedures for sample pre-treatment were followed. The procedure depended upon the individual analytical requirements of the fuel dissolver being analysed. The first procedure involved a simple 250-fold dilution of the original sample from the fuel dissolution plant in a shielded hot cell facility. This allowed the determination of the Pu:U ratio in the original sample but did not allow the accurate and precise determination of the individual actinides.The second procedure involved the use of a large-scale dried spike, added to the original sample, to calibrate the subsequent 500-fold dilution. To satisfy standard isotope dilution methodology, a parallel dilution of the original sample, without the addition of the large scale dried spike, was required. The large scale dried spike consisted of milligram Analyst, 1999, 124, 263–269 263amounts of high enrichment 235U and 239Pu.This large scale dried spike procedure allowed the precise and accurate determination of the individual actinide concentrations in the original sample. In both cases, the diluted samples were transferred to the analytical laboratory where all subsequent manipulations of the sample could be performed within a radiological standard fume-hood. Mixed plutonium and uranium working standard solutions were prepared by gravimetric dilution from stock standard solutions of the individual analytes.The first procedure required the addition of a mixed tracer to the diluted sample. An aliquot (ca. 0.200 cm3) of the diluted sample was transferred into a small beaker and an aliquot of a mixed isotope dilution spike was added (233U, 242Pu). The sample was taken to dryness under a lamp. The second procedure, i.e., the large scale dried spike method, did not require the addition of a tracer. Simple aliquots (0.200 cm3) of the diluted samples were transferred into small beakers and taken to dryness under a lamp.The dried residues were then treated to ensure isotopic equilibrium and to fix the analytes in the required oxidation states. An aliquot of hydroxylamine hydrochloride in dilute HCl was added to the dried residues to ensure reduction of all Pu to the +3 state. The resultant solution was then taken to dryness, the residue was treated carefully with concentrated nitric acid, the solution was re-evaporated to dryness, the residue was treated again with concentrated nitric acid and the final solution was taken to dryness.This process ensured that the spike was equilibrated isotopically with the sample and that all Pu was fixed in the +4 redox state. Separation Procedures Two methods based upon EiChroM columns were investigated: (a) single column (UTEVA) with Pu removed selectively by complexation with oxalic acid and (b) single column (UTEVA) with Pu removed selectively by reduction with ascorbic acid.The column consisted of an extractant, dipentylpentyl phosphonate (DPPP), supported on a poly(methyl acrylate) resin. A variety of separation schemes were used during method development and validation and are summarized in Table 1 (oxalic acid case) and Table 2 (ascorbic acid case). Validation was achieved by comparison with the current analytical method. Dual column methodology—the current method The existing method for the separation of Pu and U from spent fuel solutions is a combination of anion chromatography and extraction chromatography.The sample solution was pretreated, the residue was dissolved in nitric acid (8 mol dm23) and the solution was loaded on an anion exchange column. Plutonium(iv) was retained strongly on this column and, after washing to remove U(vi) and fission products, the Pu was stripped from the column using hydroxylamine hydrochloride in dilute hydrochloric acid.This isolate was taken to dryness and the residue was treated with concentrated nitric acid to destroy the reductant and convert the analyte of interest into the nitrate form. The anion column effluent, containing U(vi) and fission product species, was then loaded on an extraction column consisting of D2EHPA supported on an inert polymeric substrate. This column retains U(vi) strongly but has little affinity for fission products. After washing to remove the latter, a pure U(vi) isolate was eluted with concentrated hydrochloric acid.This isolate was taken to dryness and the residue was dissolved in concentrated nitric acid and re-evaporated to dryness to convert the analyte of interest into the nitrate form. Table 1 Elution programmes for the separation of uranium and plutonium on UTEVA using oxalic acid reagent Condition Load Wash 1 Pu elution 1 Wash 2 Wash 3 U elution U elution [HNO3]/ mol dm23 3 3 3 3 3 3 0.02 3 [Oxalic acid]/ mmol dm23 3 3 3 100 100 3 100 3 A 3 3 2 cm3 2 cm3 4 3 2 cm3 10 3 2 cm3 — — 10 3 2 cm3 — Destinationa Waste Analysis Analysis Analysis — — Analysis — B 3 3 2 cm3 2 cm3 4 3 2 cm23 3 3 2 cm3 2 3 2 cm3 — 4 3 2 cm3 — Destinationa Waste Waste Waste Analysis Waste — Analysis — a Destination refers to the fate of the column effluent collected during the specified operation. Waste = column effluent diverted to a waste stream.Analysis = column effluent collected and submitted for analysis by TIMS, ICP-MS radiometric counting, etc.Table 2 Elution programmes for the separation of uranium and plutonium on UTEVA using ascorbic acid Condition Load Wash 1 Pu elution 1 Pu elution 2 Pu elution 3 Wash 2 Wash 3 U elution 1 U elution 2 [HNO3]/mol dm23 3 3 3 3 3 3 3 3 0.02 3 [Ascorbic acid]/ mmol dm23 3 3 3 100 10–100 10 10 3 3 3 [HCl]/mmol dm23 3 3 3 3 3 3 3 3 3 20 C 3 3 2 cm3 2 cm3 4 3 2 cm3 3 3 2 cm3 Destinationa Waste Waste Waste Analysis D 3 3 2 cm3 2 cm3 4 3 2 cm3 4 3 2 cm3 — — — — — — Destinationa Waste Waste Waste Analysis — — — — — — E 3 3 2 cm3 2 cm3 4 3 2 cm3 — 3 3 2 cm3 — — — — — Destinationa Waste Waste Waste — Analysis — — — — — F 3 3 2 cm3 2 cm3 4 3 2 cm3 — — 5 3 2 cm3 — — — — Destinationa Waste Waste Waste — — Analysis — — — — G 3 3 2 cm3 2 cm3 4 3 2 cm3 — — 3 3 2 cm3 2 3 2 cm3 3 3 2 cm3 4 3 2 cm3 4 3 2 cm3 Destination Waste Waste Waste — — Analysis Waste Waste Analysis Analysis a See Table 1. 264 Analyst, 1999, 124, 263–269Single column methodology using oxalic acid complexant SPE was based upon the extraction of the nitrate complexes of U(vi) and Pu(iv) from dilute nitric acid (3 mol dm23) into DPPP supported upon an inert polymeric substrate.After washing to remove fission products, the Pu(iv) was stripped selectively from the column using oxalate as a complexant. Uranium(vi) was eluted by lowering the nitrate concentration on column by washing with very dilute nitric acid. Recovery of U(vi) was aided by the use of oxalate.Evaporation of the fractions to dryness, followed by dissolution in concentrated nitric acid and hydrogen peroxide and re-evaporation to dryness resulted in samples suitable for measurement using TIMS. The oxalic acid solutions were prepared freshly for each batch of samples. An elution programme was designed to optimize the separation and clean-up (Table 1, separation A). The results of this optimization were then applied to the final analytical separation design (Table 1, separation B).The use of a vacuum manifold was investigated to augment the eluent flow and provide a rapid separation (Visiprep Manifold, Supelco, Poole, Dorset, UK). In general, flow augmentation was used for all but the preliminary investigation. Destruction of oxalic acid residue The oxalic acid residue was destroyed by the careful addition of concentrated nitric acid (3 cm3) and 30% v/v hydrogen peroxide (0.5 cm3). The solution was warmed gently to initiate the reaction, whereupon hydrogen peroxide was either consumed or decomposed thermally.When this reaction was complete, the solutions were taken to dryness under a lamp. The hydrogen peroxide-containing solution was not evaporated directly because, without the preliminary digestion, a vigorous reaction occurred with resultant ‘foaming over’ of the sample. Single column extraction using ascorbic acid for selective stripping of Pu SPE was based upon the extraction of the nitrate complexes of U(vi) and Pu(iv) into DPPP supported upon a poly(methyl acrylate) resin.After washing to remove fission products, the Pu(iv) was stripped selectively from the column. This was accomplished by reduction of the Pu(iv) using ascorbic acid. The ascorbic acid was destroyed by wet ashing with fuming nitric acid. Uranium(vi) was eluted by washing with very dilute nitric acid–oxalic acid (0.02 mol dm23 nitric acid, 0.1 mol dm23 oxalic acid). The U fraction was taken to dryness and the residue was treated with concentrated nitric acid–hydrogen peroxide to destroy the oxalic acid as described previously.Alternatively, U(vi) could be eluted with very dilute hydrochloric acid (0.02 mol dm23). The U(vi) fraction was taken to dryness under a lamp and this residue was converted to the nitrate form by dissolution in concentrated nitric acid and evaporation to dryness. The various separation schemes involving ascorbic acid are detailed in Table 2 (separations C–G).Destruction of ascorbic acid residue The Pu isolate (3 mol dm23 nitric acid, 0.01 mol dm23 ascorbic acid) was taken to incipient dryness, two drops of concentrated nitric acid were added and the residue was taken to dryness under a high-power lamp to induce charring of the residue. This residue was treated with fuming nitric (1 cm3) and warmed to oxidize and dissolve the partially charred residue. This solution was transferred into a coned beaker in three small portions.Each portion was taken to incipient dryness before the next portion was added. When taken to dryness, a small amount of charred residue could remain, this was destroyed by addition of a few drops of fuming nitric to the warm beaker. If necessary, this process was repeated. This procedure resulted in a sample that, when mounted on a TIMS filament, left no visible residue. Thermal ionization mass spectrometry The pure isolates were dissolved in dilute nitric acid (3 mol dm23) and mounted, by evaporation, on the side filament of a triple mass spectrometry bead (Cathodeon, Cambridge, UK). The bead consists of an Re centre and W outer filaments.The samples were transferred to the TIMS instrument, a double focusing system fitted with nine Faraday collectors (VG Sector-54; MicroMass UK, Altrincham, UK). The centre filament was heated to produce a 187Re ion current of 2 pA and the side filament to produce ion currents of 30 and 10 pA for 238U and 239Pu, respectively.Quantification was obtained by standard isotope dilution methodology. Results Single column using oxalic acid complexant. Uranium and plutonium standards. Elution profiles for the separation of U(vi)/Pu(iv) from fission products and the separation of Pu(iv) from U(vi) were determined using elution programme A (Table 1). The fission product elution profile was obtained using a highly diluted sample of a spent fuel digest. The U and Pu elution curves were produced using aqueous standards of the relevant analyte.The results are given in Table 3 for the removal of (a) fission products, (b) Pu and (c) U. The majority of the b,g activity was washed off the column within the first 5 cm3 of dilute nitric acid. Similarly, most of the Pu(iv) and U(vi) was removed within 8 and 5 cm3 of their respective stripping solutions. These results were used to design elution programme B (Table 1). To ensure that all b,g activity was removed, the column was washed with a total volume of 8 cm3 of nitric acid (3 mol dm23).Two batches of six U/Pu mixed standards were separated according to elution programme B (Table 1). One batch was separated using gravity fed columns. The second batch was separated with augmentation of the eluent flow using a vacuum manifold. The complete separation was accomplished in less than 1 h including changing of the solvent guides on the manifold, preparation of the stripping solutions and conditioning of the columns.The oxalic acid stripping solutions were prepared freshly for each batch by dissolution of solid oxalic acid in the appropriate matrix. The results of the determination are shown in Table 4 for (a) the gravity column and (b) the vacuum column. The values for the U and Pu content, derived using the gravity and augmented flow columns, were indistinguishable statistically at the 99% confidence level from the true value (t-test). Similarly, the precision of the determinations, using either the gravity or augmented flow columns, was indistinguishable at the 99% confidence level (F-test). It was therefore concluded that the use of a vacuum manifold did not introduce any bias or degrade the precision of the determination.Determination of uranium to plutonium ratio in dissolved spent fuel. The results from the analysis of the U/Pu standards were so encouraging that it was decided that there was sufficient Analyst, 1999, 124, 263–269 265confidence in the method that the analysis of simulates would not be required.It was therefore decided to test the method on a sample derived from one of the on-site re-processing dissolvers. This sample was subjected to a 250-fold dilution in a shielded hot cell facility before transfer to the laboratories. All manipulations of this diluted sample could then be performed at a radiobench. It should be noted that the required measurement was the ratio of the U and Pu contents of the original dissolver solution and therefore an accurate dilution was not required.If absolute concentrations were required, addition of an appropriate mixed tracer prior to dilution would be required, i.e., the large scale dried spike procedure described in the Experimental section. The samples were spiked, equilibrated, redox conditioned and supplied as a dry residue. A pair of U/Pu standards were prepared and run within each batch as a quality control measure.The results are given in Table 5 and the precision of the method, as applied to samples or standards, was indistinguishable at the 99% confidence level. The accuracy of the single column method was tested by comparison with the existing separation method. Parallel determinations were performed on the same sample solution. The samples were run in duplicate and the results are given in Table 6. The precisions quoted for these determinations are the 95% confidence intervals and are derived from instrumental and method control charts.These results demonstrate clearly that there was no detectable bias between the two separation methods. The method as developed demonstrated one major problem, which only became apparent as the method was applied to a greater range of “real” samples. The Pu fraction was contaminated with U, and although not present at levels that would adversely affect the analysis, did result in a significant ion current in the mass spectrometer multi-collector head channel positioned to detect ions of m/z 238.This was undesirable with respect to the detector lifetime. The efficiency of the separation was estimated by loading the column with a known mass of U and processing the column according to the relevant separation procedure (Table 1, procedure B). The U was determined in the Pu isolate fraction (Table 1, procedure B, Pu elution 1) using ICP-MS. The decontamination factor (DF) was defined as the ratio of the mass of U added to the column to the mass of U recovered in the Pu fraction.The results of this determination indicated that Table 3 Elution profiles (a) Elution profile for fission product activity from UTEVA using oxalic acid (0.1 mol dm23) in nitric acid (3 mol dm23)— Wash volume/ cm3 b-Activity/Bq cm23 g-Activity/Bq cm23 0 (5.14 ± 0.095) 3 104 (2.10 ± 0.015) 3 104 2 (24.3 ± 0.47) 3 104 (9.63 ± 0.032) 3 104 4 (0.288 ± 0.006) 3 104 (0.117 ± 0.005) 3 104 6 40 ± 9 0 ± 30 8 14 ± 7 0 ± 30 (b) Elution profile for Pu(IV) from UTEVA using oxalic acid (0.1 mol dm23) in nitric acid (3 mol dm23)— Wash volume/ cm3 Total a-activity/ Bq cm23 Uncertainty/ Bq cm23 Activity (%) 2 181 13 7.6 4 1090 242 45.8 6 1040 232 43.7 8 54 2 2.3 10 5 0.7 0.2 12 3.5 0.8 0.15 14 0.9 0.4 < 0.1 16 1.07 0.4 < 0.1 18 0.8 0.3 < 0.1 20 1 0.4 < 0.1 (c) Elution profile for U(VI) from UTEVA using oxalic acid (0.1 mol dm23) in nitric acid (3 mol dm23)— Wash volume/ cm3 U/mg cm23 RSD (%) Recovery (%) 2 0 96 0 5 142 0.86 93 7.5 0.91 17.6 < 1 10 1.1 6.8 < 1 12.5 0.99 8.8 < 1 15 1.35 6.6 < 1 17.5 2.14 5.2 < 1 20 1.44 18.2 < 1 22.5 1.25 14.5 < 1 25 1.7 14.2 < 1 Table 4 Separation of U(vi) and Pu(iv) standards Pu U Sample 103 Pu/U [Pu]/mg g21 Recovery (%) [U]/mg g21 Recovery (%) (a) Using a gravity fed column— 1 2.5998 2.2640 100.12 0.8705 100.12 2 2.5989 2.2616 100.01 0.8702 100.08 3 2.6038 2.2643 100.13 0.8696 100.01 4 2.5964 2.2654 100.18 0.8725 100.35 5 2.6020 2.2645 100.14 0.8703 100.09 6 2.5952 2.2602 99.85 0.8709 100.16 Mean 2.5994 2.2633 100.07 0.8707 100.14 s 0.0032 0.00199 0.0009 RSD (%) 0.125 0.088 0.106 (b) Using an augmented flow column— 1 2.6009 2.2659 100.08 0.8712 100.12 2 2.6011 2.2656 100.01 0.8710 100.08 3 2.5965 2.2579 100.13 0.8696 100.01 4 2.5881 2.2599 100.18 0.8732 100.35 5 2.5951 2.2598 100.14 0.8708 100.09 6 2.6037 2.2657 99.85 0.8702 100.16 Mean 2.5976 2.2619 100.024 0.8702 100.17 s 0.0056 0.0033 0.0011 RSD (%) 0.217 0.146 0.130 266 Analyst, 1999, 124, 263–269eluting Pu with an oxalic acid strip resulted in a decontamination factor of only 100 for U in the Pu fraction.Single column methodology using ascorbic acid reductant for Pu elution The selective removal of Pu from the UTEVA, using an oxalic acid strip, did not yield an isolate of sufficient purity. The oxalate based separation depended upon: selective complexation of the Pu with respect to U and the competition between the Pu oxalate and Pu DPPP complexes favouring the oxalate species.To improve the purity of the Pu isolate, an alternative selective stripping procedure was considered. This was based upon manipulation of redox states as follows: the UTEVA complexant has little affinity for Pu(iii); reduction of Pu(iv) to Pu(iii) could be accomplished selectively using ascorbic acid; and repeated wet ashing of the ascorbic acid residue yielded an isolate suitable for filament mounting.An elution profile for Pu(iv) was determined (Table 2, separation procedures C + D). The results of this experiment suggested that Pu was eluted quantitatively within the first 6 cm3 of the stripping solution, i.e., recovery = 91 ± 1% for 3 3 2 cm3 of eluent and recovery = 91 ± 1.3% for 4 3 2 cm3 of eluent. The clean-up of the Pu fraction, with respect to U, was determined for a separation based upon an ascorbic acid strip (Table 2, separation procedure C).A procedural blank was run in which the ascorbic acid solution (0.1 mol dm23 ascorbic acid, 3 mol dm23 nitric acid) was replaced with dilute nitric acid (3 mol dm23). The results of these determinations yielded decontamination factors of 1 3 104 (ascorbic acid) and 1.6 3 104 (procedural blank). These decontamination factors were at least two orders of magnitude better than were observed with an oxalic acid reagent (100) and suggested that the presence of ascorbic acid in solution had little effect upon the retention of U(vi).It was believed that this improvement in the purity of the Pu isolate, with respect to 238U, would be sufficient to attenuate the unwanted mass spectral response. This procedure was repeated and the isolate was taken to dryness and treated repeatedly with fuming nitric acid until the majority of the ascorbic acid residue was destroyed. This residue was mounted on a filament and treated as a Pu containing sample for TIMS. No 238U was detected, indicating that the clean-up of the Pu isolate was now fit for the purpose.The product of the initial oxidation of ascorbic acid was surprisingly indifferent to further oxidation using either fuming nitric acid or nitric acid–hydrogen peroxide. The complete destruction of the ascorbic acid was required to prepare clean filaments for TIMS. Therefore, the inability to perform an efficient and timely wet ashing of that compound was a threat to the successful implementation of this separation procedure.It was found that repeated evaporations with fuming nitric acid, coupled with a preliminary charring of the residue under a lamp, would eventually destroy the majority of the residue. The time required for this procedure dominated the separation procedure and removed the competitive advantage of the rapid EiChroM separation. To avoid the lengthy treatment of the Pu isolate, two approaches were identified: the use of an alternative reductant, e.g., hydroxylamine, and optimization of the amount of ascorbic acid added to the column.The use of an alternative reductant was investigated by modification of separation procedure C (Table 2). The ascorbic acid strip solution was replaced with 3 3 2 cm3 of nitric acid– hydroxylamine (3 mol dm23 nitric acid, 0.1 mol dm23 hydroxylamine) and the Pu content of the column effluent was determined using ICP-MS. The results of this experiment indicated that no Pu was recovered from the column using a hydroxylamine strip.Note that hydroxylamine, when employed as a reductant, would normally be prepared in dilute hydrochloric acid. From the known characteristics of UTEVA,1 retention of U on the UTEVA column would require HCl concentrations of ca. 2–3 mol dm23. This would be incompatible with one of the stated objectives of the method development, i.e., compatibility with off-the-shelf automation. The optimization of the ascorbic acid concentration of the strip solution was achieved according to separation procedure E (Table 2).The strip was accomplished using 3 3 2 cm3 portions of nitric acid–ascorbic acid; the concentration of nitric acid was maintained at 3 mol dm23 but the ascorbic acid content was varied between 0.01 and 0.1 mol dm23 (procedure E, Table 2, Pu strip 2). The results are given in Table 7 and suggest that Pu was recovered efficiently over the entire range of ascorbic acid concentrations.The ascorbic acid concentration was set to 10 mmol dm23 for all future experiments. It was believed that the variation in recovery with respect to ascorbic acid concentration was a kinetic effect and the flow through the columns was reduced by a factor of ca. 2 in subsequent experiments. An elution profile for Pu, using a stripping solution of nitric acid (3 mol dm23) and ascorbic acid (10 mmol dm23), was determined (Table 2, separation procedure F).The recovery of Pu within the first 6 cm3 of the strip was > 91%. This was accomplished only by slowing the eluent flow rate by a factor of ca. 2. These lower concentrations of ascorbic acid could be wet ashed with relative ease using fuming nitric acid during transfer into a coned 5 cm3 beaker prior to mounting on a TIMS filament. This yielded a filament with no visible residue and Table 5 Determination of plutonium to uranium concentration ratio with the oxalic acid method Sample 103 [Pu]/[U] Sample 103 [Pu]/[U] Sample 1 (batch 1) 7.128 Sample 1 (batch 2) 7.122 Sample 2 (batch 1) 7.126 Sample 2 (batch 2) 7.121 Sample 3 (batch 1) 7.147 Sample 3 (batch 2) 7.109 Sample 4 (batch 1) 7.164 Sample 4 (batch 2) 7.104 Sample 5 (batch 1) 7.161 Sample 5 (batch 2) 7.123 Sample 6 (batch 1) 7.132 Sample 6 (batch 2) 7.122 Mean 7.143 Mean 7.116 s 0.017 s 0.0083 RSD (%) 0.24 RSD (%) 0.121 Recovery (%) Recovery (%) [U]/ [Pu]/ [U]/ [Pu]/ mg g21 mg g21 U Pu mg g21 mg g21 U Pu Standard 1 0.8690 1.932 99.94 99.84 Standard 3 0.8700 1.935 100.1 100.0 Standard 2 0.8685 1.935 100.1 100.0 Standard 4 Analyst, 1999, 124, 263–269 267could be completed with minimal additional effort compared with the normal transfer procedure.In practice, an intermediate wash with ascorbic acid (Table 2, separation procedure G, wash 2) was used to ensure the complete removal of any residual Pu(iv) from the column prior to elution of the U(vi) fraction. These results were used to generate a single column separation of U and Pu (Table 2, separation procedure G) using a UTEVA column and an ascorbic acid strip.An intermediate nitric acid wash (Table 2, separation procedure G, wash 3) was used to remove residual ascorbic acid from the column before elution of the U fraction. This avoided the need for the U(vi) fraction to be treated with fuming nitric acid to destroy any ascorbic acid carried over from the Pu elution. Two distinct procedures were used to elute the U(vi) fraction, i.e., elution with nitric acid–oxalic acid (Table 2, U elution 1) or with very dilute hydrochloric acid (Table 2, U elution 2).The latter elution procedure required less processing of the U(vi) isolate to provide a sample suitable for TIMS, i.e., a simple evaporation to dryness followed by dissolution and evaporation with a small volume of concentrated nitric acid to convert the chloride to the nitrate. In contrast, elution with nitric acid–oxalic acid required wet ashing with concentrated nitric acid–hydrogen peroxide to destroy the oxalic acid reagent.The separation chemistry had been designed with future automation as a key parameter and a nitric acid–oxalic acid reagent may be more compatible with a simple robotic sample processor than the very dilute hydrochloric acid reagent. For manual operations, the hydrochloric acid reagent was preferred for elution of the U(vi) fraction because of the simpler post-separation treatment. This procedure was tested by comparison with samples analysed using the existing anion exchange method.The samples were received as isotopically spiked, equilibrated and redox conditioned residues. The pure Pu isolate was treated to destroy ascorbic acid during transfer into a coned beaker prior to mounting for TIMS. The separation was tested for both analytical requirements, i.e., determination of Pu:U ratios and determination of absolute concentrations of the individual actinides.The analyses were conducted over several days, using a variety of samples with distinct and separate addition of spikes. The results are given in Tables 8 (Pu:U ratios) and 9 (absolute concentrations). The uncertainties on the method are quoted as 95% confidence intervals and were generated from instrumental and method control charts. The differences between values derived using the two column (anion exchange) separation and the single column (UTEVA) separation are not significant statistically, i.e., the results agree within the quoted 95% confidence interval.It was therefore concluded that the accuracy and precision of the single UTEVA column method were validated by comparison with an accepted and quality assured method. This accepted method has been in routine, daily use for ca. 30 years. The single column UTEVA based separation has now been accepted by the routine analytical laboratories at this site.The method is perceived as offering at least a 50% improvement in Table 6 Comparison of separation techniques for the determination of U and Pu (oxalic acid method): procedures performed simultaneously Method Pu/mg g21 U/mg g21 Existing 1.399 ± 0.008 203.4 ± 1.1 UTEVA 1.400 ± 0.008 203.6 ± 1.1 Large scale dried spike methodology applied. Concentrations apply to original sample. Table 7 Optimization of ascorbic acid concentration Ascorbic acid concentration/mmol dm23 100 75 50 25 10 Pu recovery (%) 74 74 68 67 60 Table 8 Determination of plutonium to uranium concentration ratio with the ascorbic acid method Isotopic abundance Separation Replicate Pu-238 Pu-239 Pu-240 Pu-241 Pu-242 103[Pu]/[U] Existing 1 0.249 68.033 25.5791 4.885 1.254 2.818 2 0.248 68.063 25.543 4.895 1.249 2.848 Mean 0.249 68.050 25.560 4.890 1.251 2.833 EiChroM 1 0.243 68.022 25.584 4.889 1.254 2.800 2 0.249 68.048 25.568 4.884 1.251 2.828 Mean 0.246 68.030 25.580 4.887 1.253 2.814 Existing 1 0.260 67.360 26.018 5.039 1.322 2.854 2 0.260 67.360 26.011 5.047 1.321 2.872 Mean 0.260 67.360 26.010 5.043 1.322 2.863 EiChroM 1 0.260 67.359 26.018 5.041 1.322 2.857 2 0.260 67.332 26.027 5.057 1.324 2.870 Mean 0.260 67.350 26.020 5.049 1.323 2.864 Table 9 Determination of plutonium and uranium concentrations with the ascorbic acid method Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7 Sample 8 Method Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 Pu/ mg g21 U/ mg g21 EiChroM 1.260 179.0 1.274 180.4 1.162 188.6 1.051 173.3 1.189 180.4 1.159 184.9 1.136 174.2 1.161 182.4 Existing 1.259 178.8 1.280 181.0 1.166 188.5 1.054 173.5 1.900 180.5 1.160 185.2 1.138 174.5 1.162 182.6 Differencea 0.001 0.2 20.006 20.6 20.004 0.1 20.003 20.2 20.001 20.1 20.001 20.3 20.002 20.3 20.001 20.2 Uncertainty on Pu determination = ±0.007 mg g21.Uncertainty on U determination = ± 1.0 mg g21. Uncertainties on individual determinations, based on 95% confidence interval. Concentrations refer to original samples. a Difference = (result from EiChroM) 2 (result from Existing). 268 Analyst, 1999, 124, 263–269turn-around of the analysis, similar savings in the cost of the analysis, generates 50% less aqueous wastes and potentially reduces operator dose uptake. These benefits were achieved without sacrificing analytical performance. Conclusions Methods have been developed for the determination of U and Pu in solutions of spent nuclear fuel, based on extraction chromatography and utilizing a single column with selective elution of Pu and U as pure isolates. These methods are as accurate and precise as the existing analytical scheme. The preferred method uses a reductant (10 mmol dm23 ascorbic acid, 3 mol dm23 nitric acid) to strip selectively the strongly retained Pu(iv) from the column as Pu(iii). This method is significantly faster than the existing two column method. At least a 50% reduction in the time required for sample preparation has been demonstrated. This method is fully compatible with off-the-shelf automation and this option is being pursued. Acknowledgements The authors acknowledge the valuable aid in the preparation of samples and standards and for the thermal ionisation mass spectrometry of D. Kegg, E. Watters, R. Jenkinson, M. Armstrong and A. Owen. References 1 E. P. Horwitz, M. L. Dietz, R. Chiariza, H Diamond, S. L. Maxwell, III and M. R. Nelson, Anal. Chim. Acta, 1995, 310, 63. 2 E. P. Horwitz, M. L. Dietz, R. Chiariza and H. Diamond, Anal. Chim. Acta, 1992, 266, 25. 3 S. L. Maxwell, III and M. R. Nelson, Inst. of Nuc. Mat. Management, Naples, FL, 1994 (MS194). 4 A. G. Adriaens, J. D. Fasset, W. R. Kelly, D. S. Simons and F. C. Adams, Anal. Chem., 1992, 64, 2945. 5 J. R Cadieux, Jr. and S. H. Reboul, Radioact. Radiochem., 1996, 7, 30. 6 J. H. Kaye, R. S. Strebin and R. S. Orr, J. Radioanal. Nucl. Chem., 1995, 194, 191. Paper 8/09219B Analyst, 1999, 124, 263–269 269

ISSN:0003-2654    

DOI:10.1039/a809219b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of uranium and thorium in geological materials using extraction chromatography Helen E. Carter,a Peter Warwick,a John Cobbb and Geoff Longworthb a Environmental Radiochemistry Research Group, Chemistry Department, Loughborough University, Leicestershire, UK LE11 3TU b Analytical Services Group, AEA Technology plc, 551 Harwell, Didcot, Oxon., UK OX11 0RA Received 16th December 1998, Accepted 18th January 1999 A procedure has been developed for the determination of uranium and thorium in geological samples using extraction chromatography. Following sample preparation, uranium and thorium are pre-concentrated by precipitation with iron(iii) hydroxide and then separated using UTEVA resin.The separated uranium and thorium are electrodeposited onto stainless-steel discs and then measured by alpha spectrometry. The procedure was evaluated using uraninite ore, coral and granite reference materials. The uranium and thorium concentrations and the 234U/238U and 230Th/234U activity ratio values determined for the reference materials were in good agreement with the certified values.The presence of plutonium was found to interfere with the separation, but the inclusion of a reduction step using iron(ii) sulfamate eliminated the problem. Chemical recoveries for the procedure are similar to those for an anion-exchange procedure, but the extraction-chromatography procedure provides a more rapid separation using less reagents. 1. Introduction Measurements of the relative abundances of naturally occurring radionuclides, such as the isotopes of uranium, thorium and radium, in the natural decay series originating with 238U, 235U and 232Th, have been used to study a wide variety of problems in, for example, geology, hydrology and archaeology.1 Natural separation of parent and daughter nuclides, due to differing geochemical behaviour, leads to radioactive disequilibrium (indicated by a daughter/parent isotope activity ratio not equal to one) which may be used as an indicator of the timescale of past geochemical events.An example of such use is in the dating of calcites such as speleothem.2 When dating, it is assumed that thorium is not incorporated into the calcite on formation and the rate of the resulting ingrowth of 230Th from its parent 234U, governed by its half-life, is used to determine the age of the calcite. Alpha spectrometry has been used to measure the activities of 238U, 234U, 230Th, 232Th and 228Th.Chemical separation of uranium from thorium is required prior to measurement due to overlap between the alpha energies of the 234U and 230Th. Many methods have been reported for the separation of uranium and thorium from geological materials and are usually based on liquid–liquid extraction3–7 or anion-exchange chromatography. 8,9 Disadvantages associated with these procedures include the use of large volumes of organic solvents and acids resulting in large volumes of generated waste, and the procedures are labour intensive and time consuming.In the early 1980s, Horwitz and co-workers at Argonne National Laboratory developed new reagents for liquid–liquid extraction of the actinides.10,11 Reagents were then developed for the separation of actinides by extraction chromatography (reversed phase partition chromatography).12 This technique uses an inert polymeric support impregnated with a selective liquid extractant to form a solid sorbent which can then be loaded into a column.The use of extraction chromatography reduces both the volume of reagents used and the separation time. The UTEVA resins are commercially available from EIChroM Industries Inc. (8205 S, Darien, IL 60561, USA). This paper reports the use of an extraction chromatography resin comprising diamyl amyl phosphonate as the extractant to separate uranium and thorium from geological materials and compares its performance to an anion-exchange procedure.9 Previous work using the UTEVA resin has shown that uranium is extracted by the resin at nitric acid concentrations of 1 mol l21 and greater.13 The uranium can then be eluted from the resin using 0.01 mol l21 nitric acid.The separation of thorium from uranium has also been investigated using a nitric acid load solution to separate the uranium and thorium from matrix interferences present in environmental samples due to, for example, aluminium, iron or calcium.Thorium and uranium are then eluted from the column using 6 mol l21 hydrochloric acid and 0.025 mol l21 hydrochloric acid, respectively. Horwitz et al.14 also used the UTEVA resin in conjunction with other extraction chromatography resins to separate thorium, neptunium, uranium, plutonium and americium for the characterization of nuclear waste solutions. Uranium and thorium separation schemes for water and soil samples have been published by EIChroM Industries.15,16 These, along with the separation method for americium, plutonium and uranium in water,17 were used to develop the uranium and thorium separation method described in this paper. 2. Experimental 2.1. Reagents All reagents used were of analytical-reagent grade and 18 MW deionized water was used throughout. The following reagents were used: hydrochloric acid (9, 8, 5 and 0.02 mol l21), nitric acid (density = 1.42 g cm23, 6 and 3 mol l21), ammonia solution (density = 0.88 g cm23), 10 mg cm23 iron carrier solution [iron(iii) nitrate in 5% v/v nitric acid], 0.1 mol l21 ammonium oxalate.The extraction chromatography resins used were prepackaged UTEVA resin columns with a bed volume of 2 cm3. Analyst, 1999, 124, 271–274 271The radioactive tracer solutions used were: 229Th (192 mBq g21), 236U (167 mBq g21) and 232U/228Th (147 mBq cm23 of each radionuclide in secular equilibrium). The solutions were in approximately 7 mol l21 nitric acid. Interference studies were conducted using 208Po (133 mBq cm23), 210Pb/210Po (433 mBq cm23), 241Am (60 mBq cm23), 243Am (1.573 Bq cm23), 242Pu (98.2 mBq cm23) and 239Pu (43.08 mBq cm23). 2.2. Instrumentation Alpha spectrometry measurements were carried out using a 7401 model alpha spectrometer (Canberra Packard Ltd., Pangbourne, Berkshire, UK) fitted with passivated ion implanted planar silicon detectors (PIPS) having an active surface area of 450 mm2. The spectrometers were connected to a personal computer fitted with a TRUMP-8k-W3 multi-channel analyser plug in card (EG&G Instruments, Wokingham, Berkshire, UK). The electrodeposition of uranium, thorium, plutonium and americium onto stainless-steel discs was carried out using a procedure adapted from Kressin.18 A twelve position electrodeposition rig (ESI Ltd., Woodstock, Oxfordshire, UK) capable of maintaining a constant current of 1 A was used for the electrodepositions. Polonium was autodeposited onto silver discs from dilute hydrochloric acid solution in the presence of ascorbic acid.Gamma spectrometry measurements were carried out using germanium detectors (EG&G Instruments) coupled to a computerized analytical system. Stored spectra were analysed using Super Sabre software (AEA Technology plc, Didcot, Oxfordshire, UK) for photopeak identification and quantification. 2.3. Method development Initially, the elution behaviour of uranium and thorium from the UTEVA resin was investigated qualitatively using the procedure described by EIChroM.15 Load solutions comprising 10 cm3 of 3 mol l21 nitric acid spiked with either 236U or 229Th were prepared. These solutions were loaded onto separate UTEVA columns pre-conditioned with 5 cm3 of 3 mol l21 nitric acid.The following solutions were passed through each column: 5 cm3 of 3 mol l21 nitric acid (fraction 1); 5 cm3 of 9 mol l21 hydrochloric acid (fraction 2); five 3 cm3 aliquots of 5 mol l21 hydrochloric acid (fractions 3 to 7); five 3 cm3 aliquots of 0.02 M hydrochloric acid (fractions 8 to 12) and 10 cm3 of 0.1 M ammonium oxalate (fraction 13).Each separate column fraction was collected and prepared for electrodeposition. The resulting sources were measured by alpha spectrometry. The results showed that the 229Th was present in the 5 mol l21 hydrochloric acid (fractions 3 to 7) and the 236U was present in the 0.02 mol l21 hydrochloric acid (fractions 8 to 12). The experiment was repeated quantitatively using a resin load solution containing 232U and 228Th in secular equilibrium.The 15 cm3 fractions of 5 mol l21 and 0.02 mol l21 hydrochloric acid were collected and spiked with either 229Th or 236U, respectively, prior to electrodeposition to determine the plating and counting efficiencies of thorium and uranium. The results showed that the recoveries of uranium (101.8 ± 2.9%) and thorium (90.1 ± 2.6%) were almost quantitative in their respective fractions. 2.4. Interferences The proposed sample pre-concentration step was to coprecipitate uranium and thorium with iron(iii) hydroxide from a solution of the dissolved sample. Therefore, the effect of the presence of iron in the load solution was investigated by adding 1, 2, 3 or 5 cm3 of iron(iii) carrier solution. The 5 mol l21 and 0.02 mol l21 hydrochloric acid fractions were quantitatively analysed for uranium and thorium, respectively. The elution profile of a load solution containing 10 mg of iron(iii) showed uranium (109.6 ± 3.2%) was recovered from the column in fraction 8 and thorium (100.8 ± 3.1%) was recovered in fractions 3, 4 and 5.Table 1 shows uranium and thorium recoveries when varying amounts of iron(iii) were added to the load solution. These results show that the chemical recoveries of uranium and thorium were unaffected by the presence of between 10 and 50 mg of iron. The behaviour of several potential radionuclide interferences was investigated.Load solutions were prepared spiked with either 226Ra, 210Po, 241Am or 239Pu. These were loaded onto the column which was then sequentially eluted with 5 cm3 of 3 mol l21 nitric acid, 5 cm3 of 9 mol l21 hydrochloric acid, 15 cm3 of 5 mol l21 hydrochloric acid, 15 cm3 of 0.02 mol l21 hydrochloric acid and 10 cm3 of 0.1 mol l21 ammonium oxalate. The combined load and nitric acid fraction, the three hydrochloric acid fractions and the ammonium oxalate fraction were analysed for the presence of the radionuclide interferences. 226Ra was measured directly by gamma spectrometry whereas 210Po, 241Am and 239Pu were measured by alpha spectrometry following spiking with 208Po, 243Am and 242Pu, respectively. Table 2 shows the recoveries of the radionuclides in each of the analysed fractions. The 226Ra, 210Po and 241Am activities were recovered quantitatively in the combined load and nitric acid wash (fraction 1). The 239Pu was present in both the uranium and thorium eluates (fractions 3 and 4).This indicated that the column procedure would not be suitable for environmental samples that may contain plutonium isotopes. However, for many geological samples plutonium is unlikely to be present. Co-elution of plutonium with uranium and thorium from the UTEVA resin can be prevented by its reduction to Pu(iii) using iron(ii) sulfamate.17 This was investigated using load solutions comprising 239Pu and varying amounts of iron(iii) carrier solution, to which 0.6 g of sulfamic acid and 0.2 g of ascorbic acid were added.Table 3 shows that the 239Pu was present in the combined load and nitric acid wash (fraction 1; F1). This Table 1 The effect of the presence of iron(iii) on U and Th recoveries Amount Fe/mg U recovery (%) Th recovery (%) 10 103.4 ± 2.7 102.5 ± 3.7 10 105.5 ± 3.7 105.9 ± 7.0 20 98.4 ± 3.9 110 ± 3.5 30 98.5 ± 3.4 107.5 ± 3.6 50 102.2 ± 3.3 100.2 ± 3.0 50 103.5 ± 3.4 106.1 ± 3.1 Table 2 Percentage recoveries of Po, Ra, Am and Pu Fraction Substance F1a F2b F3c F4d F5e Po 101.1 ± 3.6 1.4 ± 0.1 < 1 1.4 ± 0.2 < 1 90.5 ± 3.1 1.3 ± 0.3 2.7 ± 0.7 < 1 2.9 ± 0.5 100.6 ± 2.2 2.3 ± 0.3 1.1 ± 0.2 < 1 2.0 ± 0.2 93.3 ± 5.0 2.5 ± 0.6 1.4 ± 0.3 < 1 3.2 ± 0.7 Ra 100 < 1 < 1 < 1 < 1 Am 110.0 ± 1.9 < 1 < 1 < 1 < 1 110.0 ± 2.3 < 1 < 1 < 1 < 1 Pu < 1 < 1 34.1 ± 1.5 56.3 ± 2.1 9.7 ± 0.5 4.5 ± 0.6 < 1 22.5 ± 0.9 49.0 ± 2.3 21.8 ± 1.1 a F1 = 10 cm3 of 3 M nitric acid load solution and 5 cm3 3 M nitric acid wash.b F2 = 5 cm3 of 9 M hydrochloric acid. c F3 = 15 cm3 of 5 M hydrochloric acid. d F4 = 15 cm3 of 0.02 M hydrochloric acid. e F5 = 10 cm3 of 0.1 M ammonium oxalate. 272 Analyst, 1999, 124, 271–274indicated that the reduced plutonium did not interact with the resin. The effect of the presence of the sulfamic acid and ascorbic acid on the uranium and thorium separation was investigated using the uraninite ore solution to which 1 cm3 of iron(iii) carrier solution was added.The results showed that the thorium and uranium were recovered quantitatively in their expected fractions. Therefore, iron(ii) sulfamate could be used for separation of uranium and thorium from samples where plutonium is suspected to be present. 2.5. Analytical procedure After dissolution of the geological material (see section 2.6), 1 cm3 of iron(iii) carrier solution was added to the sample solution.Uranium and thorium were co-precipitated from the solution with iron(iii) hydroxide by the addition of ammonia solution (density = 0.880 g cm23). The sample was warmed to coagulate the precipitate, which was then isolated by centrifuging. The precipitate was then dissolved in 5 cm3 of 6 mol l21 nitric acid and the resulting solution was diluted to 10 cm3 with deionized water. For samples expected to contain plutonium, 0.6 g of sulfamic acid and 0.2 g of ascorbic acid were dissolved in the resulting solution.A UTEVA resin column was pre-conditioned with 5 cm3 of 3 mol l21 nitric acid. The sample solution was loaded onto the column and allowed to drain through. A 5 cm3 aliquot of 3 mol l21 nitric acid used to rinse the centrifuge tube was passed through the column. The column was then converted to the chloride form by passing through 5 cm3 of 9 mol l21 hydrochloric acid. Thorium was eluted from the column with 15 cm3 of 5 mol l21 hydrochloric acid and then uranium was eluted from the column with 15 cm3 of 0.02 mol l21 hydrochloric acid.The uranium and thorium fractions were prepared for counting by alpha spectrometry by electrodeposition onto stainless-steel discs. The electrodepositions for uranium and thorium were carried out at 1 A for 3 and 5 h, respectively. 2.6. Procedure test The procedure was tested using the following materials: a solution prepared from a uraninite ore, a coral (RKM5), and a granite (SARM-1).The uraninite ore and coral were used in the Uranium Series Intercomparison Project.19 2.6.1. Sample preparation of uraninite ore. The uraninite ore sample was prepared by diluting 0.5 cm3 of the uraninite solution to 100 cm3 using 0.1 mol l21 nitric acid. 0.5 g of 236U and 229Th internal standards and 1 cm3 of the iron(iii) carrier solution were added to the solution, which was then left to equilibrate overnight. 2.6.2. Sample preparation of coral. Approximately 2 g of the coral (RKM5) were added to a 250 cm3 glass beaker and then 100 cm3 of distilled water were added. The calcium carbonate was then slowly dissolved by the gradual addition of nitric acid (density = 1.42 g cm23), such that the solution was maintained at pH 1.The sample was centrifuged to remove any acid insoluble residue. Suitable levels of 236U and 229Th internal standards and 1 cm3 of the iron(iii) carrier solution were added to the supernate, which was then left to equilibrate overnight. The solution was boiled prior to precipitation of iron(iii) hydroxide to ensure removal of dissolved carbonate. 2.6.3. Sample preparation of granite. 0.6 g of 236U and 0.6 g of 229Th internal standards and 1 cm3 of the iron(iii) carrier solution were added to approximately 0.5 g of the granite. The granite was ashed at 450 °C and digested in 20 cm3 of hydrofluoric acid for 48 h in a Teflon dish. A 5 cm3 aliquot of nitric acid and 5 cm3 of perchloric acid were added and the resulting solution was refluxed at 150 °C for 3 h.The solution was evaporated to dryness and the residue was treated twice with 10 cm3 of nitric acid (density = 1.42 g cm23) and evaporated to dryness. The final residue was dissolved in 20 cm3 of 8 mol l21 hydrochloric acid and the resulting solution diluted to 100 cm3 with deionized water. 2.7. Ion-exchange method for separation of uranium from thorium To compare the results from the UTEVA resin method, thorium and uranium were also separated from similar samples using a method based on the use of an ion-exchange resin.9 3.Results 3.1. Uraninite Table 4 shows the results, and the repeatability, of the measurements on the uraninite reference material. The errors in the measurements are based on counting statistics alone (1 s). The value of the 230Th/234U ratio shows that, within experimental errors, the radionuclides are in equilibrium. Comparison of the mean values of the 234U/238U (0.996 ± 0.02) and 230Th/ 234U (1.021 ± 0.03) ratios for the UTEVA resin method with those for the anion-exchange method (234U/238U, 0.968 ± 0.013; 230Th/234U, 1.016 ± 0.025) shows that the methods gave statistically identical results. 3.2. Coral Table 5 shows the results of the measurements on the coral reference material. The results show that the measured values of uranium (ppm ww) compare well with the reference value of 3.25 ± 0.10 ppm uranium. The measured 234U/238U activity ratios compare well with the certified value of 1.098 ± 0.016 and the 230Th/234U measured ratio compares well with the reference value of 0.709 ± 0.025.Since this ratio is less than 1, the radionuclides are not in equilibrium and the age of the coral reference material can be calculated.20 The measured age of the coral compares well with the reference value of 131 000 ± 9000 years. 3.3. Granite Table 6 shows the results of measurements on the granite reference material. The uranium concentration is 15 ppm (ww) Table 3 Percentage recoveries of Pu with 10, 30, 50 and 70 mg of added Fe(iii) Fraction Amount of Fe/mg F1a F2b F3c F4d F5e 10 105.1 ± 3.8 < 1 < 1 < 1 < 1 30 99.5 ± 2.5 < 1 < 1 < 1 < 1 50 102.4 ± 2.9 < 1 < 1 < 1 < 1 50 101.1 ± 4.4 < 1 < 1 < 1 < 1 70 100.9 ± 2.5 < 1 < 1 < 1 < 1 70 96.8 ± 3.0 < 1 < 1 < 1 < 1 a F1 = 10 cm3 of 3 M nitric acid load solution and 5 cm3 3 M nitric acid wash.b F2 = 5 cm3 of 9 M hydrochloric acid. c F3 = 15 cm3 of 5 M hydrochloric acid. d F4 = 15 cm3 of 0.02 M hydrochloric acid. e F5 = 10 cm3 of 0.1 M ammonium oxalate. Analyst, 1999, 124, 271–274 273which compares well with the mean UTEVA measurement of 17.544 ± 0.626. The certified thorium concentration of 51 ± 3 ppm (ww) also compares well with the mean UTEVA measurement of 47.098 ± 4.326 ppm (ww). 4. Conclusions The proposed method, which uses UTEVA resin to separate uranium from thorium in geological materials, has been evaluated using coral, uraninite and granite reference materials.The results of the measurements have been compared to those obtained by a method based on the use of an ion-exchange resin. The UTEVA resin results compared favourably with both the ion-exchange results and the certified values. The use of the UTEVA resin provides a quicker separation which uses less reagents than the ion-exchange resin method.Acknowledgements The authors are grateful to Richard Ku, University of Southern California for provision of the dated coral (RKM5). One of us (HC) thanks the EPSRC and AEA Technology plc, Harwell for supporting this work with a Case Award. References 1 M. Ivanovich and R. S. Harmon, Uranium-Series Disequilibrium. Applications to Earth, Marine, and Environmental Sciences, Oxford Science Publications, Oxford, 2nd edn., 1992. 2 E. L. Wild and I. Steffan, Radiochim. Acta, 1992, 57, 153. 3 A. A. Nemodruck and I. E. Varotimtskaya, Zhur. Anal. Chim, 1962, 17, 481. 4 K. Bril and S. Holzer, Anal. Chem, 1961, 33, 55. 5 V. Pfeifer and F. Hecht, Mikrochim. Acta., 1960, 3, 378. 6 D. F. Wood and R. H. McKenna, Anal. Chim. Acta, 1962, 27, 446. 7 D. Ishii and T. Takenchi, Jpn. Anal., 1961, 10, 1125. 8 L. R. Bunney, N. E. Ballon, J. Pascual and S. Foti, Anal. Chem, 1959, 31, 324. 9 A. E. Lally and J. D. Eakins, Some Recent Advances in Environmental Analyses at AERE, Harwell; Symposium on Determination of Radionuclides in Environmental and Biological Materials, Paper 12, CEGB, Sudbury House, London, 1978. 10 E. P. Horwitz and D. G. Kalina, Solvent Extr. Ion Exch., 1984, 2, 179. 11 E. P. Horwitz, D. G. Kalina, H. Diamond, G. F. Vandergrift and W. W. Schultz, Solvent Extr. Ion Exch., 1985, 3, 235. 12 E. P. Horwitz and M. L. Dietz, Anal. Chim. Acta, 1990, 238, 263. 13 E. P. Horwitz, M. L. Dietz, R. Chiarizia, H. Diamond, A. M. Essling and D. Graczyk, Anal.Chim. Acta, 1992, 266, 25. 14 E. P. Horwitz, M. L. Dietz, R. Chiarizia, H. Diamond, S. L. Maxwell and M. R. Nelson, Anal. Chim. Acta, 1995, 310, 63. 15 EIChroM Industries, Inc., Analytical procedure/UK, ACW06 method, U, Th in water, 1994. 16 EIChroM Industries, Inc., Analytical procedure/UK, ACS06 method, U, Th in soil, Rev 1.1a, 1994. 17 EIChroM Industries, Inc., Analytical procedure/UK, ACW03 method, Am, Pu and U water, 1995. 18 I. K. Kressin, Anal. Chem., 1977, 49(6), 842. 19 M. Ivanovich, T-L. Ku, R. S. Harmon and P. L. Smart, Nucl. Instrum. Methods Phys. Res., 1984, 223, 466. 20 M. Ivanovich and R. S. Harmon, Uranium-Series Disequilibrium. Applications to Earth, Marine, and Environmental Sciences, Oxford Science Publications, Oxford, 2nd edn., 1992, p. 71. Paper 8/09781J Table 4 Results of the uraninite ore analyses Sample 238U/mBq g21 234U/mBq g21 230Th/mBq g21 234U/238U 230Th/234U U recovery (%) Th recovery (%) 1 154.2 ± 3.3 156.9 ± 3.4 154.5 ± 3.1 1.018 ± 0.018 0.984 ± 0.029 85 85 2 151.4 ± 2.8 149.7 ± 2.8 153.8 ± 3.9 0.989 ± 0.017 1.028 ± 0.032 81 80 3 156.8 ± 2.9 150.4 ± 2.8 158.4 ± 3.6 0.959 ± 0.016 1.053 ± 0.031 82 69 4 151.9 ± 3.4 153.4 ± 3.4 154.7 ± 3.2 1.009 ± 0.021 1.008 ± 0.031 74 79 5 151.8 ± 3.8 152.4 ± 3.9 156.9 ± 4.3 1.004 ± 0.020 1.030 ± 0.038 92 85 Mean 153.2 ± 2.3 152.6 ± 2.8 155.7 ± 1.9 0.996 ± 0.020 1.021 ± 0.030 83 ± 6.5 80 ± 6.5 Certified value 151 151 151 1 1 — — Table 5 Results of the coral (RKM5) analyses Sample [U] (ppm; ww) 234U/238U 230Th/234U U recovery (%) Th recovery (%) Age/ka 1 3.40 ± 0.08 1.072 ± 0.026 0.719 ± 0.026 83 80 134+12 29 2 3.30 ± 0.09 1.078 ± 0.027 0.720 ± 0.025 72 71 135 ±1 0 3 3.33 ± 0.07 1.075 ± 0.020 0.724 ± 0.026 70 61 136+12 29 4 3.16 ± 0.07 1.136 ± 0.023 0.705 ± 0.025 61 68 128 ± 9 5 3.23 ± 0.07 1.102 ± 0.026 0.708 ± 0.021 90 80 130+10 28 Mean 3.28 ± 0.09 1.093 ± 0.03 0.715 ± 0.008 75 ± 11 72 ± 8 132.6 ± 3.44 Certified value 3.25 ± 0.105 1.098 ± 0.016 0.709 ± 0.025 131 ± 9 Table 6 Results of the granite (SARM-1) analyses Sample [U] (ppm; ww) [Th] (ppm; ww) 234U/238U 230Th/234U 1 16.706 ± 0.558 51.467 ± 2.076 1.144 ± 0.035 0.903 ± 0.047 2 17.821 ± 0.401 48.900 ± 1.381 1.044 ± 0.021 0.913 ± 0.033 3 17.963 ± 0.565 44.820 ± 2.082 1.059 ± 0.031 0.720 ± 0.041 4 17.065 ± 0.539 40.694 ± 1.434 1.053 ± 0.031 0.809 ± 0.038 5 18.164 ± 0.661 49.612 ± 1.170 1.108 ± 0.037 0.837 ± 0.036 Mean 17.544 ± 0.626 47.099 ± 4.326 1.082 ± 0.043 0.836 ± 0.078 Certified value 15 51+4 23 274 Analyst, 1999, 124, 271–274

ISSN:0003-2654    

DOI:10.1039/a809781j

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Surface partitioning studies of N-methylcarbamate-treated post-harvest crops using SFE-HPLC-postcolumn reaction-fluorescence Iain A. Stuart,† Ray O. Ansell, John MacLachlan and Peter A. Bather Department of Physical Sciences, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA Received 15th October 1998, Accepted 8th January 1999 The partitioning characteristics of selected carbamate insecticides (carbaryl, aldicarb, bendiocarb and pirimicarb) on five fruit and vegetable types were investigated.Post-harvest samples were surface-saturated with a methanolic–aqueous mixed carbamate spiking solution for a number of time periods. Samples were taken at 3, 7, 10 and 14 d, and extracted using supercritical CO2 at pressure = 300 atm modified with 10% dimethyl sulfoxide. Extracts were analysed by HPLC-postcolumn reaction-fluorescence detection at lex = 330 nm and lem = 450 nm for N-methylcarbamates and at lex = 315 nm and lem = 380 nm for pirimicarb.The relative partitioning of each insecticide between sample skin and flesh was investigated. This included the determination of both half-life and normalised matrix metabolic rate studies with respect to each carbamate. Multilinear regression (MLR) was applied to a number of insecticide and matrix-based variables to develop regression models for carbamate partitioning for each matrix type studied. Experimentally derived carbamate half-lives ranged from 3.6 d (carbaryl in pear flesh) to 8.0 d (bendiocarb in banana skin).Determinations of normalised metabolic rates were based on calculating the time period from the point of sampling through to the point where carbamate concentration was reduced to 5% of its initial value. These values ranged from 16.2 d (bendiocarb in potato skin) to 34.7 d (bendiocarb in banana skin). Although no practicable MLR partitioning models were obtained, it was found that the models created indicated that carbamate solubility in water (and hence log P) and the number of days in contact with the spiking solution were the most important parameters in model construction.Introduction As a consequence of the practice of increasing pesticide loadings onto crops to counteract developing insect resistance, there exists a potential for significant quantities of material to pass into the foodchain through uptake in roots, leaves and, more importantly, the crop itself. It is this latter area that has been of increasing interest in recent studies as workers have attempted to model the partitioning effects of many hydrophilic pesticides on fruit and fruit-like vegetables.Yoshida et al.1 have indicated a partitioning ‘order of importance’ for pesticide cross-over into crop flesh by determining the ratio between [pesticide]flesh to [pesticide]skin. For ratios of the most common classes of pesticides, this order was shown to be: organochlorides, organophosphates, carbamates and chrysanthemic acid derivatives in ascending order.This is in good agreement with the mean hydrophobicities of each pesticide class. Since Noble first suggested that pesticides with log P values > 4.0 could be classed as being ‘fat-soluble’,2 the above order is also as might be expected. The work reported here provides further information on these partitioning effects by using the inert extraction medium of modified supercritical CO2 3 while also utilising the highly selective carbamate detection method of HPLC-postcolumn reaction-fluorescence.4–17 The increase in partitioning of each compound in individual skin and, subsequently, flesh samples has been monitored over a period of 14 d continuous carbamate contact.Samples were taken at 3, 7, 10 and 14 d, and extracted using dimethyl sulfoxide (DMSO)-modified supercritical CO2. The compounds for this study (carbaryl, bendiocarb, pirimicarb and aldicarb) were chosen for their variation in methods of crop uptake, crop host and group class. Additional work has also determined the relative rate of insecticide degradation exhibited by each fruit or vegetable studied through the calculation of both carbamate half-life and carbamate longevity in each matrix.A Multilinear regression (MLR) technique was used to develop regression models between relevant parameters in the partitioning process. Experimental Instrumentation and reagents All extractions were completed on a SFE-723M Supercritical Fluid Extraction (SFE) system (Dionex, Camberley, UK) using two 16 cm3 capacity extraction cells (Keystone Scientific, Bellefonte, PA, USA) in parallel (1200 cm3 linear restrictors were used to maintain back-pressure and flow rate) for each extraction run. 99.99% SFE-grade CO2 was used as the primary solvent, supplied with a 110 bar He over-pressure (BOC Speciality Gases, Guildford, UK). An extracting fluid density of 0.795 g cm23 unmodified was used, modified with 10% mol/v DMSO using the instrumental conditions as described in Table 1.Procedure All carbamate standard and spiking solutions were made up from certified insecticides: carbaryl, carbofuran, aldicarb, bendiocarb and pirimicarb (Promochem, UK). Chromato- † Present address: Strathclyde Institute for Drug Research, University of Strathclyde, Royal College, 204 George Street, Glasgow, UK G1 1XW. Analyst, 1999, 124, 275–280 275graphic solvents used were of HPLC-grade (Sigma-Aldrich, Poole, Dorset, UK) pumped via a LC 9012 Solvent Delivery System (Varian, Walnut Creek, CA, USA) with a 10 ml injection loop.Detection was completed by postcolumn reactionfluorescence on a scanning wavelength detector (model 9070, Varian) at lex = 330 nm and lem = 450 nm for carbaryl, bendiocarb, carbofuran (internal standard) and aldicarb isoindole derivatives. Pirimicarb was detected at lex = 315 nm and lem = 380 nm without the presence of the reagents. A linear gradient of 100 + 0 H2O–THF to 30 + 70 H2O–THF in 20 min was used to complete carbamate elution at a flow rate of 1 cm3 min21.All determinations were completed on a 150 mm 3 4.6 mm C18 column at 42 °C (Pickering Laboratories, Mountain View, CA, USA) contained in the postcolumn reaction module (PCX 5100, Pickering Laboratories). Fluorescence reagents, o-phthaldehyde (OPA), NaOH hydrolyser (0.3% at 100 °C), OPA diluent (0.3% boric acid) and Thiofluor (N,N-dimethyl-2-mercaptoethylamine hydrochloride) were of chromatographic grade (Pickering Laboratories).Crop sample preparation. Washed whole banana, pear, onion, potato and apple samples were immersed in a MeOH– H2O (20 + 80) mixture of the carbamates (50 mg cm23 with the exception of aldicarb at 25 mg cm23). All samples were subsequently refrigerated for 3, 7, 10 and 14 d (4 °C) in order to simulate any partitioning/contact processes involved between the crop and the carbamates. A 10 cm3 sample of the spiking solution was removed prior to the addition of the fruit/vegetable sample and stored under the same conditions. This aliquot was analysed on each day of sampling with the resultant carbamate response being used to normalise carbamate recovery, independent of carbamate stability in the spiking solution as a function of time.Sample extraction and determination. On completion of the required time period of contact, each sample was washed thoroughly in a deionised water bath to remove any surfaceadhered residues (adsorption).Samples were divided into skin and flesh portions with the inner surface of each skin portion being scraped to remove the lower dermis layer which was bulked with the flesh fraction. In addition, only the outermost 0.5 cm thickness of sample flesh was used. Individual flesh and skin samples were then chopped into individual 2 g samples, air-dried for 24 h in a fan-assisted oven (no heat) and stored in a desiccator on the day of use. The samples were mixed with an equal amount of Celite to ensure uniform cell packing.In addition, cells were packed at each end with methanol-cleaned glass wool to prevent end-cell frit blockage. On completion of extraction, each extract was filtered as necessary and spiked with 1 cm3 of an aqueous 50 ppm carbofuran internal standard. The extract was then reduced to near dryness and reconstituted with 1 cm3 of methanol and analysed by HPLC-postcolumn reaction-fluorescence using the conditions as described earlier.Study of the metabolic action of crops on carbamates. The sample spiking procedures for this work were identical to those used previously. Individual carbamate levels were determined in the skin and flesh of all five samples after a period of 10 d in contact with the carbamate spiking solution. These ‘Day 10’ samples were removed from the spiking solution, washed and refrigerated. Samples were taken on days 3, 7, 10 and 14 postremoval from the carbamate spiking solution, mixed directly with Celite and extracted using the conditions as described in Table 1.Both relative carbamate half-life and longevity values were calculated on the data obtained from this part of the work. Due to the dehydrating effect of refrigeration, these samples were periodically moistened with a water spray to maintain biological activity at an, albeit, reduced capacity due to the incapacitating effect of the reduced temperature on crop metabolic rate.The additional benefit of refrigeration is that it is possible to mimic the storage method adopted by many fruit and vegetable producers/vendors and so obtain a more ‘real world’ carbamate half-life value under these conditions. Results and discussion Carbamate concentration against matrix types (Fig. 1) When comparing carbamate concentrations in both parts of each crop type, it is possible to rank which crops are susceptible to greatest pesticide cross-over.Those crops that possess high partitioning characteristics in both parts of the crop are naturally of greatest concern. It was noticeable that the partitioning trends for the aromatic carbamates, e.g., bendiocarb, pirimicarb, and carbaryl, show that the magnitude of partitioning across all matrices tends to rank in order of decreasing hydrophobicity, e.g., carbamates with higher relative log Pcarbamate values possess a higher affinity for partitioning than those of more hydrophilic nature. However, this correlation is not maintained in recovery trends for aldicarb (aliphatic) which is absorbed to a greater level in all cases than its relative hydrophobicity would suggest.One possible reason for this apparent anomaly is that the lipophilic 2-methylthiopropyl tail may orientate the molecule in such a way as to maximise absorbed aldicarb onto the dried, lipidic surface. This is further confirmed in Fig. 1(a) as absorbed aldicarb is shown to be the second highest in the entire data set when determined from onion skin (the most lipophilic substrate investigated).From the viewpoint of matrix type, all carbamates are strongly absorbed by all matrices studied. One important observation is that in matrices that can be classified as being lipophilic, such as the crop skin samples, residue partitioning is considerably higher in comparison with matrices of high moisture content (flesh samples). Within this, there is a definite ranking of partitioning in both forms of the crop.It is most noticeable that, for all carbamates, onion skin demonstrates significantly greater partitioning kinetics in comparison with either other skin samples or the entire data set as a whole. Additionally, banana flesh also indicates elevated concentrations of carbamates across the data set. Difficulties in completely drying flesh samples rapidly prior to extraction without incurring excessive thermal damage to the matrix studied can cause errors in determining the true concentration of individual carbamates.Inaccurate carbamate concentrations can also be obtained through the presence of greater inherent moisture in some matrices as a combination of moisture and heat also maintains enzymatic activity within the host matrix thus decreasing the carbamate concentration more Table 1 Supercritical fluid extraction conditions Parameter Level/setting Extraction temperature 70 °C Extraction pressure 300 atm Total extraction time 60 min Restrictor temperature 70 °C Restrictor volume 500 ml Flow rate (gas state)a 860 ml min21 Modifiers used DMSO (10%) Solvent collection Liquid collection in vial (15 ml) Solvent collection temperature 1 °C Extraction cell geometry 14 mm 3 100 mm (16 ml capacity) Cell packing Methanol-cleaned glass wool with Celite wet support a Mean flow rate. 276 Analyst, 1999, 124, 275–280rapidly. Consequently, future work may focus on using freeze drying at this stage.Equally as important, competitive solvation between bound water within the matrix and the extracting fluid causes additional limitations in carbamate quantification from the flesh matrices. Generally, it appears that the aromatic carbamates studied have a greater affinity to lipophilic substrates, again implying that the attraction between the site of partitioning and the carbamate is directed by the hydrophobic property of the carbamate backbone.In comparing partitioning values obtained for ‘Day 10’ and ‘Day 14’, it is apparent, by varying degrees, that the relative concentrations of carbamate have decreased between these times. This implies that the equilibrium between carbamate partitioning and carbamate metabolic degradation has shifted towards carbamate degradation. This indicates that the surface has become saturated with carbamate. As no further partitioning is possible due to both the build-up of sorbed carbamate degradation residues and the continuing action of crop enzymes, this is as might be expected.Multilinear regression modelling of carbamate partitioning The technique of MLR modelling has been applied to the recovery data obtained for carbamate partitioning onto both skin and flesh samples of selected fruit and vegetable types. Similarly to simple regression, MLR is capable of identifying the relevance of particular variables considered in the development of a model.In addition, it is also possible to determine which variables, either controlled or independent, have the greatest effect on the model and which variables can be discarded. Fig. 2 displays a typical plot obtained for carbamate partitioning characteristics against both matrix type and time (for pear samples) illustrating both partitioning and metabolic processes of carbamate build-up and decay. In this context, MLR is used to individually model these concepts with a view to quantifying their respective effect on carbamate persistence in the foodstuffs selected for study.In this study, days of carbamate contact, sample matrix, carbamate type, log Pcarbamate and Sol.H2Ocarbamate predictors were investigated for their respective relations to [carbamate] skin, [carbamate]flesh and their ratio (flesh/skin). Statistical outputs were obtained using MINITAB v10 software (MINITAB, PA, USA) with Table 2 illustrating the relations investigated in the first instance.It became evident from the models obtained that carbamate partitioning/recovery data did not produce models that were statistically suitable to be used for partitioning prediction. It was found that due to low r2 adj values and poor fitting illustrated by low F-values for significance, the models created only partially explained the data set used in their construction. The most relevant models produced involved the investigation of the variables examined in tests 10–13 (Table 3).On consideration of the relative P values obtained, Table 3 also ranks all predictors in order of relevance. From Table 3, it can be concluded that carbamate solubility in water, carbamate log P and carbamate–matrix time of contact are of most importance in surface partitioning of the carbamates to the matrices studied. This suggests that the models developed in tests 10 and 12 are most relevant and that their respective models of both skin and flesh concentration are the most practicable.This conclusion is further confirmed as these tests produced the two highest r2 adj values for all tests conducted. In addition to this, these tests also return the lowest number of unusual observations‡ implying a high level of model fit to the data. ‡ Unusual observations are determined from large standard residuals calculated and compared to a predetermined level in model construction. Fig. 1 Matrix type against time. (a) Aldicarb partitioning; (b) pirimicarb partitioning; (c) bendiocarb partitioning; and (d) carbaryl partitioning.Analyst, 1999, 124, 275–280 277 On cancellation and simplification: t k 1 2 0 693 = . Half-lives were determined by pre-calculation of the rate constant, k12, for the decay and each constant was used in the final determination of t12. Calculated carbamate half-life values in each matrix type are included in Table 4. Carbamate longevity calculation From the carbamate recovery data obtained for the metabolic rate study, exponential decay regression equations were developed for each carbamate from each matrix studied.These expressions were applied in the determination of carbamate longevity in each matrix type. This was completed by setting an arbitrary �esafe level�f of carbamate concentration at 5% of the total incurred carbamate concentration determined from the extraction and quantification of carbamate concentration from �eDay 10�f samples. Using exponential fits for each carbamate concentration plotted as mg g21 of sample against post-sampling time, it was possible, through algebraic manipulation of each expression, to extrapolate the carbamate decay rate to the 5% level and determine the time period required to reach this level.For illustration, only full plots for banana skin and flesh, Fig. 3(a) and (b), respectively, are included. Calculated values for the above are also included in Table 4. Calculated carbamate half-life and longevity values Data obtained for the half-life calculations for each carbamate (Tables 4 and 5) indicate that, in general, carbamate stability is low in each matrix type in comparison with previous insecticide generations, e.g., the organochlorides.�� In matrices that contain relatively high moisture values, e.g., the majority of the matrix flesh samples (t12: 3�]4 d), the carbamate half-lives are seen to be lower in comparison with matrices of low inherent moisture, e.g., banana skin (t12: 4.02�]8.01 d).It would also appear that the sole aliphatic carbamate, aldicarb, is the most stable under these conditions with carbaryl being the least stable. Values obtained for �e5%�f levels (Tables 4 and 5) all indicate that detectable levels of carbamate residue exist in all matrices over periods of up to 1 month, e.g., bendiocarb and carbaryl in �� Kushwaha et al. determined the half-life of aldrin in soil at 49 and 177 d (25 and 35 ��C, respectively).18 Table 4 Observed and calculated recovery data for carbaryl, pirimicarb, bendiocarb and aldicarb from selected fruit and vegetable samples Matrix Portion Carbamate [Day 0]/ mg g21a [Day 14]/ mg g21a Exponential fitb Half-life (t12)/d �e5%�f level/dc Banana Skin Aldicarb 48.47 4.34 y = 46.73e20.1743x 4.0 17.0 Pirimicarb 15.90 2.67 y = 14.99e20.1210x 5.4 24.3 Bendiocarb 80.23 23.92 y = 79.41e20.0867x 8.0 34.7 Carbaryl 339.41 95.76 y = 342.45e20.0909x 7.6 33.0 Flesh Aldicarb 131.47 29.60 y = 131.27e20.1079x 6.5 27.7 Pirimicarb 40.51 5.81 y = 35.25e20.1350x 4.9 21.2 Bendiocarb 191.41 38.99 y = 185.38e20.1108x 6.1 26.7 Carbaryl 680.06 86.55 y = 691.84e20.1439x 4.7 20.9 Potato Skin Aldicarb 45.46 7.01 y = 39.02e20.1308x 5.2 21.7 Pirimicarb 25.47 4.46 y = 21.69e20.1218x 5.7 23.3 Bendiocarb 81.07 5.73 y = 87.95e20.1898x 3.6 16.2 Carbaryl 379.92 31.74 y = 404.51e20.1801x 3.9 26.1 Flesh Aldicarb 20.21 3.91 y = 18.07e20.1173x 5.9 24.6 Pirimicarb 16.12 3.14 y = 14.84e20.1124x 5.9 25.9 Bendiocarb 37.13 5.17 y = 37.53e20.1395x 4.9 21.5 Carbaryl 143.77 16.20 y = 140.14e20.1546x 4.4 19.2 Pear Skin Aldicarb 16.16 2.39 y = 12.93e20.1349x 5.1 20.5 Pirimicarb 31.27 4.09 y = 27.36e20.1406x 4.7 20.3 Bendiocarb 30.29 3.78 y = 25.27e20.1451x 4.6 19.4 Carbaryl 115.50 15.07 y = 114.10e20.1474x 4.7 20.2 Flesh Aldicarb 38.54 4.23 y = 35.77e20.1601x 6.7 18.2 Pirimicarb 17.78 2.91 y = 16.08e20.1273x 5.3 22.7 Bendiocarb 48.42 4.06 y = 49.43e20.1738x 3.9 17.3 Carbaryl 235.35 16.23 y = 250.67e20.1912x 3.6 16.0 Onion Skin Aldicarb 233.65 21.13 y = 246.74e20.1701x 4.0 17.9 Pirimicarb 85.42 8.66 y = 72.90e20.1649x 4.2 17.2 Bendiocarb 328.89 30.04 y = 365.75e20.1716x 4.0 18.1 Carbaryl 964.14 76.91 y = 841.36e20.1734x 3.8 16.5 Flesh Aldicarb 19.52 3.74 y = 16.35e20.1133x 5.8 24.9 Pirimicarb 24.32 4.02 y = 21.28e20.1265x 5.3 22.6 Bendiocarb 20.57 3.46 y = 16.68e20.1210x 5.4 23.0 Carbaryl 124.58 11.47 y = 125.40e20.1689x 4.0 17.7 Apple Skin Aldicarb 22.57 3.03 y = 18.81e20.1412x 4.8 31.1 Pirimicarb 16.22 3.92 y = 14.15e20.1008x 6.8 28.4 Bendiocarb 44.33 5.18 y = 41.19e20.1474x 4.5 19.8 Carbaryl 190.03 16.08 y = 170.11e20.1785x 3.9 16.2 Flesh Aldicarb 77.90 10.38 y = 69.00e20.1382x 4.8 20.8 Pirimicarb 38.54 5.18 y = 32.99e20.1391x 4.8 20.4 Bendiocarb 126.90 12.82 y = 124.35e20.1616x 4.2 18.4 Carbaryl 433.53 50.28 y = 409.19e20.1517x 4.5 19.4 a Carbamate concentrations on the first and final day of sampling, respectively.b Best exponential fit for each data set over the five sampling days. c The time taken for each carbamate residue to decrease to within 5% of the respective total day 0 sample value. Analyst, 1999, 124, 275�]280 279banana skin. As expected, matrices of high moisture content metabolise each carbamate at a greater rate, with typical ‘5%’ level time periods of approximately half that of matrices of lower moisture content, e.g., bendiocarb and carbaryl in pear flesh.Conclusions In this work it has been shown that it is possible to estimate both pesticide half-life and potential longevity in a variety of matrices. The potential of carbamate cross-over into the foodchain has been demonstrated as a function of resident time within the food sample. The success in developing practical regression models using MLR has been limited. This result only strengthens the argument that the application of linear modelling techniques to active biological systems such as matrix partitioning could be inappropriate.This is due mainly to the introduction of large uncertainties from many, normally uncontrolled, non-linear systems active within the experiment. As it may be possible to improve model relevance by increasing experimentation replicates, it would be more prudent to focus this solely on one matrix type. Although no practical models were developed, it was possible to determine which predictors had the greatest effect on model construction.It was found that, of all predictors used, only three were consistently statistically relevant in the 13 tests conducted, namely, log[Sol. H2O] (log P), [Days of contact] and [Carbamate type]. References 1 S. Yoshida, H. Murata and M. Imaida, J. Jpn. Soc. Biosci. Biotech. Agrochem., 1992, 66(6), 1007. 2 A. Noble, J. Chromatogr., 1993, 642, 3. 3 I. A. Stuart, J. Maclachlan and A. McNaughtan, Analyst, 1996, 121, 11R. 4 H. A. Moye, S. J. Scherer and P. A. St. John, Anal. Lett., 1977, 10, 1049. 5 Y. Tsumura, K. Ujita, Y. Nakamura, Y. Tonogai and Y. Ito, J. Food Prot., 1994, 571, 1001. 6 S. Chiron and D. Barcelo, J. Chromatogr., 1993, 645, 125. 7 H. Hiemstra and A. de Kok, J. Chromatogr. A, 1994, 667, 155. 8 R. J. Argauer, K. I. Eller, M. A. Ibrahim and R. T. Brown, J. Agric. Food Chem., 1995, 43, 2774. 9 Y. Tsumura, K. Ujita, Y. Nakamura, Y. Tonogai and Y. Ito, J. Food Prot., 1995, 58, 217. 10 K. M. S. Sundaram and J. Curry, J. Chromatogr. A, 1994, 672, 117. 11 M. S. Ali, J. D. White, R. S. Bakowski, E. T. Phillippo and R. L. Ellis, J. AOAC Int., 1993, 76, 1309. 12 M. S. Ali, J. D. White, R. S. Bakowski, N. K. Stapleton, K. A. Williams, R. C. Johnson, E. T. Phillippo, R. W. Woods and R. L. Ellis, J. AOAC Int., 1993, 76, 907. 13 V. A. Simon, K. S. Pearson and A. Taylar, J. Chromatogr., 1993, 643, 317. 14 H. Frister, H. Meisel and E. Schlimme, Fresenius’ Anal. Chem., 1988, 330, 631. 15 S. S. Simons and J. Johnson, J. Am. Chem. Soc., 1976, 98, 7098. 16 R. T. Krause, J. Chromatogr. Sci., 1978, 16, 281. 17 R. T. Krause, J. Chromatogr., 1979, 185, 615. 18 K. S. Kushwaha, H. C. L. Gupta and V. S. Kavadia, Ann. Arid. Zone, 1978, 17, 200. Paper 8/08013E Fig. 3 Graphical representation of carbamate decay in (a) banana skin and (b) -sampling time. Table 5 Reduced summary statistics of Table 4 calculated values Carbamate Mean halflife/ d RSD (%) Mean ‘5%’ level/d RSD (%) Aldicarb 5.2 0.9 22.4 4.6 Pirimicarb 5.3 0.7 22.6 3.1 Bendiocarb 4.9 1.3 21.5 5.5 Carbaryl 4.5 1.1 20.5 5.3 280 Analyst, 1999, 124, 275–280

ISSN:0003-2654    

DOI:10.1039/a808013e

出版商:RSC    

年代:1999

数据来源: RSC