摘要:

The AnalystThe Analytical Journal of The Royal Society of ChemistryAdvisory Board*Chairman: J. D. R. Thomas (Cardiff, UK)*J. F. Alder (Manchester, UK)D. Betteridge (Sunbury-on-Thames, UK)E. Bishop (Exeter, UK)A. M. Bond (Australia)R. F. Browner (USA)D. T. Burns (Belfast, UK)G. D. Christian (USA)"N. T. Crosby (Teddington, UK)"L. Ebdon (Plymouth, UK)*J. Egan (Cambridge, UK)L. de Galan (The Netherlands)A. G. Fogg (Loughborough, UK)"H. M. Frey (Reading, UK)*C. W. Fuller (Nottingham, UK)T. P. Hadjiioannou (Greece)W. R. Heineman (USA)A. Hulanicki (Poland)I. Karube (Japan)*D. L. Miles (Wallingford, UK)"J. N. Miller (Loughborough, UK)E. J. Newman (Poole, UK)T. B. Pierce (Harwell, UK)E. Pungor (Hungary)J. R6iiCka (USA)"R. M. Smith (Loughborough, UK)W.I. Stephen (Birmingham, UK)M. Stoeppler (Federal Republic of Germany)J. M. Thompson (Birmingham, UK)K. C. Thompson (Sheffield, UK)J. F. Tyson (Loughborough, UK)A. M. Ure (Aberdeen, UK)A. Walsh, K.B. (Australia)J. Wang (USA)G. Werner (German Democratic Republic)T. S. West (Aberdeen, UK)"G. M. Telling (Bedford, UK)*Members of the Board serving on the Analytical Editorial BoardRegional Advisory Editorsf o r advice and help to authors outside the UKProfessor Dr. U. A. Th. Brinkman, Free University of Amsterdam, 1083 de Boelelaan, 1081 HVProfessor Dr. sc. K. Dittrich, Analytisches Zentrum, Sektion Chemie, Karl-Marx-Universitat,Dr. 0. Osibanjo, Department of Chemistry, University of Ibadan, Ibadan, NIGERIA.Dr.G. Rossi, Chemistry Division, Spectroscopy Sector, CEC Joint Research Centre,EURATOM, lspra Establishment, 21020 lspra (Varese), ITALY.Professor K. Saito, Coordination Chemistry Laboratories, Institute for Molecular Science,Myodaiji, Okazaki 444, JAPAN.Professor M. Thompson, Department of Chemistry, University of Toronto, 80 St. GeorgeStreet, Toronto, Ontario M5S I A I , CANADA.Professor P. C. Uden, Department of Chemistry, University of Massachusetts, Amherst,M A 01003, USA.Professor Dr. M. Valcarcel, Departamento de Quimica Analitica, Facultad de Ciencias,Universidad de Cordoba, 14005 Cordoba, SPAIN.Professor Yu Ru-Qin, Department of Chemistry and Chemical Engineering, Hunan University,Changsha, PEOPLES REPUBLIC OF CHINA.Professor Yu.A. Zolotov, Vernadsky Institute of Geochemistry and Analytical Chemistry,USSR Academy of Sciences, Kosygin str., 19, 1 17975, GSP-1, Moscow V-334, USSR.Amsterdam, THE NETHERLANDS.Talstr. 35, DDR-7010 Leipzig, GERMAN DEMOCRATIC REPUBLIC.Editorial Manager, Analytical Journa IsJudith EganEditor, The AnalystJanet DeanAssistant EditorsPaul Delaney, Mandy Mackenzie, Harpal MinhasEditorial Office: The Royal Society of Chemistry, Thomas Graham House, Science Park,Milton Road, Cambridge CB4 4WF. Telephone 0223 420066. Telex No. 818293 ROYAL.Fax 0223 423623.Advertisements: Advertisement Department, The Royal Society of Chemistry, BurlingtonHouse, Piccadilly, London, W I V OBN. Telephone 01-437 8656. Telex No. 268001.The Analyst (ISSN 0003-2654) is published monthly by The Royal Society of Chemistry,Burlington House, London W I V OBN, England.All orders accompanied with payment shouldbe sent directly to The Royal Society of Chemistry, The Distribution Centre, Blackhorse Road,Letchworth, Herts. SG6 1 HN, England. 1989Annual subscription rate UKf200.00, Rest ofworld€230.00, USA$460.00. Purchased with AnalyticalAbstracts UK f432.50, Rest of World €490.00,USA $963.00. Purchased with Analytical Abstracts plus Analytical Proceedings UK f510.00,Rest of World f580.00, USA$I 142.00. Purchased with AnalyficalProceedingsUK€254.00, Reslof World f292.00, USA $584.00. Air freight and mailing in the USA by Publications ExpeditingInc., 200 Meacham Avenue, Elmont, NY 11003.USA Postmaster: Send address changes to: The Analyst, Publications Expediting Inc., 20CMeacham Avenue, Elmont, NY 11003.Second class postage paid at Jamaica, NY 11431. Allother despatches outside the UK by Bulk Airmail within Europe, Accelerated Surface Posloutside Europe. PRINTED IN THE UK.Information for AuthorsFull details of how to submit material forpublication in The Analyst are given in theInstructions to Authors in the January issue.Separate copies are available on request.The Analyst publishes papers on all aspects ofthe theory and practice of analytical chemistry,fundamental and applied, inorganic andorganic, including chemical, physical, biochem-ical, clinical, pharmaceutical, biological,environmental, automatic and computer-basedmethods. Papers on new approaches to existingmethods, new techniques and instrumentation,detectors and sensors, and new areas of appli-cation with due attention to overcoming limita-tions and to underlying principles are all equallywelcome.There is no page charge.The following types of papers will be con-sidered:Full papers, describing original work.Short papers: the criteria regarding origin-ality are the same as for full papers, but shortpapers generally report less extensive investi-gations or are of limited breadth of subjectmatterCommunications, which must be on anurgent matter and be of obvious scientificimportance. Rapidity of publication is enhancedif diagrams are omitted, but tables and formulaecan be included. Communications receive pri-ority and are usually published within 5-8weeks of receipt.They are intended for briefdescriptions of work that has progressed to astage at which it is likely to be valuable toworkers faced with similar problems. A fullerpaper may be offered subsequently, if justifiedby later work.Reviews, which must be a critical evaluationof the existing state of knowledge on a par-ticular facet of analytical chemistry.Every paper (except Communications) will besubmitted to at least two referees, by whoseadvice the Editorial Board of The Analyst will beguided as to its acceptance or rejection. Papersthat are accepted must not be published else-where except by permission. Submission of amanuscript will be regarded as an undertakingthat the same material is not being consideredfor publication by another journal.Regional Advisory Editors.For the benefit ofpotential contributors outside the United King-dom, a Panel of Regional Advisory Editorsexists. Requests for help or advice on anymatter related to the preparation of papers andtheir submission for publication in The Analystcan be sent to the nearest member of the Panel.Currently serving Regional Advisory Editors arelisted in each issue of The Analyst.Manuscripts (three copies typed in double spac-ing) should be addressed to:The Editor, The Analyst,Royal Society of Chemistry,Thomas Graham House,Science Park,Milton Road,CAMBRIDGE CB4 4WF, UKParticular attention should be paid to the use ofstandard methodsof literaturecitation, includingthe journal abbreviations defined in ChemicalAbstracts Service Source Index. Wherever pos-sible, the nomenclature employed should fol-low IUPAC recommendations, and units andsymbols should be those associated with SI.All queries relating to the presentation andsubmission of papers, and any correspondenceregarding accepted papers and proofs, shouldbe directed to the Editor, The Analyst (addressas above). Members of the Analytical EditorialBoard (who may be contacted directly or via theEditorial Office) would welcome comments,suggestions and advice on general policy mat-ters concerning The Analyst.Fifty reprints of each published contribution aresupplied free of charge, and further copies canbe purchased.@ The Royal Society of Chemistry, 1989. Allrights reserved. No part of this publication maybe reproduced, stored in a retrieval system, ortransmitted in any form, or by any means,electronic, mechanical, photographic, record-ing, orotherwise,withouttheprior permissionofthe publishers

ISSN:0003-2654    

DOI:10.1039/AN98914FX021

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALAO 1 14(6) 653-756 (1 989)The AnalystJune 198965366366767 567968368969569970370771 171 571 972372773 1735739743747749753755The Analytical Journal of The Royal Society of ChemistryCONTENTSResearch and Development of Biosensors. A Review-Frieder Scheller, Florian Schubert, Dorothea Pfeiffer, RainerHintsche, Ina Dransfeld, Reinhard Renneberg, Ulla Wollenberger, Klaus Riedel, Mariana Pavlova, Manfred Kuhn,Hans-Georg Muller, Pham minh Tan, Werner Hoffmann, Werner MoritzOxygen-sensitive Reagent Matrices for the Development of Optical Fibre Chemical Transducers-Phillip Y. F. Li,Ramaier NarayanaswamyIsotopic Determination of Selenium in Biological Materials With Inductively Coupled Plasma Mass Spectrometry-BillT. G.Ting, Christine S. Mooers, Morteza JanghorbaniInductively Coupled Plasma Mass Spectrometric Determination of the Absorption of Iron in Normal Women-Paul G.Whittaker, Tom Lind, John G. Williams, Alan L. GraySecondary Ion Mass Spectrometric Determination of Impurities in Aluminium Oxide-Hisashi Morikawa, YoshinoriUwamino, Toshio lshizukaReconstruction of Constituent Spectra for Individual Samples Through Principal Component Analysis of Near-infraredSpectra-Ian A. Cowe, James W. McNicol, D. Clifford CuthbertsonLinearity Range of Gran Plots from Logarithmic Diagrams-Carl7 MaccaPlatinum-dispersed Nafion Film Modified Glassy Carbon as an Electrocatalytic Surface for an Amperometric GlucoseEnzyme ElectrodeHari Gunasingham, Chee Beng TanLiquid Chromatographic Determination of Tetracycline Antibiotics at an Electrochemically Pre-treated Glassy CarbonElectrodeWeiying Hou, Erkang WangWater-promoted Formation of Phenylboronates of 1,S-Diols During Gas - Liquid Chromatographic Analysis:Application t o the Assay of Meprobamate-R.J. Flanagan, M. W. J. ChanDetermination of Trace Amounts of Molybdate in Soil by Ion Chromatography-Harish C. Mehra, William R.Frankenberger, Jr.Continuous-flow Chemiluminescence Determination of Acetaminophen by Reduction of Cerium(lV)-loanna I. Koukli,Antony C. Calokerinos, Themistoceles P. HadjiioannouFlow Injection Spectrophotometric Determination of Selenium Based on the Catalysed Reduction of Toluidine Blue inthe Presence of Sulphide Ion-Carmen Martinez-Lozano, Tomas Perez-Ruiz, Virginia Tomas, Concepcion AbellanSpectrophotometric Determination of Oxytetracycline in Pharmaceutical Preparations Using Sodium Tungstate asAnalytical Reagent-Milena Jelikic-Stankov, Dragan VeselinovikSpectroscopic Investigation of the Equilibria of the Ionic Forms of Sinapic Acid-Bogdan Smyk, Regina DrabentTitrations in Non-aqueous Media.Part XVIII. Observation of the Bentidine Rearrangement Reaction in A c e t o n i t r i l eTurgut Gunduz, Esma KiliG, Adnan Kenar, S. Gul OztaSSHORT PAPERSSeparation and Determination of Inorganic Anions by High-performance Liquid Chromatography Using a MicellarThiamine - Reineckate Liquid Membrane Electrode for the Selective Determination of Thiamine (Vitamin B,) inComplementary Study on the Use of the Potassium Reinecke's Salt as a Chemical Actinometer-Jerzy Szychlinski,Photometric and Fluorimetric Methods for the Determination of Bromate in Bread-F. Garcia Sanchez, A. Navas Diaz,Mobile Phase-Biswanath Maiti, Arvind P. Walvekar, T. S. KrishnamoorthyPharmaceutical Preparations-Saad S. M. Hassan, Eman ElnemmaPiotr Bi I ski, Kazi m i erz Ma rtu szewski, Je rzy Blaiej owskiM. Santiago NavasCOMMU NlCATlONRapid Determination of Nitrite in Water by Flow Injection With Chemiluminescence Detection-Alexander R. Thornton,BOOK REVIEWSERRATUMCUMULATIVE AUTHOR INDEXJosef Pfab, Robert C. MasseyTypeset and printed by Black Bear Press Limited, Cambridge, Englan

ISSN:0003-2654    

DOI:10.1039/AN98914BX023

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST. JUNE 1989, VOL. 114 653 Research and Development of Biosensors* A Review Frieder Scheller, Florian Schubert, Dorothea Pfeiffer, Rainer Hintsche, Ina Dransfeld, Reinhard Renneberg, Ulla Wollen berger, Klaus Riedel, Mariana Pavlova, Manfred Ku hn and Hans-Georg Muller Central Institute of Molecular Biology, Academy of Sciences of the GDR, 1 175 Berlin-Buch, GDR Pham minh Tan and Werner Hoffmann Central Institute of Nuclear Research, Academy of Sciences of the GDR, 8101 Dresden-Rossendort GDR Werner Moritz Humboldt University Berlin, Section of Chemistry, 1040 Berlin, GDR Summary of Contents Introduction Types of biosensor State of development of test strips and enzyme electrodes Dry reagent tests Generations of biosensors Substrates measured with biosensors Analytical parameters of enzyme electrodes Application in the clinical laboratory Fermentation Co m mercia I isat ion M u It i -f u n cti o na I sensing Mi n i at u risat ion Inducible recognition systems G roup-effect sensing Receptrodes I m mu nosensors Research fields Conclusions References Keywords: Bissensors; research and development; review Introduction Selectivity is a problem of paramount importance in analytical chemistry and it is gencrally achieved by means of selective c h e ni i c a 1 reactions .Extra() r di r i a r i 1 y se 1 e c t i ve and versa t i 1 e reagents are provided by nature in the form of enzymes, anti b odi e s an d rece p t o rs . Enzyme technology has exploited the use of biological components in analytical systems by the development of i m m o bi 1 i se d enzymes an d i m m unworn pone n t s which may be re-used.The invention o f enzyme test strips by Free and Free1 has opened the way for simplification of the analytical procedure: the sample is merely dropped on to the test strip, thus avoiding the need to dilute or take aliquots of the sample and reconstitution of the reagents. However, test strips are one-way materials and the reagents may not be re-used after enzyme immobilisation. Another breakthrough in the applica- tion of immobilised enzymes to analytical chemistry was made by Clark and Lyons’ in 1962. Based on the principle of their membrane-covered oxygen electrode, an enzyme solution was entrapped by a semi-permeable membrane placed in front of the indicator electrode. ’This step provided a reusable enzyme and hence represented the birth of the first biosensor.B i o s e n s o rs a re devices u’ h i c h use i ni m o b i 1 is e d b i o m o 1 e c u 1 e s combined with transducers to detect or respond to specific interactions with environmental chemicals.’ In order t o achieve optimal signal transfer, the immobiliscd biocom- ponent is in close physical contact with the transducer unit (Fig. 1). Types of Biosensor Basically there are two types of biosensor differing in the mode of signal generation (Fig. 1). Direct bioaffinity sensors utilise a binding event to detect substances. The binding of the incoming analyte to the complementary immobilised bio- ligand results in a change in the conformation of the bioniolecule and/or physical changes in the imrnobilisation medium such as changes in charge, thickness, temperature or optical parameters (colour or fluorescence), By analogy with Sample Receiver Trans- Electronics ducer Invited lccturc prcccnted at the International Bio Symposium.Nagoqa XX, Nagoba, Japan. 1&13 March. 1988. V Biosensor Fig. 1. ii biosensor Schematic representation of the composition and function of654 ANALYST, JUNE 1989, VOL. 114 affinitb chromatography, well known complementary pairs of analyte and bioligand are used for molecular recognition in bioaffinity sensors (Table 1). The second basic sensor type is the enzymatic’metabolic biosensor. Here recognition of the substrate by the immobilised receiver (enzymes on different levels of integration, i.e., purified enzymes, organelles, micro-organisms or tissue slices) is followed immediately by chemical conversion to the corresponding product, which is detected.A combination of both principles, binding of the analyte without its chemical conversion and signal generation by con\ erting an auxiliary substance, is realised in “catalytic” biosensors for the determination of prosthetic groups and inhibitors, and in enzyme immunosensors. State of Development of Test Strips and Enzyme Electrodes Dry Reagent Tests Providing a reagent i n dry form for rapid use is by no means novel; litmus paper, which dates back to the 19th century, is an example. The first major impact of dry chemistry on clinical testing was the introduction in the 1950s of the Ames Clinistix4 for the indication of urinary glucose.The further development of dry reagent chemistry has been the cumulation of several technologies; typically, a reagent zone consisting of several different layers is fixed above a reflective zone constructed with pigments such as Ti02 or BaS04 on a support layer. For removing corpuscular particles, e.g., red blood cells, the sample is dropped on to a separation layer at the top of the test strip. The chemical reactions are monitored typically by diffuse reflectance, whereas measurements of light trans- mission or fluorescence are less frequent occurrences. With the first approach, any light not absorbed by the products in the reaction layer is reflected to the detecting system from the underlying reflective zone. The analyte concentration is derived relative to known (reflectance) standards with a non-linear concentration dependence.As the enzyme layer and the optical transducer are separate entities, test strips cannot be classed as biosensors. However, they may be considered to form the basis of optoelectronic biosensors (optrodes). Dry reagent test systems have been described for serum metabolites, enzymes, serum electrolytes and therapeutic drugs.4 Dry reagent assays for nearly all commonly measured blood metabolites. namely glucose, blood urea nitrogen, uric acid. cholesterol, triglycerides, creatinine and bilirubin, are available commercially. In their highly complex, layered reaction systems. conversion of the analyte either releases hydrogen peroxide or causes a change in pH. Subsequently, a co 1 o u r - p r o du c i n g pe rox id as e -ca t a 1 y se d reaction or the a bso r - bance change of a pH-indicator is evaluated.Sophisticated dry reagents are also available for the determination of serum enzymes such as creatine kinase, lactate dehydrogenase, transaminases, y-glutamyltransferase, Table 1 . Types of biosensor Analytc Substrate analogues . . Haptens ” . ’ . . I Antigens Viruses . . . . . . Glycoproteins . . . . . . Prostheticgroups . . . . Inhibitors . . . . . . Enzyme-labelled antigens Cofactors . . . . . . Substrates , . . . . . . . . . Cells . . . . . . : : 1 . . . . Bioaffinity ligand Enzymes Antibodies Lectins Apoenzynies Enzymes Antibodies Enzymes Organelles Microbes &-amylase and alkaline phosphatase. Ion-selective mem- branes, combined with either potentiometric or spectrometric detection, are used for the determination of K+, Na+, C1- and CO? in serum.The K+ electrode consists of a base support and an Ag - AgCl layer, which is covered with a hydrophobic membrane containing the ionophore valinomycin. The sample is deposited on the ion-selective layer and the potential difference is measured versus an identical reference electrode system loaded with a KCI solution of known concentration. Therapeutic drugs are accessible by the competitive protein binding assay based on substrate-labelled fluorescence immu- noassays. Using this principlc, theophylline can be deter- mined in the concentration range 1-5 pg ml-1. Generations of Biosensors Whereas dry chemistry research focuses more on commercial applications, it is the phenomenological and fundamental aspects that are under intensive investigation in the field of biosensors.From an integrational point of view, biosensors may be subdivided into three generations.5 The simplest approach is based on membrane-entrapped or membrane- bound biocomponents (first generation). Clark and Lyons’ used a highly concentrated enzyme solution held in a reaction chamber by a membrane placed in front of the slightly recessed indicator electrode and thus created the enzyme electrode. This represents the unit consisting of dialyser, reactor and electrochemical transducer and has led to con- siderable simplification of the experimental set-up compared with that of conventional analysers. The direct adsorptive or covalent fixation of biocomponents at the transducer surface leads to the elimination of inactive membrane layers.Addi- tional covalent coupling of the cosubstrateh is a pre-condition of reagentless measuring regimes (second generation). In this variant no reagent has to be added to the sample. This approach has been continued by immobilising the cofactor7 or electron relays8 in the neighbourhood of the active centre of the enzyme. A more recent principle of sensor integration is the combination of molecular recognition and signal process- ing at the biocomponent level.9 This results in “high quality” chemical information at the transducer surface. Finally, immobilisation of the biocomponent directly on an electronic element (third generation), e.g., on the gate of a field-effect transistor (FET), which directly senses and amplifies changcs in surface properties, integrates biospecific recognition and clectronic signal processing. 10,’ This permits differential elimination of disturbances and statistical evaluation with multi-gate sensors.Various patents and scientific publications describe biosen- sors of the first and second generation for the determination of about 120 different parameters including substrates, cofac- tors, prosthetic groups, hormones, antigens and enzyme activities. Among biosensors, biospecific electrodes are at the forefront with respect to the number of measurable sub- stances, the body of published data and the degree of commercialisation. Signal-generating step Sensor type Binding of the analyte Binding sensor (Pure) i Conversion o f an auxiliary substance I Catalytic biosensor Chemical conversion Mctabolicienzymatic of the analyte sensor IANALYST, JUNE 1989.VOL. 114 65 5 Substrates Measured With Biosensors Amino acids. D-Alanine, I -arginine, L-asparagine, id-aspar- tic acid, L-cysteine, L-glutamine, L-glutamate, glutathione, J -histidine, I - and D-leucine, L-lysinc, L- and D-mcthionine, N-acetylmethionine, L- and D-phenylalanine, sarcosine, serine, L-tyrosine, L-tryptophan and D-valine. Gases. NH3, H?, CH4, SO2, NO, 0 2 , CO and CO?. Cofactors. Adenosine monophosphate (AMP), adenosine triphosphate (ATP), pyridoxalphosphate (PLP), pyridine nucleotides (NADPH), flavin adenine dinucleotide (FAD), H202. thiamine pyrophosphate (TPP) and phosphoenolpyru- vate (PEP).Carbohydrates. Amygdaline, lactose, galactose, maltose, glucose, sucrose, glucose 6-phosphate, starch and fructose. Amines, amides and heterocyclic compounds. Adenosine, aminopyrine, aniline, aromatic amines, acetylcholine, chol- ine , phosphatidylcholine, creatinine, creatine , guanidine, guanosine, penicillin, spermine, uric acid, urea, xanthine and hypoxanthine . Organic acids or their salts. Acetate, formatc, malate, gluconate, glyoxylate, D-isocitrate, J,- and D-lactate, maleic acid, glycolate, nitrilotriacetic acid, oxalate, oxaloacetate, pyruvate and succinate. Alcohols and phenols. Bilirubin, catechol, cholesterol, cholesterol esters, ethanol, glycerol, glycerol esters, methanol, phenol and 2,4-dinitrophenol. Inorganic ion&. Fluoride, nitrite, nitrate, phosphate, sul- phate, sulphide, sulphite, Hg'+, Zn2+ and Cu2+.Enzymes and proteins. Thyroxine, human albumin, pro- insulin, immunoglobulin G, a-fetoprotein, human choriogon- adotropin, peroxidase, a-amylase, glucoamylase, cholineste- rase, creatine kinase, pyruvate kinase, lactate dehydrogenase, transaminases and pullulanase. Miscelluneous. Antibiotics, blood groups, assimilable sub- stances, volatile substances, biochemical oxygen demand, meat freshness, mutagenicity , vitamins, peptides, t heophyl- line and testosterone. Analytical Parameters of Enzyme Electrodes Linear calibration graphs of enzyme electrode based analysers usually extend over 2-3 decades of concentration, with a detection limit of 10-100 PM. This sensitivity is sufficiently high to determine the previously mentioned compounds, using 10-100-pl sample volumes.The precision of the measurements is reflected by coefficients of variation of about 1%. The most evident advantage of enzyme electrodes is their negligible enzyme consumption; it is typically in the order of 10-3 U per sample. At present, the sample throughput of enzyme electrode based analysers is limited to a maximum of about 100 samples h-1. For example, the application of a glucose electrode containing the enzyme in a polyurethane layer in a stirred measuring cell permits a sampling frequency of 120 samples h-1.12 In a flow injection (FI) system for which the measuring time at 1% carry-over is only 12s, a throughput of 300 samples h-1 is possible. The response is linear from the detection limit of 10 VM up to 100 mM glucose.The relative standard deviation for 25 successive injections of 1 mM glucose solution is 0.5%. This highly effective FI system has also been extended to the measurement of lactatel3; 200 lactate samples h-1 can be measured with good precision and negligible carry-over. Such a high sample throughput is rcalised even with double-membrane type sensors. Therefore, application of biosensors of thesecond or third generation will speed up the measuring process further. Actually, amperometric enzyme microelectrodes for glucose and enzyme FETs for urea characterised by response times for the steady-state signal of only 3-4 s have been developed by Japanese workers. 11-14-15 The combination of these biosensors with optimised FI devices should permit the determination of 700samples h-1.avalue recentlyreported for anFIsystemusing soluble enzyme reagents. 16 Both for measuring undilcted samples and for continuous monitoring of biologically active substances (e.g., in the blood-stream, fermenters or wastewater), the measuring range of biosensors has to be adapted to the respective analyte concentration. In blood or fermentation broths the concentra- tion of metabolites generally exceeds the upper limit of the calibration graph of enzyme electrodes (Fig. 2). For continu- ous in vivo monitoring of metabolites such as glucose or lactate, the linear measurement region has to be extended to higher concentrations; this can be done by decreasing the diffusion-controlled substrate flow to the enzyme layerl7.18 or by exploiting mediator chemically modified electrodes.19-23 On the other hand, the concentration of many hormones, drugs or toxins is several orders of magnitude below the limit of detection of enzyme electrodes. Immunoelectrodes permit the specific determination of these substances in the picomolar range. However, so far, the applicability of many immunosen- sors is restricted by their extremely long measuring time. If substrate concentrations in the nanomolar range are to be detected, the sensitivity of enzyme electrodes can be enhanced using substrate amplification by two enzymes.' Operational conditions have to be adjusted in such a way that one enzyme catalyses regeneration of the substrate of the second enzyme (Fig. 3). This has been achieved by coupling dehpdrogenases with oxidases or transaminascs, or by coupling kinases.For a lactate sensor, using lactate oxidase (LOD) and lactate dehydrogenase (LDH) immobilised in poly(viny1 chloride), Mizutani et ~ 1 . 2 4 obtained a maximum amplification factor of 250 for the steady-state current. Using a very thin enzyme layer of co-immobilised 1,OD and LDH, a much higher amplification, up to 4100-fold, is possible.25 Theoretical considerations demonstrate that the maximum amplification obtained in the latter system is a realistic value for the parameters of the enzyme membrane used, namely, the characteristic diffusion time and the first-order rate constant. However, application of these recycling systems to "real" samples is restricted by their sensitivity to both substrates of the enzymatic cycle, in this instance lactate and pyruvate.For example, the determination of lactate in plasma using this method requires the endogenous pyruvate to be removed first. By combining glucose oxidase (GOD) and glucose dehydro- Sensitivity Immuno- Substrate Enzyme Mediator chemically electrodes modified electrodes electrodes recycling 10-12 10-9 10-6 10-3 100 mol 1 - 1 Blood Hormones, drugs, enzymes Substrates Fermenter Wastewater Toxins Metabolites Substrates "BOD" Fig. 2. concentrations Measuring range of biusensors compared with typical analyte Glucose Lacto n e ikonic acid NAD+ NADH Fig. 3. conversion for signal amplification in biosensors Diagram of two-enzymc cycles for the increase in cosubstratc656 ANALYST, JUNE 19x9, VOL.114 genase the sensitivity of a glucose sensor has been increased by a factor of 10.26 This enzyme system seems more promising for, in physiological media, the second cubstrate, gluconolac- tone, is not present; therefore, the enhanced signal is only related to the glucose concentration. For signal amplification in the determination of the co- factors adenosine triphosphate (ATP) and adenosine di- phosphate (ADY), hexokinase (HK) and pyruvate kinase (PK) were co-immobilised with LDH and lactate mono-oxy- genase(LM0).’7 The last two were used to transduce pyruvate formation sequentially into oxygen consumption. In the presence of an excess of glucose and phosphoenolpyruvate, the nucleoside phosphate is shuttled between the kinases. Pyruvate is formed in much higher amounts than the amount of cofactor in the enzyme layer.The linear range of the sensor extends up to 6 VM ADP with a 200-fold increase in sensitivity compared with the unamplified signal in the absence of glucose. By combining the amplification system for ADP with that for pyruvate using LOD and LDH, “double amplification” for ADP should bc possible. In this system the excess of pyruvatc formcd in the kinase cycle enters the second cycle, where it is amplified further. Recently a patent was taken out for an amplification scheme differing from those discussed so far in that the recycling is effected by only one enzyme? LDH normally catalyse5 the establishment of the reaction equilibrium between pyruvate and nicotinamide adenine dinucleotide (NADH) to give lactate and the oxidised cofactor.On the other hand, in the presence of NAD+, LDH catalyses the oxidation of thc unusual substrate glyoxylate to oxalate with concomitant formation of NADH (Fig. 4). If both reactions are conducted simultaneously, i . r . , in the presence of pyruvate, glyoxylate and limiting amounts of NADH, the cofactor is shuttled between them. Thus a 170-fold increase in lactate is formed over the reaction without glyoxylate, leading to a detection limit for NADH of 50nM. This type of “monoenzymatic recycling” might facilitate the design of novel and simple enzymatic amplification systems. Application in the Clinical Laboratory The measurement of glucose dominates the application of enzyme electrodes. Significant problems in the “prc-analytics” of blood glucose determination are the complete inhibition of the glucose-consuming glycolytic reactions and the glucose contained in erythrocytes.The latter problem has been neglected by several manufacturers of analysers prescribing the use of undiluted blood samples. The Gluco 20 A analyser (Fuji, Tokyo, Japan), using 20 p1 of undiluted whole blood, gi\es good correlation with the hexokinase method only for serum. The glucose values measured in whole blood are 13% too low.”’ The same tendency has been reported for the Auto Stat GA-1100 (Daiichi, Kyoto, Japan) and for the analyser from Yellow Springs Instruments, Yellow Springs, OH, USA.-IO A comparison between undiluted blood and 1 + 10 Glyoxylate NADH Pyruvate 0 2 Fig. 4. (LDH) - glyoxylate - pyruvate system Principle of cofactor recycling in the lactate dehydrogcnase pre-diluted samples has been conducted using the Glukometcr GKM manual analyscr (ZWG, Berlin, GDR).31 With direct injection of undiluted blood the glucose values are 18.8% lower than those obtained for pre-diluted samples.This difference shows that the transport of glucosc from the red blood cells into the bulk solution is incomplete during the response time (4s). In contrast, the results obtained with pre-diluted samples agree well with those of the GOD - peroxidase method. Recently the problem of glucose dcgrada- tion during sample storage and transport has been solved. Neither storage at 4°C nor addition of the aldolase inhibitor fluoride resulted in acceptable stability. However, blood dilution (1 + 50) in a hypotonic buffer gave excellent stability and also good correlation with established methods.31 The problems that have been discussed also present serious challenges to the application of test strips.It is well known that the values measured with aqueous standard solutions, serum and blood differ substantially. Therefore the procedure for calibration of the strips for blood glucose measurement is laborious. The first biosensor of the second generation has been made commercially available by Genetics International, Abingdon, U K 3 This blood glucose sensor is based on the ferroccne- mediated glucose oxidation developed by groups i n Cranfield and Oxford. The device consists of a pen-sized instrument coupled with a small, disposable strip electrode. The patient has to apply a drop of fresh blood to the sensor, push a single button and wait for the rcsult, which appears after 30s.Comparison with an automatic analyser yields the regression line y = 1 .04x + 4.9 mg dl-1. The scrial precision for capillary or venous blood is reflected by coefficients of variation of 8.1% at 53.9mgdl- 1 and 3.9% at 85.8mg d1-I. Obviously these values do not compete with those of laboratory analysers. By comparison with alternative devices for measur- ing blood glucose at home the sensor has the following advantages: (i) a response time of only 30 s; (ii) a much simpler operating procedure; and (iii) small size and portability. The sensor will allow diabetics to control their diet, insulin injection and intakc o f oral antidiabetic drugs.An important parameter for the diagnosis of kidney disease and for the control of artificial kidney dialysis is the concentration of urea in serum. For the determination of urea, ureasc-bascd potentiometric sensors are well established. Further, an amperometric urea sensor based on the pH- dependent anodic oxidation of hydrazine has been des- cribed.33 The advantages of this method are a linear calibra- tion graph, in contrast to the logarithmic response of potentiometric sensors, excellent reproducibility, a sample throughput o f 40 samples h-1 and a measurement range of 1-80 mmol - 1 in the sample. Combined with the Glukometer this amperometric urea sensor has been applied successfully to urea monitoring in dialysis patients. Such process control results in a considerable decrease in cost and a reduction in the patient’s discomfort.For the sequential determination of lactate and pyruvate, an enzyme electrode with co-immobilised LDH and lactate mono-oxygenase has been developed.34 Owing to optimum enzyme loading and the equality of the diffusion coefficients of the two substrates, this sensor is characterised by having identical sensitivities towards lactate and pyruvate. Therefore it is suitable for the sequential determination of lactate and pyruvate, the ratio of which is important diagnostically. The same electrode has also been used to measure glutamate pyruvatc transaminase activity in blood serum and pyruvate kinase activity in red blood cells.35 Direct measurement of cholinesterase activity in serum samples is possible using the Glukometer.The rate of formation of thiocholine from butyrylthiocholine iodide is evaluated electroehemically3~ and thc esterase activity is indicated only 20s after injection of the sample. Also, inhibitors of this enzyme are easily measured: 40 s afterANALYST. JUNE 1989, VOL. 114 6.57 addition of the respective compound to the pooled serum sample of a given cholinesterase activity the degree of inhibition is reflected by the remaining activity. This inherent kinetic measuring principle is an important advantage over other methods, e.g., the use of test strips, where a given time regime has to be followed very closely. By analogy with the optimisation of exercise control in sportsmen, the fitness of race horses has been checked by measuring the blood lactate level .37 The lactose content in the urine of cows is of economic importance.During the acute phase of the udder disease mastitis the lactose concentration exceeds 10 mM38 (Fig. 5 ) . Therefore, typically, lactose concentrations are monitored to screen for this disease. Fermentation Biotechnologists already take advantage of process control in setting the conditions for fermentations: temperature, stirrer speed and pH are controlled by closed-loop circuits. In addition to these parameters the concentrations of several nutrients, e.g., carbohydrates and amino acids, and, most importantly, those of the final products have to be controlled. The cost of growth media is extremely high, particularly for cell cultures. Therefore the consumption of the key nutrients, such as glucose.glutamine and other amino acids, has to be monitored in order to reconstitute the medium by feeding the appropriate substances. To control nutrient consumption and lactate formation in a cell fermenter producing human growth hormone (hGH) the Glukometer analyser has been used either in conjunction with an enzyme electrode for glucose or lactate, respectively, or with a microbial sensor (Fig. 6). The increase in the number of cells is paralleled by a decrease in the nutrient concentration as reflected by the signals of the 7 I - 5 15 E E , W 0 ,+ g 10 - .I- C 0 .- Y 2 5 c C a, c 0 0 0 5 10 Ti rneid Fig. 5. Variation of lactose concentration in cows‘ urinc with time. A. Healthy con; B, cow with acute udder disease; and C.recovery of normal lactose levcl after mastitis 50 r I - - E E 1 $ 30 c m 0 a, - L s - = 10 c3 1.5 I 0 1.0 ; 7 -. - a, Y- 0.5 6 z 5 4 r I - 3 m E I , 2 z 1 0 1 2 3 4 5 Timeid Fig. 6. Transient curve o f the concentration of nutrient (glucose) .ind of metabolite (lactate) measured by enzyme electrodes and of human growth hoimone (hGH) produced by fermentation during the cultiLation cycle o f human cells glucose electrode and the microbial sensor. On the other hand, lactate in the medium is observed to increase. Whereas the determination of high relative molecular mass fermentation products (e.g. , hGH) using biosensors is not yet possible, such instruments have been developed for measuring the concentration of lysine, ethanol, gluconic acid and penicillin in diluted samples of the fermentation broth.?Y The number of reports in the field of in Jitu fermenter control is relatively small owing to the following peculiarities: (i) sterilisation of the fermenter denatures the biocomponent of the sensor; (ii) elevated temperatures during operation decrease functional stability; and (iii) the analyte concentra- tions exceed the dynamic range of the sensor.Nevertheless, the in situ application of enzyme electrodes has been studied for the measurement of penicillin, glucose and ethanol. An autoclavable enzyme electrode for penicillin was de- veloped by Enfors and Nilsson.4(j After the heat-sterilisation process a 13-lactamase solution is pumped into the reaction chamber placed in front of a flat glass electrode. This arrangement also allows the enzyme to be exchanged during the fermentation process; hence an extended sensor lifetime is achieved.In order to eliminate disturbances caused by a drift in pH in the fermenter, an enzyme-free pH electrode is combined with the enzyme electrode in the differential mode. The requirements for operation in a suitable concentration range are partly fulfilled. Measurement in the fermentation broth yields a linear response up to 3 0 m ~ , whereas in penicillin, fermentation concentrations of up to SO mM are produced. The concentration of carbohydrates, such as glucose, in the fermentation broth generally exceeds the measuring range of “normal” enzyme electrodes. Cleland and Enfors31 extended the linear response of a GOD-based sensor by electrolytic generation of oxygen in the en7yme layer and internal dilution of the sample inside the sensor housing.Thus the linear response has been extended up to 0.5 M glucose and the electrode can be used in completely anaerobic media. Using an oxygen-independent ferrocene-modified GOD electrode Brooks et a1.42 have been able to extend the measuring range to 70 mM glucose. With continuous use of the electrode its half-life is about 6 d. To study the short-term kinetics of alcoholic fermentation by aerobic yeast an alcohol oxidase based electrode was developed by Verduyn et al.43 The dissolved oxygen tension in the fermenter was kept between 9.5 and 100% by vigorous stirring and aeration. Fluctuations in this value were compensated for by an enzyme-free reference oxygen electrode.In the future, continuous measurement of biomacromolecules by immu- nosensors will offer an attractive alternative to the existing discontinuous methods for the detection of gene products such as interferons, insulin and monoclonal antibodies. Commercialisation Dry chemistry has made great inroads in clinical laboratories. Instrumentation ranges in size from hand-held to large free-standing devices. Examples of hand-held instruments are the Glucometer reflectance photometer (Ames Division, Miles, Elkhart, I N , USA) and the Reflocheck (Boehringer Mannheim, Mannheim, FRG). Bench-top manual instrumen- tation is available for handling more than one analyte, e.g., the Seralyzer system from Ames Division with a specific module for each test. Here the analysis time ranges from 30 s to 5 min.Fully automated analysis is possible with the Ektachem 400 (Eastman Kodak, Rochester, NY, USA), which is capable of 500 tests h-1 and has ten different assays available at present. The operator only has to provide the sample and select the desired tests. Because biosensors perform nondestructive measurements on a “real-time” basis their application results in almost revolutionary advantages of cost-saving automation and data658 ANALYST, JUNE 1989, VOL. 114 Table 2. Enzyme electrode based analysers Company Model Yellow Springs Instruments, Yellow Springs, OH. USA , . 23A 23L 27 Zentrum fur Wissenschaftlichcn Geratebau (ZWG). Berlin, GDR Glukometer Fuji Electric, Tokyo. Japan . . Gluco 20 UA-300A Daiichi, Kyoto, Japan . . . . Auto Stat Radelkis.Budapest, Hungary . . OP-GL-7110s Ferment, Vilnius, USSR . . . . ExAn La Roche, Bask, Switzerland . . LA640 Omron Tateisi, Kyoto, Japan . . HER-100 Seres. Aix-en-Provence, France . . Enzymat GA-1120 Tacusscl. Lyon, France . . . . Glucoprocesscur Priifgerate-Werk (PGW), Medingen, GDR (Eppendorf, FRG) . . ADM 300 ECA 20 (ESAT 6660) Analyte Glucosc Lactate Ethanol Lactose Galactose Sucrose Glucose Uric acid Glucose a- Amylase Uric acid Glucose Glucose Glucose Lactate Lactate Glucose Choline L-Lysine D-Lactate Glucose Glucose Glucose Lactate Uric acid 1 Range of measurement/ mmollkl 1-45 (L15 M0 0-5s 0.5-50 0.1-1.2 0-27 1 4 0 1.7-20 2.5-30 0.5-12 0-8.3 0.3-22 1.0-29 0.1-2 0.5-20 0.05-5 1- 100 0 . M 0 1-30 0.1-1.2 Sample throughput/ samples h-1 40 40 20 20 60-90 40 80-90 30 61)-120 40 20 2C30 60 60 60 60 90 80 120-130 120 80 50-60 Coefficient of variation, % Stability <2 300 samples <2 2 1.5 > 1000 samples 2 10 d 1.7 >500 samples 4-5 3 1 5- 10 240 d 5 <S 40 d <5 10 d <2 >2000 samples <2 >2000 samples 4 .5 10 d <2 14 d <2 10 d Table 3. Analytical characteristics of routinely used biospecific electrodes Analyte Lactate . . . . Pyruvate + lactate Glucosc . . Lactose . . Maltose . . Sucrose . . Glutamate Uric acid Urea . . Phosphate I . . . . . . . . . . . . . . . . . . . . . . . , . . . . . . . . . . . Lactate dehydrogenase Pyruvate kinase . . Crcatinc kinase . . Acetylcholine esterase Range Sample throughput/ 1 Enzyme* mmoll- . . LMO 1-40 pmoll-' s-l samples h-1 60 - . . LDH-LMO &7 .. GOD 0.4-44 . . . . . . . . . . . . . . . . . . GOD- [3-gal. 1-150 GOD - GA 1-50 GOD - MR - 1N 1-44 GLOD 0.04-40 Uricase 0.1-1.2 Urcase 0.8-50 GOD - AcP 2-24 LMO - LDH - LMO - PK - LDH - LMO - - 20 - 120 - - 100 60 40 40 40 - - - - - 40 12 1-20 15 - 1-14 15 5-80 10 20-400 40 Alanine aminopcptidase . . - - 0.1-1 - CV, % 1 2 2 1.5 2 1 .5 1 .5 1.6 1 2 2 3 - 2 - Lifetime/ d 55 55 10 20 14 5 10 10 15 - 55 55 14 21 - Correlation r = 0.998 ( n = 30, plasma) y = -0.1 + 1 . 0 3 ~ - y = -0.015 + 1.003X r = 0.99 ( n = 196. blood) y = 0.32 + 1.047~ r = 0.990 ( n = 100, serum) y = 0.149 -f- 1.105 + (0.958 k 0.009)~ r = 0.995 ( n = 20, urine) - - - - 1' = 0.0198 + 1 9 . 8 ~ Y = 0.995 ( n = 30) y = 0.125 + 0.9912~ Y = 0.997 ( n = 67, dialysate) y=0.29+ 1 . 1 1 ~ r = 0.999 ( n = 30, serum) y = 0.002 + 1.003~ Y = 0.984 ( n = 20) = 1.70 + 0.986~ r = 0.993 ( n = 205, serum) y = 0.064 + 0.973~ r = 0.994 (n = 32, serum) - - glucose oxidase; GLOD = glutamate * LMO = lactate mono-oxygenase; LDH = lactate dehydrogcnase; PK = pyruvate kinase; GOD - oxidasc; (3-gal.= (3-galactosidase; GA = glucoamylase; MR = mutarotase; IN = invertasc; and AcP = acid phosphatase (potato slice). handling. In particular, monitoring of therapeutic drug levels, office testing and implantable devices will benefit from the development of biosensors. A recent study44 estimated the 1986 US market for biosensors at $70 million and predicted that US sales would rise to $500 million by 1990 with major applications in the clinical laboratory and to drug monitoring.Further, it was concluded that biosensors would become the dominant sensor technology by the end of the century. Candidates for wide clinical use include sensors for glucose (artificial pancreas), urea (haemodialysis) and, perhaps, creatinine and cholesterol. To date, self-contained enzyme electrode based devices for the determination of 11 different analytes including glucose, sucrose, lactate, uric acid, lysine and a-amylase have been made commercially available (TableANALYST. JUNE 1989. VOL. 114 659 2), starting in 1976 with the La Roche lactate analyser (Basle, Switzerland) and the YSI Model 23A (Yellow Springs Instruments). These analysers use the enzyme immobilised in the membrane and as such represent the first generation of sensors. In 1981 development of the Glukometer analyser was completed.Currently about 300 such devices are being used in clinical laboratories for whole blood glucose measurement. The analyser has also been applied to the determination of ten other substrates and six enzyme activities (Tablc 3). Thc enzymc membranes used contain up t o three interacting enzymes. Thcy arc incorporated in gclatin, which generally provides ex ce 11 en t functional \ t a bi I it y ,J5 or i n synthetic car r i c: r material such as polyurethane. These enzyme membranes ensure high stability and the short diffusion time allows a sampling rate of 120 samples h-1. Based on thc routine application of thc Glukomcter and results of recent flow injection expcrimentslzT13 the ECA 20 Enzyme Chemical Analyzer (PG W, Medingen, GDR) was developed.I t uscs an optimiscd flow cell and a GOD - polyurethane membrane. Operation and maintenance are extremely straightforward; a built-in microcompiiter conducts the measuring process including calibration and printing of thc measured values. The largest immediate markct for biosensors is for clinical a n d he a I t h care a p pl i cat ions . I m port ant eco n o m i c adv ant ages , however, may be gained by exploiting novel sensors in process engineering. Owing to the cxtrcmely high cost of growth media for human cell cultivation, process control is in high demand. Control Equipment (Lowell, MA, USA) uses the Yellow Springs glucosc sensor in a flow injection device for process control in cell cultivators. The sample is injected automatically into thc carrier strcam aftcr filtration.Analo- gou\ process analysers have been developed by Nippon General Trading (Tokyo, Japan) for the measurement of cight diffcrcnt metabolitcs and by Scrcs for the parallcl determina- tion o f glucose, lactate and glutamine.4" Research Fields In order to achicvc broad acccptance in routine analytical operations the following fields of sensor re\earch are under investigation on a world-widc scale. Multi-functional Sensing Until now, cnzymc electrode based analysers have been one-parameter deviccs. However, it would be desirablc to mea\ure several different substances in parallel using a single analyscr, without changing thc electrode mcmbranc. One such approach involves the simultaneous application of different enzyme sensors to the same sample solution.Mascini47 installed three enzyme electrodes (for glucosc, lactate and pyruvate) in a flow line of an artificial pancrea\. Hence the levels of these important metabolitcs in the blood of critical-care patients could be controlled simultaneously during trcatment. An alternativc approach towards simultaneous multi- parameter determinations is the integration of several independently working enzymes in one membrane, c'.g., glutamate, tyrosinc and lysinc oxidases used for the simul- taneous determination of glutamate, tyrosine and lysine, respectively. By sweeping the potential bctween -600 and +600 mV, a current - potential graph with three distinct steps is obtaincd; these reflect oxygen consumption by tyrosinase, hydroquinone formation, which indicate5 the concentration of glutamate, and the formation of hexacyanoferrate(II), which is related to the concentration of lysinc.All three analyte concentrations can be quantificd by comparing the three sIgnals.Js This example demonstrates that the measuring procedure can be simplified significantly as the determination of x \ c r a l parameters is possible without changing any part o f the enzyme sensor. Alternatively, different enzymes, e.8.. for glucosc, urea and lipids, have bccn tixed at thc sensitive regions of a multiple sensors chip. I Thus multi-functional sensing has bccn rcalised using fabrication technologies for the mass production of in i croc 1 cc t ro n i c c I e m e n t s . Further, several substances are accessible using only one biosensor and coupled enzyme reactions.Glucose 6-phos- phate (G6P), glucose, NADP+, ATP and fructose have all been quantified using a hexokinase - glucose 6-phmphate - dehydrogenase (HK - G6P - DH) sequence sensor.+) Both glucose and fructose are converted by HK to the respective monosaccharide phosphates and G6P i\ subsequently de- hydrogenated by G6P - DH in the presence of NADP+ to produce NADPH. Therefore the NADPH oxidation current at a modified electrode reflects the rate of glucosc conversion. At a constant glucose concentration, the addition of fructose results in a decreased NADPH signal hith the current decreasing linearly up to a fructose concentration of 1 .0 mht. The 5equential reaction of HK and G6P - DH is the basis of the response of the sensor to glucose, NADP+ and ATP and.by changing the experimental conditions, five different sub- stances can thus be measured using two enzymes (Fig. 7). Miniaturisation Thc goals of miniaturising biosensors arc reduction of the required sample volumc, multi-component analysis ot corn plex chemical substances by mu1 tiple microsensors and cost reduction by mas\ production. Three basic types of niicrocen- sor have bccn used so far: (i) ion-sensitive field effect transistors (ISFE'Ts); (ii) gas-sensitive metal oxide \emicon- ductor (MOS) capacitors; and (iii) thin-film electrodes. These sensors arc now fabricated by microelectronic production technologies, which allow the production of low-cost, repro- ducible, small-scale electronic devices. Such technologies, C J .~ . , thick- and thin-film deposition, photolithography and chemical and plasma etching, permit well delineated pattern- ing of metallic, insulating and 4emiconducting surface layers. Hcncc arrays of identical or different electrochemical sensors can be produced, enhancing the reliability, repeatability and versatility of the \ensor. Most problcms ari\e from the manufacturing techniquc used for the formation of the i m m o bil i se d b i c) m c m b rane . Using en7yme cntrapment in a polyurethane layer, enzyme FETs for glucosc and urea have been developed. Thc basic scnsor is a pH-sensitive FET possessing a multiple Si3N4 gate 4tructin-e. These microbiosensors are charactericed by a typical responsc time ot 1 min for the steady-statc valuc, a lifetime of about 30 d and a calibration graph slope (in 0.1 [TIM phosphate buffer) of 30 mV pcr concentration decade (Fig.8). In order to avoid interferences due to changes in thc sample buffer capacity, a tluoridc-sensitive FET based on LaFi ha\ been used. This FET ir covered by an enzyme layer containing x c / 0 1 2 3 4 Con cent rat ionlm M Fig. 7. Concentration dependencies of the cofactors NADP+ and ATP and of the substrates glucose 6-phosphate and glucose using the hevokinase - glucose 6-phosphate - dehydrogenase sequence and N-methylphcnazinium in combination with an oxygen electrode. A, NADP+; B. glucose: C. glucose 6-phosphate; D, ATP; and E. fructose660 ANALYST, JUNE 1989, VOL. 114 4 0 1 1 -80 :::I \ 100 c I - 1 2 o t , , , I 10 5 10 10 3 10-2 c,,,,lmol I Fig.Is. Concentration dependencies for a Si3N4 pH-xn\itive field effect transistor covered with glucose oxidase and urease, respectively peroxidase (POD) and GOD. In the presence of the POD substrate pentafluorophenol, glucose can be determined at levels of between 0.05 and 1.0 mM with a sensitivity of 42 mV per decade. For the microbiosensor to realise its potential, it is necessary to develop an enzyme-layer fabrication method which meets the following requirements: (i) the membrane must be deposited precisely on to the sensitive region of the basic electronic element; (ii) the deposited layer must not peel off in use; and (iii) the enzyme covering should be compatible with the integrated circuit (IC) process. Two alternative procedures have been proposed. Naka- moto et ~ 1 .~ 0 masked the passive region of the wafer with a positive photoresist. The enzyme molecules are fixed in the uncovered area by typical immobilisation methods, e.g., a combination of aminopropyl-silanisation of the surface and enzyme adhesion by glutaraldehyde. The enzyme layer has a typical thickness of 1 pm and the minimum width and line separation are both 1.5 pm. I n contrast, Shiono et ul. l 5 deposited the enzymes only at the sensitive region, c’.g., at the gate of ISFETs, from an enzyme solution in a negative photoresist. Thus ISFET- biosensors for glucose, urea and lipids with a response time of 3 s and a sensitivity of 85% after 2000 measurement\ have been developed. When this enzyme FET is in a flow-through cell, which ensures both isolation and electrical connection of the conductive fields to the measuring device, it can be used without any polymeric encapsulation or wire bonding. This technique promises to reduce the cost per sensor.Inducible Recognition Systems Typically. the optimum operating conditions are specific for each enzyme; hence, with multi-enzyme sensors, a compro- mise is necessary which deviates from the standard opera- tional parameters of a simple enzyme sensor. Therefore it seems obvious to use organelles, whole cells or tissue sections from animal or plant sources as biocatalytic packages. These structural entities contain all the necessary components in an environment optimised by evolution. Almost 60 biosensors of this type have been reported51 but the possible number is considerably higher in view of the thousands of strains and tissues available. Microbial sensors offer several advantages over conven- tional enzyme electrodes, e.g., the enzyme preparation step is eliminated, the lifetime is increased compared with isolated enzymes and processes requiring enzyme sequences or cofac- tors are easy to achieve.However, one disadvantage o f microbial sensors is their low specificity in comparison with enzyme electrodes. This is especially true for glucose (the main nutrient for microbes in fermentations), which interferes with the determination of other substances by microbial sensors. To improve the selectivity of microbial sensors undesired metabolic pathways and transport mechanisms might be blocked or inhibited whereas appropriate metabolic activities might be induced.The latter was demonstrated with a microbial glutamic acid sensor using Bacillus suhtilis.52 When a sample solution containing glutamic acid or glucose was injected into the measuring cell, the substrate was taken up by the micro-organisms. The respiration rate therefore increased and oxygen consumption by B. suhtilis resulted in a decrease in the dissolved oxygen signal; the electrode current decreased with time until a steady state was reached. It is not possible to determine glutamic acid in the presence of glucose with this sensor. A sensor using B. ,subtilis cells grown in a medium without glucose shows a higher, although not sufficient, specificity for glutamic acid over glucose. A decrease in the glucose signal, without changing the glutamic acid signal, is obtained by treatment of R..suhtilis cells with a relatively low concentration of chloromercuriben- zoate (CMB) for 20 min. The effect of CMB is irreversible. A further reduction in the glucose signal is obtained using NaF, which is an inhibitor of the enzyme enolase. However, the action of NaF at pH6.8 is reversible and, therefore, the solutions being measured must always contain NaF. The decrease in the glucose signal by CMB is probably due to the uptake of glucose being blocked. The concept described here for the determination of glutamic acid opens up possibilities for the development of microbial sensors with higher specificities. Further, to improve the sensitivity of mierobia! sensors, special cell systems, responsible for transport and assimilation of the substance of interest, can be induced by adding the respective substrate to the growth medium.For example, the rate of assimilation of maltose or aspartame by B. subtilis has been increased substantially by inducing the. respective enzyme systems.53 The “induced bacteria” are used for the determina- tion of a-amylase activity or the dipeplide sweetener aspar- tame. By analogy, the sensitivities of microbial sensors using Nocardiu erythropolis~4 or Hansenula anornula55 for choles- terol and lactatc, respectively, have been improved consider- ably when the respective enzymes, cholesterol oxidase and cytochrome bl, have been induced. By using tolerant plant structures that possess induced enzyme systems, herbicides and pesticides might be measured.56 Group-effect Sensing As already outlined, microbial sensors suffer from the multi-receptor behaviour of intact cells, which results in decreased substrate specificity.On the other hand, this ability to recognise a group of substances is exploited in sensors for complex variables such as the sum of biodegradable com- pounds. This method of determining the biological oxygen demand (BOD) has provided an impetus to environmental control.4 The conventional BOD test takes 5 d and, therefore, it is unsuitable for process control. Rapid BOD determina- tions are only possible using microbial sensors, in which i mm o bil ised m i c ro- org a n i s in s and di sso 1 ve d ox y ge n e 1 e c t r ode s are combined. Different microbial BOD sensors using acti- vated sludges and cells of Trichosporon cuturzeiim, Hansenulu atlomala, Clostridium butyricum, P.seudomorias sp.and Es- cherichiu coli have been developed. 574)0 These sensors have response times of 15-20 min. A rapid-response BOD sensor consisting of B. subtilis or T. cutaneum cells immobilised in poly(viny1 alcohol) and based on the measurement of the acceleration of respiration or of the change i n current, i.e., on a kinetic measurement regime, has also been developed.6’ rhe change in current is linearly related to a glucose - glutamic acid standard in the range 0-22mg 1 1 of BOD using B. suhtilis and in the range 0-100 mgl-1 of BOD using T. cutanrum. The limits of detection are 2 and 4 mg I-’ of BODANALYST. JUNE 1989, VOL. 114 Trans- ducer Membrane Micro-orgar Dialysis membrane iisms Gas- permeable mem brane Substrates Volatile compound Fig.9. Diagram of microbial sensors used for the determination of ( [ I ) volatile substances and ( h ) the biological oxvgen demand where the separation of substances is governed m a h y by the type of membrane, whereas the rnicrobes generate an almost unspccified oxygen consumption with the B. suhtilis and T. cutuneum sensors, respectively. The signal is reproducible to within 5% for a series of ten standard solutions containing 22 mg 1- 1 of BOD. The main advantage of these senscm is the short rcsponsc times (15-30 s). Other types of microbial sensor (Fig. 9) can indicate the carbohy- drate content, the mutagenicity of substances, the presence of volatile compounds, over-all toxicity and vitamin action.jT51JJ2 Receptrodes The chemical senses of living organisms provide exceptional selectivity, sensitivity and response dynamics.For example, certain insects respond to just a few molecules of stimulant and several marine animals can sense amino acids in water at below picomolar concentrations. As electrochemical processes are involved in the neuronal signals of the chemical senses, it is possible to apply the principle of chemoreception t o a further type of biosensor, the receptrode. Two com- plementary directions are currently being pursued, these being the development of sensors using isolated receptors and sensors based on complex chemoreceptor structures. The purified form of the nicotinic acetylcholine receptor from the electric organ of Torpedo culiforniu has been used in an acetylcholine sensor.63 'Thc molccule consists of five sub-units containing both the agonist binding sites and the ion channels.In the absence of acetylcholine the channels are closed whereas on binding of the agonist the channels open and a flux of sodium ions is started. A purified receptor preparation and the lipid lecithin have been deposited on thc gate of a metal insulator semiconductor FET (MISFET). The Si3N4 layer of the gate is sensitive to protons, sodium ions, potassium ions and surface charge and the receptor-covered MISFET exhibits a specific response to acetylcholine and sodium ions. Obviously, the lipid membrane prevents sodium ions in the solution from reaching the Si3N4 gate and, therefore, a concentration gradient is formed.The ion flux through the receptor channel is reflected by the change in the gate potential on the addition of acetylcholine. This effect is only found with the active receptor preparation whereas other proteins, ~ . g . , bovine serum albumin or the heat-denatured receptor, are ineffective. This principle may be extended t o other receptors or channels. Because isolated receptors appear to retain their function only for a very short time such investigations arc far from yielding actual biosensors. On the other hand, the isolation, reconstitution and immobilisation of receptors are eliminated when intact receptor structures are used as the recognition elements of the biosensor. The antennules of blue crabs were excised and the neural fibres coupled to micropipette elec- trodes in such a manner that the impulses generated in the recognition process could be rccorded and accumulated .OJ 66 1 This arrangement was adapted from typical neurophysio- logical techniques and it is useful for detecting extremely low concentrations of amino acids, down to ~ O - I ~ M , with surprising selectivity. Immunosensors The bulk molecular structure of an antibody is fairly constant. Local differences, however, result in specific antigen binding and, consequently, sensor? embodying a wide range of antibodies could use a common transduction mechanism to detect very different analyte3.Immunological interactions of the antibody - antigen or coin pl e men t - mediated type do not usual I y in vol ve e 1 e c t ro - active reactants.Hence enzyme labelling"SJj() or ionophores'b have been used to mediate between immunological and electrochemical processes in the biosensor. Enzyme immuno- sensors are prepared by attaching antibodies to the surface of a Clark oxygen electrode. An enzyme such as catalase or GOD is attached either to a second antibody (two-site test) or to an antigen (competition configuration). The assay involves either the attachment of the enzyme to, or its displacement from, the electrode in the presence of the antigen. After washing, a substrate is added causing a change in current which is either directly or inversely related to the antigen concentration. Sensors of this type have been described for insulin and albumin,"7 immunoglobulin,^^ thyroxine,hq human chorionic gonadotropin,70 cv-fetoprotein71 and the hepatitis B surface antigen.72.73 Because the operation of these enzyme immunosensors necessarily involves the use of reagents admixed with the analyte they cannot be seen as stages in the development of "probe" immunosensors.Nevertheless the detection of sub- nanomolar concentrations of antigens within scconds is a realistic goal. Currently an extremely wide range of pure monoclonal antibodies can be produced each of which is to the first approximation selective for an individual analyte (antigen). Monoclonal antibodies are especially attractive as molecular recognition elements for antigen monitoring. Bush and Kcchnit~7~ devcloped a completely reversible biosensor, using monoclonal antibodies so that a potentiometric signal was produced as a result of the antigen - antibody interaction.In this sensor, monoclonal antibodies to the hapten 2,4-dinitro- phenol (DNP) are entrapped in front of the sensing tip of a DNP-responsive membrane electrode by a collagen mem- brane. This membrane has a molecular cut-off that retains the antibodies but allows the passage of low relative molecular mass antigens. Hence, the entire antibody - antigen reaction takes place in the chamber in front of the sensor tip. Competition between binding of the entrapped antibody to either the constant level of antigen at the surface of the censing membrane or to the changing samplc antigen concentration results in variations in the tranc-membrane potential. The dynamic rcsponse of the probe (the steady state is achieved in approximately 15min) is limited by the slow antibody - antigen reaction.Dissociation of the complex, achieved by immersing the sensor in DNP-free buffer between measure- ments, restores the base-line potential within 15-20 min. After more than 50 cycles of binding and dissociation there is found t o be no loss of sensor response, the lifetime of the sensor being about 17 d. Hence the sensor is a reusable probe requiring only small amounts of the expensive monoclonal antibody. It seems likely that this concept will be extended to other small haptens. Conclusions So far the practical application of biosensors has been limited almost completely t o oxidase-based amperometric enzyme electrodes. However, in dry chemistry, hydrolytic and coupled enzyme reactions have also been widely used and hence it seems plausible that the coupling of enzyme reaction4 could662 ANALYST, JUNE 1989, VOL.114 also lead to an extension of applicability and result in improved analytical performance of the biosensor. Important results might be obtained using chemically and genetically modified enzymes and, further, the “site to site” oriented fixation of artificially coupled enzymes could improve the characteristics of sequential substrate conversion. In this way the cffectiveness of evolution-optimised natural systems could be surpassed. Also the biosensor field might be expanded by the adaptation of complex biological recognition elements. As electrochemical processes arc involved in the neuronal signals of the chemical senses, it is possible to apply chemoreception to recognition, processing and signal trans- duction in biosensors.Potentially, chemical signals related to olfaction or taste might be quantified using these receptrodes. Finally, the gap between metabolic and binding biosensors might be closed by the design of enzymatically active antibodies. 1. 2. 3. 4. 5 . 6. 7. 8. 9. 1 0 . 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. References Frec, A. H., and Frec, H. M., Gastroenterologia, 1953,24,414. Clark, L. C., and Lyons, C., Ann. N . Y . Acad. Sci., 1962,102, 29. Aizawa, M., in Seyama, T., Fueki, K., Shiokawa, J., and Suzuki, S . , Editors, “Proceedings of the International Meeting of Chemical Sensors, Fukuoka, 1983,” Elsevier, Amsterdam, 1983, p.683. Walters, B., Anal. Chem., 1983, 55, 498A. Schcller, F., Schubert, F., Renneberg, R., Miillcr, H.-G., Janchen, M., and Weise, H., Biosensors, 1985, 1. 135. Blacdel, W. J., and Jcnkins, R. A., Anal. Chem., 1976, 48, 1240. Torstcnsson, A., Johansson, G., Hansson, M.-O., Larsson, P., and Mosbach, K., Anal. Lett., 1980, 13, 837. Hcller, A., and Degani. Y . , J. Electrochem. Soc. Rev. News, 1987. 134, 494C. Schcller, F., Schubert, F., Pfciffer, D., and Remeberg, R., Enzyme Eng., 1987, 8, 240. Janata, J., 1. Am. Chem. Soc., 1975, 97, 2914. Karubc, I., Sci. Technol. Jpn., 1986, July/Scptember, 32. Olsson, B., Lundback, H., Johansson, G., Scheller, F., and Ncntwig, J . , Anal. Chcm., 1986, 58, 1046. Schellcr, F., Schubert, F., Olsson, B., Gorton, L., and Johansson. G., Anal.Lett., 1986, 19, 1691. Ikariyama, Y., Yamauchi, S . , Yukiashi, T., and Ushioda, H., Anal. Lett., 1987, 20, 1791. Shiono, S . , Hanazato, Y., Nakako, M., and Macda, M., GBF Monogr., 1987, 10, 291. Mottola, H. A,, Analyst, 1987, 112. 729. Mullen, W. H., Churchousc, S. J., and Vadgama, P. M., Analyst, 1985, 110, 925. Scheller, F.. and Pfeiffcr, D., Z. Chem., 1978, 18, 50. Ccnas, N . K . , and Kulys, J. J., Bioelectrochem. Bioenerg., 1981, 8, 103. Frcw, J. E., and Hill, H. A. O., Philos. Trans. R. Soc. London, Ser. B, 1987, 316, 95. Albery, W. J., Bartlett, P., and Cass, A. E. G., Philos. Trans. R. Soc. London, Ser. B , 1987, 316, 107. D’Costa, E. J., Higgins, 1. J . , and Turner, A. P. F., Biosensors, 1986, 2, 71. Ikeda, T., Miki, K., Fushimi, F., and Senda, M., Agric.Biol. Chem., 1987, 51, 747. Mizutani, F., Yamanaka, T., Tanabe, Y . , and Tsuda, K., Anal. Chim. Acta. 1985, 177, 153. Schellcr. F . , Wollenberger, U., Schubcrt, F., Pfeiffcr, D., and Bogdanovskaya, V. A., GBF Monogr., 1987, 10, 39. Schubcrt. F., Kirstein, D., Schroder, K.-L., and Scheller, F., Anal. Chim. Acta, 1985, 169, 391. Wollenberger, U., Schubcrt, F., Schcller, F., Danielsson, B., and Mosbach, K., Anal. Lett., 1987, 20, 657. Schubcrt. F., and Schcller, F., DDR Pal., WPC 12Q/315 7391, 1988. Niwa, M., Itoh, K . , Nagata, A., andOsawa, H., Tokai.1. Exp. Clin. Med., 1981, 6, 403. Mason, M., paper presented at Biotec 87, Dusseldorf, 1987. Hanke, G., Schellcr, F., and Ycrsin, A.. Zentralbl. Pharm., 1987, 126, 445.32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. McCann, J., World Biotech. Rep., 1987, 1, Part 2, 41. Kirstein, D., Kirstein, L., and Scheller, F., Biosensors, 1985,1, 117. Wcigelt, D., Schubert, F., and Scheller, F., Fresenius Z. Anal. Chem., 1987, 328, 259. Weigelt, D., Schubert, F., and Scheller, F., Anal. Lett., 1988, 21, 225. Gruss, R., and Scheller, F., 2. Med. Lab.-Diagn., 1987, 28, 333. Turner, A. P. F., papcr presented at the Ciba Foundation Meeting, “Biosensors,” London, 1986. Pfeiffer, D., Ralis, E. V., Makower, A., Meiske, C., and Scheller, F., J. Chem. Technol. Biotechnol., 1989, 43, in the press. Schellcr, F., Stud. Biophys., 1987, 119, 221. Enfors, S.-O., and Nilsson, H. J., Enzyme Microh. Technol., 1979, 1, 260. Cleland, V., and Enfors, S.-O., Anal. Chem., 1984, 56, 1880. Brooks, S . , Ashby, R., Turner, A. P. F., Calder, M., and Clarke, D. J., Bimensors, 1987/88, 3,45. Vcrduyn, C., Zomcrdijk, T., van Dijkcn, J . , and Scheffcrs, W., Appl. Microhiol. Biotechnol., 1984, 19, 181. “Biosensors: Today’s Technology, Tomorrow’s PrQducts,” Technical Insights Incorporated, Lee, NJ, USA. Bertermann, K . , Scheller, F., Pfciffer, D., Janchen, M., and Lutter, J., Z. Med. Lab.-Diagn., 1981, 22, 83. Romctte, J. L., GBF Monogr., 1987, 10, 81. Mascini, M., GBF Monogr., 1987. 10, 87. Pfciffer, D.. Wollenberger, U . , Scheller, F., Risinger, L., and Johansson, G., Stud. Biophys., in the press. Schubert, F., Kirstcin, D., Scheller, F., Abraham, M., and Boross, L., Anal. Lett., 1986, 19, 2155. Nakamoto, S . , Kimura, J., and Kuriyama. T., GBF Monogr., 1987, 10, 289. Karubc, I . , and Suzuki, S., Ion-Sel. Electrode Rev., 1984,6,15. Riedel, K., and Scheller, F., Analyst, 1987, 112, 341. Renneberg, R., Riedel, K . , and Schellcr, F., Appl. Microbiol. Biotechnol, 1985, 21, 180. Wollenberger, U . , Schellcr, F., and Atrat, P., Anal. Lett., 1980, 13. 825. VinckC, B. J . . Devleeschouwer, M. J., and Patriarche, G. J., Anal. Lett., 1985. 18, 593. Rechnitz, G. A., GBFMonogr., 1987, 10, 3. Karubc, I., Matsunaga, T., Mitsuda, S . , and Suzuki, S . , Biotechnol. Bioeng., 1977, 19, 1535. Hikuma, M., Suzuki, H., Yasuda, T., Karube, I., and Suzuki, S., Eur. J . Appl. Microbiol. Biotechnol., 1979, 8, 289. Kulys, J. J., and Kadziauskiene, K.-V., Biotechnol. Bioeng., 1980, 22, 221. Strand. S. E., and Carlson, D. A., J. Water Pollut. Control Fed., 1984, 56, 464. Ricdel, K., Renneberg, R . , Kleine, R., Kriiger, M., and Scheller, F., Appl. Microbiol. Biotechnol., 1988,28, 316. Ricdel, K., Renneberg, R., Wollenberger, U., Kaiser, G., and Scheller, F., J . Chem. Technol. Biotechnol., 1989, 44, 85. Gotoh, M., Tamiya, E., Momoi, M., Kagawa, Y., and Karube, I., Anal. Lett., 1987, 20, 857. Belli, S. L., and Rechnitz, G. A . , Anal. Lett., 1986, 19, 403. Aizawa, M., Philos. Trans. R. Soc. London, Ser. B, 1987,316, 121. Aizawa, M., GBFMonogr., 1987, 10, 217. Mattiasson, B., and Nilsson, H., FEBS Lett., 1977, 78, 251. Aizawa, M., Morioka, A., and Suzuki, S . , J. Membrane Sci., 1978, 4, 221. North, J., Trends Biochem. Sci., 1985, 3, 180. Robinson, G. A., Cole, V. M., Rattle, S. J., and Frost, G. C., Biosensors , 1986, 2,45. Aizawa, M., Morioka, A., and Suzuki, S . , Anal. Chim. Actu, 1980, 115, 61. Boiticux, J. L., Desmet, G., and Thomas, D., Clin. Chem., 1979, 25, 318. Boitieux, J. L., Thomas, D., and Desmet, G., Anal. Chim. Acta, 1984, 163, 309. Bush, D. L . , and Rechnitz, G. A., Anal. Lett., 1987,20, 1781. Paper 8104772C Received December 2nd, I988 Accepted December I6th, I988

ISSN:0003-2654    

DOI:10.1039/AN9891400653

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST, JUNE 1989, VOL. 114 663 Oxygen-sensitive Reagent Matrices for the Development of Optical Fibre Chemical Transducers Philip Y. F. Li and Ramaier Narayanaswamy" Department of Instrumentation and Analytical Science, University of Manchester Institute of Science and Technology, P.O. Box 88, Manchester M60 IQD, UK A number of chemically-sensitive media were investigated for their suitability in the development of an oxygen-sensitive optical fibre transducer. The reagent phase is the heart of the transducer, where oxygen is measured by collisional quenching of the immobilised fluorescent indicator by oxygen. Fluorescence quenching is a rapid and reversible process and can be appropriately incorporated in an optical fibre oxygen transducer. The analytical system employed for measuring the fluorescence response to oxygen is described and the underlying reasons behind the selection process for the most suitable reagent matrix are discussed.Keywords: Chemically-sensitive reagent phase; oxygen measurement; optical fibre sensors; fluorescence quenching; coumarin dyes A suitable reagent phase for the direct measurement of oxygen levels is needed for the development of an oxygen- sensitive optical fibre transducer. The reagent phase should ideally be sensitive to fluorescence quenching by oxygen and have a high quantum yield such that the fluorescence emission is sufficiently intense to provide a measurable signal. Optical sensors have been developed for the measurement or determination of a wide variety of chemical variables or species, such as pH,' metal ions2 and glucose,3 and also for the measurement of oxygen in blood4 and as bioreactors.5 Optical fibres are used for sensor fabrication because they are small in physical size, flexible, low in cost and easy to fabricate, which would enable these sensors to be considered as disposable.The optical nature of transducers offers immunity from electrical interference and their passiveness resulting from their electrical isolation means that they are inherently safe. Remote measurements in hazardous environments are pos- sible owing to their rugged construction and reliability. These advantages can be incorporated in an optical fibre oxygen transducer through a suitable method of oxygen detection. The principle of oxygen measurement is based on the efficient oxygen quenching of the fluorescence from a large number of fluorescent organic indicators.This phenomenon was used here for the development of a suitable reagent phase which can be incorporated on an optical fibre for the detection of oxygen. The reagent phase consists of a polymeric support matrix on which an oxygen-sensitive fluorescent indicator is physically adsorbed. The three fluorescent indicators selected for the studies were the coumarins 1, 102 and 153 and this choice was based on their availability, sensitivity and stability, respectively. These indicators were immobilised on organic polymeric support matrices (XAD resins) and on an inorganic adsorbent (silica gel). The important areas of investigation were the interaction of the immobilised indicators with oxygen and the subsequent evaluation of their performance characteristics. In this paper the results of initial studies on the use of oxygen-sensitive media for the development of an optical fibre oxygen transducer are reported. Fluorescence quenching is a photochemical process in which a chemical quenching species can interact with a fluorophore by decreasing its fluorescence intensity. The quenching process involves a collisional encounter between the fluorescent indicator and the oxygen molecule.The degree of collisional quenching of a fluorophore is expressed as a * To whom correspondence should be addressed. u plotter Luminescence spectrometer blender Inlet Flow cell Fig. 1. Schematic diagram of the analytical system linear dependence on the oxygen level present by the Stern - Volmer equation6 /(JI= 1 + kt[Q] = 1 + K[Q] .. . . ( I ) where Zo and I are the fluorescence intensities in the absence and presencc of the quencher (Q), respectively, k is the bimolecular quenching constant, t is the lifetime of the fluorophore in the absence of the quencher, [Q] is the concentration of the quencher and K = kt. Experimental Instrumentation A schematic diagram of the analytical system used is shown in Fig. 1. The Perkin-Elmer Model LS-5 luminexence spec- trometer was equipped with a xenon discharge lamp, pulsed at a line frequency of 50 Hz, as the excitation source. The source produces a band of energy with a width at halt the peak intensity of less than 10 p. The incident monochromatic light enters through the silica glass window and irradiates the reagent phase contained inside a powder sample holder which is mounted on the front surface accessory.The powder $ample holder was modified into a flow cell which allowed the controlled passage of gases. The fluorescence signal was measured at the etnission wavelength (Table 1) by the detector system. When operating in the fluorescence mode, a second gating period occurs shortly before the next pulse of light. The second emission signal is subtracted from the first to correct for any contribu- tion from dark current and from any phosphorescence with a lifetime greater than 20 ms. Fluorescence spectra were recorded on a Hitachi Model 057 X - Y recorder.664 Gaseous oxygen b Needle standard ANALYST, JUNE 1989, VOL.114 m Table 1. Excitation and emission wavelength maxima of the various re agent phases anal yse d Indicator Coumarin 1 Coumarin 102 Coumarin 153 A,, i A,,,, I A,, I A,,,,l A,, i A, I Matrix nm nm nm nm nm nm XAD-4 . . . . 373 405 385 422 405 460 XAD-8 . . . . 375 415 385 422 415 468 Silicagel . . 335 434 281 336 310 392 Reagents The fluorescent indicators coumarin 102 and coumarin 153 were purchased from A.G. Electro-optics. Coumarin 1 was supplied by Aldrich. The coumarin indicators are a group of widely used laser dyes which fluoresce in the blue - green region of the spectrum. Coumarin dyes are derived from coumarin by substitution with either an amino or a hydroxy group in the 7-position.' The inorganic support matrix used was silica gel and thc organic polymeric support matrices employed were Amberli te XAD-4 and XAD-8 resins.Whereas Amberlite XAD-4 is a styrene - divinylbenzene copolymer matrix and is hydrophobic in nature, XAD-8 and silica gel are hydrophilic adsorbents; XAD-8 is composed of a polymethacrylate copolymer. Preparation of the XAD resins for use involved washing the resins consecutively with dilute acid, water and methanol followed by drying. The inorganic adsorbent was used as supplied. The 5 x 10-5 M indicator solutions were prepared by dissolving the appropriate mass of coumarin dye in methanol (HPLC grade). Immobilisation Procedure The method of immobilisation involved adding 5.0-ml aliquots of a 5 X 10-5 M solution of the coumarin derivatives in methanol to 1 .0 g of the polymer support matrices.A period of up to 16 h was allowed for the indicator to be adsorbed on the support matrices. Excess of the indicators was removed and the samples were washed thoroughly with doubly distilled water before being dried. Gas Blending System The oxygen gas standards used for calibrating the oxygen responses from the reagent phase were generated by a gas blender (Fig. 2). Oxygen and the inert diluent gas nitrogen were supplied from cylinders. The gas standards were produced by controlling the ratio of the flow-rates of the two gases entering the mixing chamber. The gas mixture was divided into two streams, one stream passing into a Munday reference cell oxygen analyser (Servomex, Model OA500) and the second into the flow cell containing a sample of the reagent phase.500 r 0 50 100 Oxygen, O/O Fig. 3. Fluorescence response of the reagent phases to oxygen in the 0-100% range. A, Coumarin 1 - XAD-8; B, coumarin 102 - XAD-8; C , coumarin 153 - XAD-8; D, coumarin 102 - XAD-4; E, coumarin 1 - XAD-4; F, coumarin 153 - XAD-4; G, coumarin 1 - silica gel; H, coumarin 102 - silica gel; and I, coumarin 153 - silica gel Sample Analyses Samples of the reagent phase were introduced into the powder holder in amounts sufficient to cover the fused-silica window of the powder holder. The powdered sample was sandwiched between the window and the screw-cap with the gas inlet and outlet tubing. The gas inlet receives the gas mixture supplied from the gas blender. The sample holder was mounted on the front surface accessory of the luminescence spectrometer.Fluorescence from the sample was measured at different oxygen levels. The excitation wavelength maximum of the adsorbed indicator was isolated in order to irradiate the sample. The fluorescence thus produced was measured at the emission wavelength maximum. Results and Discussion The responses of the immobilised indicators to oxygen levels, introduced in increments of approximately lo%, were obtained by measuring the fluorescence intensities at the emission maxima (Fig. 3 and Table 1). A calibration graph can be used to relate the degree of fluorescence quenching to the oxygen level present. The pore sizes and surface areas of the support matrices are the important physical factors affecting the performance of the reagent phase.# The results in Fig.3 demonstrate that theANALYST, JUNE 1989. VOL. 114 665 0 50 100 Oxygen, O/O Fig. 4. Stern - Volmer plot of the reagent phases investigated, showing their sensitivities to oxygen. A. Coumarin 102 - silica gel; B, coumarin 1 - silica gel; C, coumarin 153 - silica gel; D, coumarin 102 - XAD-4: E, coumarin 1 - XAD-4: F, coumarin 153 - XAD-4; G. coumarin 102 - XAD-8; and H, coumarin 1 - and 153 - XAD-8 500 , 0 60 120 Time/min Fig. 5 . Fluorescence intensity monitored from the reagent phases over a period of 2 h in the absence of oxygen. A , Coumarin 1 - XAD-8; B, coutnarin 102 - XAD-8; C, coumarin 153 - XAD-8; D, coumarin 102 - XAD-3; E. coumarin 1 - XAD-4; F, coumarin 153 - XAD-3; G, coumarin 1 - silica gel; H, coumarin 102 - silica gel; and I, couinarin 153 - silica gel emission intensity from the adsorbed indicator in the absence of an oxygen quencher is related to the range of pore sizes found in the support matrices (ranging from 4.0 to 25.0 nm for the XAD matrices).The larger the pore size in the matrix, the greater is the fluorescence emission and this observation is in agreement with the results of a similar study reported recently.9 The highest unquenched emission intensities were measured for the three indicator dyes adsorbed on XAD-8 and were compared with the values obtained with the other adsorbents. The presence of silanol functional groups on the surface of the silica gel will reduce the amount of hydrophobic indicator being adsorbed and would account for the lower emission signal observed. The oxygen sensitivity of the reagent phase was generally higher for indicators adsorbed on silica gel than for those adsorbed on XAD matrices.These findings are illustrated by the Stern - Volmer plots in Fig. 4. A linear Stern - Volmer relationship is generally indicative of a single class of fluorophore population, the members of which are all equally accessible to oxygen quenching. However, the experimental quenching data in Fig. 4 show a negative deviation from the Stern - Voimer relationship, which does not conform to the linear Stern - Volmer expression shown in equation (1). The negative deviation observed has been rcportcd to be due to the presence of two fluorophore populations in the reagent phase, in which a fraction of the tluorophore is unquenched duc to its inaccessibility to oxygen, while the remaining fraction of fluorophores is quenched in the normal manner.10 The surface area of the organic matrices can be correlated with the oxygen sensitivity of thc reagent phase.The support matrix XAD-4 has a larger surface area (725 m2 g-l) than XAD-8 (140 in2 g- 1). The results showed a correspondingly higher oxygen sensitivity for the indicator adsorbed on XAD-4. The increase in surface area exposcs a proportionately larger amount of the immobilised fluorescent dye to collisional quenching by oxygen and causes an apparent increase in sensitivity. Oxygen sensitivity is also influenced by the type of coumarin indicator dye immobilised. Coumarin 102 showed thc highest oxygen sensitivity on the support matrices studied (Figs.3 and 4). The unquenched fluorescence intensity from the samples was measured periodically over a period of approximately 2 h (Fig. 5 ) . The degree of instability observed was dependent on the intensity and wavelength of the excitation radiation, the period of irradiation, the species of coumarin indicator and the support matrix used. Generally, the most stable response was observed for the immobilised coumarin 153. The largest drift in signal with time was observed for coumarin 102 adsorbed on XAD-4, where a 25% decrease in the fluorescence intensity was recorded over a 2-h period of continuous irradiation. The photochemical stability of the coumarin dyes in solution has been studied by several workers. 1 1.12 The 4-trifluoromethyl group in coumarin 153 was reported to stabilise the molecule by resisting photo-oxidation, whereas the 4-methyl group present in coumarins 1 and 102 is susceptible to photo-oxidation.1 1 The 4-methyl group in coumarin 1 was reported to be oxidised photochemically to give five products, of which one was a carboxylic acid responsible for the absorption at the excitation wavelength, thus causing an over-all reduction in the fluores- cence intensity.” This study has demonstrated that sensitivity, high signal intensity and stability cannot be obtained from a single indicator - matrix combination. What has emerged is the necessity to select a reagent phase that is a compromise between these desirable parameters. The collisional qucnching of the immobiliscd indicator by oxygen involves no chemical reaction and is completely reversible, without the need to replenish or regenerate the reagent phase.The above consider- ations have indicated that coumarin 102 adsorbed on the XAD-4 support matrix would be the reagent phase most suitable as an oxygen-sensitive mcdium for the development of a sensitive and reversible optical fibre oxygen sensor. The results of these studies will be reported in a subsequent paper. 13 We thank the Science and Engineering Research Council, the Department of Trade and Industry and a consortium of industrial members of the Optical Sensors Research Unit (OSRU) for the financial support of this work.666 ANALYST, JUNE 1989, VOL. 114 References 1. 2. 3. 4. 5 . 6. 7. 8. Kirkbright, G . F., Narayanaswamy, R., and Welti, N. A., Analyst, 1984, 109, 1025. Zhang, Z., and Seitz, R. W.,Anal. Chim. Acta, 1985,171,251. Shultz, J. S., Mansouri, S., and Goldstein, I. J., Diabetes Care, 1982, 5 , 245. Peterson, J. I., Fitzgerald, R. V., and Buckhold, D. K., Anal. Chem., 1984, 56, 62. Kroneis, H. W., and Marsoner, H. J., Sensors Actuators, 1983, 4, 587. Stern, 0.. and Volmer, M., Phys. Z., 1919, 20, 183. Kubin, R. F., and Fletcher, A. N., Chern. Phys. Lett., 1983,99, 49. Pirotta, M., Angew. Makromol. Chem., 1982, 109/110, 197. 9. 10. 11. 12. 13. Wyatt, W. A.. Poiricr, G. E., Bright, F. V., and Hicftje, G. M., Anal. Chem., 1987, 59, 572. Lakowitz, J . R., “Principles of Fluorescence Spectroscopy,” Plenum Press, New York, 1983, p. 257. Schimitschek, E. J . , Trias, J . A., Taylor, M., and Celto, J. E., IEEE J . Quantum Electron., 1973, QE-9, 781. Winters, B. H., Mandelburg, H. I., and Mohr, W. B., Appl. Phys. Lett., 1974, 25, 723. Li, P. Y . F., and Narayanaswamy, R.. Analyst, 1989, in the press. Paper 8104607G Received November 21st, 1988 Accepted February 14th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400663

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST. .JUNE 1989, VOL. 114 667 Isotopic Determination of Selenium in Biological Materials With Inductively Coupled Plasma Mass Spectrometry Bill T. G. Ting, Christine S. Mooers and Morteza Janghorbani" Clinical Nutrition Research Center, Department of Medicine, The University of Chicago, Chicago, IL 60637, USA A method for the isotopic determination of selenium in biological matrices is described. The method is based on hydride generation inductively coupled plasma mass spectrometry (ICP-MS). The development is specifically related to the requirements of stable isotope tracer studies in human subjects. The method is based on isotope dilution using 82Se as the in vitro spike and can quantify the 74Se and 77Se contents of samples. It involves wet oxidation (HN03 - H202 or HN03 - HC104) of the 82Se-spiked matrix, reduction to selenite by boiling with HCI followed by measurement of the isotope ratios (82SePSe and 74SePSe) in the gas stream (H2Se) generated from on-line reduction of the sample selenite with NaBH4.Compared with the isotopic signal resulting from a selenite solution containing 5 ng ml-1 of Se, the total sample blank contributions at m/z = 74,77 and 82 were less than 5% of the respective isotope signal. Worst-case absolute detection limits were 0.2-0.9 ng of Se, depending on the isotope used. Ion beam intensity ratios were measured with an over-all precision [relative standard deviation (RSD)] of 1% for both isotope pairs. Measured ratios (MR,) were stable during a given day's operation within the expected precision of the measurements but varied for different days.The magnitude of MRaIb was generally independent of the nature of the matrix. Highly linear relationships were found between ion beam intensity ratios (MRa;b) and the corresponding true isotope ratios for calibration solutions whose isotope ratios had been altered by as much as one order of magnitude. The precision/accuracy of the isotopic analysis was established by replicate measurements of the Se content of several biological matrices [National Bureau of Standards Standard Reference Material (NBS SRM) 1577a Bovine Liver, human plasma, red blood cells and human urine], and comparison of the results with independent measurements obtained using hydride generation atomic absorption spectrometry (AAS). The following data were obtained (mean k SD, n = 3-5; first result, hydride generation ICP-MS; second result, hydride generation AAS): NBS SRM 1577a Bovine Liver, 0.697 -t 0.002, 0.69 k 0.01 pg 9-1; plasma, 0.098 k 0.001,0.135 k 0.008 pg 9-1; red blood cells, 0.21 1 t 0.002, 0.216 k 0.012 pg 4-1; and urine, 0.0473 k 0.0003,0.0489 k 0.0003 pg ml-1.It was concluded that the proposed method could be used as the measurement method for studies of Se metabolism in human subjects using the concept of stable isotope tracers. Compared with other available methods of isotopic analysis, this method possesses the added advantage of requiring no chemical separation steps as the hydride generation is siifficient for removal of any potential matrix-related interferences. Keywords : Stable isotopes; selenium; inductive1 y coupled plasma mass spectrometry; h ydride generation Since the first report of the feasibility of the \table isotope tracer approach to investigations of selenium metabolism in man,' a number of successful applications have been re- ported.'.' At the heart of this approach is the need for analytical methods permitting quantitative measurement of the relevant stable isotopes of Se in matrices derived from human metabolic studies (foods, faeces, urine and blood) with the required precision and accuracy.J At present, two methods are available: radiochemical neutron activation analysis (NAA)" and gas chromatography - mass spectrometry (GC - MS).S Both of these methods have been applied to a number of metabolic problems.2J.h Inductively coupled plasma mass spectrometry (ICP-MS) has been shown to be an important tool for stable isotope tracer investigations.7.8 Conditions have been established for its routine application to metabolic tracer studies of Fe,q Zn and C U , ~ ~ ' .~ ' IAi,12 Brl-3 and Mg.14 A number of metabolic applications have also been reported. 15-17 The development of a suitable approach using ICP-MS for stable isotopes of Se has been hampered owing to two major difficulties: a lack of sufficient ion beam intensity for the available sample sizes; and the relatively large ion beam backgrounds resulting from the argon plasma in the region where these measurements are made. Preliminary observa- 1 o whom correspondence 4hould bc dddre\sed. Present address. Depdrtnient of Mcd:cme, Box 223, Thc Uni\cr\ity of Ctaicdgo, 5841 South Maryland Avcnuc, Chicago, TI, 60637.USA tionsl7 have indicated that of the six stable isotope\ of Sc. three (74Se, 77Se and "Sc) could putcntially be amcnablc to this approach a\ far as thc argon pla\ma background i\ concerned. In another paper.lY we have compared thew t u o issues (background and sensitivity) for two modes of sample introduction, w z . , pneumatic nebulisation (the standard method employed with commercial instruments) and hydride generation. The data demonstrated that whereas the pneu- matic nebulisation method provided sufficient i o n beam intensities for the three stable isotopes 74Sc. 77Se and "Se in matrices such as human urine. it did not pos\ess the neccssarq sensitivity for other matrices such as plasma tor i n those situations resulting from Se deficiency experiments.In con- trast, the hydride generation approach possessed worst-case absolute detection limits ( 3 \ / z ) within the range 0.6-1.8 ng of elemental Se.lX This paper describes the uce of hydride generation ICP-MS for the routine a n d accurate isotopic determination of the three stable isotopes 74Se, 77Se, and q2Se in relation to the rcquirernentr; of stahlc iwtope tracer investigations in human subjects. Ex per i mental Instrumentation Hydride gcnerirtion ICf'-.WS The ICP-MS instrument employed i n the\e studie4 wa\ a n Elan Model 250 system (SCTEX. Thornhill. Ontario, Canada). The dixtmce from the load coil to the satnplcr wit>668 ANALYST, JUNE 1989, VOL. 114 fixed at 27 mm.The length of the outer coolant tube of the torch (distance from the end of the auxiliary gas tube to the end of the outer coolant tube) was 36 mm. The torch was placed in a fixed position as close to the sampler as possible. Vertical and horizontal positions of the torch with respect to the sampler were adjusted to provide maximum ion beam intensity, using a solution of 5 ng ml-1 of Se (77Se, abundance Argon gas was obtained from liquified argon (industrial high purity. A & R Welding Supply, Alsip, IL, USA) and supplied all the requirements of the instrument via stainless- steel and high-density polyethylene tubing. The hydride generation system consisted o f a two-channel peristaltic pump (Rabbit, Kainin Instrument Co., Woburn, MA, USA) which pumped both the analyte and the reagent (NaBH4) solutions into the mixing chamber of a commercial hydride generator (PS Analytical, Questron Corporation, Princeton, NJ, USA).The output of the mixing chamber flowed into the gas - liquid separator of the hydride generator. The output of the latter was connected to the water-jacketed (25 "C) spray chamber of thc Elan spectrometer (nebuliser removed ) by means of approximately 120 cm of PTFE tubing 2.5 mm i.d.). All the tubing used between the reagent - sample containers and the gas - liquid separator consisted of flexible PTFE with an i.d. of <1 mm, with the exception of the sinall sections of plastic tubing that were neccssary for the two channels o f the peristaltic pump. Reagent - sample flow-rates were regulated by the settings of the peristaltic pump, which had been calibrated previously.The three argon stream flow-rates (plasma, auxiliary and carrier gases) were moni- tored using either the flow-meters installed with the instru- ment (plasma and auxiliary gases) or a mass flow-meter (carrier gas, Model 8200, Matheson Gas Products, E. Ruther- ford, NJ, USA). Data acquisition was in the (peak hopping) isotope ratio mode, using the multi-channel data acquisition capability of the system. details of which have been described previously. lo For the investigations reported here, the following settings of the software parameters were employed: resolution, M; measurements per peak, 3; scanning mode, I; measurement mode, M; measurement time, 1 .OOO s; repeats per integration, 5-10: dwell time, 10 ms; and cycle time.0.200 s. A schematic diagram of the hydride generation TCP-MS system employed has been given previously. 18 The optimised operating parameters used in this investigation are sum- marised in Table 1. 7.38 Yo ) . HJ dri de genera tion atomic absorption spectrometry In order to establish the accuracy of isotopic analyses with hydride generation ICP-MS, we compared our results (exprevxd in elemental terms) with those obtained using a hydride generation atomic absorption spectrometry (AAS) system. This syctem consisted of a batch-type hydride genera- tor (MHS-10, Perkin-Elmer, Norwalk, CT, USA) and an atomic absorption spectrometer (Model 5000, Perkin-Elmer). Hydrogen selenide (H2Se) was generated by introducing a Table 1. Optimised operating parameters for the hydridc generation ICP-MS system Gas tlou-rates .. . . Plasma gas, 1 1 .O: auxiliary gas, 2.3; and ICP settings . . . . . . Incident power. 1100 W; spray chamber Interface parameters . . See text Mass spectrometer Hydride generation carrier gas, I .4 I min--' temperature, 2.5 "C pat;anietcrs . . . . . . Adjusted for maximum intensity paranietcrs . . . . . . Rcagcnt flow-rate, 3.0 ml min-I; sample solution flow-rate, 10.0 nil min- I solution of NaBH4 (1% mlm, 0.25°/o NaOH) at a constant rate into a 10-ml aliquot of the sample digest. Argon gas at 27 Ib in-2 was used for this purpose. The H2Se stream was carried into the heated quartz cell (165 mm long, 12 mm i.d.) which was fitted over the standard burner head. A Se electrodeless discharge lamp was used.The other parameters of this system were: absorption wavelength, 196.0 nm; slit width, 2.0 mm; and air and acetylene flow-rates, 15.5 and 2 1 min-1, respectively. Data were acquired with a PRS-10 printer with a 0.25 cycle time starting at the time of the introduction of NaBH4 into the reaction vessel and continuing until five data points had been obtained corresponding to the plateau of the absorption plot. The five data points were averaged and used for calculations. All determinations employed the method of standard additions with four points (three spike levels). Certified solutions (Fisher) of Se standards were used for spiking purposes. Chemicals All chemicals used were of analytical-reagent grade and wcrc used without further purification.The stable isotopes employed were purchased as the clemcntal powder (Oak Ridge National Laboratory, Oak Ridge, TN, USA), which was dissolved in the minimum volume of HN03. The resulting selenite solution was added incrementally to solutions of selenite of natural isotopic composition in order to provide working standards of known isotope ratio covering the range of interest. The preparation of such standard solutions has been described previously for a number of other trace clements.11-14 These standard solutions will be referred to as stable isotope ratio calibration colutions. Sodium tetrahydro- borate(lI1) solution (1% m/V) was prepared in 1-2-1 batches before use (2.5 g of NaOH plus 10 g of NaBH4 diluted to 1 1). The resulting solution was filtered to remove any undissolved matter.Chemical Procedure The sample preparation procedure employed for both the hydride generation ICP-MS and the hydride generation AAS systems was as follows. Samples were wet-ashed according to previously described procedures-J using HNOl - H202 (plasma, red blood cells, faeces and foods) or HN03 - HCIOJ (urine) as the oxidant mixture, followed by boiling for 1 0 min with concentrated HC1 in order to convert the Se content into selenite. Any insoluble residue was filtered off and the Table 2. Details of Experiment 1 Run timeimin Sample 0 4 10 14-35 39 43-52 566.5 69-77 81 85- 102 106 110-119 123 127 131-148 152-160 164-169 173 178- 194 199-208 2 12-22 1 225-233 Standard, 2 ng ml 1 of Sc Standard, 5 ng ml 1 of Se Procedural blank Standard, 2 ng ml- I o f Se Plasma NBS SRM 1577a Bovine Liver Red blood cells Standard, 2 ng ml- 1 of Se Urine Standard, 2 ng ml-1 of Se Is0 tope calibration st an duds Spiked plasma Standard, 2 ng ml I of Se Spiked NBS SRM 1577a Bovine Liver Spiked red blood cells Spiked urine 10% HCI 10% HCl 10% HC'I 10?"' HCl 10% HCI 10% HCI No.o f replicates 1 1 1 6 1 3 3 1 2 1 3 1 1 3 2 1 5 3 3 3 I -ANALYST, JUNE 1989, VOL. 114 669 resulting solution was diluted to the final volume (25-50 ml) with 10% HCI. The sample sizes used were: urine, 2-10 ml; plasma, 1-5 ml; red blood cells, 0.5-2 ml; faeces, 0.1% of daily output: and foods, 0.1% of composite daily intake. These were equivalent to 100-500 ng of Se. If only isotope ratio measurements were required, no spike was included, other- wise the sample was spiked with 82Se032- (level of spike, 3-5 times the expected sample content of the isotope).For the hydride generation AAS system, aliquots of the digest were spiked with accurate amounts of Se from certified solutions of Se standard (four points). For background correction for the hydride generation TCP-MS system, the use of both 10% HCl and a true procedural blank was investigated. The background corrected ion beam intensity ratios (MR71 77 and MRg2 77.L) were converted to the expected true isotope ratios (MIR7477 and MIR,? 7 7 , idm basis) by means of stable isotope ratio calibration standards. Individual Experiments In order to investigate various aspects of this work, several experiments were carried out. These are described briefly below.Basic expi~rirn en ts A number of basic experiments were performed with the hydride generation ICP-MS system in order to determine its performance parameters (signal to background levels, ion beam intensity stability, precision of isotope ratio measure- ments, detection limits, etc.). For these experiments, solu- tions o f Se (2-5 ng ml-1) or stable isotope ratio calibration solutions were employed. Experiin en t I The purpose of this experiment was to establish the analytical characteristics of the system for the matrices of interest. Various samples (Table 2, both unspiked or spiked with X2Se) were processed as described under Chemical Procedure. Six complete procedural blanks were also prepared; these were analysed together with standard solutions of Se and 10% HCI in the sequence given in Table 2.Experiment 2: inter-method comparison Sub-samples from a well mixed pool of the biological matrices of interest [National Bureau of Standards Standard Reference Material (NBS SRM) 1577a Bovine Liver, human plasma, red blood cells and urine] were processed for quantitative isotopic analysis (hydride generation ICP-MS; isotope dilution analy- sis) or elemental analysis (hydride generation AAS: method of standard additions) as described under Chemical Pro- cedure. Results and Discussion Human metabolic investigations employing the concept of stable isotope tracers impose certain requirements on the analytical method of isotopic measurement. A clear under- standing of these requirements is essential for the appreciation of the measurement methodology.In the simplest experi- ments of this type. a single isotope (highly enriched 74Se is the preferred choice, referred to as the in vivo tracer) is administered to the subject in an appropriate manner (orally or intravenously). Timed samples (urine, blood and faeces) are then obtained. Generally, the samples are analysed quantitatively for two stable isotopes (74Se and a reference isotope, in this instance 77Se). In order to perform quantitative isotopic analysis with ICP-MS the preferred method is based on it? vitro stable isotope dilution (in vitro SID). For this, enriched QSe is used in the form of selenite. Terminology We have adopted the following terminology: MRaih is the measured ion beam intensity ratio for an isotope pair N and h ; MR,,h.c is the corresponding ratio corrected for background contributions; and MIRUlh is the corresponding true isotope ratio in the sample (pg of isotope a to vg of isotope b ) .A stable isotope ratio calibration plot is a plot of the observed relationship between MRCllh,, and MIR,h for a set of stable isotope ra.tio calibration standards. Factors Affecting Analytical Performance The over-all objective of the measurement method is the accurate determination of 74Se and 77Se in the sample. This requires accurate measurement of two isotope ratios (MIR74)77 and MIRX2,77). Therefore, the fundamental measurement criteria are related to the accuracy with which MIR74177 and MIR82,77 can be determined. Two types of performance criteria were evaluated for these analyses: the fundamental analytical performance of the hydride generation ICP-MS instrumentation and the aspects related to its application to biological materials.Fundamental Aspects of Hydride Generation ICP-MS We have previously compared the basic characteristics of the hydride generation ICP-MS system employed in this work with pneumatic nebulisation, which is the standard mode of sample introduction with the current ICP-MS device5.18 The optimised parameters leading to the lowest detection limit for the three stable isotopes of Se are given in Table 1 . The fundamental factors determining the over-all capabilities and limitations of this method for the accurate measurement of stable isotopes of Se in relation to tracer studies are 4gnal to background considerations, time stability of ion beam ratios, achievable precision of measurement for isotope ratios, matrix-related systematic biases in the ion beam intensity ratios, memory effects and factors related to isotope calibra- tion methods.Signal to background considerations The signal to background intensities obtained in Experiment 1 were plotted as a function of time for all matrices investigated; the results are shown in Fig. 1. Data for the ion beam intensities 177 were normalised to a Se concentration of 1 ng ml-I based on a knowledge of the true Se concentration of the samples (determined using the in vitro SID procedurs; spiked samples, Table 2). These data clearly demonstrate a reasonably stable ion beam intensity during the 4 h o f this run. The ion beam intensities (177) for different matrices were [ions s-1, mean f 1 standard deviation (SD), normalised to 1 ng ml-1 o f Se]: standard solution, 5250 k 460; plasma, 6600 f 490; NBS SRM 1577a Bovine Liver, 6490 -t 170; red blood cells, 7600 f 650; and urine, 5600 2 210.The intra-matrix variations [relative standard deviation (RSD)] were within the G 1 n2 .1 0 100 2 00 Run ti m e 'mi i n Fig. 1. Ion beam intensities for all samples, normalised to a Se content of 1 ng m1-1, and blank contributions for the three stable isotopes plotted against run time (see Table 2 for run detail\). 0. All samples; +, I,, (all blanks); 0. I,7 (all blank\); and ., 174 (all blanks)670 ANALYST, JUNE 1989, VOL. 114 Table 3. Signal to background data for the hydride generation ICP-MS system. Each data point corresponds to the mean i 1 SD for ten sequential me abure nic 11 t s Dats 173 I77 18, 174 I77 182 1% iihin-dup rrieusure?ncnts*- 10% MCl Procedural blank 27/6/88 ( 1 ) .. lh(lk27 624 -t 54 1146 k 57 163 * 49 458 k 80 784 i 87 27i6i88 (2) . . 149 k 30 797 i 275 1347 I 357 140 * 10 555 i 74 960 i 124 276i88 (3) . . 139k25 574 * 67 1044 i 64 133 L 8 374 * 55 727 k 56 27/6/88 (4) . . 128 k 25 474 i 32 - 149 i 23 397 i 42 783 +- 81 - 190 k 21 554 k 16 938 & 20 27i6'88 (6) . . 112 k 18 431 t 16 27/6/88 (5) . . 113 i 23 424 * 20 - 110 L 40 312 k 10 437 k 26 Brtt$wti -day vuriutiori 5- 10n/o HC1 2ngmlk'ofSe 1914i88 . . . . 87 i 2 395 k 10 1279 k 26 1137 t 13 9431 k 97 14 009 ? 223 20/4'88 . . . . 101 i 14 1331 i 209 1435 2 260 1257 k 28 1 1 031 i 458 1s 122 t 255 11'5'88 .. . . 97 k 28 38O+ 140 7 1 9 i 170 1303 t 48 11 201 k 424 16 332 k 801 30/5,88 . . . . 110 +_ 13 676 * 02 1732 k 117 2977 rfi 9 25 409 k 273 35 291 k 84 20i6 88 . . . 72 2 3 327 i X 810 i 21 1713 -t 24 14 306 ?c 240 19 9 14 ?c 490 ' Ion beaiii intensities for 2 ng nil--' of Se on 27!6i88: 1490, 177 11 700 and I,, 15 500. ~ rartgc 3 (SRM 1577a)-9% (standard solution or red blood cells). In comparison, the mean value of I,, (normalised to 1 ng ml-1 of Sc) for each matrix differed from that for the standard solution within the range 7 (urine)-44% (red blood cells). reflecting potential matrix effects on the generation of H7Sc . w 2 2 An understanding of the background intensities is im- portant i n the light o f the betwecn-day variations in the instrument parameters and other uncontrollable factors.We have investig:ited this by observing both within-day variations. using either 10% HCI solutions or complete procedural blanks. and between-day variations (Table 3). Background intensities (combined data for all 10% HCI and procedural blanks) during the entire run on a particular day (27/4/88) (Experiment 1) were (mean ?c 1 SD): mlz = 74, 141 i 24; In/: = 77, 498 k 131; and mlz = 82, 907 & 264 counts SKI. This corresponded to 10, 4 and 5% of the ion beam intensities resulting from a 2 ng ml-1 solution of Se. Although the background count rates at m l z = 82 were higher for the 10% HCI solutions compared with procedural blanks, there did not seem to be a marked difference for the other two iwtopes.Comparative data for the belween-day variations for 10% HCI solutions and a standard solution of Se (2 ng ml-I) on a number ot occasions are summarised in Table 3. Although between-day variations in the background were observed, the background count rate was, on all but one occasion (20/4/88 for 177), less than 10% of the corresponding count rate for a solution of 2 ng ml-1 of Se. Calculated detection limits for Se based o n the experimental SD of the six measurements carried out during a 4-h period [only three measurements for /x2.10%, tlcI (Table 3)] were 0.9, 0.30 and 0.2 ng of Se for 10Y0 HCI and 73Se, 77Se and **Se, respectively. The corresponding values using the procedural blank data were 0.9, 0.3 and 0.3 ng of Se. For these calculations, we have asIumed a total solution requirement of SO ml.Such a solution volume would readily permit the acquisition o f ten isotope ratio data points and provide sufficient soluticn for the necessary wash-out to minimise any memory effects due to a previous sample.18 Hence, these detection limits correspond to conservative estimates. The data presented here for background count rates and their between-day variations clearly demonstrate that the background can rcadily be reduced to a few per cent. of the signal intensity for the three stable isotopes if the Se concentration of the analyte solution is about 5 ng ml-1. This would necessitate 250 ng of Se under the conditions of 50-ml solution volume requirements. The most size-limited matrix in human metabolic studies corresponds to plarma for which the normal Se concentration is about 100 ng ml-1.Therefore, it is clear that this method readily meets the sample size require- ments of these investigations. We have previously investigated the effects of instrument operating parameters (1OPs) such as argon gas flow-rates or incident power on the magnitude of the signal to background ratio. 19 Over the range of optimum IOPs for stable isotopes of Se using the hydride system,18 the signal to background ratio does not vary markedly. However, other potentially import- ant factors such as the spray-chamber temperature could play a more marked role and these require further investigation. Precision and time stability of MRdIb The measured ion beam intensity ratios vary under different operating conditions.12-14 These variations can be accounted for by the use of appropriate calibration procedures. 12-14 However, within a particular day's run, unacceptably large variations cannot be tolerated. Therefore, it is necessary to establish that the measured ion beam intensity ratios are sufficiently constant during any particular day of operation such that the stable isotope ratio calibration procedure can be used. That this is true for stable isotopes of Se, as has been shown for a number of other isotopes,9.11-'4 can be seen from the stability data given in Fig. 2. These data were obtained by continuous measurements on a single solution of a Se standard (5 ng ml -I), each data point corresponding to the mean of ten sequential data acquisitions. The data demonstrate a more stable ion beam intensity over the 4 h of observation than is characteristic of the current design of our instrument .c ) * l The data given in the bottom part of Fig. 2 indicate that the measured ion beam intensity ratios did not deviate outside the expected random variations corresponding to a 21% RSD. There were no systematic changes in these values, consistent with previous observations for other stable isotopes even under conditions of much less stable ion beam intensities than those observed here.9.11-14 Therefore, it is evident that the within-day stability of the ion beam intensity ratios is sufficient to permit the use of stable isotope ratio calibrations for the calculation of the expected true ratios. Matrix effects Data relating to MRUlb for various matrices used in Experi- ment 1 are summarised in Table 4.For all the matricesANALYST, JUNE 1989, VOL. 114 1 - 67 1 + investigated, the intra-matrix precision (2 1 RSD) was better than 1.4% for MR74/77,c and better than 1.0% for MR82/77.c. Similarly, the absolute value of MR74/77,c for all the matrices investigated was within 1.1% of the value for the standard solution (2 ng ml-I), whereas that of MR82/77,c varied from the standard solution value by less than 2.7%. Although the intra-matrix precision of the ion beam intensity ratios was consistently about 1-1.4% for both isotope ratios, a small systematic bias was apparent in the MRX2/77,c for NBS SRM 1.577a Bovine Liver (+2.7%) and red blood cells (+2.2%). The reasons for this are not yet understood, but are most likely due to a positive bias in 182,c. This could originate from molecular interferences specific to the nature of these matrices, similar to previously reported situations,13 or could result from changes in the argon-related background at mlz = 82 arising from unknown matrix influences.Memory effects We have previously reported on the over-all “memory effect” of ICP-MS using the pneumatic nebulisation sample introduc- tion system for Li12 and Br.13 The results indicated that this effect is not a limiting factor in the practical application of the method to accurate isotopic analysis. In a previous paper, comparing the pneumatic nebulisation and hydride generation methods for the isotopic determination of Se,lS we found that the memory effect could be significant for the hydride method if sequential solutions had large differences in isotope ratios and if the solutions were run in the order of decreasing isotope ratio.Fig. 3 shows the speed with which ion beam intensities are reduced when a solution highly enriched with respect to QSe (corresponding to a natural Se concentration of 20 ng ml-1) is replaced with 10% HC1. The data show that about 10 min after the introduction of HC1, 182,c was less than 1% of its original value. Because of the desirability of short wash-out times in routine isotopic analysis, the effect of relatively short wash-out times on the accuracy of isotope ratio measurements - 2 1 1 I 1 0 100 200 300 Time’min Fig. 2. Data establishing time invariance of MR,,), for stable isotopes of Se.Data are for continuous measurement on a Se standard solution ( 5 ng ml-I of Se). 0, MR82,77; and +, MR74,77 was studied. The results are presented in Fig. 4. The data given in Fig. 4 (filled diamonds) were obtained by switching (time 0) from a solution of Se of natural isotopic composition ([Se] = 2 ng ml-1, MIR082,77 1.291) to a solution of similar Se concentration but enriched to a small extent with respect to We (MIR82177 1.3.51). The data shown as open squares were obtained when these two solutions were run in the reverse sequence (prior to running the solution with MIR82/77 = 1.3.51, a solution of higher isotopic enrichment, MIR82177 = 1.410, had been run).lS A 60-s wash-out with 10% HC1 had also been incorporated in the latter run, as shown in Fig.4. Several important points should be noted from the data given in Fig. 4 and from the more extensive information provided else- 105 104 c I, 103 2000 1 0 -10 0 10 20 - 30 40 Ti meimi n Fig. 3. Decay pattern of I,, for a solution of 20 ng ml-1 of Se replaced with 10% HCl. Fig. ( h ) shows the expanded vertical scale for the same data 1.57 c 1 HCI start Sample start I I -200 -100 0 100 200 300 400 Time!s Fig. 4. “Memory effect” when solutions with different isotope ratios are run (see text for details). +, Order of increasing ratio; and 0, order of decreasing ratio Table 4. MR,,h,c for different biological matrices (data are from Experiment 1) No. of Matrix replicates MR74/77.c * A , %o t MRX2/77 ,c * A,%+ Standard, 2 ng ml-I of Se . . . . 5 0.1176 k 0.0017 0 1.284 k 0.0074 0 Plasma .. . . . . . . . . 3 0.1176 2 0.0015 0 1.302 k 0.012 +1.4 NBS SRM 1577a Bovine Liver . . 3 0.1189 t 0.0016 + 1.1 1.319 k 0.013 +2.7 Red blood cells . . . . . . . . 3 0.1176 k 0.0005 0 1.313 t 0.0049 t-2.2 Urine . . . . . . . . . . 3 0.1165 t 0.0006 -0.9 1.285 k 0.0040 0 * Mean k SD for n independently prepared sub-samples. ‘i Percentage difference from that for 2 ng ml-l of Se.672 ANALYST, JUNE 1989, VOL. 114 where. 18 First, comparing the measured ion beam intensity ratios (MR82,77) with the corresponding true ratios ( M I R s ~ , ~ ~ ) , the two values are not identical. This is a consistent feature of ICP-MS and has been observed on numerous occasions both for Se and other elements.7~”1~-]4 The extent of the deviation of MR,//, from the corresponding true ratios depends on the specific settings of the IOPs.This highlights the need for suitable isotope calibration procedures (see next section). Second, when the order of analysis is from low to high isotope ratios (filled diamonds) the sequential data obtained for the high-ratio solution reach a plateau reasonably rapidly. However, when the order of sample introduction is reversed (high to low ratio, open squares) there is a consistent decrease in the measured ratio for the low-ratio solution, indicating a significant memory effect of the previous (high-ratio) solu- tion. Although the actual reasons for this are not clear, they probably relate to residual Se from the previous solution contaminating the low-ratio solution. We are currently attempting to pin-point the location of this memory effect.Therefore, in the current design of the system and until the source of this problem has been located and suitable design modifications implemented, accurate isotopic analysis necessitates that samples be run in the sequence of increasing isotope ratio. Linearity of isotope ratio calibration plots Fig. 5 shows the relevant data for stable isotope ratio calibration solutions used to convert the ion beam intensity ratios into the estimates of true isotope ratios. The data presented are for X’SeP7Se (left-hand side) and 7JSe/77Se (right-hand side). For each isotope pair, both ion beam intensities and their ratios (corrected for blank) are presented. The data given in Fig. S(c) and (d) for each isotope pair show the degree of variability in the ion beam intensities that are sometimes observed (cf.Fig. 2). Tn this particular run (actual run time 3 h), 177 varied by about 40%. This is also reflected in the deviations from linearity observed for both 174 and Zxz [Fig. S(a) and ( h ) ] . The ion beam intensity ratio (MR,,lh.c) did not show any of these fluctuations [Fig. 5 ( e ) and ( f ) ] , as expected. Highly linear correlations were obtained for both isotope pairs over the entire range of ion beam intensity ratios involved. The linear regression equations for the two calibration plots were MR7, 77 = -0.0023 + 1.071 (MTR,, 77) *o 4 1 0 c 3 I t F l b ln c 0 m ._ 2 P ,P 0.4 0 0.3 h r. TI n” 0.2 2 0 4 8 12 0.1 0.2 0.3 0.4 MI R82 77 MlR74 77 Fig. 5. for stable isotope ratio calibration solutions Ion beam intensities and their background-corrected ratios (Y’ = 0.9982) and MRX2,77.c = 0.098 + 1.036(M1R82,77) (r2 = 0.9992).The linear regression parameters for a single set of calibration solutions obtained on several occasions are sum- marked in Table 5. These data clearly demonstrate the highly linear nature of these plots. The intercept is close to zero in all instances, but the slope does not appear to be invariably unity. Whether this is related to the effect of operating parameters on mass discrimination is not yet understood; this point needs to be investigated further. PrecisiordAccuracy and Practical Aspects of Routine Isotopic Analysis In order for these methods to be suitable for routine isotopic analyses in relation to metabolic investigations, two important criteria must be satisfied: (1) the precision/accuracy in relation to the requirements of the experiment; and (2) a sample throughput consistent with the requirements of the investiga- tions.Precisionlaccuracy Two types of isotopic analyses may be involved in metabolic investigations: those requiring only isotope ratio measure- ments; and investigations requiring quantitative isotopic analyses. Precision requirements of isotopic measurements vary widely for different chemical elements and for various aspects of the metabolic issues being addressed.23 With specific reference to studies with Se, an isotope ratio measurement precision of 1-5% has been shown to be adequate for exploring many of the issues of current interest.1 Both NAA4 and GC - MSS has been reported to possess the necessary capability in this respect. The data given in Fig. 2 and Table 4 demonstrate the capabilities of the present method in relation to this require- ment. Hence, it is clear that this method can provide isotope ratio data for a variety of biological materials with a measurement precision of about 1 Yo for routine analysis, including measurement of the ratio involving the least abundant stable isotope (74Se). Tn particular circumstances, it may be possible to achieve a measurement precision of better than 1 70, but this is not yet possible for routine analyses. The data given in Table 4 indicate that for MR71/77 no inter-matrix bias is present. The measurement ratios for various biological matrices are within the expected value for standard solutions of selenitc. In contrast, the data given for MRX2/77 in red blood cells and NBS SRM 1577a Bovine Liver, although precise to the cxtcnt of lo/, show a positive bias of 2.2 and 2.7%, respectively.The reasons for this have not yet been identified, but are probably related to positive bias from some as yet unknown interferent in Zxz. Such biases have been reported previously for 74Se (FeO+) in red blood cells using the pneumatic nebulisation method of samplc introduction7 Table 5 . Observed betwcen-day variations in regression parameters for the stable isotope ratio calibration plots: MRulh = a(MIR,,h) + b Date of run a b rz MR8y77 versus MIR82/77- 28/1/88 . . . . 0.916 0.134 0.998 17/2/88 . . . . 1.162 0.162 0,9991 11/3/88 .. . . 1.371 -0.174 0.9999 15/3/88 . . . . 1.073 0.180 0.9990 16/3/88 . . . . 1.050 0.172 0.9990 27/6/88 . . . . 1.036 -0.098 0.9992 MK71/77.c versus MIR7J177- 28/1/88 . . . . 1.057 -0.010 0.9997 17/2/88 . . . . 1.099 -0.01 1 0.9999 11/3/88 . . . . 1.165 -0.0 1 6 0. ’$999 27/6/88 , , . . 1.071 -0.002 0.9982ANALYST, JUNE 1989, VOL. 114 Table 6. Inter-isotope comparison for isotope dilution analysis Seipg g-1 Sample n 82SePSe 8?3ei74Se Red blood cells . . 3 0.199 k 0.008 0.196 ? 0.008 Plasma . . . . . . 3 0.0857 k 0.005 0.0864 k 0.005 Urinc . . . . . . 3 0.0411 ? 0.0005* 0.0402 k 0.0004* * Values in pg ml-1. Table 7. Intcr-method comparison between hydride generation ICP-MS and hydride generation AAS Selug g- Hydride generation Hydride generation Matrix ICP-MS" AAS'r NBS SRM 1577a Bovine Liver .. . . 0.697 ? 0.002 0.69 ? 0.01 Human red blood cells . . 0.21 1 k 0.002 0.216 k (1.012 Human urine . . . . . . 0.0473 k 0.00033 0.0489 k 0.0003< Human plasma . . . . 0.098 ? 0.001 0.135 k 0.008 I I = 5: the isotope pair employed in the calculations was " S C / ~ ~ S ~ . t . n = 3 . $ Values in pg ml 1 . and for X1Br (S03H+) in urine.'3 Because of the complex nature of the biological materials and the ICP process, these interferences are likely to be much more significant than has been realised so far. The over-all accuracy of quantitative stable isotope analysis was investigated by carrying out in vitro SID (spike: f%e) on thc samples of interest and on NBS SRM 1577a Bovine Liver. For the isotope pair used in the calculations, both X'SeP7Sc and "Se/75e were employed.In addition, independent elemental analyses on sample replicates were performed with hqdride generation AAS using the method of standard additions. The results of these analyses, expressed in terms of elcmcntal content (to permit ready comparison), are sum- marised in Tables 6 and 7. The data given in Table A show the degree of agreement obtained in the irz ciitro SID procedure using either X%e/77Se or QSe/7%e as the isotope pair (SZSe as spike for both experi- ments) for the analyses. Although such an agreement would not necessarily indicate accuracy, it does show that no systematic biases are introduced into the analyses because of isotope-specific interferences. Comparative data for hydride generation ICP-MS and hydride generation AAS, indicating the true accuracy of the two methods, are summarised in Table 7.The Se content of NRS SRM 1577a Bovine Liver obtained with both methods agreed well with the certified value of 0.7 & 0.07 ug g-1. Inter-method agreement to within their combined measurement precision was observed for all the matrices, except plasma for which a difference of 30% was found. As expected, the in vitro SID method provided analytical data with considerably better precision than the hydride generation AAS procedure. The over-all precision of the hydride generation ICP-MS method for quantitative analysis was generally similar to the expected measurement precision of the ion beam intensity ratios (Fig. 2 and Table 4), demonstrating the validity of the assumption that this method has greater analytical precision because it relies primarily, but not exclusively, on the measurement precision of the underly- ing ratios.Practical aspects of isotopic analysis A major feature of the proposed method is its accuracy under the conditions of high sample throughput. All the methods of 673 isotopic analysis reported for mineralitrace element~4~S~"1~ require chemical separation, the extent of which varies considerably depending on the specific method, the isotopes in question and the nature of the matrix. Methods based on thermal ionisation mass spectrometry generally require the most extensive chemical separations, and this requirement constitutes one of the major limiting factors in the develop- ment of this technique for stable isotope tracer studies.In this respect, ICP-MS has allowed a considerable degree of simplification, but in all the instances reported to date some degree of chemical separation has been required .7,4-14 For Se, both NAA4 and GC-MSS require the use of separation techniques. We have previously attempted to deve!op the method of ICP-MS for Se isotopes in the pneumatic nebulisation sample introduction mode and have found that it also requires multi-step chemical separation procedures with potentially serious interference problems.7 The proposed hydride generation ICP-MS method obviates this requirement completely. The method only requires that Se be present as selenite prior to its reaction with NaBH4. a common requirement for all current methods of isotopic analysis for this element."S The removal of the need for chemical separation is, of course, due to the selective separation capability inherent in the hydride generation process. The method described here permits the preparation of at least 20 samples per day up to the point of mass spectrometric analysis.If the measurement precisions reported here are sufficient for a specific application, about 100 analyte solu- tions can be readily processed per 8-h working day for the two isotope ratios, including the necessary blanks and calibration standards. Therefore, it is clear that this method represents an important addition to the analytical methodology that is currently available for stable isotope tracer investigations. This work was supported by NIH R01-CA38943 and DK26678-09.1. 2. 3. 4. 5 . 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. References Janghorbani, M., Christensen, M. J . , Nahapetian. A . , and Young, V. R.. Am. J. Clin. Nutr.. 1982, 35, 647. Janghorbani, M., and Young, V. R . , in Combs, G . F., Jr., Spallholz, J. E., Levander, 0. A., and Oldfield, J . E., editor^, "Selenium in Biology and Medicine. Third International Symposium," Van Nostrand Reinhold, New York. 1987, pp. 45C471. Swanson, C. A . , Reamer, D. C., Veillon. C . , King, J . C.. and Levander, 0. A., Am. J . Clin. Nutr., 1983, 38, 169. Janghorbani, M., Ting, B. T. G . , and Young, V. R . , Am. J . Clin. Nutr., 1981, 34, 2816. Reamer, D. C., and Veillon. C., J . Nutr., 1983, 113, 786. Swanson, C . A., Reamer, D. C., Veillon, C., and Levander. 0. A., .I. Nutr., 1983, 113, 793. Janghorbani, M., in Date, A. R.. and Gray, A. L., Editors, "Applications of ICP-MS." Blackie, London, 1989, pp. 115- 141. Houk, R. S . , Mass Spectrom. Rev., 1988, 7. 425. Janghorbani, M.. Ting, B. T. Ci.. and Fomon, S. J.. Am. J. Hcmatol., 1986. 21, 277. Serfass, R . E., Thompson, J . J . , and Houk, R . S., Anal. Chirn. Acta, 1986, 188, 73. Ting, B. T. G., and Janghorbani, M., Spectrochrm Acta, Part B , 1987, 42, 21. Sun, X. F., Ting, B. T. G . , Zeisel, S. H . , and Janghorbani, M., Analyyt, 1987, 112, 1223. Janghorbani. M., Davis, T. A., and Ting, B. T. G.. Analyst, 1988. 113, 405. Schuette, S., Vereault, D., Ting, B. T. G . , and Janghorbani, M., Analyst, 1988, 113, 1837. Woodhead, J . C., Drulis. J . M., Rogers, R. R., Zieglcr, E. E., Stumbo, P. J . , Janghorbani, M., Ting, B. T. G.. and Fomon, S. J . , Pediutr. Res., 1988, 23, 495. Fomon, S. J., Janghorbani, M., Ting, €3. T. G., Ziegler. E. E., Rogers, R. R., Nelson, S. E.. O5tedgaard, L. S . . and Edwards. R R . Prdintr. Roc., 1988, 24. 20.674 ANALYST, JUNE 1989, VOL. 114 17. Janghorbani, M., Ting, B. T. G., and Zeisel, S. H . , in Prasad, A. S . , Editor, “Essential and Toxic Trace Elements in Human Health and Disease,” Alan R. Liss, New York, 1988, Janghorbani, M., and Ting, B. T. G., Anal. Chem., 1989, 61, 701. Ting, B. T. G:, and Janghorbani, M.. J. Anal. At. Spectrom., 1988. 3, 325. Welz. B., and Melcher, M., Analyst. 1984, 109, 569. pp. 545-556. 18. 19. 20. 21. 22. 23. Thompson, M., Pahlavanpour, B . , Walton, S. J . , and Kirk- bright, G. F., Analyst, 1978, 103, 705. Pettersson, J . , Hansson, L., and Olin, A . , Tulanta, 1986, 33, 249. Janghorbani, M., Prog. Food Nutr. Sci., 1984, 8, 303. Paper 810381 8J Received September 27th, 1988 Accepted January 11 th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400667

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST, JUNE 1989, VOL. 114 675 Inductively Coupled Plasma Mass Spectrometric Determination of the Absorption of Iron In Normal Women Paul G. Whittaker and Tom Lind University Department of Obstetrics and Gynaecology, Princess Mary Maternity Hospital, Newcastle, Tyne and Wear NE2 35D, UK John G. Williams and Alan L. Gray ICP-MS Unit, Department of Chemistry, University of Surrey, Guildford GU2 5XH, UK The determination of iron isotope ratios in blood, without prior sample preparation, using inductively coupled plasma mass spectrometry (ICP-MS) with sample introduction by electrothermal vaporisation (ETV) is described. Following oral administration of 5 mg of enriched 54FeS04 and intravenous administration of 200pg of 57FeS04 to non-pregnant women, the 54Fe:56Fe and 57Fe:56Fe isotope ratios in serum were measured reliably within 20 min per sample in quintuplicate.Changes in the fractional absorption of iron during human pregnancy could therefore be assessed. Keywords: Inductively coupled plasma mass spectrometry; iron; stable isotopes; women; absorption Assessment of the extent to which oral iron absorption increases during normal pregnancy can help to resolve the issue of whether normal pregnant women need routine iron supplements. This is important not only because oral iron can cause gastrointestinal upsets in many pregnant women but also becauw the absorption of other elements, such as magnesium and zinc, may be affected. The resulting therapy is also a considerable financial burden on the National Health Service, costing millions of pounds annually.Although mass spectrometric methods have been available to investigate the absorption of trace minerals,13 there are few reports of this technique having been used to study human pregnancy.5 Until inductively coupled plasma mass spec- trometry (1CP-MS) became commercially available, the deter- mination of stable isotope ratios was carried out using techniques such as neutron activation analysis or thermal ionisation mass spectrometry, where measurement was both complex and time consuming and sample preparation often difficult . This paper describes the use of ICP-MS to determine iron isotope ratios with sample introduction by electrothermal vaporisation (ETV) without prior sample preparation. The use of two enriched isotopes was considered most suitable for the studies relating to pregnancy, and because of polyatomic ion interferences ETV was used as a means of sample introduction.Experimental Chemicals Enriched 54Fe and 57Fe were obtained as iron(I1) sulphate from the UK Atomic Energy Authority at Harwell; a mass analysis is shown in Table 1. Both iron(l1) sulphate prepara- tions were dissolved in doubly distilled, de-ionised water with 3mgml-1 of ascorbic acid (final pH, 2.9) and then 5-ml aliquots of each solution were sealed under nitrogen in glass ampoules. The ampoules for oral administration contained Table 1. Mass analysis of enriched isotope preparations At.-?" Enriched isotope 54 56 57 58 s4FeS04 95.6 4.31 0.014 0 s7FeS04 0 3.00 95.10 1.9 5.23 mg of total iron, equivalent to 5.01 mg of 54Fe and those for intravenous (IV) administration contained 196 pg of total iron, equivalent to 187 pg of 57Fe.Patients Four normal healthy women were recruited to assist in the study. All women had the routine haematological and biochemical tests performed to assess their iron status and all were found to be without evidence of anaemia or other medical disorders. All were non-smokers and none took oral contraceptive preparations. Ethical approval was obtained from the Newcastle District Health Authority. Patient Protocol Subjects attended the Newcastle research unit at 9.00 a.m. having fasted overnight for at least 10 h. An 8-ml basal blood sample was obtained by venepuncture and the IV cannula was left in situ. The IV dose of 57Fe was then administered, followed 5 min later by the oral dose, washed down with 60 ml of tap water.Further blood samples were obtained from the other arm 15, 30, 45, 60, 90, 120, 150, 180, 240, 300 and 360 min after the IV dose. A light breakfast was given after the 30-min blood sample and lunch after the 180-min sample, so that any effects of the increase in nausea and hunger during pregnancy could be minimised. The blood samples were allowed to clot at room temperature and the serum was separated and stored at -20 "C until required for analysis. ICP-MS Isotope Ratio Determination The combination of an atmospheric inductively coupled plasma and a quadrupole mass spectrometer was first des- cribed by Houk rt aZ.6 The sampling of ions from the plasma into the mass spectrometer is achieved with differentially pumped vacuum regions.The ICP-MS system used for these studies is described in detail elsewhere.7 The technique has been used successfully to carry out isotope ratio determina- tions of several elements in a variety of materials, for example, Br, Li, Fe and Zn in biological materials,8-11 Pb and Re in geological materials'2,13 and U in nuclear materials. 14 The determination of the entire range of iron isotope ratios by ICP-MS with sample introduction by conventional nebuli- sation is prone to significant error as 54Fe and "Fe suffer from severe polyatomic ion interferences by doArl4N+ and 40ArlhO+, respectively. Fig. 1 shows the blank spectrum of676 ANALYST, JUNE 1989, VOL. 114 this mass region, with 10 ng ml-1 of "Mn as a reference.As '-'Fe and "Fe were two of the three isotopes under investiga- tion (the third being "Fe), a method was required to determine isotopes of Fe in the absence of these interfering ions which are produced as the result of ion - molecule reactions between the plasma gases (Ar, H, 0 and N) in the interface region. Each, apart from Ar, is derived mainly from the water introduced with pneumatic nebulisation. Sample introduction with ETV instead of the normal nebuliser - spray chamber arrangement12 (Fig. 2) significantly reduces the levels of certain polyatomic ion interferences as samples are introduced in the absence of any accompanying 80 I 1 53 54 55 56 57 Massiu Fig. 1. Iron spectrum of the solution blank with I0 ng ml-* of -S'Mn. Integrated counts: 54 u , 3900; 55 u, 5300; 56 u, 4300; and 57 u.5 To ICP Additional flow (0.2 I min I ) I Volts and WE er out Volts and wi Carrier flow (0.5 I min-l) Fig. 2. Schematic diagram of the electrothermal system er in solvent. Fig. 3 shows the blank spectrum of the mass region 53-58 u in the absence of water, with the peaks at 54 and 56 u reduced almost to background levels. An ETV system allows gentle heating of the sample on a graphite rod, causing the solvent content of a sample to be driven off, leaving only a dry coating of the sample on the rod. Further ashing, at a higher temperature, can be carried out to "burn off" the organic matrix, followed by a 5-s high- temperature vaporisation to drive off the remaining material, containing the elements of interest. As vaporisation com- mences, data acquisition is initiated, in the knowledge that the peaks occurring at 54 and 56 u can only be due to iron.In addition, an ETV sample introduction system allows only very small volumes of sample to be introduced, in this instance 5 p1. This is particularly useful in biomedical studies, where there is often only a minimal volume of sample available. The TCP-MS operating conditions are summarised in Table 2. With a dwell time of 100 ps (100 sweeps), 1024 channels of the multi-channel analyser (MCA) were used; the scanning mode was in the range 53-58 u. Following data acquisition on the MCA, the data were stored and peak integration was carried out on an IBM-PC portable computer using "in- house" software. Isotope ratio calculations, including blank subtractions.were carried out using commercial spreadsheet software. The ICP-MS system (ion optics, plasma-sampling cone alignment) was optimised for a dry aerosol by first monitoring the signal for 12C+ (which could easily be generated from a blank firing of the ETV, but which was also present as a COZ impurity in the argon) , then by fine tuning the ion optics using the signal for 114Cd+, which was volatilised slowly from the rod.'* In the absence of a system for measuring the rod temperature, the applied voltage settings for drying and ashing were determined by trial and error. The optimum vaporisation temperature was determined by carrying out repeated iron standard vaporisations, each at a high rod setting, until the maximum signal response was obtained. Isotope ratio accuracy was assessed with aqueous solutions of natural iron (1 pgml-1) and the solutions of known enrich- ment from Harwell.On each day of analysis, the natural ratios were checked with a basal serum sample from each patient. hence enabling the level of background interferences to be assessed. In addition, before each sample was analysed, the interference levels were determined from a blank rod vapori- - - ~ ~ 'Table 2 . ICP-MS opcrating conditions Plasmapower . . . . . . Kcflectedpower . . . . . . Coolant Ar flow-rate . . . . Auxiliarytlow-ratc . . . . E'TV unit . . . . . . . . By-pass . . . . . . . . Sampling cone orificc diameter Skimmer cone orificc diameter Carrier flow-rate- . . 1300 W . . <low . . 14lmin-1 . . 0.5Imin-1 . . 0.5Imin-' . . 0.2lmin-' .. l m m . . 0.7 mm E 0 F L g 20 8 ul C 3 4- 10 t m L 0 0 54 55 56 57 Massiu Fig. 3. Iron spectrum of the graphite rod blank. Intgrated counts: 54 u, 315; 56 u, 378; and 57 u , 7ANALYST, JUNE 1989, VOL. 114 677 2.8 1 Table 3. Assay reproducibility. y1 = 10 L a, 2.0 - c.l 8 3 8 1.6 - 2 0.4 0 i l 54 55 56 57 Massiu Fig. 4. Iron spectrum of the basal serum. Integrated counts: 54u. 23 100; 56 u, 321 000; and 57 u , 10300 UY 8 1.2 m 0 3 0.8 + c 2 0 0 0.4 - a, 54 55 56 57 L 0 Massh Fig. 5. Iron spectrum of the enriched serum at f = 60 min. Integrated counts: 54 u, 88200; 56 u, 205000; and 57 u , 18400 sation. Each 5-pl serum sample was Subjected to at least five replicate analyses and between each sample the graphite rod was "cleaned" by performing several maximum-voltage vaporisations.Each set of samples for a given patient test occasion was analysed within 1 d in order to reduce variability. A basal quality control serum sample ( i e . , with natural Fe isotope ratios) was analysed each day to assess the across- assay variability. Pharmacokinetic Analysis The commonly used method for calculating absorption is to calculate the total area under the curve (AUC) of the logarithm of the enrichment in isotope ratio versus time15 for the oral "Fe : '6Fe (AUC,,,,) and the intravenous 57Fe : ShFe (AUCIV). The ratio of these areas after intravenous and oral administration of equal doses of a compound is equivalent to the fraction absorbed.16 When different doses are given, then oral absorption = (AUC,,,,/AUCIV)(doseIV/dose,,,,,) Results and Discussion For the solution blank the polyatomic peak at 54 u has a count equivalent to about 50 pg (Fig.1) and the use of ETV reduces this background by one order of magnitude (Fig. 2). The spectrum from a basal (i.e., not enriched) serum sample is shown in Fig. 4 and that of an enriched sample in Fig. 5 . These signals were generated from 3-5 ng or iron in 5 pl of serum; in basal samples this corresponds to about 2-300 pg of 54Fe. From Fig. 4 it can be seen that the isotope ratios are 0.0705 for 5AFe : '6Fe and 0.0306 for 57Fe : "Fe. Sixty minutes into the S4Fe S6Fc 57Fe : S6Fc CV* within-sample, YO . . 2.9 3.8 CV across-assay, YO . . . . 5.1 8.7 Limit of dctcction of cnrichmcnt (3SD) . . . . 0.0060 0.0033 * CV = coefficient of variation.f -1.2 +d r- I / $ -2.0 Y -2.8 /- I 5 I I 1 I 11 0 100 200 300 400 Timeimi n Fig. 6. 'Time course of Fc ratio enrichment in serum after A. oral '"c and B , intravenous "Fc administration to a non-pregnant woman test 54Fe :56Fe has increased to 0.4044 and 57Fe:"Fe to 0.1004. The assay reproducibilities and limits of detection are shown in Table 3. The precision of the measurement of isotope ratios was calculated over ten assays. For 54Fe : "Fe the average within-sample coefficient of variation (CV) was 2.9% [0.9% standard deviation (SD)]. Hence, for the average basal or natural sample 54Fe : "Fe was 0.0694 with an SD of 0.0020, and a sample with an isotope ratio >3SD, i.e. , above 0.0754, was significantly different. The limit of detection of enrichment was, therefore, 0.0060.For 57Fe : 5hFe, the aver- age basal ratio was 0.0289, the within-sample CV 3.8% (1.5% SD) and the limit of detection of enrichment 0.0033. Across- assay variation was 5.1% for 54Fe : "Fe and 8.7% for 57Fe : 56Fe. For a clinical assay, the precision was good. Calculation of the effect of within-sample variation showed that typically it would give a CV of 6.4% in the final absorption measurement. Fig. 6 shows the time course of the enrichment in Fe ratios over the period of study in one non-pregnant individual. The absorption of a 5-mg oral dose of iron(I1) sulphate over 6 h was 9.0% in this subject and 8.4, 8.9 and 17.2% in three other non-pregnant women (geometric mean 10.4%). A further nine women have now been studied throughout pregnancy and the results are still being analysed. Conclusion Past studies using radioactive isotopes17 have suggested that the absorption of iron increases during human pregnancy.However, reports differ widely about the degree of absorp- tion, probably because of the variation in clinical and analytical techniques, the doses used and the small number of subjects studied. The use of a whole body counter 2 weeks after the administration of "Fe and 59Fe suggested that the mean oral absorption of 3 mg of the iron(I1) salt was 35% (but with a range of 1745%) in eight healthy non-pregnant women. 18 Ethical considerations now prevent repetition in healthy subjects even using low doses. Stable isotope studies had until recently involved costly and complicated methodolo- gies but the advent of ICP-MS promises to simplify these important investigations.The existence of polyatomic ion interferences particularly at 54 and 56 u had initially restricted the scope of these studies. Janghorbani et al. 10 investigated the ingestion of one stable isotope of iron (SXFe) to determine iron availability in infants and children. However, the relatively rare 58Fe isotope is particularly expensive and requires a long sample counting time to achieve adequate precision. Further,678 ANALYST, JUNE 1989, VOL. 114 although the patient protocol was simple, requiring only one sample of red blood cells 2 weeks after ingestion, erythrocyte analysis requires assumptions to be made about the level of oral iron incorporation in erythrocytes, which may not be applicable during pregnancy.This study has chosen to use the double-isotope approach to assess short-term iron absorption, two concomitant tracers being held to give greater accuracy ‘9 and blood samples allowing a more simple patient protocol than the faecal collection used by Ting and Janghorbani.20 This work shows that the interfcrenccs from polyatomic ions can be overcome, hence opening up the potential for dual tracer techniques for the study of iron absorption. A further consideration is the dose required for successful tracer studies. The IV dose had to be sufficiently low not to harm the patients as a result of saturating their iron binding capacity, while being sufficiently high to be detected accurately. This work shows that 200 pg of 57Fe given intravenously will increase the natural 57Fe : 56Fe ratios from 0.030 to 0.100 with a limit of detection at an enrichment of 0.003; samples were still above this level 6 h post injection.Five milligrams of 54Fe is well within the average daily dietary intake of our normal subjects (about 10-1Smg) and, with a non-pregnant mean of 10% absorption in our 6-h study (about 500 pg absorbed into the serum), again allows a precise determination of isotope ratio enrichment. This new adaptation of ICP-MS offers excellent scope for studies of iron absorption during normal pregnancy and those complicated by anaemia and malabsorption. The authors are grateful to the charity Birthright for financial support of the clinical project. The ICP-MS unit (Surrey) is an NERC analytical facility.J. G. W. acknowledges additional support from the MOD (DQA). References I . King, J. C.. Raynolds, W. L., and Margen, S . , Am. J . Clin. Nurr., 1975, 31. 1198. 2 3. 4. 5 . 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. Miller, D. D., and van Campcn, D., Am. .I. Clin. Nutr., 1979, 32, 2354. Janghorbani, M., Ting, B. T. G., and Young, V. R.. J . Nutr., 1980. 110, 2190. Johnson, P. E., J . Nutr., 1982, 112, 1414. Dycr, N. C., and Brill, A. B . , “Nuclear Activation Techniques in the Life Sciences,” IAEA, Vienna, 1972, pp. 469477. Houk, R. S . , Fassel, V. A.. Flesch, G. D., Svec, H. J . , Gray, A. L., and Taylor, C. E., Anul. Clzem., 1980, 52, 2283. Datc, A . R . , and Gray, A. L., Anulyst, 1983, 108, 159. Janghorbani, M., Davies, T. A . , and ‘ring, B. T. G., Analyst, 1988, 113, 405. Sun, X. F., Ting, B. T. G., Zeisel, S. H., and Janghorbani, M.. Anafyvt, 1987, 112, 1223. Janghorbani. M., Ting, B. T. G., and Fomon, S . J., Am. J . Hemutol., 1986, 21, 277. Serfass, R. E., Thompson, J . J . , and Houk, R. S . , Anal. Chim. Acra, 1986, 188, 73. Date, A. R., and Cheung, Y . Y., Analyst, 1987, 112, 1531. Linder, M., Leich, D. A . , Borg, R. I., Russ, G . P . , Razan, J . M., Sirnons, D. S . , and Date, A. R.. Nuture (London), 1986, 320, 246. Russ, G . P., Bazan. J . M., and Date, A. R., Anul. Chem., 1987, 59, 984. Gibaldi, M., and Perrier, D., “Pharmacokinetics,” Marcel Dckker, New York, 1982, pp. 444-449. Gibaldi, M., and Perrier, D., “Pharmacokinetics,” Marcel Dekker. New York, 1982, pp. 169-175. Heinrich, H. C., Bartcls, H., Hcinsch, B., Hausmann, K., Kuse, R., Humke, W.. and Mauss, H. J . , Klin. Wochenschr., 1968, 46, 199. Sandberg, B., Acta Ohstet. Gynerol. Scund., Suppl.. 1975,48, 43. Werner, E . , Hansen, C., Wittmaack, K., Roth, P.. and Kaltwasser, J. P., Znserm. Symp. Ser. (Puris), 1983, 113, 201. Ting, B. T. G., and Janghorbani, M., Anul. Chem., 1986, 58, 1334. Paper 8/04872J Received December 19th, 1988 Accepted February 16th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400675

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST, JUNE 1989, VOL. 114 679 Secondary Ion Mass Spectrometric Determination of Impurities in Aluminium Oxide Hisashi Morikawa, Yoshinori Uwamino and Toshio lshizuka Analytical Chemistry Division, Chemistry Department, Government Industrial Research Institute, 1- I Hirate-cho, Kita-ku, Nag0 ya 462, Japan Impurities in aluminium oxide (alumina) were determined by secondary ion mass spectrometry (SIMS), which enabled the sample powder to be analysed without time-consuming pre-treatment stages. The difficulty of acquisition of standard samples associated with quantitative SIMS was overcome by preparing chemically the standard samples in powder form and applying the calibration graph method. The four elements (Ca, Fe, Ga and Ti) in an alumina reference material and commercial alumina samples were determined successfully and the results agreed well with those of inductively coupled plasma atomic emission spectrometry.The procedures are described and experimental results presented. Keywords: Secondary ion mass spectrometry; determination of impurities; aluminium oxide Aluminium oxide (alumina) is one of the advanced ceramics which are currently of importance in the progress of materials technology. The physical , electrical , thermal , mechanical and optical properties of the material enable it to be used widely in various fields of industry.' As most of these properties are affected significantly by impurities present in the alumina powder, their determination is important in the characterisa- tion of alumina. The impurities present in alumina have been determined successfully by several methods, particularly atomic absorp- tion spectrometry'-5 and inductively coupled plasma atomic emission spectrometry (ICP-AES) .'+g Both these methods, however, require the sample powder to be converted into solution, requiring special care and much time. On the other hand, other instrumental techniques, e.g., neutron activation analysis9 and X-ray fluorescence spectrometry,lO which do not require sample dissolution, have been reported for alumina.Secondary ion mass spectrometry (SIMS) has the potential to analyse a solid sample directly with high sensitivity and the capacity to detect all the elements from H to U while also providing lateral and in-depth distributions. 11 In quantitative analysis by SIMS, there are several quantifi- cation algorithms including a Calibration graph method, a thermodynamic approach based on a local thermal equilib- rium model and a matrix ion species ratio method based on the calculation of a series of sensitivity factors of elements, generated as a function of the sample chamber oxygen pressure. 12 Each method has been applied successfully in certain instances.13-16 The calibration graph method is recog- nised as a valuable technique. However, the acquisition of standards having the same matrix as the sample to be analysed is the limiting factor for practical applications owing to the matrix effects on the secondary ion yields of the elements. Hence, the preparation of standards for the calibration was studied first. The operating conditions, relative sensitivities of the elements, detection limits, analytical precision and appli- cations will be discussed in the following sections.Experimental Apparatus The quantitative analysis of alumina was conducted using an Hitachi IMA-2 secondary ion mass spectrometer equipped with a duoplasmatron primary ion source. High-purity oxygen was used in the ion source, producing the primary 02+ ion beam. The ions were accelerated to 7 keV relative to the sample potential and focused to 170-1000 pm in diameter on the sample surface. Instrumental parameters are summarised Table 1. Instrumental parameters Primary ion- Source . , . . . . Energy . . . . Current . , . . Spotsize . . . . Polarity . . . . Accelerating voltage Vacuum conditi0n.s- Primary ion column Samplechamber .. Electron multiplier Secondary ion- Detector- . . O,+ . . 7keV . . 2-4pA . . 17CL1000pm . . Positive . . 3kV . . 5 x 10--1Pa . . 3 x 10-5-4 x 1 0 - 5 in Table 1. Charge build-up effects at the surface due to ion bombardment were eliminated by mixing spectroscopic grade graphite powders (SP-1, National Carbon) with the sample powders. Reagents A stock solution of A1 (2% m/v> was prepared by dissolving A1 metal (99.999% pure, Soekawa Chemicals) in 1 M nitric acid. Commercial standard solutions for atomic absorption spectrometry (1 mg g-1 , Wako Pure Chemical) were also used as the stock solutions for Ga and Fe, and Titrisol (Merck) for Ti. The stock solutions of other elements were prepared from their nitrates. All other chemicals were of analytical-reagent grade.Samples The National Bureau of Standards Standard Reference Material (NBS SRM) 699 Alumina was used as a reference standard. Kanto Chemical (01 173), Katayama Chemical (01-2810) and Wako Chemical (012-0196s) alumina powders were analysed. Preparation of Calibration Standards The calibration standards were prepared by two distinct methods, i. e., chemical and physical. The chemical prepara- tion was as follows: mixtures of the stock solutions of the elements and the A1 stock solution were evaporated to dryness and then ignited at 1200 "C in alumina crucibles (99.9%) for conversion to the corresponding oxides. The physical stan-680 ANALYST, JUNE 1989, VOL. 114 .$ 103 = 102 t EJ 10' e 100 >z 'v, 105 E 104 103 c 4- ._ 106 4- c - 102 10' 100 10 20 30 40 50 60 70 ml z Fig. 1.Secondary ion mass spectra of (a) graphite powders and ( h ) NBS SRM 699 Alumina mixed with graphite powders. Alumina content is given in Table 2 dards were obtained by thorough mixing of the oxide powders of the elements with the alumina powders in a mixer/mill (Spex 8000). Two series of standards were prepared to prevent the interference from CaO on Fe at m/z 56. The first series consisted of Ca, Cr, Cu, Fe, Ga, Mg and Mn, each at concentrations of 10, SO, 250 and 500 c(g g-1, and the second of Fe and Ti, each at concentrations o f 1.25, SO and 250 pg g-1. Procedure The sample was powdered with an agate pestle in an agate mortar after which it was mixed well with graphite powder (sample - graphite, 2 + 1 m/m) and the mixture pressed at 200 kg cm-2.A sheet of PTFE (1 mm thick) was placed between a tungsten platen and the sample powder to prevent contact of the powder with the platen. The sample was obtained in the form of a pellet 10 mm in diameter and about 1 nim thick. This was then mounted on a sample holder and introduced into the sample chamber. Prior to analysis, pre-sputtering was performed for 10 min to clean the sample surface. This removed contaminants of monolayer coverage efficiently. Results and Discussion Background Signals The presence of a background signal in SIMS, as in any form of spectrometry, is a nuisance. As the sample was mixed with graphite powder, the intensities of the cluster ion, Cz+, and the molecular ion, AIC+, were not negligible and these ions produced backgrounds which interfered with the Mg+ and K+ signals, respectively (Fig.1). No other cluster ions or molecular ions contributed to the background because their mass numbers did not correspond to those of the elements of in tcres t . Homogeneity of the Prepared Samples for Standards The homogeneity of the standard samples was investigated and the deviations of a few ion intensity ratios obtained for chemically and physically prepared standards and for the NBS SRM were compared. The elements measured were Ca and Fe owing to their higher content in the NBS SRM. Each sample was analysed between five and eight times at five different points. The ion intensity ratios Ca/Al and Fe/Al are shown in Fig. 2. The matrix ions Al+, AlO+, A12+, AI20+ and their combinations were measured and AI+ was selected for Ca r, Fe I 0 1 Fig.2. Reproducibility of the ion intensity ratios Ca/Al and Fe/AI in the alumina powders. ( a ) NBS SRM 699 Alumina; ( b ) chemically prepared alumina; and (c) physically prepared alumina. Amount of Ca: ( a ) , 257 pg g-'; (6) and (c), 250 pg g-l. Amount of Fe: (a), 91 pg g-l; ( b ) and ( c ) , 250 pg g-l $0 I. I O L l;O 330 67 C 1000 Spot size urn Fig. 3. Kclative secondary ion intensities of Al, Ca and Fe as functions of the spot size o f the primary ion beam. Error bars represcnt the range ot values from five mcasurements. 0. Fe; 0, Al; and A, Ca normalisation purposes because of the minimum fluctuation in its signals; the other ions were relatively more sensitive to slight variations in the operating conditions.The relative standard deviations of the Ca/AI intensity ratios were 9.1, 0.4 and 60.6% for the NBS SRM 699 Alumina, the chemically prepared and the physically prepared samples, respectively. As a result of tests on the lateral homogeneities of NBS SRM Glass and Clay powder samples (89, 91, 93a and 98), Morgan and WernerlJ reported that a relative secondary ion current was reproducible to better than 20%. This indicates that the chemically prepared sample is as homogeneous as the NBS SRM 699 Alumina and hence suitable for use as the standard for quantitative SIMS. Spot Size of the Primary Beam The current transmission of an ion optical system and the maximum transmittable ion current in an optimally matched secondary ion mass spectrometer depend o n the spot size of the primary beam.It has been shown17 that a high trans- mission can easily be achieved for a small primary beam diameter. However, the smaller spot size results in a lower sensitivity owing to the smaller area analysed and the lower primary ion current. In this work the relationship between the spot size of the primary beam and the secondary ion intensity was studied. A spot size of 670 pm was found to be optimum for collecting secondary ions efficiently and gave the most reproducible ion intensity ratios u7ith our instrument (Fig. 3). The ion intensity ratios were found to be independent of the spot size of the primary beam. Relative Sensitivities The chemically prepaied sample containing 250 pg g- of each element was bombarded with a 7-keV 02+ beam.Fig. 4 showsANALYST, JUNE 1989, VOL. 114 68 1 > c Mg C T i Y o Cr Mn 0 o Fe Ga o 1 u l o *I0 20 30 Atomic number Fig. 4. Relative sensitivities of scvcral clcmcnts in alumina. Aluminium is taken as unity Concentration yg g Fig. 5 . a1 urn ina Calibration graphs for 0, Ca; 0. Fe; U, Ga; and 0, Ti in a plot of relative sensitivity against atomic number for eight elements (Mg, Ca, Ti, Cr, Fe, Mn, Cu and Ga). Each value was corrected for the natural abundance of the isotope, and the value for Al was taken to be unity. These results are similar to those for NBS SRM 466 Low-Alloy Steel.’* Detection Limits The detection limits for Ca, Cr, Cu, Fe, Ga, Mg, Mn and Ti were 4, 10, 10,23, 1, 14,3 and 1 pg g-’, respectively, and were defined as the concentrations producing signals equal to three times the background fluctuations.The detection limits for Cr and Fe were higher than might have been expected from the relative sensitivities comparcd with the other elements studied (Fig. 3). This may be due to background formations arising from sputtering of instrument components (stainless steel) by scattered ions (O+, 02+, etc.) or the neutral molecule 02. Calibration Graphs The relationships between the concentrations of the elements in alumina and the ion intensity ratios M/Al were examined. Fig. 5 shows the calibration graphs obtained for Ca, Ti, Fe and Ga; these were linear in the concentration range 1-500 pg g-1. The linear range extends over a further two orders of magnitude of concentration. This wide dynamic range is a further advantage of SIMS.Analytical Precision When the sample surface was sputtered continuously, changes occurred in the surface topography, surface electrical states and the intensity of the primary ions. The ultimate reproduci- bility [coefficient of variation (CV)] of the secondary ion Table 2. Analytical rcsults for NBS SRM 699 Alumina Element Ca . . . . Cr . . . . Cu . . . . Fe . . . . Ga . . . . Mg . . . . Mn . . . . Si . . . . Ti . . . . SIMS*/ t-lg g-‘ 237 f 8 < 10 < 10 113 f 14 72 k 7 < 14 <3 L - 5 f 1 ICP-AES*/ g-’ 257 k 1 <1 1.5 f 0.1 96 f 0.6 75 * 0.6 2.7 f 0.02 3.3 k 0.02 69 f 7.5 3.6 f 0.1 Certified value/yg 8-1 257 1.4 4 91 74 3.6 3.9 65 6 * Mean content k standard deviation for five determinations. t Not determined. Table 3.Analytical results for practical alumina samples Element Ca . . . . Cr . . . . c u . . . . Fe . . . . G a . . . . M g . . . . Mn . . Ti . . . . Kanto*/u.g g-’ Katayama*/pg g-I 27 k 4 (29) 45 * 2 (49) < I 0 (< 1) <10 (<1) <10(<1) <10 (<1) 68 k 4 (67) 72 * 8 (64) 71 k 4(78) 55 k 8 (51) < I4 (0.6) <14 (2.6) <3 (0.4) <3 (0.4) 17 * l(16) 21 +2(19) Wako*/pg g-’ 29 * 1 (29) <10(<1) < 10 (< I ) 62 k 2 (62) 55 k 4 (49) <14(1.7) <3 (0.2) 24 k 4 (18) * Mean content k standard dcviation for five determinations. Values in parenthescs wcre obtained by ICP-AES. intensity ratios was calculated to be within 35% ; this allows for some sample inhomogeneity and instrumental variations over the time during which data were collected.Even allowing for such changes, however, the deviation (CV) in secondary ion intensity ratios was <lo% for a cycle of measurements. Applications The NBS SRM 699 Alumina was analysed by the proposed method; the results are given in Table 2 with the reference values determined by ICP-AES. The good agreement between these values provided some evidence that, in addition to ICP-AES, the SIMS method is also reliable. Impurities in the commercial samples of alumina were also determined by the proposed method. The results are given in Table 3, from which it can be seen that the values for some of the elements in the sample powders were in good agreement with ICP-AES measurements. Conclusion The comparison between SIMS and ICP-AES shows that SIMS is the inferior technique in terms of accuracy, sensitivity and reproducibility.However, this study has emphasised one of the advantages of SIMS, namely its capacity to analyse intractable materials without the difficulties associated with sample dissolution or decomposition. References 1. Gitzen, W. H., Editor, “Alumina as a Ceramic Material,” Amcrican Ceramic Society, Columbus, OH, 1970. 2. Khavezov, I., and Tamnov, B . K . , Freseniuh Z. And. Chem., 1978, 290, 299. 3. Palmer, T. A . , and Winkler, J. M., Anal. Chim. Acta, 1980, 113. 301. 4. Nikolova, B., and Iordanov, N., Talanta, 1982, 29. 861. 5 . van der Walt. T. N., and Strelow, F. W. E., Anal. Chcrn., 1985, 57, 2889. 6. Tshizuka. T., Uwamino, Y., Tsuge, A , , and Kamiyanagi, T.. Anal. Chim. Acta, 1984, 161, 285.682 ANALYST, JUNE 1989, VOL. 114 7. 8. 9. 10. 11. 12. 13. 14. Morikawa, H., Iida, Y., Ishizuka, T., and Yokota, F., Bunseki Kagaku, 1986, 35, 636. Harada, Y., and Kurata, N., Bunseki Kagaku, 1986, 35, 641. Foner, H. A . , Analyst, 1984, 109, 1469. Bennett, H., Oliver, G . J., and Holmes, M., Trans. J. Br. Ceram. Soc., 1977, 76, 11. Grasserbauer, M. , Stingeder, G. , Wilhartitz, P., Schreiner, M., and Traxlmayr, U., Mikrochim. Actu, 1984, 111, 317. Benninghoven, A . , Rudenauer, F. G., and Werner, H. W., Editors, “Secondary Ion Mass Spectrometry,” Wiley-Inter- science, New York, 1987, p. 287. Ishizuka, T., Anal. Chern., 1974, 46, 1487. Morgan, A. E., and Werner, H. W., Anal. Chem., 1977, 49, 927. 15. 16. 17. Ganjei, J. D., Leta, D. P., and Morrison, G. H., Anal. Chem., 1978, 50, 285. Andersen, C. A , , and Hinthorne, J . R., Anal. Chern., 1973,45, 1421. Benninghoven, A., Rudenauer, F. G., and Werner, H. W., Editors, “Secondary Ion Mass Spectrometry,” Wiley-Inter- science, New York, 1987, p. 455. Tsunoyama, K., Ohashi, Y., and Suzuki, T., Anal. Chem., 1976, 48, 832. 18. Paper 8J04136 1 Received October 18th, 1988 Accepted January 13th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400679

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST. JUNE 1989. VOL. 114 683 Reconstruction of Constituent Spectra for Individual Samples Through Principal Component Analysis of Near-infrared Spectra Ian A. Cowe, James W. McNicol and D. Clifford Cuthbertson Scottish Crop Research Institute, Mylnefield, Invergowrie, Dundee DO2 5DA, UK Compound component weights for a specific constituent were combined with near-infrared spectra of individual samples to produce a form of spectral reconstruction which highlighted the influence of individual absorbance bands in the estimation of a fitted value for the constituent. The technique is illustrated using a set of wheat flour spectra with corresponding values for protein and moisture. For moisture, all samples exhibited variation a t two points in the spectrum where absorbance bands for moisture are known.There was little variation between samples in the relative responses of these bands. For protein, however, individual samples exhibited as few as two and as many as six sources of variation which were used by the model to estimate the protein content of the sample. Not all these sources of variation related to known protein bands, indicating that the model was sensitive to the presence of other constituents such as starch. ”Null” points, where adjacent absorbance effects were always in balance, were identified for both moisture and protein. Variation in particle size of samples was shown to distort reconstructed spectra. A simple algorithm using ”null” points for protein was shown to reduce this distortion and enabled absorption effects to be more clearly observed.Keywords: Principal components; near infrared; graphics; particle size; wheat flour Data compression techniques, such as Fourier analysis,l.‘ partial least squares3--’ and, i n particular. principal component -7 have all been shown t o be useful for relating near-infrared diffuse reflectance (NIR) spectra to the compo- sition of samples. Each of these techniques involves a transformation of the original data to new variables which have better properties than the raw spectra. While all involved increased computation, the time required to produce a regression model that adequately predicts composition may in fact be reduced, as fewer combinations of variates need be examined. In addition, many of these compression techniques provide additional information which can give greater con- fidence that the model is indeed predicting the constituent of interest, rather than some other correlated variate. In previous publications7-” we have shown that principal component weights can be plotted in the same way ;is spectra.Using such plots. absorption bands relating to individual constituents can be identified. a s can effects relating to physical factors such as particle size variation. Taken one stage further. component weights can be combined with regression coefficients to produce compound component weiights,l!’ a form of spectra 1 re cons t r uc t i on that ex hi bits character i s t ics similar to the spectrum of the constituent to which the regression is related. Compound weights are derived from an analysis of the spectral variation of all the samples in a calibration set.They represent an “average” response. A spectral analogue pro- filing the response of an individual sample, with respect to a specific constituent such as moisture or protein, would allow a study to be made of the degree by which samples vary in the expression of absorbance bands relating to that constituent. This paper presents a method that attempts to achieve these objectives using combinations of compound weights and centred spectra. Experiment a1 Derivation of Sample Specific Compound Weights The theory of principal components has been described elsewhere, 1 1 Previous papers have discussed the nature of principal component weights.’ principal component scores7 and compound weights10 and their interpretation with respect to NIR spectra: such discussions are not repeated here. To define the variates plotted however.it is necessary to repeat the definition of compound weights. In reference 10, equation (4) defined a regressiori model using principal components a s follows: Y = B() + ClEl + G& + . . . . . . . . + G-i.()&(’,) (1) Here Y is the fitted chemical value, E l to E700 represent centred spectral data, where the centred value at any wavelength is the actual spectral value minus the mean Ldue at that wavelength taken over all samples, and B,) is the intercept coefficient; C , to C;700 are compound weights. where the compound weight for the nth wavelength is defincd by (1 G,l = c R,C,,, . . . . . . (2) i = I where B I to B, are regression coefficients relating principal component scores to composition of samples and Cl,, to Cq,, are principal component weights for q selected components.Compound weights (G) therefore describe, in terms of all the samples, the relative importance of the data at each wavelength to the fitted value for the constituent of interest. Equation (1) presents a principal component regression model in its equivalent wavelength form; G I to Gj(lo are the regression coefficients that would be obtained if all wavelengths werc included i n the equation In this paper we concentrate on an examination of the behaviour of sample specific compound weights, which are denoted by ( G E ) . From equation ( 1 ) it can be seen that the deviation of the estimatcd value from the population mean (Bo) is a summation of the regression coefficients (G) times the centred energy values ( E ) over all wavelengths.As principal components arc derived from centred energy values (the difference between the actual and mean response at any wavelength), centred values must be used i n this reconstruc- tion. By plotting GiE, against the corresponding wavelength i. the contribution at each wavelength to the fitted value for an individual sample can be observed. Each set of compound weights (GE) i s specific t o a single sample, as each saniple is defined by its spectrum (El lo E700), and to a specific regression model for one constituent as defined by the wavelength regression coefficients (GI to G7()!)). A different constituent, or regression model, would involve a different set of regression coefficients.Hence for any constituent. we are :ihle to examine the responses of individualANALYST, JUNE 1989. VOL. 114 684 4.00 2.00 0 2 -2.00 c L 0 .- a, -4.00 2 1.74 a, u) c .7 = 1.09 2 0.44 -0.21 - 0.86 6.17 I c3 3.53 2 3 0.89 u m a, .- a, u) 3 c - 1.75 '5 a -4.39 1100 1300 1500 1700 1900 2100 2300 2500 Wavelengthinm Fig. 1. Reconstructed NIR spectra for A, water and B, protein derived from samples of wheat flour. ( a ) Effects of dividing each set of compound weights by the standard deviation of the constituent values. ( h ) Compound weights scaled individually samples to the regression model and determine the extent to which absorbance bands relating to the constituent are expressed in that sample. Samples A set of 39 spectra of samples of wheat flour obtained from the Flour Milling and Baking Research Association (Chorley- wood, UK), with analytical values for moisture and protein was used.All spectra were ccllected as 700 data points at 2-nm intervals from 1100 t o 2498 nm. Before derivation of the components, the first and last ten data points were discarded to avoid the influence of edge effects from the optics and a five-point moving average was used to smooth the spectra. Adaptation from Published Method Earlier publications detailed the shapes of the first six principal components for this sample set,' and of compound weights derived for two regression models, 1 0 the first predict- ing moisture content the other, protein. The graphs shown in reference 1 0 were scaled in terms of deviations from the mean concentration of a specific constituent and no effort was made to produce a common scale for graphs of compound weights.In fact, the scaling was chosen to emphasise the variation in weight within each graph. For data reported in this paper, compound weights for all constituents are now scaled in the same units. To standardise the compound weights we divided, in each case, by the standard deviation of the constituent values. This standardisa- tion produces compound weights scaled in terms of standard deviations, enabling a direct comparison of the reconstructed spectra for different constituents or for the same constituent in different samples regardless of the inherent range of concen- trations found in the sample set.Results and Discussion Reconstructed Sample Spectra for Moisture and Protein Fig. l ( n ) shows reconstructed NIR spectra (compound weights G) for water and protein derived from 39 samples of wheat flour. In reference 10, compound weights for moisture were derived using the first three principal components and those for protein using components one, three, four and five. Here, the first five components were used to derive compound 'Table 1. Refcrence oven dricd moisture and Kjeldahl protein values together with fittcd values by NIR for four of the original wheat flour samples. Regression model for moisturc used components 1 and 2: for protein components I . 3 and 4 werc uwd Moisture, YO Pro t e i n , (Yo Sample Reference Fitted Reference Fittcd 1 12.40 12.53 13.80 13.74 3 14.30 14.36 8.40 8.84 12 14.90 14.75 13.90 14.37 16 12.00 11.83 8.90 8.51 0.05 - % - 0.02 0 - - 8 0 + -0.02 2 m n -0.05 0.20 0.10 0 l- Lc 0 -0.19 .?? -0.20 P G 0 ._ w $ 0.20 0.10 m 0 -0.1c -0.2c lb) I I I I I I I 1100 1300 1500 1700 1900 2100 2300 2500 Wavelengthinm Fig.2. ( u ) Centred NIR \pectra, ( h ) reconstructed moisture spectra and ( c ) protein spectra for tour typical samples o f wheat tlour. A, sample 1 ; B, sample 3; C. sample 12: and D. sample 16 weights for both moisture and protein. This combination was used to illustrate that the regression coefficients, rather than the combination of components determine the shapes of sample specific compound weights ( G E ) . Comparing Fig. l ( u ) with Fig. l(h) shows the effects of standardising the scales.Four representative samples were selected to illustrate features of sample specific compound weights. Table 1 shows values for actual and fitted moisture and protein concentra- tions for these samples. Fig. 2(a) illustrates how the four centred spectra differ from the mean spectrum (shown as the centre line) derived using all 39 samples. In these plots, the same absorbance band may appear either as a local maximum or as a local minimum depending on whether, for that sample, the band intensity was greater or smaller than average. All four samples exhibited increasing variation towards the upper end of the spectrum. Sample 3 appeared to lack any clearly defined absorbance bands. The only absorbance effects observed in samples 12 and 16 related to water (1940 nm).For sample 1, this water band was less well defined and other bands were present in the upper part of the spectrum. Fig. 2(6) shows sample specific compound weights (GE) for moisture. The dominant feature in these plots was the clearlyANALYST, JUNE 1989. VOL. 114 685 defined absorbance band centred at 1940 nm. The lower half of the spectrum was almost featureless with only small absorbance effects visible in the regions 1400-1550 and lhS(L1800 nm. For all samples, the sign of the deviations in the first of these regions was always the same as that for the band at 1940 nm. In the second region the sign was always opposite to that found at 1940 nm. From 2048 to 2390 nm there was a broad region with at least two bands that acted in opposition to the band at 1940 nm.As the fitted value for any sample is the sum of all the GiEi values across the spectrum, it can be seen to be a weighted average of all these effects. While the sample with the highest moisture content (sample 12) showed the greatest positive response at 1940 nm, and the sample with the lowest moisture (sample 16) had the greatest negative response, the correlation of the sample specific compound weights (GE) at 1940 nm with the reference moisture values for all 39 samples was low (Y = 0.35). Similarly at 1450 nm, the correlation was 0.23. However, multiple regression using 1940 and 1450 nm gave a correlation of 0.83. This indicates that these two known moisture bands]’ rep- resent much of the essential information required to predict moisture.A much more complex pattern was observed for protein [Fig. 2(c)]. At least five distinct regions. most incorporating more than one band, were identified. The two largest effects were centred around 1980 and 2200 nm. both known protein bands.” The high protein samples (samples 1 and 12) both had local maxima at these points, while the low protein samples (samples 3 and 16) exhibited local minima. At 2100 nm, a band known to relate to starch,l’ high protein samples exhibited local minima. while low protein samples had local maxima. This indicates that the amount of starch present in the samples is negatively correlated with protein and is used in the prediction of protein content. Between 1400 and 1700 nm two adjacent regions of the spectrum also acted in opposition to the protein bands at 1980 and 2200 nm, as did the region between 1750 and 1900 nm.Not all bands were present in all 39 samples. Although all the bands were highly intercorrelated (Y > 0.95), the relative responses of bands did change from sample to sample. The four samples used in these graphs illustrate this point. Of the two high protein samples, sample 1 had the greater response at 2200 nm, but exhibited a smaller response than sample 12 at 1980 nm. For the two low protein samples the responses at 2200 nm were similar, but at 1980 nm sample 16 showed a much greater negative response than sample 3. For the protein reconstructed spectra, the correlation of peak height with the reference protein content was uniformly low ( Y between 0.34 and 0.09) for all the bands.An “all possible pairs” search using the data at bands found in these reconstructed spectra showed that some combinations had correlations around Y = 0.72. However, one combination (1980 + 2180 nm) had a multiple correlation of 0.99. The band centred on 2200 nm appears to consist of overlapping absorbance bands in the range 218&2210 nm and any of these in combination with 1980 nm produced a high multiple correlation with reference values for protein. As the bands centred on 2200 and 1980 nm represent the largest sources of variation, and together are the only two bands which accurately predict protein in a two-term regression equation, we conclude that they provide two independent sources of information relating to protein. One interesting feature of Fig.2(b) and (c) was the set of points for which all 39 samples had values equal to the mean. These “null” points lie between bands where the signs of the correlations with the constituent of interest are opposite. At these points in the spectrum. opposing absorbance effects are always in balance, in effect cancelling each other out. Where two bands had the same correlation (e.g., the broad bands centred on 1460 and 1580 nm) there may be a local minimum but no “null” point occurs. In Fig. 2(b), the “null” points on either side of the moisture band at 1940 nm were 1830 and 2048 nm. Although the latter point almost coincides with 2050 nm, a known protein band,lz any absorbance effects relating to protein are absent. Protein “null” points [see Fig. 2(c)] were found at 1356, 1682, 1746, 1930,2066,2144 and 2398 nm.Of these, 2144 nm is particularly interesting in that it lies between the protein band at 2180 nm and the starch band at 2100 nm. Previously, Norris and Williams13 used first and second derivatives of log 1/R ( R = reflectance) in the region 2170-2210 nm to predict the protein content of hard red spring wheat. These derivatives characterise the slope of the spectrum between the protein band at 2180 nm and the “null” point at 2144 nm, and will be influenced both by the protein and starch in the samples. Effects of Particle Size on Reconstructed Spectra Close examination of all 39 reconstructed ( G E ) spectra for both moisture and protein indicated that the magnitude of the response for any band, regardless of the sign of that response, related more to particle size influences than any absorbance effect.Whereas the compound weights (G) were relatively free of particle size influences, due to the fact that the first principal component has a relatively flat shape and low regression coefficient, the centred spectra ( E ) were influenced significantly by particle size [see Fig. 2(a)]. Sample spectra which deviated most from the mean produced reconstructed spectra ( G E ) with the largest responses at any of the absorbance bands, irrespective of the concentration of either constituent of interest. Other workers have suggested methods to compensate for particle size variation. 13-15 The presence of “null” points suggests a simple method of adjusting for this effect.If a constituent can be found that is highly correlated with particle size and then widely separated “null” points selected where the absorbance of this constituent does not vary, then at these points in the spectrum variation between samples should relate primarily to particle size. A straight line through these points will describe approximately the influence of particle size on each individual sample. Adjusting the sample spectra so that all these straight lines are collinear will minimise particle size influences. For these 39 wheat samples, protein was a suitable constituent for this purpose as it was highly correlated with the first component ( Y = 0.70), which in turn related to particle size. We wished to identify “null” points where the opposing effects due to absorbing constituents were likely to occur in all samples, and not just this data set.The two highest protein “null” points were 2398 and 2144 nm. The upper point was discarded as the causes of the opposing effects were unclear, although edge effects from the optics may be involved. The next “null” point (2144 nm) appeared to be ideal as it related to a balance between starch and protein absorption bands. As these constituents form most of the dry matter in wheat, the negative relationship between their concentrations is unlikely to change for any set of wheat flour samples. There were no “null” points at the lower end of the spectrum; 1120 nm was therefore selected as the lower correction point as all samples tended to have fairly constant absorbance values at this point.Fig. 2(u) shows that, apart from absorbance effects relating to water, the slopes of centred spectra are relatively linear. If the slope of each centred spectrum is defined by a straight line through the energy values at 1120 and 2144 nm, then all centred spectra can be adjusted so that all slopes are superimposed on the slope of the mean sample spectrum. Similarly, individual values at each wavelength are adjusted by a factor representing the degree of rotation and translation required to superimpose the slopes. The equation used is given by E:ll = (E,-&) - Bi . . (3)686 ANALYST, JUNE 1989, VOL. 114 m 0.02 0 - - 8 0 T II $ -0.02 I) Q - 0.05 0.05 0.02 z o + 0 g -0.02 .- c m .- b -0.05 -0 p 0.08 m -0 6 0.04 5 0 -0.04 t I t I I 1 I -0.08 ’ ’ ’ 1100 1300 1500 1700 1900 2100 2300 2500 Wavelengthinm Fig.3. ( u ) Ccntred NIK spcctra. ( h ) reconstructed rnoisttirc spcctra and ( c ) protein spectra after correction for particle size intluences for f o u r t).pical samplcs of wheat tlour. A , samplc I ; B. samplc 3: C , sarnplc I?: and D, sample 16 Ivhere B = ( 1 5 ( ~ - E ~ ) / s . given that E , , and E, are the adjusted and original energy values at the ith data point, E l is the energj value at the lower correction point (1 120 nm). E, is the energy value at the upper correction point (2144 nm), B i\ the \lope o f the regression line (or the unit increace i n absorbance due to particle size) and 5 is the number of data points between the upper and lower correction points. Reconstructed Sample Spectra for Moisture and Protein After Correction for Particle Size Effects Fig.3 shows equivalent plots to Fig. 2, but after correction for particle size. The most dramatic change is in the centred spectra [Fig. 3(c()]. No longer is there any consistent increase i n \,ariation from the mean across the spectrum. The water bands at 1910 and 1450 nm are still the dominant features, but arc much more obvious. The reconstructed moisture spectra [Fig. 3(h] are very similar to those of Fig. 2 ( h ) , except that the variation above 2050 nm is greatly reduced. If any distortion was introduced as a result o f correcting the spectra, then moisture would be the most likely constituent to be affected and would serve as a check of the distortion. Whereas the four samples still did not rank in order of moisture content, samples 1 and 3 were closer to their anticipated positions. The correlation between the individual sample compound weights (GE) at 1940 nm and the reference moisture values increased after correction for particle size ( Y = 0.65 compared with r = 0.35) and at 1450 nm ( Y = 0.32 compared with I‘ = 0.23).The multiple corrclation of 1940 and 1150 nm with reference moisture content values also increased (from Y = 0.83 to 0.94) while the intercorrelation between values at 1940 and 1450 nm dropped from I’ = 0.99 to 0.92. For protein, the shapes of the reconstructed spectra changed considerably. No longer were the greatest responses, whether positive or negative, around 2200 nm. This band virtually disappeared. Instead, the dominant band was the protein band at 1980 nm.Whereas the band centred at 1460 nm was relatively unchanged from Fig. 2(c), the adjacent band centred at 1580 nm did change. Correcting tor particle size reduced the intercorrelation of the bands. Although some adjacent bands still correlated highly, there was now no general trend to high correlation across the spectrum. An interesting feature o f this data was the behaviour of two pairs of protein bands (1980 and 2050 nm, 2200 and 2346 nm). The band at 2050 nm appeared as a shoulder on the strong band at 1980 nm and as a consequence the two were highly correlated ( r = 0.91). The other pair, although further apart, were also highly correlated (Y = 0.86). There was, however, no correlation between the pairs ( r = 0.17 between 1980 and 2200 nm was the highest correlation).Further, bands in the lower half of the spectrum could be separated into two groups, being correlated highly with one pair, but never with both. For example. the bands at 1452 and 1582 nm were both highly negatively correlated with 1980 and 2050 nm ( I ‘ between -0.74 and -0.91,) but showed little correlation with either bands at 2200 or 2346 nm ( r < 0.34). The band at 1730 nm. however, was highly positively correlated with either of the higher pair ( Y > 0.74) and showed much lower correlations with 1980 or 2050 nm ( r < 0.43). The band at 1766 nm behaved similarly, although it uas negatively correlated with 2200 and 2346 nm. These results \upport the suggestion that the reconstructed sample spectra for protein contained two independent sources of variation relating to protein.Whether these related to different types ot protein or to different molecular groups found in all proteins is not clear. Surprisingly, now that the intercorrelation of bands had been reduced, a multiple regre,\ion using one band from each of the sources of variation had a lower correlation than before correction for particle 5ize ( r reduced from 0.99 to 0.73 tor 1980 i n combination with 2200 nm). This trend wa4 opposite to that found for moisture. A comparison of Figs. 2(h) and 3(h) shows that adjusting for particle size did not change the absorbance bands relating to moisture. For protein, however, [Figs. 2(c) and 3(c)] adjusting for particle size suppressed the bands around 2200 nm.As a consequence variation around 2200 nm in the adjusted protein spectra no longer appears to comple- ment the variation at 1980 nm. In this paper, use of a particle size correction algorithm has been considered solely in terms of the effects on individual sample reconstructed spectra. Having shown that one algorithm breaks the intercorrelation of bands, the obvious step is to apply particle size correction algorithms to raw spectra, examine how the principal components change and determine whether regression models based on these com- ponents improve the prediction of moisture and protein content of the samples. This work will be reported in a later pubiication. Conclusions Although the recon\tructed cample spectra for moisture and protein are by no means ideal solutions to the problem of revealing absorbance bands relating to constituents in the spectra of samples, they do highlight differences between samples in their response to known absorbance bands.Ucing the reconstructed spectra ( G E ) as a data set, measurements at two known absorbance bands for protein (1980 and 2180 nm) correlated highly with protein. This indicates the importance of these bands to regression models used to predict composi- tion. The identification of “null” pointc suggests one mechanism to overcome particle size variation. Although, in this instance, the correction for particle size was carried out on the basis of a wavelength chosen for its relevance to protein, it did not appear to &\tort the reconstructed moisture spectra.For protein, variation was distributed quite differently after correction for particle size.ANALYST, JUNE 1989, VOL. 114 The graphs presented in this paper still do not allow measurements at a single point in the spectrum to be directly related to either protein or moisture content of samples. For both moisture and protein there appear to be at least two distinct sources of variation which must be taken into account. The graphs do, however, provide clues as to the response of individual samples to the regression models. 1. 2. 3. 4. 5 . 6. References McClure, W. F., Hamid, A . , Giesbrecht, F. G . , and Weeks, W. W., Appl. Spectrosc., 1981, 38, 301. Giesbrecht, F. G . , McClure, W. F., and Hamid, A . , Appl. Spectrosc., 1981, 35, 210. Sjostrom, M., Wold, S., Lindberg, W., Person, J . , and Martens, H., Anal. Chim. Acta, 1983, 150, 61. Martens, H . . and Jensen, S. A., in Holas, J., and Kratochvil, J . , Editors, “Proceedings of the 7th World Cereal and Bread Congress, Prague, June 1982,” Elsevier, Amsterdam, 1983, Mark, H., Chimicaoggi, 1987, 10, 57. Robert, P., and Bertrand, D . , Sci. Aliments, 1985. 36, 505. pp. 607-647. 7. 8. 9. 10. 11. 12. 13. 14. 15. 687 Cowe, I. A . , and McNicol, J. W., Appl. Spectrosc., 1985, 39, 1 C 7 Cowe, I. A . , McNicol, J . W., and Cuthbertson. D. C., Anulyst, 1985, 110, 1227. Cowe, I. A . , McNicol, J. W., and Cuthbertson, D . C., Analyst, 1985, 110, 1233. Cowe, 1. A . , McNicol, J . W., and Cuthbertson. D. C., Analyst, 1988, 113, 269. Mardia, K. V., Kent, J . T . , and Bibby, J . M.. “Multivariate Analysis,” Academic Press, London, 1979. Osborne, B. G . , and Fearn, T., “Near Infrared Spectroscopy in Food Analysis,” Longman Scientific and Technical, Harlow, 1986. Norris, K. H., and Williams, P. C . , Cereal Chem., 1984, 61, 158. Murray, I. , and Hall, P. A . , Anal. Proc., 1983, 20, 75. Ciurczak, E . W., Torlini, R . P., and Demkowicz, M. P., Spectroscopy, 1986, 1(7), 37. Paper 8104613A Received November 2Ist, 1988 Accepted January 6th, I989

ISSN:0003-2654    

DOI:10.1039/AN9891400683

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST, JUNE 1989, VOL. 114 689 Linearity Range of Gran Plots from Logarithmic Diagrams Carlo Macca Department of Inorganic, Metallorganic and Analytical Chemistry, Universita di Padova, Via Marzolo 1, 1-35131 Padova, Italy Simple criteria for assessing the linearity ranges of Gran plots were implemented by the use of logarithmic diagrams of equilibrium concentrations. One-step precipitation, complexometric and acid - base titrations are considered. The graphical procedures do not require some of the approximations that are needed in order to obtain convenient and simple mathematical equations. Keywords: Potentiometric titration; Gran plots; logarithmic diagrams Simple criteria and procedures for assessing which part of a potentiometric titration curve can be used to obtain appre- ciably linear Gran plots’ and an accurate extrapolation to the equivalence volume for one-step precipitation, complexation, redox and acid - base titrations were discussed in a previous paper.’ Equations were derived for calculating the limiting or ”critical” values of the experimental parameters or variables that prevent systematic titration errors higher than an assigned value.In order to ensure simplicity and convenience of use of these equations, approximatiom had to be introduced, thus excluding some less obvious situations from the applicability of the final equations. Although this restriction is not substantial for most of the simple titrations dealt with previously,l it can become a real limitation for equilibria of some complexity, for example, with some weak-acid or weak-base titrations.Moreover, it can be reasonably antici- pated that for multi-step titrations (for instance, titrations of polyprotic acids or of mixtures of analytes), it would be almost impossible to obtain simple equations without introducing excessive approximations. It is well known that logarithmic diagrams of equilibrium concentrations of titration systems represent a very con- venient tool for solving complicated equilibrium problems, and provide a useful support for rationalising approximations and simplifications in the mathematical solution of the same problems.j--h The aim of this paper is to show that logarithmic diagrams can provide a simple, rational and correct solution to the problem of assessing the linearity ranges of Gran functions. Again, only one-step titrations will be considered.Nevertheless, it will be shown that the usefulness of the graphical approach, which does not necessarily resort to substantial approximations, increases as the complexity of the titrated system increases. Multi-step titrations will be dealt with i n a subsequent paper. Experimental Apparatus A Radiometer (Copenhagen) PHM 84 digital pH meter (0.1 mV, 0.001 pH), a Radiometer ABUXO motor-driven Auto- burette equipped with a B282 25.0(L2.500-ml burette and a Metrohm 6.1418.220 titration vessel with a circulation jacket containing water at 25.0 k 0 . 1 “C were used. Electrodes For the titration of calcium. a Crytur (Turnov, Czechoslo- vakia) 09-17 fluoride ion-selective electrode and a Radi- ometer K711 saturated calomel reference electrode were used.For the titration of pyridinium chloride a Radiometer GK 2301 combined glass electrode was employed. Reugerzts All reagents were of analytical-reagent grade, and water was redistilled from quartz. Principles of the Logarithmic Representation of Titration Equilibria The graphical representation with bilogarithmic co-ordinates of the equilibrium concentrations of the species taking part in a titration’ (the so-called logarithmic diagrams) has found wide acceptance and increasing application as an effective substitute for the conventional approach to the solution of equilibrium problems.46 When used in combination with the titration error equation (according to the original approach of Bjerrum”), logarithmic diagrams can account for all the important properties of the titration.’ The extensive literature on the subject can be consulted for details of the principles, construction (by hand on graph paper) and general use of logarithmic diagrams, but only for a few basic concepts.The logarithm of the equilibrium concen- tration or, when convenient, of the conditional concentrations of any ith species taking part in the titration equilibria, log[i], is plotted against pH for acid - base titrations, against p[titrant] for complexometric and precipitation titrations and against pE = EFI(RTlnl0) for redox titrations. The only assumptions are negligible dilution during the titration, i.e., a constant total analyte concentration (usually taken to be equal to the initial concentration CO) and constant activity coeffi- cients and side-reaction coefficients.8 Both assumptions can be satisfied in practice.In some instances, the solution of specific problems can be made easier by introducing auxiliary curves.5 Linearity Criteria for Gran Plots Extrapolation through two selected points of the Gran plot for a given titration, viz., the initial point (titrant volume V = 0. titration ratio f = V/VO = 0) and a second point (the “critical point”) satisfying the conditions represented by the equations in column 4 o f Table 1 that are labelled B.E.P., gives a systematic (relative) error exactly equal to a pre-selected arbitrary value b.2 When more points, evenly distributed between the initial and the critical point, are used for the linearisation, the extrapolation error decreases with respect to b , and decreases further as the number of points increases.This allows us to exclude, if necessary, the points at the bcginning of the titration (which for various reasons are frequently perturbed) and yet still to expect a systematic error no higher than h when the last point to be used for the extrapolation is calculated from the above equations. The equations in column 5 of Table 1 that are labelled B.E.P. are approximate forms of the linearity conditions, which hold for a large number of points. When the part of the titration following the equivalencc volume is linearised, a two-point extrapolation through the point at the maximum titration ratio,fM (the last experimental point), and the “critical point” satisfying the conditions expressed by the equations in column 4 of Table 1 that are690 ANALYST, JUNE 1989, VOL.114 Table 1. Critical conditions for linearity of Gran functions Type of titration" Jsovalent precipitation A + TA AT., . . . . . . . Heterovalent precipitation nA + rT+ (A,,T,)% . . . . . . . . C'helomctric titration of metals M+Y+MY . . . . . I'itration of a chelating agent with a metal Y+M-MY . . . . . . . . . . Strong acid with strong base H + + O H +H20 . . . . . . . . Weak acid with strong base HB+OH-+B+H,O . . . . . . Weak base with strong acid R + H + 4 H B . . . . . . . . . . Species sensed . . A o r T . . AorT . . M . . M . . H+ . . H+ . . H+ Critical condition Part o f titrationt Higher approximation Lower approximation B.E.P. A.E.P.[TI = 6 ( C - [A]) [A] = 6{C' - [ T ] / v M - 1)) B.E.P. A.E.P. [TI = b ( r / ~ ) ((7) - [A]) [A] = 6{Co - a[T]/t(fhl- 1)) B.E.P. [Y] = 6[MY] A.E.P. [MI = b{C" - [Y]/(fM - 1)) B.E.P. [MI = 6[MY] A.E.P. [Y] = 6{ Cy' - [M]/(fhl- 1)) B.E.P. A.E.P. [OH-] = b(C" - [H+]) [H+] = h{Cy'- [OH-llVM - 1)) Initial part$ [H+] - [OH-] = b[B] B.E.P. A.E.P. [OH-] - [H+ 1 = h[B] [HB] + [H']=S{C')- [OH-]/(fM - 1)) Initial part$ [OH ] - [H + ] = 6[HB] B.E.P. A.E.P. [H' 1 - [OH-] = b[HB] [B] + [OH-] = b{ C? - [W+]/cfM - I ) ) * In some instances charges have been omitted for clarity. t B.E.P. = before the equivalence point; A.E.P. = after the equivalence point. i Negative deviations from linearity. [Y] = 6Cj [MI = 6CO [MI = b(7' [Y] = 6co [OH-] = hC0 [H'] = bCY) [H+] = b[B] [HB] = [OII-] = bCY' [OH-] = 6[HR] [H+] = hCY) [B] = 6Cy) labelled A.E.P. [obtained by substitution of the mass balances of the reactants in equation ( 5 ) of reference 21, gives a relative extrapolation error equal to -h.A large number of experimental points and/or a relatively large value offM makes the approximate equations in column 5 of Table 1 valid. (For this reason, in reference 2 "exact" final equations were only shown forfM = 2.) The logarithmic diagram of a specific titration system allows us to obtain, in a simple and direct way, the equilibrium concentrations of the species corresponding to a critical point sought by finding the abscissa value for which the ordinates on the log[i] curves satisfy the relevant equation in Table I . Therefore, as the experimentally dependent variable of the points of a titration (the electrode potential, pH or pIon) is a linear function of the logarithm of one of the equilibrium concentrations, its critical value can be obtained directly from the diagram.Hence, the value of the measured (dependent) variable at which the collection of useful titration points should stop or may begin is identified. The experimentally controlled variable is usually the added titrant volume but, for our purposes, it is conveniently substituted by the titration ratio, f f = (u/t)n'/n" or, better, by the quantity E given by the equation (a/t)nf - no no . . . . c = f - 1 = . . (1) where nt is the total added amount, in moles, of titrant T, no is the total, i.e., initial, amount of analyte A and a and tare the stoicheiometric coefficients of the titration reaction a A + tT+products .. . . . . (2) The quantity E, the relative titration error (not to be confused with a), represents the actual titration error which would result if the titration were stopped at this point. By substitution of the equations of mass balances of solutes in equation (l), the numerator of the right-hand term becomes the sum of [i] terms. By obtaining the critical [i] values from the relevant plots on the diagram, the critical value of I, denoted by E,, can be calculated. As the curve of IoglE (the logarithm of the modulus of the titration error) is easily plGtted on the logarithmic diagram,7.9 its critical value can also be obtained directly from this plot. The critical titration ratio, f,, is calculated asf, = 1 + E,.Hence, the useful titration range for linear extrapolation can be identified. The procedure for the use of logarithmic diagrams will first be discussed in detail for a very simple titration and then summarised for the other situations, Kedox titrations have been discussed previously.7 Isovalent Precipitation Titrations The procedure for calculating the linearity range of the Gran plots is conveniently illustrated with reference to an isovalent precipitation titration. Fig. 1 shows a diagram for the titration of a 1 .O x 10-3 M chloride solution with silver nitrate (pK,,, = 9.50 at 25 "C, ionic strength = 0.1 M). Before the equivalence point Exact solution. The exact solution2 requires finding the abscissa value that satisfies the condition given by the relevant equation in Table 1 , expressed in logarithmic form log[Ag+] = log(C0- [Cl-I) + log6 .. ( 3 ) For this purpose, the auxiliary curve representing log(C0 - [Cl-1) is plotted [curve A in Fig. l(a)] using the following procedure. For [CI-] less than O . O 5 P , i.e., for log[CI-] < log CO - 1.3, the curve of log(C0 - [Cl-1) is virtually coincident with the horizontal segment log CO. On increasing [Cl-] (i.e., on increasing pAg). the value of log(C0 - [Cl-1) rapidly decreases and approaches -a for [Cl-] approaching C" (i.e., for a pAg approaching the value at which precipitation ofANALYST. JUNE 1989. VOL. 114 69 1 0 -2 -4 - 6 - .... - m 0 -J r -- I k \ \ \ \ 0 2 4 6 8 PAg Fig. 1. Use of the logarithmic diagram log[i] = f(pAg) to determine the linearity limits for the Gran titration of 1.0 x 10-3 M chloride solution with silver ion (7 = 25 "C, I = 0.1 M, pKSo = 9.50).( a ) Curve A, auxiliary curve representing [i] = CO - C1-1; point 1, log(CO - [CI-1) = log[CI-] = log Co - 0.30. Curve B, /i] = C0 - [Ag+]/Cfhl - 1) for fbr = 2 (only the position of the curve with respect to the Ag axis varies with fhl). ( h ) Segment c, [i] = 6[C1-]; curve D, [i] = pAg+] + 6[C1-]: segment e , [ I = ukf - l)&[Ag+], forfh4 = 5 and h = 0.2; and curve F, i] = [Cl-l + vM - l)G[Ag+]. The twin specular curves marked 1~1 represent the modulus of the absolute titration error. IEICO. Vertical segments represent the distances -log b indicated in the text (solid segments identify exact linearity limiting conditions; the dashed segments represent approximated conditions j silver chloride begins, very near to the initial point of the titration).The downward curving part of the plot can be drawn by hand, with an approximation that is sufficient for our purpose, through the point where [Cl-] = CO - [CI-] = 0.5C0, z.e., log[Cl-] = log Cc) - 0.30 [point 1 in Fig. 1(a)]. For increased precision, we can make use of more points: for instance. at the pAg value where [CI-] = 0.2C0, i.e., log[Cl-] = log CO - 0.70. we obtain a point with ordinate log(C0 - [Cl-1) = log(O.8CO) = log C0 - 0.10: where log[Cl-] = log (0.8cC') = log C* - 0.1, we obtain a point with ordinate log(C0 - [CI-1) = log(0.2C(9 = log C* - 0.70. The solution to equation (3) is obtained by finding the pAg value for which the vertical distance between the straight segment representing log[Ag+] and the curve plotted as described above (curve A) equals log h.Alternatively, equation (3) can he written log([&'] + 6[CI-]) = log CO + log b and solved by finding the pAg value for which the ordinate of the curve of log[i] = log([Ag+] + b[CI-1) [curve U in Fig. l(h)] is equal to log @) + log h. Curve D is, perhaps, more easily plotted than curve A (by connecting the straight lines representing log [Ag+] and log [Cl-1 + log h, respectively, through a point lying log 2 = 0.3 units above their intersec- tion); however, at variance with A, it holds for a single value of b . For this reason, and for brevity, this second type of solution will not be applied cxtcnsikely in the next sections, although it can be used if desired. Approximatc rolutiorz.Thc approximate solution (Table 1 , column 5 ) simply requircs finding the point where log[Ag+] = Example. I n the titration of a 1.0 x M chloride solution, for h = l % , identical results are obtained from both the approximate and the exact equations [see thc vertical segment g in Fig. l(a)]. The critical, i.e., t h c lowest useful, pAg is 5.06; the corresponding free chloride concentration i i 10-3'O M. If the response parameters of the indicating electrode (either a silver metal electrode or a chloride- selective electrode) are known, the critical e.m.f. o f the titration cell, i.e., the value at which the collection of useful experimental points must be interrupted, can be calculated. By using the mass balances log 0' + log b.no = n(C1) + n(AgC1) and I?' = rz(Ag) + n(AgC1) (where n(i) represents the equilibrium amount, i n moles, o f the ith species, ~ ( i ) = (V" + V)[i]} the error equation can be written (note that in order to apply logarithmic diagrams, ncgligible dilution must be assumed). By substitution o f the critical concentration values obtained above, E, = -0.02 ancl,f, = 0.9s are obtained as the critical values of the titration error and titration ratio, respectively. The value o f E, can also be obtained directly from the ploi of the logarithm of the modulus o f the titration error7.') [Fig. l(a)]. If a systematic extrapolation error better than 0.2% (log b = -2.7) is sought, the critical values are [Ag+Ic = 10- [identified by means of segment h in Fig.l ( n ) and segment 11' in Fig. l(h)], [Cl-1, = 10-3.7" M , E, = -0.20 andfc = 0.80 from the exact equation, ( 3 ) , and [Ag+Ic = lO--i~'(~ M (broken vertical segment near to h ) , [CI-1, = 10-"30 M , E, = -0.16 and f, = 0.84 from the approximate equation. The difference is appreciable; however, the approximate result is still accept- able, particularly if a large number of experimental points are available for extrapolation. The distance between curve A and log[Ag+] is always smaller than 3 units; therefore, the condition for b = 0.1'% is never attained. For this value of b, the approximate equation gives E, = -0.30 andf, = 0.70. It appears that the approximate equation is unreliable whenever it gives a critical value for the titration ratio of less than 0.80-0.85.Fig. l(a) and, perhaps more evidently, Fig. l ( b ) , show that there are two values of pAg for which equation (3) is satisfied. Of course, the second graphical solution, lying at higher pAg (and much lowera is meaningless, as is the second root o f the quadratic equation which represents the solution of the corresponding equation in reference 2. After the equivulence point After the equivalence point,fL, the minimum value offwhich, by a two-point extrapolation through the point at the choxn fM, Fives a systematic error equal to - b , depends on the value of fb,. The graphical solution based on the logarithmic eq ua t i c) n requires finding either the point where the distance between log[A] and the curve of log[i] = log(C0 - [AgfjiV;, - 1)) [Fig.l ( a ) , curve B, for fx4 = 21 is equal to log b or the point where the distance between the curve of log[i] = log([CI-] + SIAg+l/cfM - l)} [Fig. l ( h ) curve F, furf,,,, = 51 and log C'" is equal to lug h.692 ANALYST, JUNE 1989. VOL. 114 4 - 6 I 0 2 4 6 PF Fig. 2. Use of the logarithmic diagram log[i] = f(pF) for the titration of 0.010 bi calcium solution with fluoride ( T = 25 “C, I = 0.02 M , pKSo = 10.0). Auxiliary curve A. [i] = Cy) - [Ca2+]; scgment h , [i] = [F-]/4; curve C. [ i] = P - [F-]/4 (the auxiliary curve fortM = 5 ) . Distance (1‘ = log(h/2) for h = 2 % ; distance r = log h for h = 0.1% The approximate equation simply rcquires finding thc point where log[Cl-l = log P + log b , independently off,. Erunzple. For the titration of a 1.0 x 10-3 M chloride solution, with!,, = 2, the cxact condition [Fig.l(a)- curve R] gives pAg, = 3.5, c, = +0.02 andf, = 1.02 as the critical values for 6 = 1% (segment i), pAg, = 3.70, I, = 0.20 andfc = 1.20 for b = 0.2% (scgmcnt j ) and no solution for b = 0.1%. The approximate solution givcs fc = 1.02, 1.16 and (not indicatcd in Fig. 1) 1.3. respectively, for the sariie values of b . With f h l = 5 , the critical values are pAg, = 3.80, F, = 0.16 andf, = 1.16 for h = 0.2% [Fig. 1(h), curve F and segment k ] and pAg, = 3.50, E = 0.32 andf, = 1.32 for 6 = 0.10/0 (not shown in Fig. 1) both from the exact and the approximate conditions. I t is clear that on increasing f,,, the approximate equation gives results that approach the exact equation. This supports the claim’ that, whereas the approximate solution before the equi- valence point is acceptable only when it gives a critical titration ratio, f,, higher than about 0.8-0.85, higher con- fidence can be placed in the approximate cquation aftcr the equiLalence point.as both the titration range and the number of points can be increased a\ desired and the resulting f L is valid alw when appreciably higher than 1.2, provided that fh, is much greater than 2. Heterovalent Precipitation Titrations The linearity range before the equivalence point of hetero- valent precipitation titrations is found by plotting the curve of log[i] = log(C0 - [A]) on the logarithmic diagram of the titrated system and identifying the abscissa value for which the distance betueen log [TI and this curve is equal to log(t/a)b (see the equation for the critical condition given in Table 1).After the equivalence point, the curve to bc plotted is log[i] = log{ Cy) - N [ T I / I ( ~ ~ ~ - I ) } and the point where its distance from log[A] is cqual to log 6 must he found. Example. The procedure is illustratcd in Fig. 2 for the titration of a 1.0 x 10-2 M solution of calcium ion, with fluoride as the titrant (a = 1. t = 2, pKqO = 10.0 andfM = 3). The titration error is 0.5n(F) - n(Ca) Y 0.51F-] - [Caz+] - @’ € = no Before the equivalence point, the exact critical condition for ?J = 1%. i.c.., log[F-l = log(C’O-[Ca’+J) - log 2i5, is never satisfied, because the distance between log( F-] and the curve A is always smaller than 1.70. The approximate solution, log[T] = log(2bP) = log 0) - 1.70 (not indicated in Fig.2), gives a value for fc of 0.75, which is too far from 1 for the approximations on which it is based to be valid. If, however, we are content with b = 2% (segment d ) , we can extrapolate, in principle, points up to pF = 3.4, where log[Ca2+] = -3.2, E = -0.04 andf, = 0.96. In contrast, after the equivalence point both the exact and the approximate conditions forfM = 2 and b = 1% (not shown in Fig. 2) are satisfied starting from pF = 3.00, log[Ca’+] = -4.00, E = 0.04 andf, = 1.04. With a higher fM, higher accuracy can be reached in principle: for instance, with fM = 5 (curve C) the condition for b = 0.1% (segment e ) is satisfied at pF = 2.45 and f c = 1.7, giving a reasonable extrapolation range. It appears that the deviations from linearity increase as the distance from the equivalence point decreases and that this happens much more rapidly before than after the equivalence point owing to asymmetry of the titration curve.Experimental titrations of 0.01 M calcium solutions with 0.1 M fluoridc solutions monitored with a fluoride-sclectivc electrode were indeed linear in the expected range after the equivalence point. In contrast, experimcntal points in the first part of the titration were useless, because of the slowness of the precipitation reaction. I t is intercsting to note that the revcrsc titration of a 0.02 M fluoride solution with calcium (a = 2, t = 1 ; note that the shape of the relcvant logarithmic diagram is different from that shown in Fig. 2) was not really feasible.In fact, in the first part of the titration thc precipitation is again too slow (although the Gran plot satisfies in principle the condition b = 1% up to f = 0.95). After the equivalence point, a systematic error les\ important than -1% is only warrantcd, with f c = 1.25, if a relatively large f M value is reached. Lower concentrations of fluoride (for instance, 0.01 M ) cannot be titrated with rcasonablc accuracy. Complexometric Titrations of Metals The exact critical condition for linearity before the equi- valence point of a complexometric titration of a metal log[Y] = log[MY] + log b . . . , (4) can be characterised on the logarithmic diagram of the titration system without plotting an auxiliary curve. There- fore, the approximate condition (see Table 1) in this particular instance offers no advantage.I n contrast, aftcr the equivalencc point the exact critical condition (see Table 1) requires that the auxiliary curve of log(i] = log{ Cr) - [Y]/(fM - 1)) bc plotted. The values of E ~ , and therefore off,, are calculated from the equation 10 Exunzpfe. In Fig. 3 the logarithmic diagram for the titration of a 2.00 x 10-3 M solution of a cation with a conditional formation constant K , = 105.(M) (for instance. titration of calcium with ethylendiaminetetraacetic acid, EDTA, at about pH 5.4) is plottcd. For b = I%, bcfore the equivalencc point, equation (4) is satisfied (segment a ) at pY = 5.00, where [MI = 10-30(1 M and f c = 0.50, whereas the approximate solution (the next broken segment) gives the unreliable value of 0.64 forf,.After the equivalence point, the exact condition does not allow a solution for fM = 2 (not illustrated). In contrast, the approximate condition gives f c = 1.5, which is not far from the “exact” value of 1.56 for fM = 5 (the solid vertical segment and the broken segment at lower pY values, respectively). Titration of Weak Acids In addition to the positive deviations in the proximity of the equivalence point shown by very weak or dilute acids, Gran plots for the titration of weak acids or weak bases generallyANALYST. JUNE 1989, VOL. 114 693 “7 0 1 3 5 7 PY Fig. 3. Titration o f a 2 x lo-’ M solution of a cation (M) with a chelating agent Y (conditional formation constant Kf = oO). Conditional conccntrations are plotted. Distance (1 and the other vertical segments represent log 6 for h = 1 % .Segment h , [ Y]/4; curve C. [i] = C7) - [Y]/4 (the auxiliary curve for f = 5 ) - 2 h n --R Y 2 4 6 8 10 12 PH Fig. 4. Titration o f 1.2 x 10-3 M pyridinium chloridc solution, p K , , = 5.25. Auxiliary curve A: [i] = [H+] - [OH-] at pH < 7 , and [i] = [OH-] - [ H I ] at pH > 7. b: Minimum allowcd pH for linear extrapolation before the equivalcnce point according to equation (6) for h = 1% and h’, for 6 = 0.1%. c: Maximum pH for linear cxtrapolation bcforc the equivalcncc point according to equation (7) f o r h = 1 o/b and c’, for 6 = 0.1’36. ti! and d ‘ , Critical conditions aftcr the equivalcncc point [equation (S)] for fM = 2, b as above show negative deviations from linearity in their initial part.’ In Table 1 the equations for critical conditions for linearity are reported for both types of deviation.The solution of the equations in logarithmic form is supportcd by auxiliary curves, which can be plotted using procedures similar to those described above. For instance, the curve of log([H+] - [OH 1) is virtually coincident with thc linear segment log[H+] at pH values lower than 6.3; at higher pH, it is represented by a curve (A in Fig. 4) that passes through a point with co-ordinates of pH = 6.85 and l o g [ i ] = log[OH-] = 7.15 (i.e., for [H+] = 2[OH-] and [H+] - [OH-] = [OH-]), approaching - at pH 7. The curve of log([OH-] - [H+]) is symmetrical with the former. Whereas in the other types of titration discussed earlier the approximated equations in the last column of Table 1 are obtained by introducing only one kind of approximation into the corresponding cxact equations, in some of the equations for weak acids (or weak bases) two different approximations can be introduccd stepwisc according to the strength and concentration of the acid (or base).On14 the lower approxi- mation equations, which do not require auxiliary curves for thcir solution, are reported in the last column of Table 1. 0.75 I 1 0 0.5 1 .o 1.5 f Fig. 5. Experimental Gran functions and linear cxtrapolation of the cyuivalence point for the litration of 1.16 x 1 0 - 3 M pyridinium chloridc solution with 1.012 X 10-2 M sodium hydroxide 5olution. Co-ordinates are normalised to -t 1 slopes of linear plot5 The error equation for calculating F, and f , is [OH-] - [HB] - [H+] C” E = .. . . ( 5 ) Examples. Solutions (1.2 X 10-1 M ) of pyridinium chloride (pK, = 5.23 at 25 “C) were titrated with 0.01 M sodium hydroxide solution. The predictions from the logarithmic diagram (Fig. 4) were as follows. For b = l % , the equation log([H+] - [OH-]) = log[B-] + log b . . (6) in which [OH-] appears to be negligible, gives a value of 5.23 for the critical pH towards the initial part of the titration (identified by segment b in Fig. 4). At this pI1 equation ( 5 ) gives E, = -0.5 andf, = 0.5. The equation log([OH-] - [H+]) = log[B-J + 10g 6 . . (7) givcs a pH of 9.1 (scgment c) as the critical pH value towards the equivalence point. As this pH is higher than the equivalericc value (pH = 8. l), the condition sought is valid’ up t o f = 1 .00.For b = 0. I YO, the first pH, is 6.0 (scgmcnt b ‘ ) , and fc = 0.85, whereas the second f L again corresponds t o the equivalence point (scgmcnt c’). The rcsulting “linearity range” appears to be fairly narrow, so that the assumptions on which cquation (6) is based (extrapolation through at least two points, the first of which is at, or near to, f = 0) could appear, at first sight, not to be valid. However, it is obvious that in situations such as this, where points near to the equivalence point can be cxploited (at lcast in the absence of perturbations of different origin), the initial deviations are much less important. After the equivalence point, the equation log([HB] + [H’]) = log withf, = 2 gives a value for pH, of 7.23 for b = 1% (segment d ) and 8.23 for b = 0.1% (segment d ’ ) .Both pH, values correspond tofc = 1. Note that all the approximations in the last column of Table 1 are valid. The above predictions were confirmed experimentally, as shown in Fig. 5 , both before and after the equivalence point. Fig. 6 represents the titration of a much weaker acid, viz., 1.0 X lo-? M ammonium nitrate. Auxiliary curves must be plotted, unlike the previous example. (Note that in this instance the simplification4 used to obtain some of the higher approximation equations for the calculation of [Ii+], and f c given in Table 1 of reference 2 are not strictly valid.) As regards initial negative deviations, Fig. 6 shows that [OH-] cannot be neglected with respect to [H+] in equation (6). Nevertheless, as the linearity condition is valid starting from a pH of 6.6 for C, = 1% (segment b) and a pH of 7 (‘j = 0.005) for 6 = 0.1% (segment h’), only the first point of the titration cf = 0) needs to be ignored.ANALYST, JUNE 1989, VOL.114 O r i \ \ I L I 6 8 10 12 14 10 PH Fig. 6. l’itration o f I .O x 10-2 M ammonium nitrate solution. pk’,, = 9.30. A . I ) . h’ and ( , c‘. 3ec Fig. 4; D and D‘, auxiliary curves [i] = C” - [OH-]/(f, - I ) for fM = 2 and 5 , respectively, with vertical segniunts characterising the critical conditions for linearity for b = 1% accordiny to equation ( 8 ) Equation (7) is satisfied with h = 1% (segment c) at pH = 9.8 atid f = 0.8, where [H+] is negligible with respect to [OH-]. However, if b = 0.1% k sought, all terms in equation (7) become important, and pH, = 7.0 and f, = 0.05 can be obtciined bq a graphical solution (segment c’).This demon- \trcttcc that the experimental Gran plot of the first part o f the titration only provides a moderate degree of accuracy. Aftcr the equivalence point, curve D in Fig. 6 \bows that with fhl = 2 equation (8) gives pH, = 11.45 andl, = 1.28 for h = 1%. With fhl = 5 , curve I)’ gives pHc = 11.30 andf, = 1.20 for the \ame h value. For either value offM, equation (8) does not provide a wlution for 6 = 0.1%. It should be noted that [H+] i\ always negligible with respect t o [HB], whereas [ OH-]ICfhl - I ) becomes negligible with respect to Cy) only for .fhl = 5 or higher (compare with the approximate equations giien in Table 2 ot reference 2). Discussion and Conclusions Information from equations’ and from logarithmic plots about the linearity range and accuracy of Gran titrations have been corifirmed experimentally for other systems, in addition to the two described here.In some instances, however, the agree- ment was po~x- or non-existent. Generally, deviations not predicted theoretically could be ascribed to exact causes, thereby negating the inadequacies of the approach. The adirerse effect of inaccurate values of the equili brium con- stants on the correctness of the estimated linearity range can be avoided by the use of reliable tables and by introducing a (rough) correction for the ionic strength. Highly inaccurate initial estimates of the unknown CO values can also cause erroneous predictions. For totally unknown or highly variable samples, a first estimate of CO obtained.for instance, by extrapolating a few points on a roughly linear portion of the titration curve (irrespective of strict linearity) is sufficient to obtain correct predictions in order to optimise the extrapola- tion procedure. Causes of deviations from linearity that are different from those inherent in the titration equilibria can either produce more restrictive limits than predicted or disprove the predic- tions. In our experiments, many types of perturbations were not unexpected. An example of a slow reaction was discussed earlier. A frequent cause of appreciable deviations was impurities in the reagents interfering with the titration equilibria. The effect of carbon dioxide on acid - base titrations is well known.11 Marked deviations from linearity in the initial part of weak-base titrations were found to originate from acid impurities. The effect of heavy metals is also apparent in the initial part of the complexometric titration of alkali earth metals. A non-linear dependence of the e.m.f. on the logarithm of the concentration of the sensed ion, due, for instance, to electroactive background species or to the lower response limit of the electrode being approached, can also have an appreciable effect. Experimental Gran plots should always be inspected carefully for deviations from the theor- etical behaviour, as these can give very useful information about the real nature of the system under investigation. From the above discussion, it can be predicted that the graphical approach should be particularly useful when par- tially overlapping equilibria are present in the titrated system. Finally, we wish to stress that there is no rivalry between logarithmic diagrams and computer programs.Diagrams cannot compete with computers for precise, accurate and complete predictions of titrations. For instance, they do not make allowances for dilution or for variations in the ionic strength and liquid-junction potential during the titration. On the other hand, they can provide (by a rapid hand-sketch) an immediate survey of the most important properties of the system to be titrated and help in taking a rapid decision about the strategy to be followed in practice. In one respect computers must be considered to be incompatible with diagrams: it is absurd to use a scphisticated computer program to plot a logarithmic diagram giving only approximate indications, when for the same cost more exact mathematical results can be obtained (computer-plotted logarithmic dia- grams should be restricted to illustrations in publications). In contrast, diagrams are compatible with computers, as they can help in the formulation of programs by providing an under- standing of the system under study. References I . 2. 3. 4. 5 . 6. 7. 8. 9. 10. 11. Gran, G., Analyst, 1952, 77, 661. Macch. C.. and Rombi, G. G., Analyst, 1989, 114, 463. Bjerrum, N., “Die Theorie der Alkalimctrischen und Azidimc- trischen Titrierungen,” Enke, Stuttgart, 1914, pp. 60-128. I-Iagg, G . , “Die Theoretischen Grundlagen der Analytischen Chemie,” Birkhiiuser, Basel, 1950. SillCn, L. G., in Kolthoff. I. M., and Elving, P. J . , Editors, “Treatise on Analytical Chemistry, Part I,” Volume 1. Wiley- Interscience, New York, 1959. Johansson. A , , and Winninen, E . , irz Kolthoff, I. M.. and Elving, P. J . , Editors, “Treatise on Analytical Chemistry, Part I,” Volume 11, Wiley-Interscience, New York, 1975. Macch, C., andBombi, G. G . , FreseniusZ. Anal. Chem., 1986, 324. 52. Ringbom, A . , “Complexation in Analytical Chemistry,“ Inter- science, New York, 1963. InczCdy, J . , J . Chem. Educ., 1970, 47, 769. Macca. C . , Analyst, 1983, 108, 395. Rossotti, F. J . C., and Rossotti, H., J . Chenz. Educ., 1965, 42, 375. Paper 810461 1 E Received November 21st, 1988 Accepted January 19th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400689

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

ANALYST, JUNE 1989. VOL. 114 695 Platinum-dispersed Nafion Film Modified Glassy Carbon as an Electrocatalytic Surface for an Amperometric Glucose Enzyme Electrode Hari Gunasingham and Chee Beng Tan Department of Chemistry, National University of Singapore, Kent Ridge, Singapore 051 I , Republic of Singapore ~ ~ ~~~ A Nafion film dispersed with platinum particles formed on a glassy carbon electrode combines the electrocatalytic activity of platinum with the background stability of glassy carbon. It serves as a selective and sensitive electrode surface for an amperometric glucose enzyme electrode. Details on the construction and the functional characteristics of the enzyme electrode are described. It is shown that the electrode is well suited to the determination of levels of glucose i n blood by flow injection with amperometric detection.Keywords: Nafion; amperometric glucose enzyme electrode; flow injection; wali-jet electrode; platinum- dispersed polymer film Amperometric enzyme electrodes for monitoring glucose usually employ thc en7ymc glucosc oxidase. There are essentially three categories of electrodes and these have been termed first-, second- and third-generation devices. In first- generation devices, the glucose concentration is related to the amperometric current arising from the consumption of oxygen or the formation of hydrogen peroxide in the enzyme reaction. l-5 Asperger et ~ 1 . ~ have shown that the monitoring of the hydrogen peroxide is to be prefcrred because of the inherently greater accuracy of this mode. In contrast to first-generation devices, second-generation glucose sensors employ a synthetic mediator in place of the naturally occurring oxygen, 1.6 whereas third-generation de- vices employ conducting organic salt electrodes .7 Second- and third-generation devices have the advantage that the operat- ing potential can be lowered significantly, obviating interfer- ence from electroactive spccics that may be prescnt in the sample.Also, they have a wider linear range because of their insensitivity to variations in oxygen tension. However, in practical flow injection (FI) applications, first-generation devices do have the advantage of faster response times and greater stability. An important objective of much of the development work on first-generation amperometric glucose enzyme electrodes has been the search for devices that combine high sensitivity with stability and selectivity.In general, however, these appear to he contradictory goals. For example, platinum has been used widely as an electrode material in the construction of glucose enzyme electrodes based on both the hydrogen peroxide and oxygen consump- tion monitoring modes. This is because of platinum's inher- ently superior electrocatalytic response to oxygen and hydrogen peroxide. Unfortunately, however, platinum is also highly susceptible to electrode poisoning due to oxide formation and the adsorption of impurities.* The practical consequence of this is a decline in the electrode response with time which results in a loss of sensitivity and precision. The other difficulty in monitoring glucose in physiological samples is that platinum is also sensitive to several electroac- tive interfering species such as ascorbic acid and paracetamol that are commonly present.Researchers in the enzyme-electrode field have bccn seeking to overcome these difficulties by the use of permselec- tive membrane films to exclude interfering species. A com- mon approach is to sandwich the enzyme betwecn two membranes: the top membrane (such as pol yearbonate) allows the passage of glucose, whereas the bottom membrane (usually cellulose acetate) placcd on the electrode allows only hydrogen peroxide through .y Recently, there has been interest in the use of a number of' novel ion-cxchange and redox polymers in applications to electrochemical analysis in general which seek to exploit specific properties of these polymers.Of particular interest has bccn the Nafion ion-exchange polymer, which has the ability to exclude anionic interferences, Nafion film electrodes have been used in flowing stream analysis and high-perform- ance liquid chromatographic detection and have been found useful, in particular for the determination of cationic neuro- transmitters. 1 0 . 1 1 In a previous paper we described the use of a Nafion film on glassy carbon containing dispersed platinum particles (de- posited into the film clectrochemically) for the determination of hydrogen peroxide.12 Although there is some contention in the way the particles are distributed, they have been shown to be dispersed three-dimen4ionally in the Nafion film.13 Similar work has been carried out with poly(4-vinylpyridine)l4 and poly(pyrole) films.15 However, where the Nafion film differs is in its ability to reject electroactive anionic interferences such as ascorbic acid and paracetamol because of its ion-exchange properties. In this paper, we describe the use of a platinum-dispersed Nafion electrode in the development of an enzyme electrode for the monitoring of glucose. The electrode i\ used in conjunction with automated flow injection for the routine determination of glucose over a wide linear range in clinical analysis. Experimental Buffer Solution Phosphate buffer of pH 7.3 was used for the prcparation of the enzyme solutions and glucose standards. It was also used as the carrier solution in the FI studies.The composition of the phosphate buffer was 52.8 mM Na2HP04 - 15.6 mM NaH2P04 - 5.1 mM NaCl - 0.01% m/V sodium azide - 0.15 mM disodium ethylcnediaminetetraacetate. Preparation of Platinum-dispersed Nafion Electrodes Nafion (Du Pont) solutions (3% m/V) were made up i n ethanol or obtained from Solution Technology, PA, USA, as solutions in ethanol. Glassy carbon clectrodes (3-mm diamcter; Tokai Carbon, Tokyo, Japan) were polished to a mirror finish with diamond paste (K-5000, Kyoto Diamond Industrial, Japan). The electrodes were wiped with an ethanol-soaked tissue, rinsed thoroughly with ethanol and distilled water and then dipped in696 0.4 0.3 Q, 2 0.2 z 0.1 Y m 2 0 - -0.1 -0.2 ANALYST, JUNE 1989, VOL. 114 - - - - - I I 1 I I I I concentrated nitric acid for 10 s.The oxidised electrodes were rinsed in distilled water and cleaned in an ultrasonic bath containing distilled water for 15 min. Finally, the electrodes were electrochemically pretreated by potential cycling using a PAR Model 273 potentiostat (Princeton Applied Research, NJ. USA) between -0.15 and + 1.1 V (scan rate. 100 mV s- l ) for 30 min in the phosphate buffer (pH 7.3). After drying the glassy carbon electrodes in an oven for 30 niin at 80 "C, the Nafion film was coated on the elcctrode surface by applying a 20-pl drop of the Nafion solution and allowing it to dry in air. Platinum particles were deposited into the Nafion film by potential cycling between +0.7 and -0.15 V using the PAR Model 273 potentiostat in an acidic solution of K,PtCI, ( 5 mM in 0.1 M H2S04) at a scan rate of 50 mV s-1.12 The loading o f the platinum was calculated from the charge consumed.13 In this work, platinum loadings of between 100 and 150 pg cm-? were used as these gave the optimum scnsitivity/stability characteristics.Preparation of Enzyme Electrodes Glucose oxidase (E.C. 1.1.3.4. from Aspergillus niger, Sigma) wa5 immobilised directly on the Pt - Nafion electrode as follows. A freshly prepared glucose oxidase solution (20 pl; 2% m/V) was placed on the electrode surface and allowed to dry at 4 "C overnight. Then, 1 0 pl of a freshly prepared mixture of equal parts of bovine serum albumin (BSA) (100% m/V) and gluteraldehyde (2.5% VIV) was added over the dried enqme and a piece of polycarbonate membrane (Nucleopore; 1 cm2, 0.03-pm pore size) placed immediately over the electrode. The membrane was held tightly in place with a cap.The enzyme electrodes were allowed to dry for at least 12 h at 4 "C before use. For the interference studies, an additional cellulose acetate layer was placed over the Pt - Nafion film. This was achieved by spin coating a 10-pl drop of cellulose acetate solution (2% cellulose acetate in 50 + 50 VIV cyclohexanone - acetone). Automated Flow Injection A large-volume wall-jet detector with a Ag - AgCl reference electrode was used for the FI studies. The sample volume was SO pl and the flow-rate was kept at 1 ml min-1. For amperometric measurements, the applied potential was main- tained at +0.7 V using a PAR 174A potentiostat. Two peristaltic pumps (Eyela Model MP-3, Tokyo Rikaki- kai Tokyo, Japan) were used to deliver the sample and carrier streams.Two pneumatically actuated four-way valves (Model 5010, Rheodyne, CA. USA) under computer control permit- ted the automatic loading and injection of the sample. Unused sample was recycled into the sample reservoir, therefore minimising wastage. The flow system permitted an unatten- ded and continuous operation. Peak currents were digitised via a standard electronic interface and data stored in Lotus 123 files (Lotus Corpora- tion) on the computer hard disk. The data could therefore be analysed and plotted as required. Results and Discussion Potential Profile Fig. 1 shows typical peak current - potential response plots for the dispersed platinum - Nafion - glassy carbon (Pt - Nafion - C C ) enzyme elcctrodc and the solid platinum enzyme electrode for injected glucose samples (2 mM).Although the response plot for the solid platinum electrode is more sharply defined, the limiting peak currents begin at about the same potential. The electrode response varies from electrode to electrode. On average, however, the limiting peak currents for the Pt - Nafioti - GC and solid platinum electrodes are comparable. For example, for 2 mM glucose, the average peak Fig. 1. Effect of applied potential on peak current for a 2 mM glucose solution. A, solid platinum enzyme electrode: B, Pt - Nafion - GC enzyme electrode. Flow-rate, 1 ml min-I; carrier solution, isotonic phosphate buffer (pH 7.3) containing O.1S-mM disoduim ethylene- diaminetetraacetate; sample volume, SO ~ i 1 0.2 0 2 4 6 8 10 12 14 Sample number (x50) Fig.2. Long-term behaviour of: A, sclid platinum; €3, Pt - Nafion - GC; and C, Pt - CA - Nafion - GC enzyme elcctrodes. Each point represents the mean of SO successive injcctions of a 2 mhi glucose solution. Working potential. +0.7 V versus Ag - AgCl. Other conditions as in Fig. 1 1.0 1 I 0.8 Q, 0.2 0 20 40 60 80 100 Sample number Fig. 3. Short-term behaviour of: A, solid platinum: B. Pt - Nafion - GC; and C, Pt - CA - Nafion - GC: enzyme electrodes. Conditions as ir, Fig. 1 current for the Pt - Nafion - GC electrode was 0.40 k 0.2 pA (rz = 7) compared to 0.44 -t 0.2 pA ( n = 7) for the solid platinum electrode. The potential profile indicates that the Pt - Nafion - GC electrode functions in most respects as a platinum electrode with regard to its electrocatalytic activity. The detection ofANALYST, JUNE 1989, VOL.114 I 697 t + c 2 3 0 0.14 pA I I Time -+ Fig. 4. using a Pt - Nafion - GC enzyme electrode F1 profile for successive injections of a 2 mM glucose solution glucose clearly depends on the current arising from the Gxidation of HzOz at the active platinum particles deposited in the Nafion film. Electrode Stability Figs 2 and 3 illustrate the long- and short-term stabilities, respectively. of the enzyme electrodes. As can be seen, in the long term, the response for the solid platinum system declines. The long-term stability is determined in terms of the over-all relative standard deviation (RSD) of the individual RSDs for every 50 injections.The over-all RSDs were 5.1 , 2.3 and 1.2% for the solid platinum, Pt - Nafion - GC and dispersed platinum - cellulose acetate - Nafion - glassy carbon (Pt - CA - Nafion - GC) enzyme electrodes, respectively. The improved stability of the Pt - Nafion - GC enzyme electrode is exemplified further in the short-term plot obtained for the first 100 sample injections. The RSDs were 3.1, 1.6 and 0.6% for the solid platinum, Pt - Nafion - GC and Pt - CA - Nafion - GC enzyme electrodes, respectively. The short-term fluctuations observed for the solid platinum electrode can be attributed in part to variations in flow as no damping was used with the peristaltic pump. The Pt - Nafion - GC enzyme electrode, however, appears to be less susceptible t o fluctuations in flow.The Pt - Nafion - GC system therefore appears to be as sensitive as a solid platinum electrode and has enhanced stability and selectivity. The stability is due in part to the Nafion film preventing electrode poisoning and in part to the underlying stability of the base glassy carbon electrode. Linear Range A calibration graph was constructed for the Pt - Nafion - GC enzyme electrode for which an additional cellulose acetate layer had been placed over the Pt - Nafion layer. The electrode showed a linear range between 0 and 30 mM, which covered the physiologically useful range for monitoring glucose levels i 11 b 1 ood . Response Time Fig. 4 shows typical FI peak profiles for successive injections of glucose standard. The FI peak is formed in less than 60 s, giving a sample throughput of more than 60 samples per hour.With smaller sample volumes a higher throughput can be achieved. Table 1. Effect of electroactive interferences Selectivity,* % Pt - Pt-CA- I nterferent Solid Pt Nafion Nafion Hydrogen peroxide (1mM) . . . . , . 100 100 100 Ascorbic acid ( I mM) . . 80.6 3.6 I .2 Uric acid (1 r n M ) . . 69.9 4.3 1 .o Paracetamol(1 mM) . . 60.1 5.8 0.9 * Mcasurcd as the molar ratio with respect to the detection of HZO:. Table 2. Comparison of the Pt - Nafion - GC enzyme electrode with the Reflolux monitor for the determination of glucose (in mM) in whole blood Pt - Nafion - GC 3.0 5.8 7.5 10.5 12.2 9,o 10.3 Reflolux 2.8 5.2 8.3 10.6 12.3 7.4 9.9 Interferences Table 1 shows the effect of some common electroactive interferences.Although the additional cellulose acetate mem- brane improves the selectivity by a few per cent., clearly Nafion is sufficient on its own. Indeed, considering the fact that normal physiological levels of interferents are much lower (e.g., ascorbic acid, ca. 0.2 mM) the cellulose acetate membrane could be considered to be unnecessary. The effectiveness of Nafion presumably arises from its ion- exchange properties which serve to exclude anionic interfer- ents. Analysis of Glucose in Blood The Pt - Nafion - GC enzyme electrode is well suited to monitoring glucose in physiological tluids as it is not prone to interferences such as ascorbic acid. This is shown in Table 2, which compares results obtained with the electrode to a standard laboratory analysis carried out with a Hoehringer Rcfolux glucose analyser that had been previously calibrated against a Beckman Astra Analyser.The results show a good correlation. Conclusion The platinum-dispersed Nafion-based enzyme electrode has been shown to combine the electrocatalytic properties of platinum with the background stability of glassy carbon. At the same time, the dispersion of the platinum in the Nafion film serves to enhance selectivity such that the electrode is largely unaffected by most electroactive interferents found in physiological samples. The platinum-dispersed Nafion system is easy to construct and reproducible results can be obtained. Also, because a relatively small amount of platinum is plated into the Nafion film, the electrode is inexpensive in comparison with a solid platinum electrode.We gratefully acknowledge grants from the National Univer- sity of Singapore and the Singapore Science Council. References I . 2. Frew, J . E., and Hill, H . A. O., And. Chem., 1987, 59, 933A. Carr, P. W., and Bowers, L. D.. “Immobilized Enzymes i n Analytical and Clinical Chemistry.” Wiley, New York, 1980.698 ANALYST, JUNE 1989, VOL. 114 3. 4. 5 . 6. Updike, S. D., and Hicks, G . P., Nature(Londonj, 1967, 214, 989. Guibault, G. G . . and Lubrano, G. J., Anal. Chim. Acta, 1973, 64. 439. Asperger. L.. Geppert, G . , and Krabisch, C. H., Anal. Chim. Acta, 1987, 201, 281. Gas, A. E. G., Davis, G., Francis, G . D., Hill, H . A. O., Aston. W. J . , Higgins, I. J . , Plotkin, E. V., Scott, L. D. L., and Turner, A. P. F., Anal. Chem., 1984, 56, 667. 7. Albery, W. J . , Bartlett, P. N., and Craston, D. H., J . Electroanal. Chem.. 1985, 194, 223. 8. Gunasingham, H., and Fleet, B . , in Bard, A. J . , Editor, "Electroanalytical Chemistry," Volume 16. Marcel Dekker, New York. 1989. Guibault, G. G., "Analytical Uses of Immobilized Enzymes," Marcel Dekkcr. New York, 1984. 9. 10. Wang, J . , Tuzhi, P., and Golden, T., Anal. Chim. Acta, 1987, 194. 129. 11. Wang, J . , and Tuzhi, P., Anal. Chem., 1986, 58, 3257. 12. Tay, B. T., Ang, K. P., and Gunasingham, H., Analyst, 1988, 113, 617. 13. Itaya, K., Takahashi, H.. and Uchida, I . . J. Electroanal. Chem., 1986, 208, 373. 14. Bartak, D. E . , Kazee, B., Shimazu. K.. and Kuwana, T., Anal. Chem., 1986, 58, 2756. 15. Holdcroft, S . , and Funt, L. B., J . Electroanal. Chem., 1988, 240, 89. Paper 8f03607A Received September 19th, 1988 Accepted January I3th, 1989

ISSN:0003-2654    

DOI:10.1039/AN9891400695

出版商:RSC    

年代:1989

数据来源: RSC