|
1. |
Front cover |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 027-028
Preview
|
PDF (547KB)
|
|
摘要:
The AnalystThe Analytical Journal of The Royal Society of ChemistryAnalytical Editorial BoardChairman: A. G. Fogg (Loughborough, UK)K. D. Bartle (Leeds, UK)D. Betteridge (Sunbury-on-Thames, UK)J. Egan (Cambridge, UK)H. M. Frey (Reading, UK)D. E. Games (Swansea, UK)S. J. Hill (Plymouth, UK)D. L. Miles (Keyworth, UK)J. N. Miller (Loughborough, UK)R. M. Miller (Port Sunlight, UK)B. L. Sharp (Loughborough, UK)Advisory BoardJ. F. Alder (Manchester, UK)A. M. Bond (Victoria, Australia)R. F. Browner (Atlanta, GA, USA)D. T. Burns (Belfast, UK)J. G. Dorsey (Cincinnati, OH, USA)L. Ebdon (Plymouth, UflA. F. Fell (Bradford, UK)J. P. Foley (Villanova, PA, USA)T. P. Hadjiioannou (Athens, Greece)W. R. Heineman (Cincinnati, OH, USA)A. Hulanicki (Warsaw, Poland)I.Karu be (Yokohama, Japan)E. J. Newman (Poole, UK)T. B. Pierce (Harwell, UK)E. Pungor (Budapest, Hungary)J. RBiiCka (Seattle, WA, USA)R. M. Smith (Loughborough, UK)M. Stoeppler (Jijlich, Germany)J. D. R. Thomas (Cardiff, UK)J. M. Thompson (Birmingham, UK)K. C. Thompson (Sheffield, UK)P. C. Uden (Amherst, MA, USA)A. M. Ure (Aberdeen, UK)P. Vadgama (Manchester, UK)C. M. G. van den Berg (Liverpool, UK)A. Walsh, K.B. (Melbourne, Australia)J. Wang (Las Cruces, NM, USA)T. S. West (Aberdeen, UK)Regional Advisory EditorsFor 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, Institute for Analytical Chemistry, University Leipzig, Linnestr.3,Professor 0. Osibanjo, Federal Environmental Protection Agency, Federal Secretariat, PhaseProfessor K. Saito, Coordination Chemistry Laboratories, Institute for Molecular Science,Professor M. Thompson, Department of Chemistry, University of Toronto, 80 St. GeorgeProfessor Dr. M. Valcarcel, Departamento de Quimica Analitica, Facultad de Ciencias,Professor J. F. van Staden, Department of Chemistry, University of Pretoria, Pretoria 0002,Professor Yu Ru-Qin, Department of Chemistry and Chemical Engineering, Hunan University,Professor Yu. A. Zolotov, Kurnakov Institute of General and Inorganic Chemistry, 31 LeninAmsterdam, THE NETHERLANDS.D-0-7010 Leipzig, GERMANY.II, 1st Floor, IKOYI, Lagos, P.M.B. 12620, Lagos, NIGERIA.Myodaiji, Okazaki 444, JAPAN.Street, Toronto, Ontario M5S I A l , CANADA.Universidad de Cordoba, 14005 Cordoba, SPAIN.SOUTH AFRICA.Changsha, PEOPLES REPUBLIC OF CHINA.Avenue, 117907, Moscow V-71, RUSSIA.Editorial Manager, Analytical Journals: Judith EganEditor, The AnalystHarpal S.MinhasThe Royal Society of Chemistry,Thomas Graham House, Science Park,Milton Road, Cambridge CB44WF, UKTelephone 0223 420066.Fax 0223 423623. Telex No. 81 8293 ROYAL.Senior Assistant EditorPaul DelaneyUS Associate Editor, The AnalystDr J. F. TysonDepartment of Chemistry,University of Massachusetts,Amherst MA01003, USATelephone413 5450195Fax 41 3 545 4490Assistant EditorsBrenda Holliday, Sheryl YouensEditorial Secretary: Claire HarrisAdvertisements: Advertisement Department, The Royal Society of Chemistry, BurlingtonHouse, Piccadilly, London, W I V OBN.Telephone 071-437 8656. Telex No. 268001.Fax 071-437 8883.The Analyst (ISSN 0003-2654) is published monthly by The Royal Society of Chemistry,Thomas Graham House, Science Park, Milton Road, Cambridge CB4 4WF, UK. All orders,accompanied with payment by cheque in sterling, payable on a UK clearing bank or in USdollars payable on a US clearing bank, should be sent directly to The Royal Society ofChemistry, Turpin Distribution Services Ltd., Blackhorse Road, Letchworth, Herts SG6 1 HN,United Kingdom. Turpin Distribution Services Ltd., is wholly owned by the Royal Society ofChemistry. 1992 Annual subscription rate EC f276.00, USA $589, Rest of World f310.00.Purchased with Analytical Abstracts EC €604.00, USA $1299.00, Rest of World f669.00.Purchased with Analytical Abstracts plus Analytical Proceedings EC f712.00, USA $1 527.00,Rest of World f791 .OO.Purchased with Analytical Proceedings EC f351 .OO, USA $749.00, Restof World f395.00. Air freight and mailing in the USA by Publications Expediting Inc., 200Meacham Avenue, Elmont, NY 11003.USA Postmaster: Send address changes to: The Analyst, Publications Expediting Inc., 200Meacham Avenue, Elmont, NY 11003. Second class postage paid at Jamaica, NY 11431. Allother despatches outside the UK by Bulk Airmail within Europe, Accelerated Surface Postoutside 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 research papers.Communications, 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. Although publication is a t thediscretion of the Editor, communications will beexamined by a t least one referee.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 a t least two referees, by whoseadvice the Editorial Board of The Analystwill 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 and North America, a Group of RegionalAdvisory Editors exists. Requests for help oradvice on any matter related to the preparationof papers and their submission for publicationin The Analyst can be sent to the nearestmember of the Group.Currently servingRegional Advisory Editors are listed in eachissue of The Analyst.Manuscripts (four copies typed in double spac-ing) should be addressed to:Harpal S. Minhas, Editor, The Analyst,Royal Society of Chemistry,Thomas Graham House,Science Park, Milton Road,CAMBRIDGE CB4 4WF, UK or:Dr. J. F. TysonUS Associate Editor, The AnalystDepartment of ChemistryUniversity of MassachusettsAmherst MA 01003, USAParticular attention should be paid to the use ofstandard methodsof literature citation, 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 either to the Editor, or AssociateEditor, The Analyst (addresses as above). Mem-bers of the Analytical Editorial Board (who maybe contacted directly or via the Editorial Office)would welcome comments, suggestions andadvice on general policy matters concerningThe Analyst.Fifty reprints are supplied free of charge.@ The Royal Society of Chemistry, 1992. 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, or otherwise, without the prior permissionof the publishers
ISSN:0003-2654
DOI:10.1039/AN99217FX027
出版商:RSC
年代:1992
数据来源: RSC
|
2. |
Contents pages |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 029-030
Preview
|
PDF (205KB)
|
|
摘要:
ANALAO 117(7) 1061-1214 (1992)The AnalystJuly 199210611071107510851093109911051111112911331137114511511157116111651169117311751179The Analytical Journal of The Royal Society of ChemistryCONTENTSCo-fluorescence Effect in Time-resolved Fluoroimmunoassays. A Review-Yong-Yuan Xu, llkka A. Hemmila, Tim0 N.-E.LovgrenQuantification of Theophylline in Human Plasma by Reversed-phase Ion-interaction High-performance LiquidChromatography and Comparison With the TDx Fluorescence Polarization lmmunoassay ProcedureM. C.Gennaro, C. Abrigo, P. BiglinoDetermination of Thallium in Cement Dust and Sediment Samples by Differential-pulse Anodic Stripping Voltam-metry: A Chemometric Approach t o Linear Calibration-Mahmoud A. Allus, Richard G.BreretonMultivariate Analysis of a Round-robin Study on the Measurement of Chlorobiphenyls in Fish Oil-Raj K. Misra, John F.Uthe, Charles J. MusialEdible Fats and Oils Reference Materials for Sterols Analysis With Particular Attention t o Cholesterol. Part 1.Investigation of Some Analytical Aspects by Experienced Laboratories-Georges Log nay, Michel Severin, AchimBoenke, Peter J. WagstaffeFocal Plane Charge Detector for Use in Mass Spectrometry-Keith BirkinshawGas Chromatographic-Mass Spectrometric Characterization of Flavanones in Citrus and Grape Juices-Colin S.Creaser, Mohammed R. Koupai-Abyazani, G. Richard StephensonDetection and Identification of Volatile Substances by Headspace Capillary Gas Chromatography t o Aid the Diagnosisof Acute Poisoning-Peter J.Streete, Manjit Ruprah, John D. Ramsey, Robert J. FlanaganComparison of Solvent Extraction and Solid-phase Extraction for the Determination of Organochlorine PesticideResidues in Water-Guan H. TanSynthesis of Tin(iv) Vanadopyrophosphate as a Novel Stationary Phase for High-performance Liquid Chromatographyand Its Application t o Amino Acid Analysis-Xing-Dong Yao, Jin-Chun Liu, Jin-Lan Xia, Jie-Ke Cheng, Yun'e ZengComparison of Ion Chromatographic Methods for the Determination of Organic and Inorganic Acids in PrecipitationSample-Venghuot CheamAnion-exchange Chromatography of Mixed Cyano Complexes: Separation and Determination of Dicyanoaurate(1)-Emmanuel 0. Otu, Campbell W. Robinson, John J. ByerleyDetermination of Osmium Abundance in Molybdenite Mineral by Isotope Dilution Mass Spectrometry With MicrowaveDigestion Using Potassium Dichromate as Oxidizing Agent-Katsuhiko Suzuki, Qi-Lu, Hiroshi Shimizu, AkimasaMasudaDetermination of Nickel in Biological Materials After Microwave Dissolution Using Inductively Coupled Plasma AtomicEmission Spectrometry With Prior Extraction into Butan-I-ol-Elisa Vereda Alonso, Amparo Garcia de Torres, JoseM.Can0 PavonDetermination of Butyltin Species in Sewage and Sludge by Gas Chromatography-Atomic Absorption Spectrometry-Y. K. Chau, Shuzhen Zhang, R. J. MaguireDetermination of Mercury in Fluorescent Lamp Cullet by Atomic Absorption Spectrometry-Ryszard Dobrowolski,Jerzy MierzwaMatrix Effect Corrections for the Quantitative X-ray Fluorescence Determination of Iron Using Scattered Radiation-Maria Teresa Garcia-Gonzalez, Maria Dolores Haro-Ruiz, Alfonso Hernandez-LagunaOxalate-catalysed Oxidimetric Assay of Thiourea With Permanganate-Alice KurianSpectrophotometric Determination of Titanium(iv) Using Chromotropic Acid and a Flow Injection Manifold-RajeshPurohit, Surekha DeviSimplex-optimized and Flow Injection Spectrophotometric Assay of Tetracycline Antibiotics in Drug Formulations-Salah M.Sultan, Fakhr-Eldin 0. Suliman, Salih 0. Duffuaa, ldeisan I. Abu-AbdounTypeset and printed by Black Bear Press Limited, Cambridge, England0003-2654(199237-1185 Light Scattering Method for the Determination of Trace Amounts of Phosphate Using a Cationic Water-solublePorphyrin-Masaaki Tabata, Keisuke Harada1189 Spectrofluorimetric Determination of Boron in Soils, Plants and Natural Waters With Alizarin Red S-Ana M.GarciaCampafia, Fermin Ales Barrero, Manuel Roman CebaPAPER PRESENTED AT THE XXVll CSI, BERGEN, NORWAY, JUNE 6-8,19911193 Mobility of Superficially Applied Caesium-134 and Strontium-85 in Apple Branches Under Precipitation-freeConditions-Gunnar B. Bengtsson1197 BOOK REVIEWS1212 ERRATUM1213 CUMULATIVE AUTHOR INDEX(i) FACSS XIXROYAL SOCIETY OF CHEMISTRYBiotransformations:A Survey of the Biotransformations of Drugs and Chemicals in AnimalsEdited by David R. Hawkins, Huntingdon Research CentreBiotransformations is an important series which has been devised to bring together all current information on the subject, in a readily accessible form.It provides up-to-date information on the biotransformation of pharmaceuticals, pesticides, food additives, and environmental and industrial chemicals inanimals and will be of great interest to chemists, biochemists and toxicologists in a wide variety of industries, as well as to regulatory authorities andlegislative bodies.Each volume reviews biotransformation pathways that have been reported in the literature in the preceding twelve months andalthough successive volumes may cover similar compounds the biotransformation pathway in each case will differ.The abstracts are arranged according to chemical classes and are assigned to key functional groups - a concept which has been developed to provideaccess to information on the biotransformations of compounds with similar structural features.Biotransformations Volume 4 is cumulatively indexed by compound, key functional group and reaction type, covering all entries in Volumes 1-4.Volume 4 Volume 3 Volume 2 Volume 1Hardcover xxx + 492 pagesISBN 0 85186 187 3 (1992)Price €99.50 Price €89.50 Price €79.50Hardcover xviii + 462 pagesISBN 0 85186 177 6 (1991)Hardcover xx+496 pagesISBN 0 85186 167 9 (1990)Hardcover xxii+511 pagesISBN 0 85186 157 1 (1989)Price €79.50Special Package Price (Volumes 1 4 ) f299.00ROYALSOCIETY OFCHEMISTRYInformationServices~~~ ~ ~~To Order, Please write to the: Royal Society of Chemistry, Turpin Distribution Services Limited, Blackhorse Road, Letchworth, Herts SG6 1 HN,UK, or telephone (0462) 672555 quoting your credit card details.We can now accept AccessNisalMasterCard/Eurocard.Turpin Distribution Services Limited, is wholly owned by the Royal Society of Chemistry.For information on other books and journak, please write to the:Royal Society of Chemistry, Sales and Promotion Department, Thomas Graham House, Science Park, Milton Road,Cambridge CB4 4WF, UK.RSC Members should obtain members prices and order from :The Membership Affairs Department at the Cambridge address above.Circle 006 for further informatio
ISSN:0003-2654
DOI:10.1039/AN99217BX029
出版商:RSC
年代:1992
数据来源: RSC
|
3. |
Co-fluorescence effect in time-resolved fluoroimmunoassays. A review |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1061-1069
Yong-Yuan Xu,
Preview
|
PDF (1152KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1061 Co-fluorescence Effect in Time-resolved Fluoroimmunoassays A Review Yong-Yuan Xu Analytical Chemistry Laboratory, Institute of Atomic Energ y, P.U. Box 275-88, Beijing 1024 13, China llkka A. Hemmila" Wallac Oy, P.O. Box 10, 20101 Turku, Finland Tim0 N.-E. Lovgren Department of Biochemistry, University of Turku, Turku, Finland Summary of Contents Introduction Time-resolved FI uoroi m mu noassays Co-fluorescence Effect 6-Di ketones Enhancing Ions Synergistic Agents Water-soluble Organic Solvents Buffers Preparation Enhancement Kinetics FI u o rescence Characteristics Time-resolved lmmunofluorimetry of FSH Double-label lmmunometric Assay of LH and FSH Quadruple-label lmmunoassay Composition of Co-fluorescence Enhancement Solution Co-f I uorescence En ha ncement Solutions Co-f luorescence En ha ncement in I m m u noassays Conclusions References Keywords: Time-resolved fluoroimmunoassa y; co-fluorescence enhancement; re view Introduction Time-resolved Fluoroimmunoassays In immunoassays the sensitivity is of prime importance, because often very low concentrations of analytes are present in the samples.In order to ensure the assay sensitivity, the antibody, antigen or its derivative has to be coupled with an easily detectable label. The conventional radioimmunoassays (RIA) and immunoradiometric assays (IRMA) utilize radio- active isotopes as the labels and are among the most sensitive and specific analytical techniques available today. The sensi- tivity of RIA, however, is still limited to a concentration range of 1 x 10-12-1 x 10-14 mol dm-3, and the relevant working range is often between 1 x 10-8 and 1 x 10-10 mol dm-3.The disadvantages of radioactive labels relate to their limited storage time and the legal problems and health hazards in their handling and disposal. The future use of radioisotopic techniques is also hindered by difficulties encountered in assay automation; automated RIA systems for routine diagnostic use have not emerged. Consequently, intensive research has been carried out with the aim of replacing the radioisotopes with stable, non-radioactive labels with a high specific activity. Fluorescent labels and fluorimetric detection represent one of the alternative non-isotopic techniques considered to have the potential to replace radioisotopic labels. The sensitivity of fluorescent labels, although theoretically very high, is, however, limited in routine immunoassay conditions by a high background interference. The background signal originates from various sources, such as fluorescent compounds in the * To whom correspondence should be addressed.sample, impurities in reagents and luminescent compounds in cuvettes and lenses, etc. For example, serum, the most widely used sample in immunoassays, causes very high background interference and, consequently, fluorimetric analysis in the presence of serum components (homogeneous assays) is limited to ymol dm-3-nmol dm-3 concentrations. The scatter- ing of the excitation light by sample constituents and solid-phase materials also gives rise to interference, especially when labels with a small Stokes' shift (less than 50 nm) are used.The high background limits the sensitivity of conven- tional fluoroimmunoassays by factors ranging from 100 to 1000.1 22 The use of delayed measurement of a long-lifetime fluores- cence makes it possible to separate the specific fluorescence of the label from most of the disturbing, unspecific background. In time-resolved fluorimetry the fluorescent label is excited by repeated short light pulses and the specific fluorescence is detected when a predetermined period of time (delay time) has elapsed after the excitation pulses. In order to be practical in routine immunoassay conditions, the fluorescent label employed should have as high a fluorescence as possible, produce emission preferably at a long wavelength and with a large Stokes' shift and, in particular, should have an excited- state lifetime clearly longer than the average duration of the background.1.3 The fluorescent properties of some tripositive lanthanide ions, and especially the chelates of Eu3+, Tb3+ and Sm3+, are particularly well suited for time-resolved fluorimetry. In these chelates the strong ion emission originates from an intra- chelate energy transfer, where the organic ligand absorbs the excitation radiation in the ultraviolet (UV) range and transfers1062 ANALYST, JULY 1992, VOL. 117 the excited energy through its triplet state to the emitting ion.4.5 The ligand field around the ion also prevents the quenching caused by coordinated water molecules which in aqueous solution tend to create an efficient deactivation route .6 The ion-specific emission appears at narrow banded lines at long wavelengths (Tb3+ 544 nm, Eu3+ 613 nm, Sm3+ 643 nm) with a long Stokes’ shift (230-300 nm).The most important feature in this context is the long fluorescence lifetime, ranging from 1 ps to over 2 ms, which makes it possible to apply time-resolved detection for the effective elimination of the background and to increase the sensitivity. The requirements set on a lanthanide chelate applied as a label in immunoassay relate to its stability, its fluorescence intensity in an aqueous environment, its hydrophilicity and the suitability for covalent coupling with antibodies. So far, the most successful method meeting these requirements is a fluorescence enhancement-based technique, marketed as DELFIA (Wallac, Turku, Finland).In this technique the lanthanide ion, generally Eu3+, is coupled to antibodies or antigens via bifunctional chelating agents to obtain stable but almost non-fluorescent labelled reagents.7.8 After completion of the heterogeneous immunoassay on a plastic surface (most often on microtitration strip wells), the label ion is dissociated by an acidic enhancement solution. The dissociated ion is simultaneously converted into a new, highly fluorescent chelate with P-diketones and a synergistic agent, trioctyl- phosphine oxide (TOPO), present in high excess in the acidic detergent solution.9~10 The enhancement solution forms opti- mum conditions for the chelate fluorescence, and a sensitivity of 1 x 10-14 mol dm-3 can be reached for Eu3+ detection in a sensitive time-resolved fluorimeter.3 A comprehensive list of the various DELFIA applications available for clinical immu- noassays, research and deoxyribonucleic acid (DNA) hybridi- zation, can be found in a recent review.” The ion-specific, narrow-banded emissions of the different lanthanide chelates, which, in addition, also show clearly different decay times, offer a tempting possibility to create double- or even multiple-label assays.10.11 The composition of the enhancement solution can be stipulated for the optimized measurement of either Eu3+ and Sm3+8710 or Tb3+,12 e.g., by changing the ligand from a fluorinated aromatic to a fluori- nated aliphatic P-diketone.By using this system (in this context called a direct fluorescence enhancement system, DFES), double-label based on Eu3+ and Sm3+*3914 or Eu3+ and Tb3+153 have been tested.In the direct system, the simultaneous use of three labels either results in a lower assay sensitivity or requires a two-step approach.17 Addition of the fourth label, Dy3+, to the direct system is hindered by its low emission quantum yield and, in particular, by its sub-micro- second decay time.” In the search for more sensitive methods enabling several analytes to be measured simultaneously with high sensitivity, the co-fluorescence type of enhancement (CFES) is potentially a very attractive approach. Co-fluorescence Effect The addition of yttrium or certain lanthanide ions can considerably enhance the fluorescence intensities of the P-diketone chelates of Eu3+ and Sm3+ when the chelates are present as a suspension in an aqueous solution.This type of fluorescence enhancement was first reported in 196418 and in 196719 in a study of Eu3+ (or Sm3+)-thenoyltrifluoroacetone (nA)-phenanthroline (Phen) and Eu3+ (or Sm3+)-7TA- collidine chelates coprecipitated in the presence of Gd3+ or Tb3+. Later it was shown that this type of fluorescence enhancement of Eu3+ or Sm3+-7TA chelates actually is an intrinsic fluorescence phenomenon named the ‘co-fluores- cence effect.’20-25 The fluorescence enhancement by CFES is based on an intermolecular energy transfer from the chelates of the emitting ion (such as Eu3+ and Sm3+). In all of the CFES systems studied both chelates have identical structures and are formed with f3-diketones (e.g., TTA) in addition to some synergistic ligands (e.g., TOPO or pyridine derivatives).The chelates formed have the structure Ln(L)3S1-2. The co-flu- orescence can be found in coprecipitates, in chelate suspen- sions and also in a micellar environment.22.24 In an actual solution, e.g., in benzene, there is no co-fluorescence en- hancement because the chelates exist as single molecules and the long distance between the chelates makes intermolecular energy transfer impossible. In an aqueous solution, however, van der Waals forces aggregate the hydrophobic chelates, which form tiny particles of controlled composition, and in the presence of a high excess of the enhancing chelates the emitting chelates are closely associated with a large number of enhancing chelates within the particles.The close contact allows efficient energy transfer from the latter to the former. The proposed energy transfer mechanism can be resonance energy transfer, collisional transfer of exciton migration. Ci and Lan24 pointed out that the energy transfer might be of exciton transfer type because it is independent of the acceptor concentration. The aggregates formed resemble micro- crystals, in which the structurally identical chelates are highly organized containing the emitting chelates as ‘impurities.’ The absorbed energy delocalizes through the chelate matrix, formed by the enhancing chelates, and ends up in the emitting chelates which produce the emission.22.24 The mechanism of the energy transfer process in CFES is illustrated in Fig.1. First, the organic ligand of both chelates (such as Ln-=A3) absorbs the excitation light and its electrons are raised from the singlet ground state (So) to one of the vibrational multiples of the excited state (S1). Subse- quently, the energy falls rapidly to the lowest level of S1 through non-radiative deactivation processes, and by inter- system crossing to the central ion-stabilized triplet state (T1). In the luminescent chelate, the excited energy at T1 is transferred by intrachelate energy transfer to the lowest resonance level of the emitting ion (e.g., Eu3+), whereafter the ion undergoes a radiative transition resulting in a characteristic line-type emission characteristic of the ion. In the aggregates most of the excitation light is absorbed by the chelates of non-emitting ions, present in high excess.The W I 2 =I s. intersvstem -+1 crossing # I-- I ntra molecu la r 1- Y -0 Inter m o lecu I a r energy transfer a t a v) a -3 7F U’ - Ligand Eu3+ Gd3+ Ligand - Fluorescent chelate Enhancing chelate . V Fie. 1 Schematic diagram of the energy flow in ligand-sensitized enhancing ion (such as Gd3+ and Tb3+) to the chelates of the direct and indirect (co-horescence) c h e k e luminesckeANALYST, JULY 1992, VOL. 117 1063 excited energy at the ligand triplet state cannot be transferred to its central ion, e.g., with Gd3+, because its resonance energy level is located at a much higher level. Because no radiative excited-state deactivation processes exist within the enhancing chelates, the stabilized triplet state transfers the energy by intermolecular energy transfer of the nearby chelates in the aggregated particles, thus producing a con- siderably enhanced luminescence intensity.Composition of Co-fluorescence Enhancement Solution The hitherto studied co-fluorescence-sensitized enhancement systems, similar to most of the direct systems, are based on P-diketones as the energy-absorbing ligands forming tris-(3- diketonato chelates with the trivalent lanthanide ions. In addition to the P-diketones, synergistic ligands, most fre- quently N-heterocyclic Lewis bases, are utilized as neutraliz- ing and water molecule-replacing ligands. Analogously to the direct systems, detergents have been found to stabilize the system and sometimes organic solvents have a positive effect on the fluorescence.P-Diketones P-Diketones have gained the widest use as the ligands to increase lanthanide fluorescence. As bidentate chelating agents, they form relatively stable chelates. The six-mem- bered ring involved in the chelate structure directly absorbs the excitation light and efficiently transfers the energy to the chelated ion. Aromatic P-diketones containing a triflu- oromethyl group are commonly used for the fluorimetric determination of Eu3+ and Sm3+, e.g., after extraction into organic solvents.26~27 These ligands are also used in most of the DELFIA-type time-resolved fluoroimmunoassays.8-11 Fluori- nated aliphatic P-diketones have the excited triplet state at a higher energy level and, therefore, can also be used for sensitizing Tb3+ and Dy3+ luminescence,11.12 and can be applied to the simultaneous detection of Eu3+, Tb3+, Sm3+ and Dy3+.Thenoyltrifluoroacetone is one of the most commonly used P-diketones to detect Eu3+ and Sm3+, used either in organic solvents or in aqueous solution .26,27 Thenoyltrifluoroacetone also promotes a strong co-fluorescence effect, and so far most of the co-fluorescence experiments have been performed using TTA as the fluorogenic ligand.19-25 We have found, however, that other P-diketones also show the co-fluorescence effect under appropriate conditions28.29 (Table 1), and by applying, e . g . , aliphatic p-diketones the co-fluorescence enhancement system can be expanded for the determination of Tb3+ and Dy3+.29330 The co-fluorescence enhancement efficiency greatly depends on the structure of the P-diketone.According to our results, the P-diketones listed in Table 1 can be classified into three major categories: (1) (3-diketones exhibiting a strong co-fluorescence effect, such as TTA, BTA and PTA; (2) P-diketones showing a less intense co-fluorescence effect, e.g., FTA, FBTA, TFMH, DPM, HFAcA, PFH, PFDMH and HFDMO; and (3) P-diketones which show only a weak interchelate energy transfer under the conditions tested, such as DBM, PFPP, DFBM, (3-NTA, AcA, TFAcA, TFTD and PFTD . Although a quantitative correlation between the structure of the 6-diketone and the strength of the co-fluorescence effect has not been verified, some qualitative consistencies have been observed. Generally, a CF3 group is needed for the enhancement of both co-fluorescence and direct fluorescence, but a higher level of fluorination did not increase the co-fluorescence efficiency.A long fluorocarbon side-chain (R1, in TFTD and PFTD) or a bulky aromatic structure (DBM, DFBM, P-NTA), however, had a negative effect on the interchelate energy transfer. Accordingly, the optimum ligand in the direct enhancement system, P-NTA,s showed only a weak co-fluorescence effect. The excitation of Eu3+ and Sm3+ is accomplished more efficiently through aromatic P-diketones, and accordingly the highest sensitivities are obtained with P-diketones such as TTA or BTA. On the other hand, multi-label assays, in which more than two analytes are to be measured simultaneously, require an enhancement system composed of aliphatic p-diketones.160 I 1 40 2 20 L L 0 20 40 60 80 100 120 140 160 PTNymol dm-3 Fig. 2 Effect of PTA on the fluorescence of 40 pmol dm-3 E d + (W) and 25 pmol dm-3 Tb3+ (A) in the PTA-Phen-Y3+-Triton-ethanol system Table 1 Structures of j3-dikctones (general formula R*COCH2COR2) with co-fluorescence enhancement effect Aromatic j3-diketones- R' Thenoyltrifluoroacetone (TTA) Benzoyltrifluoroacetone (BTA) 2-Furoyltrifluoroacetone (FTA) p-Fluorobenzoyltrifluoroacetone (FBTA) P-Naphthoyltrifluoroacetone (P-NTA) 1,l ,1,2,2-Pentafluor0-5-phenylpentane-3,5-dione (PFPP) Dibenzoylmethane (DBM) Di-p-fluorobenzoylmethane (DFBM) Pivaloyltrifluoroacetone (PTA) l,l,l-Trifluoro-6-methylheptane-2,4-dione (TFMH) Dipivaloylmethane (DPM) 1,1,1,5.5,5-Hexafluoroacetylacetone (HFAcA) 1 ,l ,1,2,2-Pentafluorohexane-3,5-dione (PFH) 1,1,1,2,2-Pentafluoro-6,6-dimethylheptane-3,5-dione (PFDMH) 1.1,1.2,2,3,3-Heptafluoro-7,7-dimethyloctane-4,6-dione (HFDMO) 1,1,1,2,2-Pentafluorotetradecane-2,4-dione (PFTD) 1,1,1-Trifluorotridecane-2,4-dione (TFTD) l,l,l-Trifluoroacetylacetone (TFAcA) Acetylacetone (AcA) Aliphatic j3-diketone.s-1064 ANALYST, JULY 1992, VOL.117 The effect of the P-diketone concentration on the co-fluor- escence enhancement is exemplified in Fig. 2, where the effect of PTA on both Eu3+ and Tb3+ emissions in the CFES system is shown. The optimum concentration range can either be wide (as for PTA in Eu3+ detection) or narrow (as for PTA in Tb3+ detection). The optimum concentration of 6-diketones is almost always sufficiently high to ensure ternary tris-chelate formation with the enhancing ion (concentration of P-dike- tone more than three times that of the enhancing ion).Enhancing Ions The enhancing ions generally applied are Gd3+, Tb3+, Lu3+, La3+ and Y3+. Under some conditions, a weak co-fluores- cence enhancement effect is also obtained with Yb3+ and Dy3+. In the co-fluorescence enhancement system the enhanc- ing ion used must not have excited 4f or 4d levels situated below the excited triplet level of the P-diketone used. Hence the energy absorbed by these chelates cannot be dissipated through these non-existing energy levels. Instead, the energy is transferred to the fluorescent ions through an intermol- ecular energy transfer. Enhancing ions are needed to provide the high molar excess of the triplet sensitizer to ensure a linear response for the acceptor detection.The high concentration is also needed to create the aggregate particles inside which the efficient energy transfer is possible. In the mA-based co-fluorescence system the fluorescence intensity of Eu3+ increased with increasing concentration of Y3+ (Fig. 3), and the maximum fluorescence intensity was obtained at a Y3+ concentration of about 5.0 pmol dm-3. The optimum concentration was not dependent on the Eu3+ concentration. A similar result was also obtained with Sm3+ (data not shown). Hence no constant molar ratio of Eu3+ (or Sm3+) to Y3+ exists in the particulate chelate structures, suggesting that there are no defined mixed chelate structures in the particles. Taking into account the bidentate nature of 107 r I r 1,106 v) 4- 105 ; 104 E 103 0 \ a 0 ii 102 I I I 0 5 10 15 20 Y3+/1.~mol dm-3 Fig.3 Effect of Y3+ on the fluorescence of 0.5 pmol dm-3 (V), 5 pmol dm-3 (A) and 50 pmol dm-3 (@) Eu3+ in the TTA-Phen-Y3+- Triton system Table 2 Enhancement factors* of Eu3+ and Sm3+ fluorescence by different enhancing ions in the Eu3+ (or Sm3+)-TTA-Phen-Ei-Triton X-100 system Fluorescence enhancement factor Enhancing ion Gd3+ Y3+ Tb3+ Lu3+ La3+ Yb3+ Dy3+ Eu3+ Sm3+ 740 766 526 81 1 493 757 504 729 135 168 45 20 15 19 * The enhancement factor is defined as the ratio of the fluorescence intensity in the presence of an enhancing ion to that in its absence. P-diketones, the formation of mixed ligand chelates would be most unlikely. Actually, the chelate stoichiometry of 1 : 3 : 1-2 (Ln : P-diketone : synergistic ligand) has been verified for co-fluorescent aggregates.22.24 The co-fluorescence effect is caused only by the intermolecular energy transfer between the two chelate types.The enhancement factors obtained with the different enhancing ions (Ei) in the Eu3+ (or Sm3+)-?TA-Phen-Ei- Triton X-100 systems are given in Table 2. The lanthanides Gd3+, Tb3+ and Lu3+ and the non-lanthanide Y3+ gave the highest enhancement factors in the TTA-based CFES studied. An additional aspect in the choice of the enhancing ion is the inherent contamination-even trace amounts of the emitting lanthanides as impurities cause a very high background. In the respect the non-lanthanide Y3+ is to be preferred. Synergistic Agents The strong fluorescence intensity of the lanthanide 6-diketone chelates requires a non-aqueous chelate environment.The three P-diketone molecules in the tris-chelate occupy only six of the nine available coordination sites of the ion, which still remains sensitive to quenching by water. Halverson et a1.31,32 introduced the concept of an ‘insulating sheet,’ in which a synergistic ligand (or a synergistic agent) was involved in the chelate structure to replace water molecules from the ligand field. It also acts as a shield, protecting the chelate from external interactions and thus efficiently reducing non-radia- tive energy degradation. The chelate formed has the structure Ln3+ (P-diket~ne)~(L)~-~. According to our investigation, the synergistic ligands applicable for co-fluorescence enhancement are 1,lO-phenan- throline (Phen) and its derivatives, e.g., 4,7-(or 5,6)- dimethyl-1 ,lo-phenanthroline, 4,7-diphenyI-l,lO-phenan- throline and 2,9-dimethyl-4,7-diphenyl- 1 , 10-phenanthroline, pyridine derivatives such as 2,2‘-dipyridyl (DP) , 2,2’-dipyr- idylamine, 2,4,6-trimethylpyridine, 2,2’ : 6’,2”-tetrapyridine and 1,3-diphenylguanidine.Trioctylphosphine oxide is the most effective synergistic agent in the direct enhancement solutions and has been applied in some CFES studies.21 In our systems, however, it almost totally hindered the interchelate energy transfer within the CFES aggregates. Usually, the fluorescence on the lanthanide chelate increases with increasing concentration of the synergistic ligand until a maximum and almost stable fluorescence is reached.The effect of the Phen concentration on the fluorescence of Eu3+ and Sm3+ chelates in BTA-Ys+-based CFES is shown in Fig. 4.28 Detergents have a positive effect both on the direct system and on the co-fluorescence-enhanced system. The micellar environment protects the chelates against non-radiative processes, provides a non-aqueous environment for the chelate and ensures an environment in which organized 2000 v) v) 4- 5 1500 s m I 0 a C a y 1000 500 L 0 3 U - 0 20 40 60 80 100 Phen/prnol dm-3 Fig. 4 Effect of Phen on the fluorescence of50 pmol dm-3 Eu3+ (V) and 3.5 nmol dm-3 Sm3+ (A) in BTA-Phen-Y3+-Triton X-100 based CFESANALYST, JULY 1992, VOL. 117 1065 crystalline stuctures can be formed. Detergents can also solubilize the particles and stabilize the solutions by prevent- ing the sedimentation of the particles.The type of detergent used in CFES varies according to the system studied and positively or negatively charged and non-ionic detergents have been tested. Ci and Lan24 reported that an increasing concentration of detergent, e.g., Triton X-100, can even Preparation a quenching effect. A typical effect of pH on the co- fluorescence of Eu3+ and Sm3+ in BTA-Phen-Y3+-Triton- based CFES is shown in Fig. 7.28 Co-fluorescence Enhancement Solutions solubilize the aggregate particles. However, the effective concentration of detergents is well below the critical micelliza- tion concentration (c.m.c.). We have studied the effect of non-ionic detergents, Triton X-100, Triton N-101 and Triton X-450, on the co-fluorescence system.The types of effect caused by the detergent on the direct fluorescence and the co-fluorescence enhancement are different. An increasing concentration of the detergent did not change the fluorescence of the Tb3+-PTA chelate after reaching the optimum level in DFES consisting of lW+-PTA-TOPO-Triton X-10O.l2 In the respective co-fluorescence system based on Tb3+-PTA- Phen-Y3+-Triton X-100, the intensity gradually decreased beyond the optimum Triton X-100 concentration of about 0.06% .29 Obviously, at the higher Triton concentrations the chelate densities within the micelles are decreased, and eventually the fluorescent chelates and enhancing chelates may even be separated into different micelles if the micelle concentration exceeds that of the chelates. Under such conditions interchelate energy transfer is impossible.A small concentration of detergent also stabilizes the co-fluorescence system. The tiny particles in the detergent buffer remain in suspension for a longer time. The effect of different Triton X-100 concentrations on the stability of the Eu3+-BTA- Phen-Y3+-Triton X-100 system is shown in Fig. 5. Water-soluble Organic Solvents Some water-soluble organic solvents, such as ethanol, pro- panol, dimethyl sulfoxide, 2-methoxyethanol and ethane-l,2- diol, can improve the co-fluorescence intensity. In the Tb3+-PTA-Phen-Y3+-Triton X-100-ethanol system, the flu- orescence of the Tb3+ chelate is enhanced 186-fold with the addition of ethanol (Fig. 6).29 It is likely that the presence of ethanol favours the formation of optimally sized particles.Buffers Unlike the direct enhancement system, which can be opti- mized both to dissociate the ion by the acidic pH and to create the optimally emitting chelates simultaneously,8 the optimum and stable enhancement in the co-fluorescence system requires careful adjustment of pH near the neutral range. Acetate, hexamine and tris(hydroxymethy1)methylamine (Tris) can be used, whereas phosphate and citrate buffers have 500 1 1 I I I 0 20 40 60 80 Ti me/m i n The final CFES is stable for only a limited period of time because the solution contains tiny particles present in the micellar environment. The optimum pH of the CFES is near neutral. Consequently, CFES cannot be applied directly in the dissociation-enhanced lanthanide fluoroimmunoassay prin- ciple.Instead, the CFES components are prepared, stored and used in two separate solutions. The first solution (E,) contains the (3-diketone, the enhancing ion, the detergent and, sometimes, a water-soluble organic solvent, solubilized in an acidic buffer (pH 3.1-3.5). The second solution (Eb) contains the synergistic ligand and a basic buffer. When CFES is employed, e.g., in immunoassays based on antibodies labelled with lanthanide chelates, the solution E, is first used to dissociate the ions and thereafter the solution E b is added (usually one tenth of the E, volume) to raise the pH to the optimum level and to allow the formation of the particulate structures necessary for the co-fluorescence enhancement.The compositions of five different CFESs are given in Table 3. The sensitivities obtained with the CFES depend not only on the signal levels obtained but also on the signal-to-noise ratio. Consequently, the fluorescence intensity should be high, the decay time long and the background level low and reproducible. The background obtained is derived partly from the instrumental background and the photoluminescence of the solid-phase material (polystyrene) and to a great extent from the purity of reagents (contamination level). In order to keep the background at a low level, the follovving precautions VJ t U 3 40 8 m O 30 0 10 20 30 40 Ethanol (%) Fig. 6 Effect of ethanol on the fluorescence of 25 pmol dm-3 Tb3+ in the PTA-Phen-Y3+-Triton X-1Okthanol system at different Triton X-100 concentrations: 0 (+), 0.006 (V), 0.03 (A), 0.06 (M) and 0.12% (0) 2000 1 v) VJ - 5 1500 0 m 0 Q, c Q, u y 1000 500 0 L L - 0 4 5 6 7 8 PH Fig.5 Stability of fluorescence, on standing, of 50 pmol dm-3 Eu3+ in BTA-Phen-Y3+-Triton X-100 based CFES with different Triton X-100 concentrations: 0 (0), 0.005 (A), 0.01 (V), 0.02 (0) and 0.04% (M) CFES system Fig. 7 Effect of pH on the fluorescence of 50 pmol dm-3 E d + (A) and 3.5 nmol dm-3 Sm3+ (V) in the BTA-Phen-Y3+-Triton based1066 ANALYST, JULY 1992, VOL. 117 Table 3 Co-fluorescence enhancement characteristics of different systems CFES Shaking time/min Composition Fluorescent ion Ea* Eb TPY TPYE BPY PPY E PDY Eu3+, Sm3+ 70 pmol dm-3 TTA 7.5 pmol dm-3 Y3+ 0.98 cm3 dm-3 HOAc 30 pmol dm-3 TTA 1.5 pmol dm-3 Y3+ 0.06% TX-100 20% ethanol 0.96 cm3 dm-3 HOAc 60 pmol dm-3 BTA 8.5 pmol dm-3 Y3+ 0.98 cm dm-3 HOAc 70 pmol dm-3 PTA 7.5 pmol dm-3 Y3+ 0.06% TX-100 30% ethanol 1.04 cm3 dm-3 HOAc 100 pmol dm-3 PTA 3 pmol dm-3 Y3+ 2 cm3 dm-3 HOAc 0.075% TX-100 Eu3+, Sm3+ Eu3+, Sm3+ 0.02% TX-100 Eu3+, Sm3+ Tb3+, Dy3+ Eu3+, Sm3+ Tb3+, Dy3+ 0.0006% TX-100 * TX-100 = Triton X-100; HOAc = acetic acid.1.75 mmol dm-3 Phen 8 0.2 mol dm-3 Tris 0.4 mmol dm-3 Phen 7 0.2 mol dm-3 Tris 0.5 mmol dm-3 Phen 0.2 mol dm-3 Tris 0.5 mmol dm-3 Phen 0.2 mol dm-3Tris 5 mmol dm-3 DP 0.375 mol dm-3 Tris 80% ethanol 1 7 15 Waiting time/min 10 10 0 10 20 need to be taken. All reagents used to prepare the CFES must be very pure. Non-fluorescent lanthanides often contain high levels of fluorescent lanthanide ions, causing a high back- ground.A trace contamination level of 0.0001% in yttrium oxide (99.9999% pure) was found to be an acceptable level. All laboratory ware used in the preparation of CFES must be made of plastic and must be thoroughly washed with a suitable washing solution before use. Frequent rinses have to be carried out, e.g., with E, solution. During use and storage, special care must be taken in order to avoid contamination of the solution. Enhancement Kinetics In the co-fluorescence system, the ternary chelate structures exist as tiny particles which tend to precipitate out from the solution after prolonged standing. Standing of ready CFES for 24 h resulted in sedimentation of the particles and disappear- ance of the fluorescence from solution. The precipitated particulate material exhibits very strong fluorescence, whereas most of the fluorescence in solution disappears.Evidently both the fluorescent chelates and the enhancing chelates are in the same particles. The total enhancement time depends on the kinetic stability of the chelate used for labelling and on the time required for aggregate formation. When N1-(p-isothiocyanatobeny1)- diethyIenetriamine-Nl,W,N3,W-tetraacetic acid is used as the chelate for labelling of the reagents with the fluorogenic lanthanide,’ the ion dissociates within 3-5 min in the E, solution. After the addition of Eb the development of the final fluorescence depends on the speed of particle formation, which varies for different co-fluorescence enhancement systems (Table 3). Fluorescence Characteristics The excitation and emission spectra of Eu3+, Tb3+, Sm3+ and Dy3+ chelates in PPYE solution are shown in Fig.8. The excitation maximum of the PTA chelate is red shifted from the 295 nm obtained in DFESl2 and shows two maxima at 300 and 315 nm. A similar red shift has been reported with Eu-TTA chelates in CFES.24 The strongest emission peaks are found at 612, 544, 647 and 574 nm for Eu3+, Tb3+, Sm3+ and Dy3+ chelates, respectively, and are assigned to the transitions of Table 4 Fluorescence properties of lanthanides in CFES Excitation Emission Decay Composition Ion (max. )/nm (max .)/nm time/ys TPY Eu3+ Sm3+ TPY E Eu3+ Sm3+ BPY Eu3+ Sm3+ PPY E Eu3+ T b 3 + Sm3+ Dy3 + PDY Eu3+ Tb3+ Sm3+ Dy3 + 365 358 365 364 333 337 315 312 315 316 312 312 312 313 612 648 612 a 9 612 647 612 544 647 574 612 545 647 574 1062 96 730 96 764 79 820 323 88 27 948 239 48 11 5D0-+7F2 of Eu3+, 5D4-+7F5 of Tb3+, 4G5/2-+6H9/2 of Sm3+ and 4F9/2-+6H13/2 of Dy3+.The fluorescence properties of the chelates in the five CFES are given in Table 4. The excitation wavelength mainly depends on the b-diketones used and the emission wavelength completely on the fluorescent ion. The different ions also exhibit their typical decay times which, to some extent, depend on the ligand field around the ion. The measurement conditions, fluorescence intensities, back- grounds and detection limits of the four fluorescent lantha- nides measured in different CFESs are listed in Table 5. The calibration graphs exhibit wide linear responses with respect to emitting ion concentrations from 1 x 10-13 to 1 x 10-7 rnol dm-3.The detection limits with TPYE CFES were 2.5 X 10-15 mol dm-3 for Eu3+ and 8.9 x 10-14 rnol dm-3 for Sm3+, which are 15 and 37 times lower than those obtained using the respective DFES (3.9 x 10-14 rnol dm-3 for Eu3+ and 3.3 X 10-12 rnol dm-3 for Sm3+).* Regardless of both the spectral and temporal resolutions of the emissions of different lanthanides, minor spectral over- lapping occurs between some of the emissions. This needs to be corrected in simultaneous determinations in which the whole dynamic range of each ion has to be utilized. The signal cross-talk figures between the four fluorescent lanthanides in PPYE-based CFES are given in Table 6. The relatively broad emission peak of Sm3+ (Fig.8) causes a high signal interfer- ence in the Dy3+ channel.ANALYST, JULY 1992, VOL. 117 1067 Table 5 Measurement parameters and fluorescence of the lanthanides in CFES Measurement parameters Composition Ion TPY Eu3+ Sm3+ TPYE E u ~ + Sm3+ BPY E u ~ + Sm3+ PBYE E u ~ + n 3 + Sm3+ Dy3+ PDY E u ~ + Tb3+ Sm3 + Dy3+ Cycling/ ms 2 1 2 1 2 1 2 1 1 1 2 1 1 1 Delay/ ms 0.5 0.05 0.5 0.05 0.5 0.05 0.5 0.4 0.05 0.05 0.5 0.15 0.05 0.02 Counting/ ms 1.5 0.15 1.5 0.2 1.5 0.15 1.5 0.5 0.2 0.1 1.5 0.4 0.2 0.1 Excitation/ nm 310-400 310-400 310-400 310-400 310-400 310-400 250-400 250-400 250-400 250-400 250-400 250-400 250-400 250-400 Emission/ nm 613 643 613 643 613 643 613 545 643 573 613 545 643 573 Fluorescence for 1 nmol dm-3 Ln3+/l@ counts Background/ S-1 counts s-1 84 130 5 700 602 180 83 900 1456 608 85 41 940 1 860 324 400 3 640 760 1 880 4 700 12 410 20 6 200 6 846 lo00 2 983 2 400 17 100 25 5 800 Detection limit/ pmol dm-3 0.015 0.12 0.0025 0.089 0.0043 0.11 0.035 0.34 7.9 46.0 0.019 0.27 3.8 20.0 Table 6 Spectral interference* in the co-fluorescence measurement of lanthanides in CFES containing PPYE Ion to be measured Interfering ion E u ~ + m3+ Sm3+ Dy3+ Eu3+ - 0.17 0.32 0.33 Tb3+ 3.36 - 0.49 1.69 Sm3 + 1.76 0.00 - 5.39 Dy3+ 0.00 0.00 0.31 - * Interference percentage is defined as the ratio of the signal of the interfering ion obtained in the measuring channel to that of the ion to be measured. Co-fluorescence Enhancement in Immunoassays Time-resolved Immunofluorimetry of FSH A solid-phase-based non-competititive time-resolved flu- oroimmunoassay of follicle-stimulating hormone (FSH) was used as a model to assess the sensitivity obtainable with a CFES.Microtitration strip wells were coated with P-FSH specific monoclonal antibodies and a-specific antibody was labelled with Eu3+ [17 Eu3+ per molecule of immunoglobulin G (IgG)]. In the first step, samples and standards were incubated in the wells for 1 h at room temperature in the DELFIA assay buffer. In the second step, after washing, the wells were incubated with 5 ng of Eu3+-labelled anti-a-FSH (1 h), then washed six times and the bound fraction of Eu label was dissociated by shaking the strips for 3 min with 200 mm3 of TPY solution E,. The fluorescence of the label was enhanced by solution Eb (20 mm3 per well) during 8 min of shaking.The solution was allowed to stand for an additional 10 min, then the fluorescence was measured with a time-resolved spectro- fluorimeter. A direct system (commercial DELFIA enhance- ment solution) was used as a reference. The dose-response curves of FSH using the two different enhancement solutions are shown in Fig. 9. In the assay CFES gave clearly higher signal levels but the sensitivity of the assay was comparable to that of DFES. In this instance the sensitivity was more dependent on the non-specific binding property of the labelled antibody than on the detection sensitivity of the label. Double-label Immunometric Assay of LH and FSH Depending on the P-diketones used, the CFESs can be applied in double-label time-resolved fluoroimmunoassays using either Eu3+ and Sm3+ or Eu3+ and Tb3+ as the labels.The simultaneous assay of luteinizing hormone (LH) and FSH has been chosen as a model for double-label a~say.15~28~29 The assays were based on (3-specific anti-FSH and anti-LH 4- '! .- [ii / \ > 4- 'z (a) a, C Q, C a, tn 1c a, .- L! 260 280 300 320 340 360 4- I I I a, a > tn a, C a, 4-4 .- 4- .- h,,/nm I e m h m Fig. 8 Tb'+ (c), Sm3+ (d) and Dy3+ (e) chelates in PPYE-based CFES Excitation spectrum (a); and emission spectra of Eu3+ (b),ANALYST, JULY 1992, VOL. 117 L 1068 107 c I v) v) w = 106 s 8 2 v) 105 . W v) 3 - Y- w LL u 104 105 Y I v) v) w 104 =I 8 . 0 C W v) 2 103 3; n 3 - W LL 102 0.1 1 10 FSH/U dm-3 Fig. 9 Dose-response curves of time-resolved immunofluorimetry of FSH measured with use of TPY-based CFES (0) and P-NTA-TOPO- Triton-based DFES (m) 0.1 1 10 100 LH or FSH/U dm-3 Fig.10 Dose-response curves of double-label time-resolved immu- nofluorimetric assay of LH with Eu3+ (round symbols) and FSH with Sm3+ (square symbols), when measured with TPY-based (solid symbols) or BTA-based CFES (open symbols) Table 7 Comparison of double-label and single-label immunoassays of LH and FSH Sample SensitivityIU dm-3 volume/ Assay Labels mm3 LH FSH RIA* TO, 1251 200 24.4 1.5-3.6 TR-IFMAT Eu3+, Tb3+ 100 0.1 1.0 TR-IFMA with TR-IFMA with CFES (TPY) Eu3+, Sm3+ 25 0.045 0.029 CFES (PPYE) Eu3+, Tb3+ 25 0.024 0.017 DELFIA (DFES)$ Eu3+ - 0.02 - DELFIA (DEFS)$ Eu3+ - - 0.01 * Ref. 34. 7 Ref. 15. 3 Ref. 35. antibodies labelled with either Sm3+ or Tb3+. The microtitra- tion plate wells were coated with a common antibody against the a-subunit of both hormones.In both of the model assays the coated wells were incubated with 25 mm3 of samples or standards in 200 mm3 of assay buffer for 1 h with constant shaking. After three washings the wells were further incu- bated either with a mixture of 25 ng of E$+-anti-P-LH and 500 ng of Sm3+-anti-P-FSH, or a mixture of 25 ng of Ed+-anti-P-LH and 100 ng of Tb3+-anti-P-FSH monoclonal antibodies (1 h at room temperature with constant shaking). The label fractions immunologically bound on the solid c I v) ~1000 0 7 100 w Z t Y- 0 1 E l o t l 5 h 10 E n 5 % 2 1 10 100 - LL 0.1 100 TSH/pU cm-3 I(b) 40 2o t 0 I 10 100 . - I I v) C 3 $100 8 0 10 2 O 1 8 m F v) *- Q) v) 8 0.1 = 10 c F 2,1000 s fn + a m 100 I 0 r .0” Y- o 10 W C 8 $ 1 W - L 17-cu-OHP/nmol dm-3 100 1000 lRT/pg dm-3 l5 10 E n 5 g 0 1 25 20 15 E f3 10 g I 5 I I I I0 10 100 1000 CK-MM/U dm-3 Fig. 11 Dose-response curves and precision profiles of quadruple- label time-resolved immunofluorimetric assay of (a) TSH with Eu3+ ; (b) 17wOHP with Tb3+; (c) IRT with Sm3+; and ( d ) CK-MM with Dy3+ surfaces were dissociated with 200 mm3 of E, and the co-fluorescence was enhanced with 20 mm3 of Eb solution. The dose-response curves of the assays obtained with TPY- or BPY-based CFES (Eu3+ and Sm3+ as the labels) are shown in Fig. 10.28 A comparison of four different double-label immunoassays of LH and FSH, one radioimmunoassay , one time-resolved fluoroimmunoassay with PTA-TOPO-based DFES, two assays with CFES and the respective single-label assays is given in Table 7.The sensitivities with CFES were considerably higher (5-50-fold) than those obtained with previous DFES-based double-label assays. Quadruple-label Immunoassay The applicability of CFES containing an aliphatic P-diketone (PPYE) has also been tested in a quadruple label assay of four analytes used for neonatal screening: thyroid stimulatingANALYST, JULY 1992, VOL. 117 1069 hormone (TSH), immunoreactive trypsin (IRT), creatine kinase MM (CK-MM) and 17-a-hydroxyprogesterone (17a- OHP) .33 Three of the assays were based on non-competitive sandwich-type assays with monoclonal antibodies and the assay of 17a-OHP on a competitive assay with secondary antibodies on the solid surface.The antibodies, anti-TSH, anti-IRT and anti-CK-MM, were labelled with Eu3+, Sm3+ and Dy3+, respectively. The antigen, 17a-OHP, was directly labelled with Tb3+. The assay was performed as a one-step assay in microtitra- tion strip wells coated with a mixture of four antibodies. The samples were blood spots dried on a filter disc (diameter 5 mm). Standardization was performed with a dried blood spot standard for TSH and liquid standards for the others. The amounts of tracers used were 50 ng of Eu3+-anti$-TSH, 100 ng of Sm3+-anti-IRT, 250 ng of Dy”-anti-CK-MM and 0.09 pmol of Tb3+-17wOHP. The primary rabbit-anti-17-a-OHP antibody was used at a dilution of 1 : 7500. After overnight incubation at 4 “C and six washings, the bound fractions of the labels were dissociated by shaking for 4 min with 200 mm3 of solution E, (PPYE) and the co-fluorescence was enhanced by adding 20 mm3 of solution Eb for 10 min.After waiting for 10 min the fluorescence was measured on the time-resolved fluorimeter using appropriate parameters (Table 5 ) . The dose-response curves and precision profiles of TSH, 17a- OHP, IRT and CK-MM are shown in Fig. 11. The sensitivity of the assay was 0.1 pU cm-3 for TSH, 3 nmol dm-3 for OHP, 2 pg dm-3 for IRT and 4 U dm-3 for CK-MM. Conclusions Time-resolved fluoroimmunoassays using lanthanides as labels have already gained wide acceptance in a variety of routine clinical and research applications. When using the direct fluorescence enhancement solution (DELFIA) based on p-NTA and TOPO, a sensitivity of 8 x 10-18 mol of Eu3+ per well has been obtained.Although the clinical range of most analytes does not require the utmost sensitivity, the improved sensitivity is needed for certain analytes. There is also a specific need to improve the sensitivities in multi-label assays, where a simple DFES can be applied only for two sensitive assays simultaneously. The co-fluorescence enhance- ment effect has offered a way to increase further the sensitivity of TR-FIA. The improvement is based on the additional energy transfer creating an amplified excitation efficiency of the (3-diketone chelates. In CFES a sensitivity as high as 5 x 10-19 mol of Eu3+ per well is achieved. The further delayed emission as a result of the additional energy transfer processes taking place prior to the emission also makes the fluorimetric measurement of Dy3+ possible in the present time-resolved fluorimetry .Hence CFES permits simple and sensitive single-, double-, triple- and quadruple-label immunoassays, a feature unique in comparison with other non-istotopic label tech- niques. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Wieder, I., in Immunofluorescence and Related Staining Tech- niques, eds. Knapp, H., Holubar, H., and Wick, G., Elsevier North-Holland Biomedical Press, New York, 1978, pp, 67-80. Soini, E., and Kojola, H., Clin. Chem. (Winston-Salem, N.C.), 1983, 29, 65. Weissman, S. I., J. Chem. Phys., 1942, 10, 214. Crosby, G., Whan, R., and Alire, R., J . Chem. Phys., 1961,34, 743. Bryden, C., and Reilley, C., Anal.Chem., 1982, 54, 610. Mukkala, V.-M., Mikola, H., and Hemmila, I . , Anal. Bio- chem., 1989, 176, 319. Hemmila, I., Dakubu, S . , Mukkala, V.-M., Siitari, H., and Lovgren, T., Anal. Biochem., 1984, 137, 335. Lovgren. T., Hemmila, I., Pettersson, K., and Halonen, P., in Alternative Immunoassays, ed. Collins, W., Wiley, Chichester, Hemmila. I., Scand. J. Clin. Lab. Invest., 1988, 48, 389. Hemmila, I., Applications of Fluorescence in Immunoassays, Wiley-Interscience, New York, 1991. Hemmila, I., Anal. Chem.. 1985, 57, 1676. Bador, R., Dechaud, H., Claustrat, F., and Desuzinges, C., Clin. Chem. (Winston-Salem, N.C.), 1987, 33, 48. Saarma, M., Jarvekiilg, L., Hemmila, I., Siitari, H., and Sinijarv, R., J. Virol. Methods, 1989, 23, 47. Hemmila, I., Holttinen, S . , Pettersson, K., and Lovgren, T., Clin. Chem. (Winston-Salem, N. C.), 1987, 33, 2281. Siitari, H., J. Virol. Methods, 1990, 28, 179. Hemmila, I.. in Fluorescence Spectroscopy, ed. Folfbeis, O., Springer, Heidelberg, 1992, in the press. Kononenko, L., Poluektov, N., and Nikonova, M., Zavod. Lab., 1964, 30, 779. Melenteva, E., Poluektov, N., and Kononenko, L., Zh. Anal. Khim., 1967, 22, 187. Yang. J.-H., Zhu, G.-Y., and Wu, B., Anal. Chim. Acta, 1987, 198, 287. Ci, Y.-X., and Lan, Z.-H., Analyst, 1988, 113, 1453. Ci, Y.-X., and Lan, Z.-H., Anal. Lett., 1988, 21, 1499. Yang, J.-H., Zhu, G.-Y., and Wang, H., Analyst, 1989, 114, 1417. Ci, Y.-X., and Lan, Z.-H., Anal. Chem., 1989,61, 1063. Yang, J.-H., Ren, X.-Zh., Zou, H.-B., and Shi, R.-P., Analyst, 1990. 115, 1505. Haddad, P., Talanta, 1977,24, 1. Yamada, S., Kano, K., and Ogawa, T . , Anal. Chim. Acra, 1982, 134, 21. Xu, Y.-Y.. Hemmila, I., Mukkala, V.-M., Holttinen, S., and Lovgren, T., Analyst, 1991, 116, 1155. Xu, Y.-Y., and Hemmila, I . , Anal. Chim. Acta, 1992, 256, 9. Xu, Y.-Y., and Hemmila, I., Talanta, 1992, in the press. Halverson, F., Brinen, J . , and Leto, J.. J. Chem. Phys., 1964, 41, 157. Halverson, F., Brinen, J . , and Leto, J . , J. Chem. Phys., 1964, 41, 2752. Xu, Y.-Y., Pettersson, K., Blomberg, K., Hemmila, I., Mikola, H., and Lovgren, T., unpublished work. Beinlich. C., Piper, J . , O’Neal, J . , and White, O., Clin. Chem. (Winston-Salem, N.C.), 1985, 31, 2014. Apter, D., Cacciatore, B., Alfthan, H., and Stenman, U.-H., J. Clin. Endocrinol. Metab., 1989, 68. 53. 1985, pp. 203-217. References 1 Soini. E . , and Hemmila, I., Clin. Chem. (Winston-Salem, N.C.), 1979, 25, 353. Paper 2100587E Received February 3, 1992 Accepted March 3, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701061
出版商:RSC
年代:1992
数据来源: RSC
|
4. |
Quantification of theophylline in human plasma by reversed-phase ion-interaction high-performance liquid chromatography and comparison with the TDx fluorescence polarization immunoassay procedure |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1071-1074
M. C. Gennaro,
Preview
|
PDF (425KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1071 Quantification of Theophylline in Human Plasma by Reversed-phase Ion-interaction High-performance Liquid Chromatography and Comparison With the TDx Fluorescence Polarization lmmunoassay Procedure M. C. Gennaro, C. Abrigo and P. Biglino Dipartimento di Chimica Analitica, Universita di Torino, Via P. Giuria, 5-1- 10 125 Torino, Italy A new method for the determination of theophylline (1,3-dimethyl-l H-purine-2,6-dione) in human plasma is described, free from interference by theobromine (3,7-dimethyl-1 H-purine-2,6-dione) and caffeine (1,3,7- tri met hyl-I H-pu ri ne-2,6-d ione). The method makes use of ion-i nteraction reversed-p hase hig h-performance liquid chromatography (octylamine-orthophosphate being the interaction reagent and a CI8 reversed-phase column the stationary phase) with spectrophotometric detection at 274 nm.The quantitative results obtained in the analysis of samples of plasma from patients undergoing treatment with theophylline were compared with those obtained for the same samples with the TDx fluorescence polarization immunoassay procedure (using the Abbot Therapeutic Drug Monitoring system), which is generally employed in hospitals and clinical laboratories. Statistical F-test and t-test for multiple samples were applied to the data obtained by the two methods. The results showed no significant difference between the two methods at the 95% confidence level. Keywords: High-performance liquid chromatography; theophylline; plasma; fluorescence polarization imm unoassa y Theophylline (1,3-dimethyI-lH-purine-2,6-dione) (see Fig.1) acts as a stimulator of the central nervous system and is employed in medicine, particularly in the treatment of acute and chronic bronchial asthma and apnoea in premature newborns and infants.1-4 The therapeutic concentration of theophylline in the serum of adults ranges between 5 and 20 ppm and at concentrations higher than 20 ppm theophylline can become toxic.' Furthermore, it has been shown that the rate of clearance varies considerably among different indi- viduals. For these reasons, during medical treatment with theophylline, in order to optimize the therapy doses1 it is necessary to monitor its concentration in plasma constantly. Gas chromatographic methods5.6 as well as reversed-phase high-performance liquid chromatography (HPLC)1,2,4,7-1" have been used for the determination of theophylline in biological specimens.A new HPLC method is proposed here, which makes use of reversed-phase ion-interaction chroma- tography, already employed in this laboratory in the devel- opment of analytical separation methods. 11-14 H (343 1 2 3 4 Fig. 1 Structural formulae for: 1, xanthine (3,7-dihydro-lH-purine- 2,6-dione), k, (40°C) = 1.19 X lo-'", k b (40°C) = 6.09 X 2, theobromine (3,7-dimethyI-lH-purine-2,6-dione) k, (18 "C) = 0.90 X 10-10, kb 1.30 x 10-14; 3, theophylline (1,3-dimethyl-lH-purine-2,6- dione), k, (25°C) = 1.69 x k b (25°C) = 1.90 X lo-14; and 4, caffeine (1,3,7-trimethyI-lH-purine-2,6-dione), k, (25 "C) = 1.0 X k b = 0.70 x By employing octylamine-rthophosphate as the interac- tion reagent, a reversed-phase CI8 column as the stationary phase and spectrophotometric detection (at 274 nm), a method for the separation of xanthine (3,7-dihydro-lH- purine-2,6-dione), theobromine (3,7-dimethyl-lH-purine- 2,6-dione) , theophylline (1,3-dimethyl-lH-purine-2,6-dione), and caffeine (l73,7-trirnethyl-1H-purine-2,6-dione) was devel- oped.The method was then employed in the quantitative analysis of theophylline in samples of human plasma obtained from patients undergoing therapy with theophylline. The results were compared, through statistical tests, with those obtained for the same samples with the so-called TDx method (using the Abbot Therapeutic Drug Monitoring system), which is a fluorescence polarization immunoassay procedure, commonly used in hospital and clinical laboratories .4 Experimental Apparatus Analyses were carried out with a Merck-Hitachi Lichrograph chromatograph Model L-6200, equipped with a two-channel Merck-Hitachi Model D-2500 Chromato-Integrator, inter- faced with a UV-UVNIS detector L-4200.For pH measure- ments, a Metrohm 654 pH meter equipped with a combined glass-calomel electrode was employed and for the evaluation of absorptivity values a Hitachi 150-20 spectrophotometer was used. Chemicals and Reagents Ultrapure water from a Millipore Milli-Q system was used for the preparation of solutions. Octylamine, orthophosphoric acid, xanthine, theophylline , theobromine and caffeine were Merck reagents and all other reagents were Fluka analytical- reagent grade chemicals.Chromatographic Conditions Merck Lichrospher 100 RP-18,5 pm, endcapped, packed in a 250 x 4 mm column was used as the stationary phase. Octylamine-rthophosphate 0.0050 mol dm-3 to be used as1072 ANALYST, JULY 1992, VOL. 117 the interaction reagent was prepared by dissolving a weighed amount of octylamine in ultrapure water and by adjusting the pH value of the solution to 6.4 f 0.4 through the addition of orthophosphoric acid. The chromatographic system was conditioned by passing the eluent through the column until a stable baseline signal was obtained (a minimum of 1 h was necessary). The repeatability of measurements, with regard to both retention time and integrated absorbance, was 2% for sequential measurements (with the same conditions of eluent prepara- tion and column conditioning) and the reproducibility (for different eluent preparations) was always within 6%.The eluent was freshly prepared each third day and the column was regenerated by passing ultrapure water (10 min, flow rate = 0.3 cm3 min-1) and then a water-methanol mixture 1 + 1 v/v (30 min, flow rate = 0.5 cm3 min-1) through it. Preparation of Standard Solutions and Samples The standard solutions of theophylline and theobromine were prepared at concentrations of 500.0 ppm in dark-glass flasks and stored in a refrigerator at 4 "C. The samples of plasma to be analysed were prepared by centrifugation at 8000g for 15 min, dilution with ultrapure water (1 + 9 v/v) and filtration through a 0.20 pm Anotop disposable syringe filter. Results and Discussion Previous workll-14 performed in this laboratory has been devoted to the development of separation methods by means of HPLC ion-interaction chromatography.The interaction reagent is prepared before the analysis from an acid and an amine (at a pH of 6.4 k 0.4, at which the acid is dissociated and the amine protonated). This represents the mobile phase, which, when flowing under isocratic conditions determines the so-called dynamic functionalization of the column, whose interaction properties can therefore be modified. This chro- matographic system is suitable for separation studies of species that are able to give rise, under the working pH 3.64 4 3 51.91 0 10 20 30 40 50 60 Time/mi n Fig. 2 Separation of a mixture of: A, xanthine (2.0 ppm); B, theobromine (1.0 pprn), C, theophylline (1.0 pprn); and D, caffeine (3.0 pprn).Stationary phase: Merck Lichrospher 100 RP-18 (250 x 4 mm), 5 pm, endcapped. Ion-interaction reagent: 0.0050 mol dm-3 octylamine-orthophosphate. Flow-rate: 2.0 cm3 min-1, 100 mm3 injected. Spectrophotometric detection: 274 nm. Recorder setting: 0.002 a.u.f.s. conditions, to ion-pairs with the anion or with the protonated amine of the interaction reagent. Previous studies1*712 have shown that analytes are retained and then released as ion-pairs. Up to now inorganic and organic acids have been investigated,llJ2 in addition to amines,12J3 amides and imidesl4 characterized by suitable pK values. This paper shows also that xanthinic structures such as xanthine, theophylline, theobromine, and caffeine can be retained.However, by considering their structure and pK, and pKb values reported in the literature15 (see Fig. l), it cannot be clearly established which is the functional group responsible for the retention at the working pH value of 6.4. Conditions were investigated for the development of a separation method capable of allowing the determination of theophylline, free from interference from xanthine, caffeine and theobromine, which can also be present, under certain conditions, in human plasma. The separation obtained for a mixture of xanthine (2.00 pprn), theobromine (1 .OO ppm) , theophylline (1 .OO ppm) and caffeine (3.00 ppm) is shown in Fig. 2. Spectrophotometric detection at 274 nm was chosen, at which wavelength all the species considered are characterized by high absorptivities, namely, E (theophylline) = (1.08 k 0.04) x 104, E (theobro- mine) = (0.99 k 0.05) x 104 and E (caffeine) = (0.96 _+ 0.05) x 104 dm3 mol-1 cm-1.The separation in Fig. 2 shows very good resolution, especially if one considers the very similar structures of the analytes and in particular the isomerism of theobromine and theophylline. The detection limit of theophylline, given as a signal-to-noise ratio of 3 : 1, was evaluated as 0.10 ppm. A standard graph was then constructed in the concentration range 0.10-3.00 pprn and good linearity [correlation coeffi- cient ( r ) = 0.99891 was obtained between peak area and standard concentration. The method was then applied to the analysis of theophylline in human plasma. The method proved to be suitable because of a very low matrix interference at this wavelength, as shown in Fig.3, which presents as an example a chromatogram recorded for a human plasma sample which is characterized by the absence of theophylline. Fig. 4, shows examples of 1.16 11.14 A 13.70 L 0 10 20 30 40 50 60 Ti me/min Fig. 3 Chromatogram recorded for a sample of human plasma (diluted 1 + 9 v/v) which does not contain theophylline. Recorder setting, 0.002 a.u.f.s. Other conditions as in Fig. 2. Peak A, theobromineANALYST, JULY 1992, VOL. 117 1073 , I- - I 0 10 20 30 lb) 2.86 2.84 4 I l l 0 10 20 30 Ti m e/m i n 0 10 20 30 Fig. 4 Chromatogram recorded for three samples of plasma (diluted 1 + 9 v/v), containing: ( a ) 2.05 ppm of theophylline; ( b ) 0.60 ppm of theophylline; and ( c ) 0.40 ppm of theophylline.Other conditions as in Fig. 2. Peaks: A, theobromine; and B, theophylline Table 1 Comparison of results (ppm) obtained for different samples of plasma using the TDx and proposed HPLC methods. Average data from triplicate measurements are reported Proposed Sample TDx method HPLC method A 4.02 3.91 B 6.87 6.23 C 18.60 18.80 D 20.39 18.20 E 20.50 20.70 F 20.89 19.42 chromatograms recorded for plasma obtained from three different patients who were undergoing theophylline-based therapy. These samples contain 2.05, 0.60 and 0.40 ppm of theophylline, respectively. In all of the chromatograms, besides the theophylline peak (peak B), two other major peaks can be observed; one of them (peak A) was identified as theobromine. Quantitative analysis of theophylline was performed at 274 nm, which corresponds to the wavelength of maximum sensitivity.The plot of peak area (y) versus concentration ( x ) , obtained by standard additions of theophylline to plasma samples showed a good linear trend ( y = -0.027 + 0.674~; r = 0.9841). The recovery of theophylline from spiked plasma was always greater than 98.8%. As mentioned, in hospital and clinical laboratories theophylline is generally determined through the TDx immu- noassay procedure, based on measurements of polarized light fluoresence.4 In order to compare the results obtained from the proposed HPLC method with those from the TDx method, the analysis was performed using the two methods for a series of six samples of plasma obtained from six different patients treated with theophylline.Each analysis was repeated in triplicate. Repeatability was within 2% and reproducibility within 6%. Reproducibility data are comparable with those obtained by the TDx method. The average data obtained for the two series of measurements are given in Table 1. A graph reporting the results obtained with the HPLC method against those from the TDx method gave a correlation coefficient of 0.9921. The F-test was then performed for the two sets of data, whose degrees of freedom are v1 and v2, respectively. The calculated Fparameter ( F = sx2/s,2) was equal to 1.06. As the tabulated value of the F-distribution at the 95% confidence level and for v1 = v2 = 5 degrees of freedom was 5.05 it can be concluded that there is no significant difference between the variances of the two methods.The data presented in Table 1 were also treated with the so-called ‘t-test for multiple samples’. This test is particularly suitable for the analysis of different samples of slightly varying composition, as is usually the case with biological fluids in which the analyte is present in narrow physiological concen- tration ranges. According to this intercalibration test, the t value is evaluated through the parameter Di, computed as the difference between the results obtained with the two methods for the same sample with regard to the sign, as where D is the mean of all the individual Di differences and sd = dc(Di - D)2/(N - 1) From the data in Table 1 a t value equal to 1.68 was calculated. The tabulated value of the t-distribution at the 95% confidence level and 5 degrees of freedom of 2.57 confirms that there is no significant difference between the two methods (at this confidence level).In conclusion, even if the results obtained with the HPLC method seem on the average slightly lower than those from the TDx method (see Table l ) , the F- and t-tests indicate that the methods are comparable at a significance level of 0.05. Reproducibility over time was also checked: results are reproducible for up to 2 d after which time the theophylline content appeared to be progressively lower. The methodology proposed here for the quantification of theophylline in plasma presents the advantage of being free from interference from xanthine, theobromine and caffeine and it does not require a pre-treatment step (apart from centrifugation and 0.20 pm filtration) or pre-column derivatization of the sample before injection.This work was supported by the Consiglio Nazionale delle Ricerche (CNR, Roma, Italia) and by the Minister0 della Ricerca Scientifica e Tecnologica (MURST, Roma, Italia). 1 2 3 4 5 6 7 8 9 10 11 12 References Matsumoto, K., Kikuchi, H . , Iri, H., Takahasi, H . , and Umino, M., J . Chromatogr., 1988, 425, 323. Wahllander, A . , Renner, A . , and Karlaganis, G . , J. Chroma- togr., 1985, 338, 369. Lee, B. L., Jacob, P., and Benowitz, N. L., J. Chromatogr., 1989, 494, 109. Blanchard, J . , Harvey, S . , and Morgan, W. J . , J. Chromatogr. Sci., 1990, 28, 303. Argoudelis, C. J., J. Chromatogr., 1984, 303, 256. Kasuya, Y., Furuta, T., and Shimoto, H . , J. Chromatogr., 1989, 494, 101. Wenk, M . , Eggs, B . , and Follath, F., J. Chromatogr., 1983,276, 341. Kinberger, B . , and Holmen, A . , J. Chromatogr., 1982, 229, 492. Hartley, R., Cookman, J . R . , and Smith, I. J., J . Chromatogr., 1984, 306, 19. Naline, E., Flouvat, B . , Advenier, C . , and Pays, M., J. Chromatogr., 1987,419, 177. Gennaro, M. C., and Bertolo, P. L., J. Chromatogr., 1989,472, 433. Gennaro, M. C., and Bertolo, P. L., J . Chromatogr., 1990,509, 147.1074 ANALYST, JULY 1992, VOL. 117 13 Gennaro, M. C., Bertolo, P. L., and Marengo E., J. Chroma- 16 The Merck Index, eds. Budavari, S . , and O’Neil, M. J . , Merck togr., 1990,518, 149. Rahway, NJ, 1989. 14 Gennaro, M. C., and Abrigo, C . , J. Chromatogr. Sci., 1991,29, 410. Paper I/05316G 15 Smith, R. M., and Martell, A. E., Critical Stability Constants, Received October 21, I991 Accepted February I I , I992 Plenum Press, New York, London. 1976.
ISSN:0003-2654
DOI:10.1039/AN9921701071
出版商:RSC
年代:1992
数据来源: RSC
|
5. |
Determination of thallium in cement dust and sediment samples by differential-pulse anodic stripping voltammetry: a chemometric approach to linear calibration |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1075-1084
Mahmoud A. Allus,
Preview
|
PDF (1340KB)
|
|
摘要:
ANALYST, JULY 1992. VOL. 117 Determination of Thallium in Cement Dust and Sediment Samples Differential-pulse Anodic Stripping Voltammetry: A Chemometric Approach to Linear Calibration Mahmoud A. Allus and Richard G. Brereton" School of Chemistry, University of Bristol, Cantock's Close, Bristol BS8 1 TS, UK 1075 bY Differential-pulse anodic stripping voltammetry (DPASV) was used to determine TI in cement kiln dust and sediment samples. The separation of TI from the sample matrix was achieved by the extraction of TP+ into diethyl ether from an HBr-Br2 medium. Initially, a mixture of HN03 and HF was used to digest the cement dust samples to cause volatilization of silica as SiF4. The Tl3+ was then reduced to TI+ using hydrazine sulfate prior to determination by DPASV. A calibration graph based on a deterministic relationship between the measured response y and the analyte concentration x (so that y = Po + Px + r ) is also discussed using classical (or reverse) regression.The precision of parameter estimates of the model (bo and b: p = bo + bx) can be achieved by display of the diagonal elements of the hat matrix. Methods for computing the standard uncertainty in the predicted response s$. and in the concentration s$ are also provided. Thallium concentrations of 700 k 3.07,395 k 2.91 and 850 k 2.59 ng 9-1 were found in the cement dust and sediment samples, respectively. Keywords: Differential-pulse anodic stripping voltammetry; chemometrics; linear calibration The quantitative determination of T1 in cement kiln dust (collected from Libya) and sediment samples is described in this paper using differential-pulse anodic stripping voltam- metry (DPASV).Thallium is concentrated as a trace element in these samples. It is a very toxic element, and local pollution problems with T1 emitted from cement producing installations have been reported.'-3 Cement production in Libya is one of the major heavy industrial projects that the country has achieved recently. The cement is produced from more than six factories in agricultural areas, namely Khomes, Zaletin, Benghazi, Sauk El-Khameasl Tarhona and Derna. The emission of trace amounts of toxic metals produced as a by-product from the roasting of the raw materials used in cement manufacture may affect the quality of the agriculture in the vicinity of these cement producing installations and thus increase environmental pollution that has already become a serious problem.4~5 In Germany the TI content of the cement dust emitted by one factory constituted an environmental hazard.The source of T1 was shown to be the pyrite cinders and slag which had been added to the raw mixture for the production of the sulfate resistant Portland cement. Pyrite cinders are the iron oxide residue which remains after roasting sulfur-containing ores for the production of sulfuric acid. When added to a cement raw mix and burned in a kiln, it combines with other materials to form a calcium aluminoferrite phase with the approximate molecular formula Ca4A1Fe. At the same time, TI is volatilized and a portion of the recirculating load in the kiln system is emitted to the atmosphere in the stack gases.1-3 The experiments described in this paper involve the determination of T1 in cement kiln dust samples and the quantification of T1 in sediment samples. The separation of Tl from the sample matrix was achieved by the extraction of Tl3+ into diethyl ether from an HBr-Brz medium. Initially, a mixture of HN03 and HF was used to digest the cement dust samples to cause volatilization of silica as SiF4. No loss of T1 by volatilization has been found by other workers using this dissolution procedure.616 The TP+ was then reduced to Tl+ using hydrazine sulfate prior to determination by DPASV. A calibration graph based on a deterministic relationship between the measured response y and the analyte concentra- * To whom correspondence should be addressed.tion x is also discussed:17-33 the precision of parameter estimates of the model can be achieved by the application of experimental design and display of the diagonal elements of the hat matrix h,. Methods for computing the standard uncertainty in the predicted response s;,, and in the concentra- tion sz are also provided. Experimental All chemicals used were of AnalaR grade (where possible). Concentrated Aristar HN03 and HF were used, whereas AnalaR H2S04 and HBr were employed unless specified otherwise. All glassware was soaked in H2S04-H20 (40 + 60) before use. Dilutions were performed using triply distilled water. Reagents and Sample Preparation The following reagents were prepared as de~cribed.6.~ Stock standard TI solution (lo00 pg cm-3).Prepared by dissolving 1.303 g of TlN03 in water, adding a suitable volume of HN03 to bring the pH to about 1 and diluting the mixture to 1 dm3 with water. TI calibration solution (10 pg cm-3). Prepared by diluting 1 cm3 of 1000 pg cm-3 TI stock solution to 100 cm3 in a calibrated flask. Base electrolyte [O. 1 mol dm-3 ammonium tartrate44 mol dm-3 ethylenediaminetetraacetic acid (EDTA) J. Prepared by dissolving 18.4 g of ammonium tartrate and 14.9 g of EDTA (disodium dihydrogen salt) in water and diluting to 1 dm3. Reducing solution. Prepared by cautiously adding 15 cm3 of H2S04 to 15 cm3 of water and then dissolving 10 g of anhydrous NaZSO4 in the mixture. The resulting solution was then saturated with hydrazine sulfate. HBr-Br2 mixture.Prepared by mixing 50 cm3 of Br, and 450 cm3 of HBr (density 1.46-1.49 g cm-3). The solution was stored in a brown glass reagent bottle. HBr (0.1 mol dm-3). Prepared by diluting 115 cm3 of HBr (density 1.46-1.49 g cm-3) to 1 dm3. The above reagents were used in the preparation of samples for analysis by DPASV.1076 ANALYST, JULY 1992, VOL. 117 The cement dust sample (10 g) was transferred into a 150 cm3 Teflon beaker and placed in an electric oven at 50°C for 48 h. A 1.0 g amount was weighed accurately from the dried cement dust sample in a 150 cm3 poly(tetrafluoroethy1ene) container. Then, HF (5 cm3) and HN03 (5 cm3) were added to the sample and the mixture was evaporated to dryness. A 10 cm3 volume of the HBr-Br2 mixture was added and the solution warmed gently until most of the bromine had been expelled.The beaker was covered after the addition of 25 cm3 of hot water and the whole digested on a hot-plate until the residue had dissolved, then cooled to room temperature. The solution was transferred quantitatively into a 100 cm3 separat- ing funnel (marked at 50 cm3) and diluted to the mark with water, after which 15 cm3 of diethyl ether were added and the whole system was shaken vigorously for 2-3 min. The aqueous layer was run off into a similar separating funnel with the addition of a further 15 cm3 of diethyl ether and the process was repeated. All the diethyl ether layers were combined and a further 15 cm3 of 1 rnol dm-3 HBr added. After shaking for 1-2 min, the acid layer was discarded and the combined organic extracts were transferred into a 100 cm3 conical glass flask and evaporated to dryness.Then, HN03 (2 cm3) and H2SO4 (2 cm3) were added to the residue and the mixture was heated until strong white fumes appeared, whereupon 2 cm3 of reducing solution were added and the contents evaporated to dryness. The residue was then dissolved in 15 cm3 of hot base electrolyte, cooled to room temperature and diluted with 2.5 rnol dm-3 soldium hydroxide until the final pH lay in the range 5 k 0.5.6 The solution was transferred into a 25 cm3 calibrated flask and diluted to the mark with water. The same procedure (described above) was used to prepare the reagent blank solutions but without any sample. Many cement dust samples were weighed out and the preparation procedure described above was followed both for these samples and for the sediment samples.For the cement dust samples, the final dilution was to 25 cm3, whereas for the sediment samples it was to 50 cm3. A 10 cm3 volume of the prepared sample solution was transferred into the polarographic cell containing a magnetic stirrer and purged with purified nitrogen for 10 min. The determination of the TI content was carried out using the experimental conditions described in Table 1. Differential-pulse Anodic Stripping Voltammetry The determination of TI was carried out using anodic stripping voltammetry involving two discrete steps.34-37 Firstly, after preparation of a solution of a T1 containing sample in the base electrolyte, TI is preconcentrated by deposition on a working electrode (which, in this instance, is a hanging mercury drop electrode) by maintaining its potential more negative than the reduction potential of TI.Secondly, the Tl is transferred back into the solution (which, in this instance, is the base electrolyte solution at pH 5 k O S ) , i.e., TI is stripped by applying a steadily increasing positive potential across the electrodes and the current generated by oxidation of the deposited TI produces a peak, the height of which is measured. The current is proportional to the concentration of TI in the sample solution and the potential at which the peak occurs is known as the element peak potential. The working electrode used was a mercury drop extruded automatically from a glass capil- From a study of the peak potentials of the elements known to be commonly present in the cement dust sample, it is clear that Cd2+, Pb2+ and Fe3+ are the most likely to interfere in the determination of T1+ by DPASV.However, complexing agents, especially chelating agents such as EDTA, can be used to separate the reduction potential of elements during the preliminary electrolysis stage or to alter the potential at which their peaks occur during the stripping process. Experiments using this method showed that Tl+ could be determined in the i a y . * 5 3 Table 1 Instrumental conditions used during sample analysis. 15~16,34 The base electrolyte was 0.1 mol dm-3 ammonium tartrate4.4 mol dm-3 EDTA (disodium dihydrogen salt) at pH 5 f 0.5 Source* Working electrode Reference electrode Model 303- Auxiliary electrode Conditioning potential Model 315- Initial potential Final potential Purge time Conditioning time Equilibration time Deposition time Model 174- Scan rate Scan direction Modulation amplitude Current range Drop time Display direction Low pass filter Operation mode Initial potential Potential scan rate Potential range Model 7040- Condition Hanging mercury drop (medium size) Ag-AgC1 (saturated with 3 mol dm-3 Platinum wire KCl) +0.20 V versus Ag-AgC1-3 mol dm-3 -0.20 V versus Ag-AgC1-3 mol dm-3 -0.20 V versus Ag-AgC1-3 mol dm-3 KCl KCI KCI 10.00 min 0 min 30.00 s 300.00 s 5.00 mV s-1 (+) 25.00 mV 10.00 pA 0.50 s (-) (Off) Differential pulse 0.0 v 5.00 mV s-1 3.00 V x-y recorder? * Model 303 (static mercury drop electrode), Model 315 (auto- mated electroanalytical controller), Model 174 (polarographic ana- lyser) and Model 305 (stirrer) were all from Princeton Applied Research.? The x-y recorder (Model 7040) was from Hewlett-Packard. presence of both Cd2+ and Pb*+ in a tartrate-EDTA medium. However, the presence of Fe3+ in the sample to be analysed had a suppressive effect on the recovery of TI+, which indicated that the separation of TI+ from the matrix would be necessary.614 In this work the determination of T1 was performed in a tartrate-EDTA base electrolyte after extraction from an HBr-Br2 medium into diethyl ether. However, the extraction from acidic medium could be carried out by using other solvents instead of diethyl ether; for instance, diisopropyl ether has been shown to be more selective for TP+, but, for safety reasons, diethyl ether was used in the extraction procedure.6,7,'5,'6.34 The DPASV analyses were carried out by transferring 10 cm3 aliquots of the prepared sample solution into the polarographic cell34 under the instrumental conditions given in Table 1.Results and Discussion Typical voltammograms for a series of reagent blanks and TI standard solutions are shown in Fig. 1. Several reagent blank solutions were prepared in the same way as for the TI standard solutions at pH 5 k 0.5 and examined using the same practical and instrumental conditions as for the TI standard solutions. It is evident from the traces shown in Fig. 1 that the response of the TI standard solution (20 ng cm-3) produces a peak height significantly greater than that of the reagent blank solution.A number of runs were made of these reagent blank samples to ascertain that the concentration of TI in the reagents used to prepare the samples to be analysed for their TI content was negligible. It is clear that the TI concentration in the base electrolyte is very low and probably not at a significant levelANALYST, JULY 1992, VOL. 117 1077 A 1- ll ' I -1.20 -0.80 -0.40 -1.00 -0.60 -0.20 PotentialN versus Ag-AgC1-3 mol dm-3 KCI Fig. 1 Voltammograms of (a) a series of reagent blank solutions and (b) TI standard solutions added to the base electrolyte. The base electrolyte is tartrate-EDTA at pH 5 k 0.5 [the symbols A, B, C, D, E and F are the observed responses for the reagent blank plus 20,40,60, 80 and 100 mm3 of T1 standard solution (each 1 mm3 = 1 ng ~ m - ~ ) , respectively] when compared with those responses obtained from the standard TI solutions (2G100 ng cm-3) under the same experimental conditions.The nature of the standard additions procedure was considered in order to validate the calibration graph for the determination of T1 by DPASV. This is because calibration in general is a fundamental step for accurate and more informa- tive quantitative trace analysis.38239 The calibration graph of T1 in this instance is based on a deterministic relationship between an independent factor x (TI concentration in the base electrolyte) and a dependent response y (observed peak height). The relationship between x and y can be described by the linear model as: y = po + px + r i) = bo + bx (1) (2) where bo and b are the best fit linear regression coefficients for the calibration model, and p are the true linear calibration coefficients and r is the error; eqn.(2) is the least-squares estimated regression for eqn. (1). The use of matrix least- squares solutions for such models has been discussed in many text-books.4w2 The precision of estimating eqn. (2) depends on the measurement process used to obtain the observed response y and the nature of the experimental design. Hence, our experiments aimed to achieve such a goal (precise estimation of bo and b ) . In this paper only approaches for classical (or reverse) calibration are considered, in which y is regressed on x , i.e., it is assumed that all the errors are in the response and none in Table 2 Results for Method A: x = standard Tl solution added to the base electrolyte, h, is as discussed in the text, y and 9 = the measured and predicted responses (peak height in cm), respectively, r = residual and s = standard error n X hrl Y 9 r S 1 20 0.7 4.1 3.80 0.30 0.517 2 40 0.3 7.8 8.15 -0.35 0.452 3 60 0.3 12.3 12.50 -0.20 0.452 4 80 0.7 17.1 16.85 0.25 0.517 Table 3 Results for Method B: x, h,, y, 9 , r and s are as defined in Table 2 n 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 X h n Y 20 0.14 4.1 20 0.14 4.6 20 0.14 3.6 20 0.14 4.0 20 0.14 4.2 40 0.06 7.0 40 0.06 8.6 40 0.06 7.8 40 0.06 8.0 40 0.06 7.6 60 0.06 12.5 60 0.06 13.0 60 0.06 11.0 60 0.06 12.7 60 0.06 12.3 80 0.14 17.1 80 0.14 17.5 80 0.14 16.7 80 0.14 17.0 80 0.14 17.2 9 3.80 3.80 3.80 3.80 3.80 8.15 8.15 8.15 8.15 8.15 12.50 12.50 12.50 12.50 12.50 16.85 16.85 16.85 16.85 16.85 r 0.30 0.80 -0.20 0.20 0.40 -1.15 0.45 -0.35 -0.15 -0.55 0.00 0.50 - 1 S O 0.20 -0.20 0.25 0.65 0.15 0.35 -0.15 S 0.625 0.625 0.625 0.625 0.625 0.603 0.603 0.603 0.603 0.603 0.603 0.603 0.603 0.603 0.603 0.625 0.625 0.625 0.625 0.625 the independent variable.There are a number of other approaches to linear calibration, including inverse (or for- ward) calibration, and a large amount of statistical literature comparing the merits of these approaches. It is, however, usual for most user-friendly microprocessor software (e.g., for graphics and spreadsheets) to be based on classical calibra- tion. It is beyond the scope of this paper to contrast the merits of these and related approaches, but the interested reader is referred to the text by Martens and Nzs,43 and references cited therein, for an up-to-date chemometric review of this large area.Four aliquots of the prepared TI standard solution (10 pg cm-3) were added to 10 cm3 of base electrolyte in the polarographic cell to give TI concentrations of 20, 40, 60 and 80 ng (3111-3 (at pH 5 k 0.5). The observed responses of these aliquots were measured (as the peak height in cm). The process was repeated five times. Hence 20 measurements were obtained for these factor levels (five at each level). In matrix terms, the design matrix D has 20 rows (n = 20) and one column (there is only one independent factor), while the model matrix X has 20 rows and two columns ( n = 20 andp = 2, where n = total number of experiments performed and p = number of parameters in the model). The measured peak heights were recorded resulting in 20 responses (five at each TI concentration; hence the observed response row vector Y has 20 rows and one column).A matrix least-squares regression was performed using the SAS (Statistical Analysis System) software on the resulting measurements. 17241 744 In order to study the influence of replication on the resulting analysis, the experiments were analysed using the following two methods. Method A . The regression was performed on only four points (the average at each concentration so that D and Y are 4 x 1 matrices; X is a 4 x 2 matrix). Method B. All 20 experimental points were included in the regression analysis so that D, Y and X are 20 x 1 , 20 x 1 and 20 x 2 matrices, respectively.1078 ANALYST, JULY 1992, VOL.117 The results of both methods of analysis are given in Tables 2 and 3. We definef= levels of T1 concentration: for method A, n = 4 , p = 2andf= 4, andformethodB, n = 20,p = 2andf= 4. The fitted model for both methods is the same in this instance and of the form: 9 = -0.55 + 0.2175~ (3) with variance s = 0.397 and 0.586 for methods A and B, respectively. The variance s (in this instance) is the square root of the mean square error for the residual, being smaller in value for method A, as some of the variability is reduced by averaging out the replicates: (4) where SS, is the sum of squares of residuals, and n - p is the number of degrees of freedom associated with SS,. The variance of the estimated parameters 60 and b is not the same: i.e., sbO = 0.4861 and sb = 0.0087 for method A, and for sbg = 0.321 and sb = 0.0059 for method B. The diagonal elements h, (often called leverage) of hat matrix H and h (where h is a parameter related to h,), defined by44-51 H = X(XTX)-lXT, h, = xn(XTX)-lxT and h = x(XTX)xT (some workers add a factor of lln to the definition of leverage, but this makes no difference to the over-all properties of this function: the definition used in this work is preferred as it has the property that h varies between 0 and 1 for experimental points) can be used to study the influence of experimental points on the regression parameters.The h, values for both designs are given in Tables 2 and 3. Graphically, h, could be displayed as a function of x as illustrated in Fig.2. The higher the value of h, the less well the experimental point can be estimated. However, in spite of the smaller value of h and estimated parameter variances for method B (n = 20), the standard variance for method B is still greater than that of method A. However, the 95% variance for method B is smaller than that of method A as compared graphically in Fig. 3(a) and (b). This is because although method A shows a low variance value for the fitted model, the degrees of freedom associated with the residual are very small indeed. Hence, as a result, the critical Fvalue will be higher [where Fis the Fisher variance with one degree of freedom in the numerator and ( n - p ) degrees of freedom in the denominator because sj.,, is based on s?, as j , = X,B when used for obtaining the desired level of confidence]. For example, at the 95% confidence level, F(l, - or F ( l , 2 ) = 18.51 when method A was used compared with F ( l , 18) = 4.41 for method B.The confidence bands for both results are also displayed graphically in Fig. 1 .o 0.8 & 0.6 2 > 3 0.4 0.2 1 1 I I I 0 20 40 60 80 100 TI concentrationhg cm-3 Fig. 2 Value of h as a function of the factor level x (T1 concentration in the base electrolyte) for the model y = Po + Px + r , Methods A and B 4(a) and (b). These are the 95% upper and lower confidence intervals for the following.40>44 1. Mean predicted response values: accounts for the variations caused by estimating the parameter coefficients only (bo and b): En = E n VP(1, n - p ) x s: x h,I 2.Individual predicted response values: these add in the variability of the error term, and hence they are wider than the former type: En = En !E VRl, n - p ) x s: x (1 + hn)l 3. Working-Hotelling confidence bands: used for predicting the entire response values: En = En k V[W x s; X hn] where W = p x F(p, - p ) and F values refer to 95% confidence intervals (of course a similar exercise can be performed using any desired level of confidence). The analysis of variance (ANOVA) for the results from both designs reveals important information. The ANOVA of both results is shown in Tables 4 and 5 from which it is clear that the regression results using both methods show a high significance for the goodness-of-fit (which is also called the significance of the regression): this means that the factor x (or T1 concentration in the base electrolyte) does have an effect on the predicted response 9.On the other hand, it is not possible to test the lack-of-fit for the results from method A: this is because there are no replicated experiments in the analysis. Unlike the results for method A, those for method B do have replicated experiments and the lack-of-fit in this instance can be tested by ANOVA as shown in Table 5. The ANOVA is not significant for the latter design results; hence the model fits the data we11.40>44,52 Further statistical analysis from both designs shows that r2 (the coefficient of determination) is 0.997 and 0.987 for methods A and B, respectively, indicating that the factor x, as I I a, CO 0.4 1 1 I 1 I .- 5 2.4 > 2.0 1.6 1.2 B 0 20 40 60 80 100 Fig.3 ( a ) Standard uncertainty s as a function of the factor level x (TI concentration in the base electrolyte) for the model y = Po + Px + r for Methods A and B. (6) 95% confidence level for s as a function of the factor level x for the model y = Po + fix + r for Methods A and B TI concentrationhg ~ r n - ~ANALYST, JULY 1992, VOL. 117 1079 0 20 40 60 80 100 TI concentrationhg ~ r n - ~ Fig. 4 (a) Confidence bands (95% level) for predicting individual value of response (outer bands), entire value of response (middle bands) and mean value of response (inner bands) (Method A). The least-squares regression line is in the centre of the seven bands. Note that there is a crossover between the inner and middle bands at high and low values of concentration.(b) Confidence bands (95% level) for predicting individual value of response (outer bands), entire value of response (middle bands) and mean value of response (inner bands) (Method B). The least-squares regression line is in the centre of the seven bands. Note that there is a crossover between the inner and middle bands at high and low values of concentration Table 4 Analysis of variance (ANOVA) for the data results presented in Table 2; * indicates a significance level of 99.9%. In these data results there is no replication in the design, hence it is not possible to calculate the analytical error; therefore, the lack-of-fit could not be tested Mean Sum of Degrees of Mean square Source squares freedom square ratio Regression on- Mean Factor effect Total regression Residual Pure error Lack-of- fit Corrected Total 426.4225 1 426.4225 2707.44* 94.6125 1 94.6125 600.71* 521.0350 2 260.5 175 1654.08* 0.3150 2 0.1575 - 0.00 0 0.00 - 0.3150 2 0.1575 - - - 94.9275 3 521.350 4 - - it appears in the fitted model: j = bo + bx, explains the data well.The coefficient of determination for method A is better than that for method B; however, it is important to realize that the coefficient of determination gives no indication of whether the lack of perfect prediction is caused by an inadequate model or by analytical error. The r2 term is not a good measure of the effectiveness of the factor x as it appears in the model, primarily because it does not take into account the degrees of freedom.From the statistical analysis of the results from both methods, it is evident that method B is more informative than method A. Hence a good method for analysis is one that allows for the analytical error to be estimated and utilized in a statistical test for the lack-of-fit, in addition to the improved Table 5 Analysis of variance (ANOVA) for the data results presented in Table 3; * indicates a significance level of 99.9%. The lack-of-fit is not highly significant although there are only 2 degrees of freedom and it is tested against pure analytical error mean square, which is very small (s& = 0.2875) because there are 16 degrees of freedom Mean Sum of Degrees of Mean square Source squares freedom square ratio Regression on- Mean Factor effect Total regression Residual Pure error Lack-of-fit Corrected Total 2132.1125 473.0625 2605.1750 6.1750 4.600 1.5750 479.2375 26211.3500 1 2132.1125 6215.06* 1 473.0625 1387.97* 2 1302.5875 3797.02* 18 0.3431 - 16 0.2875 - 19 20 2 0.7875 2.74 - - - - precision for estimating the model parameters (bo and b) and confidence bands.Having constructed the calibration graph for standard Tl solutions with decreased variance in the parameter estimates (bo and b) and narrower confidence bands, the next step is to utilize this graph for chemical analysis and in the estimation of confidence intervals for the T1 concentration in the samples to be analysed. The detection limit can be calculated from the confidence intervals as the amount of x associated with the predicted response j .The fitted model, given by eqn. (2), which describes the relationship between x and y , can be rewritten as: (j, - bo) x=- b ( 5 ) The linear regression analysis used to solve eqn. ( 5 ) assumes that there are errors associated with the dependent variable y only and that the errors associated with the independent factor x are negligible. However, there are many instances where the regression analysis is performed with errors associated with both the dependent and independent variables: for two variables these methods are ,equivalent to principal com- ponents analysis rather than conventional linear ~ a l i b r a t i o n . ~ ~ In the real world, experiments fit into this latter class as the standards are often prepared by weighing, dilution, and so on; as a result, their concentrations are rarely known exactly.31J9,44 In the case reported here, errors in the independent variable x are ignored and attention is paid only to the errors associated with the dependent variable y and the fitted model estimates y and also x with the same variance.A confidence interval for the predicted response j is found at any given x value. Similarly, a confidence interval for an estimated x concentration is found for a given j value [as shown in Figs. 4(a) and (b)]. The intersection of the regression bands and the x-axis (concentration) defines the range of response obtained for an analyte-free sample (x = 0) at a specific confidence level [sometimes responses less than the point of intersection result in a negative x value (lower interval). In this instance, the detection limit can be con- sidered as the upper value of the intervals and the negative value is rejected].Hence, estimation of the detection limit from the fitted linear calibration graph for any predicted response j or concentration x values can be achieved.26~39 It is also possible to utilize the linear calibration graph to estimate the amount of the analyte in the sample using the instrumental response obtained for the sample by interpolation and the associated confidence intervals. The standard additions method of analysis is a method where the standards are added to the sample matrix and the observed response y of the analyte plus the standard is monitored as a function of the added concentration of the standard. A graphical representation of the standard additions1080 ANALYST, JULY 1992, VOL.117 3 6 -6 - 3 0 3 Added concentrations 6 Fig 5 (a) Graphical representation of the standard additions method: each triangle (A) point represents a single response measurement, c is the concentration of the sample without addition of standards with response readings, cl, c2 and c3 are the added concentrations with c + cl, c + c1 + c2 and c + c1 + c2 + c3 response measurements, respectively. The solid line represents the regression line described by: 9 = bo + bx. The vertical broken line represents the centre of experimentation which is estimated with most confidence. (b) Estimation procedure by standard additions: the dotted line is the estimated regression line and the two enveloping curves are the 95% confidence bands for any given x value in the prediction equation shown in Fig.5(a). UL is the upper and lower confidence interval for c method and its estimation procedure is presented in Fig. 5.27 Although the standard additions method of analysis is a tool well known to electrochemists, the estimation procedure in this method, as shown in Fig. 5(a), has at least one disadvantage, viz., that the confidence interval (UL) for the unknown x concentration at the point C is large.27 However, by carefully considering the method for data analysis and calculating h,, the disadvantage of such a problem can be overcome. The cement dust samples analysed by DPASV were: (1) a cement dust sample from Zaletin, Libya (sample 1); and (2) a cement dust sample from Lepda, Libya (sample 2).Typical voltammograms for these samples are shown in Figs. 6 and 7. For the determination of T1 in these samples, 13 experimen- tal points with five replicated experiments at x = 0 and two replicated experiments at x = 20, 40, 60 and 80 were used. Using a linear model as above [eqns. (1) and (2)], the number of experiments n = 13, the number of factor levelsf = 5 and the number of parameters p = 2. This provides 11 degrees of freedom for the estimation of the residuals error with 8 and 3 degrees of freedom for the estimation of the analytical error and the lack-of-fit, respectively. The results of these two experiments are presented in Tables 6-9. The least-squares estimate regression equations are given as: Sample (1): j j = 5.237 + 0.1876~ Sample (2): 9 = 3.290 + 0.208~ (6) (7) with variance = 0.505 and 0.539, respectively.The standard error for parameter estimates is sbo = 0.201 and sb = 0.00467 for sample 1 and sbo = 0.214 and sb = 0.00499 for sample 2. I I I I I - 1.20 - 1 .oo- 0.80 - 0.60- 0.40- 0.20 PotentialN versus Ag-AgCI-3 mol dm-3 KCI Fig. 6 Current-voltage curves (voltammograms) for cement kiln dust samples from Zaletin (Libya) [the symbols are for (a): A = blank response, B = sample signal and C, D, E, F and G are sample plus 20, 40, 60, 80 and 100 mm3 of the added standard Tl concentration (10 pg cm-3 so that each mm3 = 1 ng cm-3); and for (b): A = blank response, B, C, D and E = sample responses] The h, value at x = 0 with this type of design is reduced to h, = 0.158 compared with h, = 0.284 at x, = 80.Both x, = 0 and x, = 80 points are outliers and will have the same h, value if the x, = 0 point is not replicated and treated in the same way as the x, = 80 point. This will reduce the variance for 9 and the unknown concentration will be estimated with improved precision. The variance (s) for 9 (and the estimated TI concentration in the base electrolyte as a result of sample solution) is a function of both s, and h,, as the standard variance s (or uncertainty) is defined as: s = vs; x (1 + h,), where: s2= ( - nyp) and h, = x,(XTX)-lx;T (8) The 95% confidence bands are displayed graphically in Fig 8. The standard variance for each analysed sample is smaller at x, = 0 due to the replication of this point, as demonstrated in Fig. 8(c).The ANOVA results for the cement dust samples 1 and 2 are given in Tables 7 and 9. The coefficients of determination are 0.993 and 0.994, respectively. The significance of the coefficient of determination is contained in F@ - 1, , - = F(1, 11) (ie., mean square for residuals divided by the meanANALYST, JULY 1992, VOL. 117 1081 I -1.20 -0.80 -0.40 -1.00 -0.60 -0.20 PotentialN versus Ag-AgCI-3 mol dm-3 KCI Fig. 7 Current-voltage curves (voltammograms) for cement kiln dust samples from Lepda (at Khomes, Libya): [the symbols are for (a): A = blank response, B = sample signal and (C, D, E, F and G are sample plus 20, 40, 60, 80 and 100 mm3 of the added standard TI concentration (10 pg cm-3 so that each 1 mm3 = 1 ng cm-3); and for (b): A = blank response, B, C, D and E = sample responses] Table 6 Results from the analysed cement dust sample (1) from Zaletin, Libya: x, h,, y, 9 , r and s are as defined in Table 2 n 1 2 3 4 5 6 7 8 9 10 11 12 13 X 0 0 0 0 0 20 20 40 40 60 60 80 80 h n 0.158 0.158 0.158 0.158 0.158 0.087 0.087 0.084 0.084 0.150 0.150 0.284 0,284 Y 5.5 4.9 5.7 5.2 5.8 8.7 8.9 12.5 11.5 17.0 16.4 20.5 20.5 E 5.237 5.237 5.237 5.237 5.237 8.988 8.988 12.739 12.739 16.490 16.490 20.241 20.241 r 0.263 0.463 0.563 -0.336 -0.036 -0.287 -0.087 -0.238 - 1.239 0.510 -0.090 0.258 0.258 S 0.542 0.542 0.542 0.542 0.542 0.526 0.526 0.525 0.525 0.541 0.541 0.571 0.571 square for factors effect) and in both experiments this value is 0.001 and 0.1, respectively, which is not significant.This means that the factor x explains the response well.In this case the sum of the squares due to the residuals is very small and that of the factor is relatively large. In contrast, when the factor has very little effect on the response, then the sum of squares due to the factor effect would be very small and, therefore, SS, would be large. The lack-of-fit is not significant in both cases. The method used for the determination of trace amounts of TI in cement dust samples was also applied to the determina- tion of TI in sediment samples (collected from the Haw-Wood area, Avonmouth, Bristol, UK). These samples were pre- pared in the same way as described above. The instrumental Table 7 Analysis of variance (ANOVA) for the analysed cement dust sample collected from Zaletin, Libya; * indicates a significance level of 99.9%.Both the mean response (intercept) and the factor effect (TI concentration) are highly significant as tested against the residual error (total error). The lack-of-fit is not highly significant as tested against the pure error mean square. In the original experiment: n = 13,p = 2, f = 5 Mean Sum of Degrees of Mean square Source squares freedom square ratio Regression on- Mean Factor effect Total regression Residual Pure error Lack-of-fit Corrected Total 1575.201 411.289 1986.490 2.801 1.248 1.553 414.089 1989.290 1 1 2 11 8 3 12 13 1575.201 6186.66* 411.289 1615.35* 993.245 3901 .OO* 0.255 - 0.156 - 0.518 3.32 Table 8 Results from the analysed cement dust sample (2) from Lepda, Libya: x, h,, y, 9 , r and s are as defined in Table 2 n 1 2 3 4 5 6 7 8 9 10 11 12 13 X 0 0 0 0 0 20 20 40 40 60 60 80 80 h n 0.158 0.158 0.158 0.158 0.158 0.087 0.087 0.084 0.084 0.150 0.150 0.284 0.284 Y 9 3.6 3.290 2.8 3.290 3.0 3.290 3.5 3.290 4.0 3.290 8.0 7.456 7.0 7.456 11.0 11.623 11.6 11.623 15.8 15.790 15.0 15.790 20.0 19.957 20.8 19.957 r 0.311 -0.490 -0.290 0.211 0.711 0.544 -0.456 -0.623 -0.023 0.010 -0.079 0.043 0.843 S 0.580 0.580 0.580 0.580 0.580 0.562 0.562 0.562 0.562 0.578 0.578 0.611 0.611 Table 9 Analysis of variance (ANOVA) for the analysed cement dust sample collected from Lepda (at Khomes), Libya; * indicates a significance level of 99.9%.The lack-of-fit is not highly significant at F(3, 8). The remainder of the regression is highly significant (>99.9%). Also, the original experiments are the same as in Table 10 Mean Sum of Degrees of Mean square Source squares freedom square ratio Regression on- Mean Factor effect Total regression Residual Pure error Lack-of-fit Corrected Total 1223.170 507.521 1730.691 3.199 2.284 0.951 5 10.720 1733.890 1 1223.170 4206.45* 1 4507.521 1745.35* 2 856.346 2975.90* 11 0.291 - 8 0.281 - 3 0.317 1.13 - - 12 13 - - conditions were altered slightly during the analysis of the sediment samples.The results from the analysis of the sediments together with ANOVA results are shown in Tables 10 and 11. The leverage points h, for the design used in the analysis of the sediment samples is the same as that used for the analysis of the cement dusts; hence, the h, values are the same as shown in their corresponding results. However, the equation for h, which can be calculated for the inverse matrix (XTX)--I for this design, is given as: h = 0.1579 - 0.00526~ + O.ooOo86~2 (9) This equation can be used to determine the standard error for the predicted response 9 and to determine the unknown T1 concentration in the sample solution.1082 ANALYST, JULY 1992, VOL.117 Table 10 Results from the analysed sediment samples: x , h,, y, j , rand s are as defined in Table 2 Y m a -40 -20 0 20 40 60 80 100 TI concentrationhg cm-3 0.62 0.58 L >" 0.54 I. ! I 0.50 1 I I I 0 20 40 60 80 TI concentrationhg cm-3 Fig. 8 (a) Confidence bands (95% level) for the analysed cement dust sample 1 for predicting individual value of response (outer bands), entire value of response (middle bands) and mean value of response (inner bands).The least-squares regression line and the measured response (A points) are in the centre. (b) Confidence bands 95% level) for the analysed cement dust sample 2 for predicting individual value of response (outer bands), entire value of response (middle bands) and mean value of response (inner bands). The solid least-squares regression line and the measured response (A points) are in the centre. (c) Standard uncertainty (variance) for the analysed cement dust samples (1 and 2) as a function of T1 concentration levels x for the linear model: y = Po + Px + r The concentration of T1 in the solid sample (cement kiln x x v x z dust and sediment samples) may be determined as: c = (7) (10) where c is the T1 concentration in the solid sample (ng g-I), x is the T1 concentration (ng cm-3) in the actual sample solution, V is the final volume (cm3) of the analyte solution, I is the dilution factor (if any) and rn is the mass of the sample (8).The procedure for estimating the TI concentration in 10 cm3 of sample solution transferred into the polarographic cell ( x ) involves extrapolation of the original standard additions graph n 1 2 3 4 5 6 7 8 9 10 11 12 13 X 0 0 0 0 0 20 20 40 40 60 60 80 80 h n 0.158 0.158 0.158 0.158 0.158 0.087 0.087 0.084 0.084 0.150 0.150 0.284 0.284 Y E 3.1 3.047 3.7 3.047 2.8 3.047 3.5 3.047 2.7 3.047 6.4 6.751 7.0 6.751 9.8 10.456 10.3 10.456 13.8 14.160 14.2 14.160 17.9 17.864 18.5 17.864 r 0.053 0.653 0.453 -0.247 -0.347 -0.352 0.248 -0.656 -0.156 -0.360 0.040 0.036 0.636 S 0.458 0.458 0.458 0.458 0.458 0.444 0.444 0.443 0.443 0.456 0.456 0.482 0.482 Table 11 Analysis of variance (ANOVA) for the analysed sediment sample; * indicates a significance level of 99.9%. The mean response from this experiment is 8.476, and r2 = 0.9951.The lack-of-fit is not significant, while the remainder of the regression is highly significant Mean Sum of Degrees of Mean square Source squares freedom squares ratio Regression on- Mean Factor effect Total regression Residual Pure error Lack-of-fit Corrected Total 994.4377 401.0805 1395.5182 1.9919 1.3170 0.6749 403.0723 1397.5100 1 994.4377 5491.95* 1 401.0805 2215.04* 2 697.7591 3853.49* 8 3 0.2249 1.37 12 13 11 0.1811 - 0.1646 - - - - - Table 12 Thallium concentration in the analysed cement and sediment samples T1 concentration/ T1 concentration/ Sample No.ng cm-3 S* ngg-' 1 28.00 3.07 700.0 2 15.80 2.91 395.0 3 16.50 2.59 850.0 * s is the standard error in the estimation of the concentration as discussed in the text. as illustrated above or from the fitted model by putting the predicted response value equal to 0: bo + bx = 0 The value of x given by eqn. (1 1) is minus the concentration of Tl in the sample. The standard deviation (variance or uncertainty) for each fitted model is expressed in the units of the dependent variable y (or response) because, as pointed out above, the errors in the independent variable x (or concentration) are ignored using the most commonly employed methods. However, the units of variance s could be converted from the units of y (or peak height in cm) to the units of x (or concentration in ng cm-3) as follows: (11) sr s, = - b The standard deviation (or uncertainty) could be easily calculated using eqn.(8) to calculate h, which is outside the region of experimentation in this case. The T1 content in the samples is given in Table 12. The estimated Tl concentration x (ng cm-3) from the fitted model is the concentration that was present in 10 cm3 of the sample solution originally transferred into the polarographic cell. The over-all T1 concentration in the total sample solution isANALYST, JULY 1992, VOL. 117 1083 calculated by multiplying this value by the final sample solution which, when divided by the original sample mass (m), gives the TI content (in ng g-1) of the solid sample. The standard uncertainty (s,) of the estimated T1 concentration is calculated via the standard uncertainty equation.40 Hence, by converting the units of s from s,, to s, and substituting x at i, = 0 in eqn.(8) to obtain h at that point, then the over-all s, can be easily calculated. In the analysis above, it is assumed that the errors in the data are constant throughout the experimental region. This assumption is usual in many forms of multilinear regression. In particular, standard approaches to experimental design nor- mally involve taking replicates in only one place (generally the centre of the design) and then assuming that the replicate error is constant. This assumption is often called homosced- acity of errors. Some experiments result in heteroscedacity of errors.Under these circumstances, the error distribution varies across the experimental region. In some instances the absolute errors associated with small measurements are less than those associated with larger measurements. Whether this occurs or not is related to the over-all experimental process, and the way to study this is to take replicates throughout the experimental region. In this work, in common with many practical analytical situations, the heteroscedacity of errors was not studied; instead, it was assumed that the errors are constant through- out the experimental region. It would be possible to extend the studies in this paper to obtain a fuller understanding of the errors. If such data are available, there are two main approaches to correcting for the heteroscedacity of the errors.53 (i) Use a weighted least-squares procedure for calibration, weighting each point according to the error distribution. (ii) Scale the y variable to take account of these errors; this is possible if the error distribution follows a well-defined trend, for example, if the errors are proportional to the measurements.It is easy to modify the approach described above if required. However, the analyst must carefully consider the amount of work required for the solution of any given problem. It is clearly impractical to perform a full and exhaustive study of error distributions every time a linear calibration model is computed, and the simple methods of visualizing confidence in the model as related to the design discussed above serve as a good approach.It is typical of chemometrics that there is a range of methods according to the detail required from the data. A typical example involves the classification of objects. Exploratory approaches such as principal components analysis allow graphical methods for visualizing whether there are major groups; cluster analysis will then allow the main groups to be defined better; soft modelling ( e . g . , SIMCA) provides mathematical models of the groups selected previously; hard modelling such as canonical variates analysis can be used where there is an even more rigid and well-defined class structure. No one approach is superior to any other, and the methods range from exploratory data analysis (EDA) to mathematical modelling. This paper has not presented a method for the mathematical modelling of errors, but does provide a good exploratory approach to the visualization of the influence of design on linear calibration experiments.Conclusion From the results presented here the feasibility of determining TI by DPASV has been demonstrated. The sample prepara- tion prior to DPASV analysis permits the separation of TI from the other chemical species that interfere in the analysis. Further, use of the standard additions method enhances the confidence in the estimated response. In the extraction procedure, the use of diethyl ether gave good results. However, no other solvents were used in the extraction procedure (for example, diisopropyl ether). By carefully considering the nature of the experiments and the analysis used to calibrate the observed response to the concentration of Tl, a high precision in estimating the parameters of the linear model was achieved via replicating the original experiments to provide a sufficient number of degrees of freedom to be used in the ANOVA procedure.A proper procedure involves the following. 1. Determination of the calibration graph for each experi- ment by the usual method of linear regression, ignoring the errors in the independent variable x. 2. The scatter about the calibration graph (variance or standard uncertainty) is obtained from the fitted model s,. The conversion of the variance units from that of sPn to the units of concentration s, is also demonstrated. 3. The replication error (pure error) is used to test the lack-of-fit of the model to the experimental data.4. The standard uncertainty (or variance) in the predicted response value sJn and for the estimated concentration of TI in the sample solution s, is obtained. 5. The confidence intervals and the confidence bands in the predicted response i, for each experiment can also be computed. 6. The leverage equation that can be used to calculate h should be considered: further discussion about leverage is provided elsewhere .43-48 Calibration is a very important procedure in analytical chemistry; therefore, further work is necessary in order to understand and explain the results fully: it would be of considerable interest to extend methods to situations where errors exist in both the x and y directions. In this paper, only the importance of experimental design in analytical experiments as explained by ANOVA, confidence bands and leverage has been considered.Errors in the x-axis have not been discussed. Although these are important, most designs used by experimental chemists assume that there are no errors in the x-axis. In order to incorporate these errors, signficantly different approaches need to be developed, but are outside the scope of this paper. M. A. A. thanks the Department of Chemistry, Faculty of Science, Garyounis University, Benghazi, Libya, and the Cultural Department of the Secretary of Higher Education, Tripoli, Libya, for finance for a research studentship. We are grateful to Dr. G. Nickless for discussions and advice. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 References Schoer, J., in The Handbook of Environmental Chemistry, ed.Hutzinger, O., Springer-Verlag, Berlin, 1984, pt. C, p. 143. Allus, M. A., Martin, M. H., and Nickless, G., Chemosphere, 1987, 16, 929. Dolgner, R. A., Brockhaus, A., Ewers, V., Wiegand. H., Majewski, F., and Soddemann, H., Biol. Absfr., 1983, 76, 92141. El-Hossadi, A., Allan, A., Ah, S. S., Faroqo, R., Hamid, A., and Majed, T. A., J. Radioanal. Nucl. Chem., Articles, 1984, 81/82, 359. Hirasue, M., and Kanda, K., Zem.-Kalk-Gips, 1985, 8, 411. Deane, P. G., Zem.-Kalk-Gips, 1982, 12,283. Irving, H. M., and Rossotti, F. J. C., Analyst, 1952, 77, 801. Ziegerova, L., Stulik, K., and Dolezal, L., Talanta. 1971, 18, 603. Dhaneshwar, R. G., and Zarapkar, L. R., Analysr, 1980, 105, 386. Liem, I., Kaaisser, G., Sager, M., and Tolg, G., Anal.Chim. Acta, 1984, 158 179. Gorbauch, H., Rump, H. H., Alter, G., and Schmitt-Henco, C. H.. Fresenius’ 2. Anal. Chem., 1984, 317, 236. Haynes, B. W., and Kramer, G. W., Characterisation of US Cement Kiln Dust, Information Circular, US Bureau of Mines, 1982, 8885, p. 1. Rechenberg, W., Transl. ZKG, 1982, 2, 90. Hrsak, J., and Fugas, M., Mikrochim. Acta, Part ZZ, 1981, 111.1084 ANALYST, JULY 1992, VOL. 117 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 EG & G Princeton Applied Research, Analytical Instrument Division, Basics of Voltammetry and Polarography , Application Note P-2, Princeton Applied Research, Princeton, NJ, 1982. Model 315 Automated Electroanalytical Controller Operating and Service Manual, Princeton Applied Research, Princeton, NJ, 1974.SAS User’s Guide: Statistics Version 5 Edition, SAS Institute, Cary, NC, 1985, p. 655. Currie, L. A., in Chemometrics, Mathematics and Statistics in Chemistry, ed. Kowalski, B. R., NATO AS1 Series, vol. 138, Reidel, Dordrecht, 1983, p. 115. Massart, D. L., and Hoogewijs, G., Pure Appl. Chem., 1983, 55, 1861. Friess, S. L., Am. Stat., 1977, 31, 2. Agterddenbos, J., Anal. Chim. Acta, 1979, 108, 315. Agterddenbos, J., Anal. Chim. Acta, 1981, 132, 127. Gabriel, R., Anal. Chem., 1970, 42, 1439. Currie, L. A,, Pure Appl. Chem., 1982,54, 715. Liteanu, C., and Rica, I., Pure Appl. Chem., 1975, 44, 535. Feinberg, M. H., J. Chemometr., 1988, 3, 103. Franke, J. P., de Zeeuw, R. A., and Hakkert, R., Anal. Chem., 1978, 50, 1375. Larsen, I. L., Hartmann, N. A., and Wagner, J. J., Anal. Chem., 1973,45, 1511. Mandel, J., and Linnig, F., Anal. Chem., 1957, 29, 743. Linnig, F., and Mandel, J., Anal. Chem., 1964, 36, 25A. Irvin, J. A., and Quickenden, T. I., J. Chem. Educ.. 1983,60, 711. Walters, F. H., and Rizzuto, G. T., Anal. Lett., 1988,21,2069. Hunter, W. G., in Chemometrics, Mathematics and Statistics in Chemistry, ed. Kowalski, B. R., NATO AS1 Series, vol. 138, Reidel, Dordrecht, 1983, p. 97. Randall, L., Ph.D. Thesis, University of Bristol, 1984. Willard, H. H., Merritt, L. L., Dean, J. A., Jr., and Settle, F. A., Jr., Instrumental Methods of Analysis, Wadsworth, Belmont, CA, 7th edn., 1988. Ewing, G. W., Instrumental Methods of Chemical Analysis, McGraw-Hill, New York and London, 5th edn., 1985. 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 Goulden, P. D., Environmental Pollution Analysis, Heyden, London, 1978. Sharaf, M. A., Illman, D., and Kowalski, B. R., Chemometrics, Wiley, Chichester, 1986. Wolters, R., and Kateman, G., J. Chemometr., 1990, 4, 171. Deming, S. N., and Morgan, S. L., Experimental Design: a Chemometric Approach, Elsevier, Amsterdam, 1987. Draper, N. R., and Smith, H., Applied Regression Analysis, Wiley, New York, 2nd edn., 1981. Myers, R. H., Response Surface Methodology, Allyn and Bacon, Boston, MA, 1976. Martens, H., and Naes, T., Multivariate Calibration, Wiley, Chichester, 1989. Brereton, R. G., Chemometrics: Applications of Mathematics and Statistics to Laboratory Systems, Ellis Honvood, Chiches- ter, 1990. Belsley, D. A., Kuh, E., and Welsch, R. E., Regression Diagnostics: Identifying Influential Data and Source of Col- linearity, Wiley, New York, 1980. Hoaglin, D. C., and Welsch, R. E., Am. Stat., 1978, 32, 18. Allus, M. A., Brereton, R. G., and Nickless, G., Chemometr. Intell. Lab. Syst., 1989, 6, 65. Allus, M. A., and Brereton, R. G., Int. J. Environ. Anal. Chem., 1990,38, 279. Allus, M. A., Ph.D. Thesis, University of Bristol, 1990. Wold, S., Esbensen, K., and Geladi, P., Chemometr. Intell. Lab. Syst., 1987, 2, 37. Rawlings, J. O., Applied Regression Analysis. A Research Tool, Wadsworth, Belmont, CA, 1988. Deming, S. N., and Morgan, S. L., Clin. Chem., (Winston- Salem, N. C.), 1979, 25, 840. Massart, D. L., Vandeginste, B. G. M., Deming, S. N., Michotte, Y ., and Kaufmann, L., Chemometrics: a Textbook, Elsevier, Amsterdam, 1988, p. 84. Paper 1 I01 3420 Received March 20, 1991 Accepted February 19, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701075
出版商:RSC
年代:1992
数据来源: RSC
|
6. |
Multivariate analysis of a round-robin study on the measurement of chlorobiphenyls in fish oil |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1085-1091
Raj K. Misra,
Preview
|
PDF (1080KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1085 Multivariate Analysis of a Round-robin Study on the Measurement of Chlorobiphenyls in Fish Oil Raj K. Misra and John F. Uthe Marine Chemistry Division, Department of Fisheries and Oceans, Halifax Fisheries Research Laboratory, P. 0. Box 550, Halifax, Nova Scotia, Canada B3J 2S7 Charles J. Musial C. Musial Consulting Chemist Ltd., Halifax, Nova Scotia, Canada B3L 4J2 Participants identified and measured chlorobiphenyls (CBs) in fish oil both before and after spiking with undisclosed amounts of four CBs following two clean-up methods, a common one and their own. Complete data were received from 18 of 30 participants of which only 2 correctly identified IUPAC No. 86. All correctly identified the other three CBs (IUPAC Nos. 52,101 and 153). The presence of correlations (10 of 12 coefficients of correlation were moderate to high) among concentrations of the three CBs precluded univariate analysis. Following deletion of one obvious outlier, paired difference and mean data from 17 participants were analysed by a multivariate procedure.Four data pairs were compared. No effect of clean-up on precision was found. Spike recoveries differed significantly from expected values, and the patterns of differences varied with the clean-up method employed. Multivariate analysis of paired differences and means was used t o rank laboratory performance with respect t o laboratory precision and bias, and t o identify possible outliers. Laboratory bias (between laboratory) exceeded random (within laboratory) variations in all data sets.None of the rankings based on paired differences (random error) was correlated significantly with the equivalent rankings based on paired means (laboratory bias). However, rankings for data set 1 were correlated with rankings for data set 2, as were those of data set 3 with those of data set 4 for paired difference data and for paired mean data. The multivariate method allows comparisons and laboratory ranking based on si m u Ita neous a na I ysis of correlated mu Iti ple determ i na nds. Keywords: Chlorobiphenyls; interlaboratory round robin; multivariate analysis; laboratory performance ranking; types of error Earlier, we1 described a collaborative study in which partici- pants had identified and measured chlorobiphenyls (CBs) in a spiked and unspiked fish oil.Recoveries ranged from 24 to 294% for spikes of 63-85 ng g-1 of oil. Investigation of the results (see below) showed that univariate analysis was not appropriate due to the presence of significant correlations between CB concentrations. Univariate analysis of paired differences and paired means has been extensively considered.2-3 However, a multivariate approach, which analyses a number of variables (p; p 3 2) jointly, is needed69 when correlations exist among these variables, as a series of p univariate analyses will ignore these correlations and will, therefore, be inappropriate. Given the nature of CB analysis, i.e., each set of CB measurements uses a single chromatogram, it is probably not surprising to find such correlations. Experimental Experimental details have been reported previously.’ In summary, herring (Clupea harengus harengus) oil was spiked as follows: IUPAC No.52-2,2‘ ,5,5’-tetrachlorobiphenyl = 82 ng g-1 of oil; IUPAC No. 86-2,2’,3,4,5-pentachlorobi- phenyl = 77 ng g-1 of oil; IUPAC No. 153-2,2‘,4,4‘,5,5’- hexachlorobiphenyl = 85 ng g-1 of oil; and IUPAC No. 101-2,2,4,5,5’-pentachlorobiphenyl = 63 ng g-1 of oil. Participants also received sufficient amounts of these CBs to prepare standards. Participants were asked to identify these CBs and, inter alia, determine their concentrations in both oils using their own and the common clean-up10 method. Seven- teen (Table 1) reported complete data (one other complete set was dropped as an obvious outlier). Only 2 correctly identified IUPAC No.86; thus its data have been dropped. Partial results’ were also dropped from the present study because multivariate analysis is best carried out with complete data sets .4,11 Statistical Considerations The statistical methodology242 12-14 is presented in the Appen- dix. SYSTATlS was employed for various computations. All tests were carried out at the 5% probability level ( P ) unless stated otherwise. Results for the three-way (laboratory x oil x method) multivariate analysis of variance (MANOVA) showed the need for separate analysis of the differences in the methods for the separate oils and vice versa, as explained in the Appendix. Four paired data sets were compared: set 1, unspiked oil, common method versus laboratory clean-up method; set 2 , spiked oil, common method versus laboratory clean-up method; set 3, common clean-up method, spiked versus unspiked oils; and set 4, laboratory clean-up method, spiked versus unspiked oils.These data sets were chosen for obvious reasons. Although these paired comparisons (contrasts) are not independent, Harris8 states, ‘It is convenient for descriptive purposes to choose independent contrasts, but this should never get in the way of testing those contrasts among the means which are of greatest theoretical import to the researcher’. A laboratory’s two determinations are designated Xi, and Xi2, their differ- ence (Xi, - Xi2) as Di, their sum (Xi, + Xi,) as Si, and their average as pi. The importance of analysing such paired data has been emphasized,2.3 both noting that the dominant factor accounting for variations in the sum (or average) values is systematic error (laboratory bias), whereas difference ( Di) values reflect only random variations3 (laboratory precision). It is important to compare both precision and bias among laboratories.Results Analysis of Paired Differences (Intralaboratory Precision) Means and ranges of differences are shown in Table 2. Each set was analysed the same way and only the results for data set1086 ANALYST, JULY 1992, VOL. 117 1 are tabulated. Pearson coefficients of correlation, r,, (where u, v = 1, 2, 3) were computed for each data set. Several of these, i.e., 10 of 12, ranged from 0.2 to 0.9 in absolute value. Although large sample sizes (17 is hardly large) are desirable for tests of significance of correlations, 6 of these were significant with P ranging from <0.001 to C0.04, and 4 were significant by the Bonferroni test (P = CO.001-<0.05).Bartlett’s x 2 (chi-square) test, which tests the global hypoth- esis concerning the significance of all of the correlations, was significant (P <0.001) in data sets 2 and 3 and nearly so (P = 0.06) in data set 4. These observations lead to the following considerations.4~7-9 (1) When the number (p) of variables is two or more, the covariance structure of the population is described by p variances and %p(p - 1) covariances (or, equivalently, correlations). Univariate analyses, carried out on each of the p variables separately, will take account of p Table 1 Reported concentrations (pg kg-l) of CBs in fish oil IUPAC Sample No.52 153 101 52 153 101 Common Laboratory Labora- clean-up clean-up tory* method method No. Unspiked oil- 1 34.5 2 49 3 57 5 41 6 65.1 7 104 10 137 12 41 17 78.4 19 43 20 51.8 21 53.9 22 90.5 23 183 24 154 25 30 26 48 Spiked oil- 1 76.8 2 116 3 132 5 113 6 140 7 187 10 227 12 110 17 147.7 19 75 20 130.9 21 144 22 135.5 23 246 24 200 25 55 26 102 51.9 76 69 59.7 88.1 107 163 47 77.8 85 62.9 66.9 90.3 73 188 46 64 105.0 159 136 127.9 168 204 313 118 158.5 127 135.6 176 144 141 238 95 115 47.3 124 62 48.9 92.9 92.4 130 47 64.1 67 93.4 53.6 146.2 95 154 43 75 77.9 155 130 104.6 147 184 308 108 119.3 95 146.0 134 211.3 146 193 78 124 80.7 66 80 29.5 56.8 204 210 38 68.9 41 53.6 78.4 47.0 134 158 37 49 145.2 158 184 118.9 127 292 277 108 144.4 98 127.7 127 100.2 210 220 66 135 * Laboratory numbers as defined in ref.1. 83.0 94 63 68.4 122 92.4 159 53 72.7 121 71.8 152.0 54.0 85 163 42 88 140.9 186 141 129.8 171 141 229 125 149.5 157 148.9 182 109.9 156 208 93 146 106.0 137 79 121.1 91.6 69.7 222 62 61.4 54 95.8 81.2 77.4 97 157 48 90 132.2 200 160 135 236 232 105 116.6 102 156.2 124 137.7 15 1 190 87 138 97.2 variances, but will ignore the correlations. Therefore, in the present study, a multivariate procedure, which employs the entire covariance structure ofp + Y2p(p - 1) parameters, must be used. (2) The correlation coefficient is a measure of the closeness of the linear relationship between two variables. A positive value of ruv shows that Di, and Dj, increase together, i.e., if the difference value by the two methods is large for one CB, it is also likely to be large for the other.When ruv is negative, large values of Di, are associated with small values of Di,. Thus, there also exists a ‘third’ type of error, proliferating from the covariation of the CBs, which induces additional disparity among the laboratories. Set I , unspiked oil, common versus laboratory clean-up method Multivariate normality of difference variables was assessed (see Appendix).4 For observations generated from a three- variate normal distribution, about 95% of the dj2 values should be S7.815, which is the 95th percentile of a x 2 distribution with degrees of freedom (DF) = 3. Sixteen of our 17 values, i.e., 94%, were G7.815; thus evidence against three-vanate normality is insufficient.4 A x2 plot4 of ordered di2 (abscissa) and the corresponding x 2 percentile (ordinate) yielded a coefficient of linear correlation of 0.969, giving no evidence against normality.The average difference between determinations is expected to be zero for each CB if no real difference results from the two clean-up methods. In order to test this, the null hypothesis (&) of no average difference, null vector, Q, was specified for C in I& : pD = C (see Appendix). (A vector is denoted by an underscored letter and a matrix by a letter in bold print. Prime denotes a transposed vector or matrix. An unprimed vector denotes a column vector.) The observed Hotelling’s T2 value4 was 3.6757 (P = 0.39) and, therefore, Ho was accepted, i.e., the clean-up method had no significant effect. Squared generalized distances, di2, (Table 3) ranged from 0.0948 to 10.3372 and determined laboratory ranking with respect to precision, yielding an over-all judgement of each Table 3 Squared generalized distances and probabilities of associated chi-square values for paired difference, data set 1 Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Laboratory No.20 2 12 25 26 3 17 6 24 19 1 23 5 22 10 21 7 di2 0.0948 0.1430 0.2160 0.2535 0.3258 0.4999 0.5499 1.2893 1.6170 1.9689 2.0463 2.6857 4.8745 6.4687 6.9112 7.7183 10.3372 Probability 99.20 98.55 97.39 96.74 95.40 91.85 90.77 73.55 65.98 58.28 56.66 55.53 17.95 8.94 7.35 5.13 1.59 (Yo) Table 2 Mean, minimum (Min) and maximum (Max) values of differences between treatments in recoveries of three CBs Data set 1 2 3 4 IUPAC Mean Min Max Mean Min Max Mean Min Max Mean Min Max N0.52 -10.04 -100.0 49.0 -17.68 -105.0 36.0 63.34 25.0 90.1 70.97 29.0 104.0 NO.153 -9.92 -85.1 36.3 2.76 -35.9 84.0 73.26 42.0 150.0 60.57 30.0 92.0 N0.101 -12.61 -92.0 68.8 -2.28 -54.3 76.0 60.31 28.0 178.0 49.98 -23.9 166.3ANALYST, JULY 1992, VOL. 117 1087 Table 4 Laboratory means of CBs for four data sets (minima are underlined; maxima are in bold print) IUPAC sample No. Labora- tory No. 1 2 3 5 6 7 10 12 17 19 20 21 22 23 24 25 26 52 153 Data set 1 101 57.60 57.50 68.50 35.25 60.95 154.00 173.50 39.50 73.65 42.00 52.70 66.15 68.75 158.50 156.00 33.50 48.50 67.45 85 .OO 66.00 64.05 105.05 99.70 161.00 50.00 75.25 103.00 67.35 109.45 72.15 79.00 175.50 44.00 76.00 76.65 130.50 70.50 85 .OO 92.25 81.05 176.00 54.50 62.75 60.50 94.60 67.40 111.80 96.00 155.50 45.50 82.50 52 111.00 137.00 158.00 115.95 133.50 239.50 252.00 109.00 146.05 86.50 129.30 135.50 117.85 228.00 210.00 60.50 118.50 153 Data set 2 122.95 172.50 138.50 128.85 169.50 172.50 271.00 121.50 154.00 142.00 142.25 179.00 126.95 148.50 223 .OO 94.00 130.50 101 105.05 177.50 145.00 100.90 141 .OO 210.00 270.00 106.50 117.95 98.50 151.10 129.00 174.50 148.50 191 S O 82.50 131.00 52 55.65 82.50 94.50 77.00 102.55 145.50 182.00 75.50 113.05 59.00 91.35 98.95 113.00 214.50 177.00 42.50 75.00 153 Data set 3 78.45 117.50 102.50 93.80 128.05 155.50 238.00 82.50 118.15 106.00 99.25 121.45 117.15 107.00 213.00 70.50 89.50 101 52 153 Data set 4 101 62.60 139.50 96.00 76.75 119.95 138.20 219.00 77.50 91.70 81 .OO 119.70 93.80 178.75 120.50 173.50 60.50 99.50 112.95 112.00 132.00 74.20 91.90 248.00 243.50 7 3 .0 106.65 69.50 90.65 102.70 73.60 172.00 189.00 51.50 92.00 111.95 140.00 102.00 99.10 146.50 116.70 194.00 89.00 111.10 139.00 110.35 167.00 81.95 120.50 185.50 67.50 117.00 119.10 168.50 119.50 109.15 113.30 152.85 227.00 83.50 89.00 78.00 126.00 102.60 107.55 124.00 173.50 67.50 114.00 laboratory’s ‘nearness’ to the expected value of zero. Extreme laboratories, i. e . , laboratories for which the squared general- ized distance (see Appendix) yielded values of P S0.05, are shown in bold, but were not excluded because there is no reason to suspect that they are ‘discordant’5 or belong to a different population. Laboratories ranked: 20 (best pre- cision), 2, 12,25,26,3,17,6,24,19,1,23,5,22,10,21 and 7 (worst).Table 3 also gives probabilities (expressed in per cent.) at which their xz ( x 2 with 3 DF) approximations are significant. Set 2, spiked oil, common versus laboratory clean-up method These findings were similar to data set 1, i.e., there was no significant difference between the two clean-up methods (72 = 6.3363; P = 0.18). The coefficient of linear correlation of the x2 plot was 0.946. Fifteen of the 17 di2 values (88%) were S7.815 of xz. The range of di2 was 0.1852-10.9729. Labora- toriesranked: 5 (best), 12,25,6,20,21,17,3,24,2,19,26,23, 1, 22, 10 and 7 (worst). Set 3, common clean-up method, spiked versus unspiked oils If all recoveries of added CBs were loo%, expected values for the means of differences between the spiked and unspiked oils would be 82 for IUPAC No.52, 85 for IUPAC No. 153 and 63 for IUPAC No. 101. To test recoveries (Ho: pD = C), vector C’ was specified as (82, 85, 63). The estimated T2 was 23.8699 with P = 0.004, showing significant average differ- ences in spike recoveries and/or of their linear combinations [see Appendix, eqn. (S)]. As individual variables are of interest, simultaneous 95% confidence intervals (CIS) for the three individual mean differences were estimated by T2 and also by the Bonferroni procedures. None of the three CIS included zero (in either procedure), showing, thereby, that all three CBs significantly contributed to poor recovery. The coefficient of linear correlation of the x 2 plot was 0.952. Sixteen of the 17 di2 values (94%) were a7.815 of x:. The range of di2 values was 0.1185-13.0115.Ranks of laboratories are: 23 (best), 12, 17,24,6,7,5,1,26,20, 19,22,21,3,25,2 and 10 (worst). Set 4, laboratory clean-up method, spiked versus unspiked oils Findings were similar to data set 3. Ho : F~ = C was rejected (72 = 42.7370 yielding P <0.001), showing, again, poor spike recoveries. Simultaneous 95% CIS (both T2 and Bonferroni procedures) showed that the difference variable for each CB contributed significantly to the rejection of Ho. The coefficient of linear correlation of the x 2 plot was 0.969. Sixteen of the 17 di2 values (94%) were S7.815 of x:. The range of di2 values was 0.4329-10.4723. Laboratories ranked: 1 (best), 23,12, 6, 24, 17, 22, 26,20, 10, 19, 3, 21, 2, 25, 5 and 7 (worst).Comparison of spike recoveries by the two clean-up methods Following the observation of no average difference between the two clean-up methods for either oil, differences in recoveries between the two methods for the two oils were compared. The difference between the two treatments is a specific case of the linear comparison among treatments.13 Instead of eqn. (2) (in Appendix) for laboratory i, variable Di is given by Di = (ei, - ei2)unspiked - (ei, - ei2)spiked A laboratory’s systematic error drops out and Di should contain only random errors. Null hypothesis & : F~ = 0 of no average difference between two differences was rejected (T2 = 31.9428 yielding P = O.OOl), showing that the patterns of variation for the two oils were dissimilar.No individual CB contributed significantly to the average difference, as judged by their 95% CIS (by Bonferroni and T2 procedures). The coefficient of linear correlation of the x2 plot was 0.953. Fifteen of 17 d,2 values (88%) were ~ 7 . 8 1 5 of x;. The range of di2 values was 0.2293-10.8385. Laboratories ranked: 24 (best), 23, 17, 6, 1, 22,2,3,25, 19, 12,26,5,20,21,7 and 10 (worst). Analysis of Paired Means As is usual in these studies,2J widely divergent laboratory means were found for all four data sets (Table 4). Laboratory bias effects are apparent as all minima are associated with one laboratory and 9 of 12 maxima with another. Consequently, more laboratories are expected to be identified as extreme in paired mean data than in paired difference data.Set I , unspiked oil, common versus laboratory clean-up methods The null hypothesis of equality of laboratory means was rejected ( P <0.001) by the two-way MANOVA of model (1 1). Squared generalized distances 7’2 [eqn. (13), Appendix] of laboratory means ranged from 2.5500 to 54.2206 (Table 5 ) . Laboratories ranked (Table 5 ) : 6 (best, i.e., least deviant from1088 ANALYST, JULY 1992, VOL. 117 Table 5 Squared generalized distances by Hotelling’s P and probabil- ity levels of significance for paired mean data set 1 Laboratory Probability Rank No. P (Yo ) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 6 17 3 1 26 22 21 20 19 5 2 12 25 23 7 10 24 2.5500 2.6300 3.0196 3.4576 3.6622 5.4149 5.7241 6.0008 7.2962 9.9097 10.8434 11.7925 15.9671 20.8363 23.7602 52.8010 54.2206 54.62 53.38 52.33 41.95 39.52 23.82 21.83 20.20 14.16 7.22 5.75 4.60 1.85 0.73 0.44 0.01 0.01 the over-all mean vector), 17,3,1,26,22,21,20,19,5,2,12, 25,23, 7, 10 and 24 (worst).Set 2, spiked oil, common versus laboratory clean-up methods The & of equal laboratory effects was rejected ( P <O.OOl). The Tz values ranged from 1.3308 to 114.9904. Laboratory rankings were: 3 (best), 6,5, 17,2,20, 1,21,26,12,19,7,23, 22, 24, 25 and 10 (worst). Set 3, common clean-up method, spiked versus unspiked oils The & of equal laboratory effects was rejected ( P <0.001). The T2 values ranged from 1.7390 to 477.4613. Laboratories ranked: 3 (best), 6,17,12,21,5,26,20,7,2,1,25,19,22,24, 10 and 23 (worst). Set 4, laboratory clean-up method, spiked versus unspiked oils The & of equal laboratory effects was rejected ( P <0.001). The 72 values ranged from 2.0029 to 337.8852.Laboratories ranked: 1 (best), 17,26,20,3,2,5,12,22,6,23,21,25,19,24, 10 and 7 (worst). Investigation of Ranks for Correlations Spearman’s coefficient of rank correlation14 analysis of ranks based on paired differences and those based on paired means showed: (1) For each data set, ranks based on paired differences were not significantly correlated with those based on paired means ( P values = 0.55, 0.068,0.605 and 0.210 for data sets 1,2,3 and 4, respectively). (2) Ranks based on paired differences for data set 1 were significantly correlated with those of data set 2 (P = 0.039) as were ranks for data set 3 with those of data set 4 (P = 0.016).(3) Ranks based on paired means for data set 1 were significantly correlated with those based on data set 2 (P <0.001) as were ranks for data set 3 with those of data set 4 (P = 0.048). Discussion Both Youden and Steiner2 and Wernimont3 have noted that analytical results generally vary far more between laboratories (laboratory bias) than within laboratories (precision); our data are no exception. Considerably larger variabilities between laboratories, relative to those within laboratory, were shown in the two-way MANOVAs of all four data sets. More outliers and larger individual deviations from the over-all means of all laboratories were found in the analyses of paired means than in those of paired differences. The need for chemists to develop and apply appropriate techniques for controlling both precision and bias is obvious.In this study, CBs were generally mutually correlated, i.e. , 10 of 12 coefficients of correlation between paired differences ranged from 0.2 to 0.9. Given that each measurement of a CB is not fully independent ‘from the others (all CBs are determined from a chromatogram resulting from a single injection), it is not surprising that error affects all CBs similarly. Unlike the univariate procedure, the multivariate procedure analysed the entire suite of response variables simultaneously and did not ignore these correlations. The multivariate procedure allows laboratory performance to be judged on the basis of an entire suite of determinands simultaneously. Judgements, based on a series of univariate analyses on individual determinands can lead to conflicting results, which are difficult to interpret (because these uni- variate analyses do not account for correlations among determinands), and might unfairly penalize a laboratory in specific instances. Undoubtedly, there are other data sets that would benefit from a multivariate analysis.A variety of newer analytical methods, e. g., inductively coupled plasma mass spectrometry for elements, gashiquid chromatography-mass spectrometry for families of organics, suggest that future intercalibrations will require simultaneous statistical analysis of suites of determinands in order to rate laboratory perfor- mance. The difference between a laboratory’s two determinations in paired data should only contain that laboratory’s random error (precision).However, analysis of data on paired differences showed that their variations could not be explained by random variation alone. Thus, other explana- tions are needed beyond random error. Analysis of paired difference data sets showed that additional causes are not likely to be associated with the two clean-up methods for the following reasons: ( i ) mean differences between the two methods were not significant (data sets 1 and 2); (ii) within the same clean-up method, the means of the spiked (less the added amount of each CB) and unspiked oils differed significantly (data sets 3 and 4); (iii) the differences in measurements by the two clean-up methods in the spiked oil were not similar to those in the unspiked oil; and (iv) the results of the ranking procedure were not consistent over the four data sets.This should be considered in the design of future intercomparative studies. Laboratory rankings based on paired differences were uncorrelated with rankings based on paired means, reflecting the separation of error into random (paired difference) and bias (paired mean)*,3 and reinforcing the necessity for analysts to focus quality control on both precision and accuracy. However, rankings for data set 1 were correlated with rankings for data set 2, as were those of data set 3 with those of data set 4 for paired difference data and for paired mean data, reflecting an over-all consistency in the relative magnitudes of a laboratory’s random error and bias. The multivariate approach allows for two integrated rank- ings, one for precision, one for bias, to be assigned to a laboratory based on its performance for multiple determi- nands, a better approach, in our opinion, than a series of rankings based on analysis of the individual determinands.Identification as a possible outlier serves to identify a laboratory’s performance as being significantly different from the other participants. Laboratories so identified might still be performing adequately with respect to other criteria, but should be aware of a need for improved performance.ANALYST, JULY 1992, VOL. 117 1089 Appendix Three-way MANOVA For value XjjK of a determinand reported by the ith laboratory in the jth oil by the Kth method, the three-way MANOVA model will be XjjK = p + Ai + Bj + CK + (AB)ij + (AC)iK + ( B c ) j K + (ABC)ijK where p is the over-all mean; Ai, Bj and CK are the treatment effects for the zth, jth and Kth groups of treatments, A , B and C , respectively; (AB),,., (AC)iK and (BC)jK are first-order interaction effects; and (ABC),, is the second-order interac- tion effect.Without replication, there is no way of testing interaction ABC. Three-way MANOVA was carried out by employing the main effect, the laboratory, as ‘random’ and the main effects, the oil and method, as ‘fixed’. All three first-order interactions were significant, with P in the range ~0.001-~0.05. These considerations led to the use of a simpler (than three-way) layout for data analysis. Paired Differences At the univariate level, in the general case where q determina- tions are made by laboratory i of a determinand (X), the model for Xij, the jth determination in laboratory i, is written as15 Xjj = pi + eij, j = 1, .. ., q It is assumed that the measurements Xij are N(pi, 02), i.e., normally distributed about the laboratory mean pi, which can vary from one laboratory to another, and that the variance (02) of Xij is the same in all laboratories. Participating laboratories are considered to be a sample of all laboratories that will use the method.3 This model can be written as Xij = p + ai + eij (1) where p is the over-all mean, aj is the laboratory deviation (bias) and where the random elements eij are N(0,02). When q = 2, for duplicate determinations made by n laboratories, the difference ( Di) between determinations from laboratory i is given by A laboratory’s systematic error (bias) drops out of Di and the sample of n measurements ( D j ) on the difference variable (6) should contain only random errors of both determinations.2 However, with Youden replicates, i .e . , single determinations on two similar materials, values of Dj are not necessarily trivial and merit investigation.’ We consider, for example, data set 1 where replicates are measurements from two treatments, which are the common clean-up method (M1) and the laboratory’s own clean-up method (M2) for the unspiked oil. The Dj for laboratory i will be given by Dj = M1 - M2 + ei, - ei2 (3) instead of the value shown in eqn. (2). As the number of treatments is two, data can be analysed as ‘paired’ data. From eqn.(3) we note the following: in addition to random errors of determinations, D j values reflect only the differential effects of treatments. The difference (Di) between two members of a pair (within laboratory i) is an estimate of the average difference (PO) in the effects of two treatments. To determine if a specified value (C; which is often specified as zero) is a plausible value for the mean difference between treatments, the null hypothesis, Ho: pD = C , is tested against the alternative H1 : p D # C. The appropriate test statistic is where D and sD2 are estimates of the population mean and variance, respectively, of the differences. It has a Student’s t-distribution with n- 1 degrees of freedom (DF). Equivalent to rejecting & when It1 is large is if its square is large.Thejariable t2 is the squared distance from the sample mean D to the test value C.4 Ho is rejected in favour of HI at a significance level (a) if this t2 > tn-12(0(/2) denotes the upper (100a)th percentile of the t-distribution with n - 1 DF.4 We note that the variance ratio F with 1 DF, n - 1 has the same frequency distributions as t2 with n - 1 DF. Equi- valently, the F-test can be used, instead of t2, to test &. Deviations, Di - p ~ , are assumed to be independent and normally distributed with a population mean of zero. These deviations reflect random variations of measurements, free from systematic laboratory effects (laboratory bias), and the test on the mean difference remains fairly precise.3 The error in the paired design is a measure of the failure of the pair differences to be identical with the mean difference. These deviations can be analysed for purposes such as ranking of laboratories and identification of laboratories that deviate markedly from the other laboratories.Barnett and Lewis5 point out: ( a ) a simple visual inspection of a data set is unlikely to identify an outlier; and ( b ) an outlier is revealed only when the parameters of the postulated probability model are fitted to the data set and the deviations of the observed responses from the fitted values analysed in terms of the variational properties of a random sample generated by the model. The above presentation is concerned with univariate analysis, i.e., where a data set consists of only one response variable X , e.g., recovery of one CB.Data reported in this study comprise 3 (or p ) responses (XK) where K = 1 for CB IUPAC No. 52, = 2 for CB IUPAC No. 153 and = 3 for CB IUPAC No. 101. It is also noted that measurements of the three CBs resulted from a single analysis of the oil in each case and are, generally, mutually correlated. These will generate p paired difference random variables (aK). Denoting the obser- vation in laboratory i for replicate j on response variable K by XijK, measurements on these difference variables become We denote the Pearson coefficient of correlation between two difference variables, 6, and 6, by ruv; u,v = 1, . . ., p . When two or more of these p variables are mutually correlated, multivariate analysis, which analyses p variables jointly, should be employed rather than p separate univariate analyses.4.69 The following is noted regarding the significance of the correlations.7J5 The joint distribution of the sample correlations r,,; u,v = 1, .. ., p , required for simultaneous inferences about their parameters, is not available in a closed and practical form. Multivariate normality of the distribution is realized only for large samples. SYSTAT15 will test the null hypothesis that the population correlation matrix is an identity, i. e . , the global hypothesis concerning the significance of all correlations in the matrix, by Bartlett x2 and give the matrix of probabilities associated with individual correlation coefficients. However, these probabilities do not reflect the number of correlations that are under consideration here. The Bonferroni criterion provides protection for mul- tiple tests by identifying correlations as significant with the assurance that the family comparison error rate will not exceed the critical value chosen by the user.The scope of the procedure is restricted because the multiple tests and intervals for the parameters offered by it are conservative. Bonferroni- adjusted probabilities are also provided in SYSTAT.15 A natural generalization of the univariate squared distance (t2) of eqn. (4) is its multivariate anal~gue,~ Hotelling’s 72. In brief, the multivariate extension is as follows: The difference variable (6) for a CB is replaced by a vector (tj) of difference variables al, . . . , 6p for p CBs. The multivariate analogue of the variance of one difference variable of the univariate procedure will be a p x p E matrix of variances 0,“ (u = v) and1090 ANALYST, JULY 1992, VOL.117 covariances (u # v ) ; u, v = 1, . . . , p of p difference variables. We denote a vector by an underscored letter and a matrix by a letter in bold print. Prime denotes a transposed vector or matrix. An unprimed vector denotes a column vector. Let be a random vector with p components, Dil, . . ., Dip of measurements on difference variables. The p-variate normal distribution of QIi is Np ( F ~ , Z) with parameters ~0 and Z for the mean vector and the variance4ovariance matrix, respect- ively. A sample of n independent random vectors (Qi, i = 1, . . ., n) is analysed for inferences about the vector of mean differences based on the P-statistic.To test the null hypoth- esis, & : = C against HI : F D # C (where Cis a null vector Q or a vector of specified values), P is estimated as T z = ( Q - L3 where @ and S are the estimated (from the data) mean vector and variance-covariance matrix, respectively, of Qi,S/n is the estimated variance- covariance matrix of the vector Q of mean differences and (S/n)-l is its Special tables for T2 percentage points are not required as we can test its significance by Fp,n-p which is the random variable with an F-distribution with p DF, n - p , in the following manner: reject Ho if the observed 7’2 > [(n - l ) p / ( n - p)]Fp,,.-p(a) where Fp,,-p(a) is the upper (100a)th percentile of the F-distribution.4.7 A test of significance should preferably be put in the form of a confidence interval statement.3 In the T2 procedure: ( i ) a 100(1 - a)”/.confidence region for the meag of a p-variate normal distribution is the set of all values of 8 for which n(B - g)’S-l(D - 5) d [(n - l ) p / ( n - p)]Fp,,-p(a) (6) and (ii) lOO(1 - a)% simultaneous cpfidence intervals (CIS) for the individual mean differences ( 6 K ) are given by where .sik is the Kth element, i.e., mean of the Kth difference variable, of 0.497 The following are noted about confidence interval statements.4.7 (1). It is possible that Ho: pD = Q is rejected by the P procedure, and yet each of thep CIS of eqn. (7) includes zero (thereby showing that the mean difference is not significant for any individual CB). This can happen easily when variables are correlated.T2 will reject Ho if at least one of the several possible linear combinations, is significant. The CIS of eqn. (7) are only p particular combinations, viz., those corresponding to choices where one WK = 1 and every other wK is zero. The Bonferroni approach will yield narrower (and thus, more precise) CIS than will the simultaneous 7’2-intervals when the number of linear combi- nations is small. By the Bonferroni method, CIS with an over-all confidence level bl - a for the p components of individual mean differences ?jK are given by eqn. (7) following the replacement of the coefficient of (SDt/n)t in the equation by t(d2p) with n - 1 DF (2) CIS of linear combinations estimated by T2 are ideal for ‘data snooping’, i.e., we can choose values of wK of the linear combinations based upon examination of the data without changing the confidence coefficient, 1 - a.I f n and n - p are large, T2 will behave approximately like a chi-square random variable, x Z p , with p DF, even if the underlying population of differences is not normally distri- buted.4 When the population is normal, this approximation is acquired rapidly.4 For all practical purposes, the distributions encountered in chemical analysis conform to the normal distribution.3 The following are noted4.5.12 pertaining to the checks for multivariate normality and for possible multivariate outliers. (1) Unlike an outlier in a unidimensional data set, an outlying observation in a sample of multivariate data does not have a simple manifestation as an observation which ‘sticks out at the end’ of the sample as the sample has no ‘end’.Vector Qi for laboratory i might be an outlier for various reasons, e.g., because of correlation distortion or a gross error in one of its components or systematic mild errors in all its components. (2) The multivariate case is too complex to provide a unique, unambiguous form of total ordering for data and to express extremeness of observations. (3) A reduced sub-ordering, i.e., less than total ordering, principle employs the distance measure to represent multi- variate observations. Squared generalized distances for indi- vidual laboratories from their mean vector are given4 by (9) (4) For the multivariate normal model, this distance measure has broad statistical support and practical appeal in terms of constant probability density contours.(5) Employing the generalized distances (d:) as approxi- mately distributed as x2, is useful to check for multivariate normality and for possible suspect multivariate outliers. Johnson and Wichern4 state, ‘Although these distances are not independent or exactly chi-square distributed, it is helpful to plot them as if they were’. When estimated values of p-vector of means and of p x p variance-covariance matrix are used instead of their true values, this procedure ‘retains a measure of informal propriety and appeal’.5 Multivariate normality of difference variables is assessed as follows? (a) for observations generated from a three-variate normal distribu- tion, we would expect about 95% of these di2 values to be S7.815, which is the critical value of x; at the 5% probability level; and ( b ) a chi-square plot of the ordered distances (abscissa) and the corresponding chi-square percentiles (ordi- nate) would yield a straight line for a normal distribution.(6) These notes indicate that the x2 approximation of T2 can be meaningfully employed for comparing laboratory per- formance. The chi-square procedure has the advantages of familiarity, simplicity and availability of tables of probabil- ities. Laboratory ranking and distance will not change with use of 72 and x2 procedures, only its outlying status. The 7’2 procedure adds nothing of practical importance to the findings based on the x2 procedure. Here, analysis for outliers was carried out to note a laboratory’s extremity, not to discard it.This is because, although a laboratory might be extreme, there is no reason to consider it either as ‘discordant’5 or as belonging to a different population. Paired Means With respect to the univariate analysis of a paired data set, Youden and Steiner2 and Wernimont3 note that the concept of blocks is very general and that a paired experimental design is a special case of a randomized complete blocks design with two treatments. Here, laboratories are considered as blocks. For the two criteria of classification, laboratories and treat- ments, the model for ANOVA without replication can be written as Xij = p + Ai + Bj + eij, i = 1, . . ., n a n d j = 1 , 2 (10) where p is the over-all mean, Ai (= pi - p) is the ith laboratory effect (bias), and Bj is the jth treatment effect.The null hypothesis Ho : A l = . . . = A , = 0, (that the laboratory means are all equal, or, equivalently, the hypothesis of no laboratory effects) is tested by comparing the laboratory mean square (MSA) with the error (discrepancy) mean square (MSE). If the ratio MSA : MSE d F, - l , n - (a), Ho is accepted. Rejec- tion of & will lead to the conclusion that not all laboratory means are equal, showing the existence of systematic labora- tory effects. Two-way ANOVA also provides for comparisons of treatment means. However, tests for these comparisons are also provided in the Paired Differences section by defining the null hypothesis in terms of Di values. In order to judge theANALYST, JULY 1992, VOL. 117 1091 over-all difference between processes, the procedure that uses differences between results within blocks is the equivalent of that which uses the interaction between blocks and treat- ments.3 The need for multivariate analysis when p variables are generally mutually correlated has been demonstrated.4,”9 In identifying these correlations, the MSE for model (10) provides an unbiased estimate of the error variance, 02, for each determination.4 The variance of the differences, Di [eqn. (2)], is an estimate of 202, as these differences include random errors of both determinations.2 As these two variances differ only by a constant, i.e., 2, correlation matrices of e values in the two-way multivariate classification of CBs will be the same as those of D values.For the p = 3 response variables, X K , where K = 1 for IUPAC No.52, = 2 for IUPAC No. 153, and = 3 for IUPAC No. 101, we denote the observation in laboratory i by treatment j on response variable K by Xi_.. The observed megn vgctor for_laboratory i is denoted by xi with components Xi,, Xi2 and X3 for means of individual CBs, and the over-all average observation vector by X. A two-way MANOVA will test the null hypothesis (H,) that laboratory mean vectors are all equal against HI that at least two mean vectors are not equal. For a vector response consisting of p components, the generalization of the ANOVA model (10) will be where vectors are all p-dimensional. Rejection of & is followed by an examination of the pattern of variations of laboratory mean vectors.Each laboratory can have its own systematic error, which will offset its result from the correct result.2 At the univariate level, a specific rational way to compare laboratory means based on the t-statistic has been recommended.3 For laboratory i and response variable XK, this t is given3 as: It measures the distance of the mean Z K , K = 1, . . ., p for laboratory i from the over-all mean xK in distance units expressed in terms of the standard error of the mean, Sx,. In using this to rank laboratories and to identify possible outliers, we note: ( i ) MSE provides an unbiased estimate of error variance for each determination, so that SyiK = (MSE/r)l where r is the number of deter-minations in laboratory i; (ii) considering that XK includes XiK, S x K could be adjusted by multiplying it by [(n - l)/n]i. Obviously, this adjustment can affect only the outlying status of a laboratory, not its rank; and (iii) again, the use of t2 provides an equivalent (to t) approach. The variable t2 is used as a measure of the squared distance of a laboratory mean at the univariate level .4 The multivariate analogue of the squared distance is the squared generalized distance, given by Hotelling’s 72,4 i.e., where S is the error variance-covariance matrix of the two-way MANOVA.Again, special tables of probabilities associated with 72 are not required as we can test its significance using F. The authors acknowledge the review and commentary on the manuscript received from J. M. Bewers, T. King and J. van der Meer. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 References Uthe, J. F., Musial, C. J . , and Misra, R. K.,J. Assoc. Off. Anal. Chem., 1988,71, 369. Youden, W. J . , and Steiner, E. H., Statistical Manual of the AOAC, Association of Official Analytical Chemists, Arlington, VA, 1975. Wernimont, G. T., Use of Statistics to Develop and Evaluate Analytical Methods, Association of Official Analytical Chemists, Arlington, VA, 1987. Johnson, R. A . , and Wichern, D. W., Applied Multivariate Statistical Analysis, Prentice Hall, Englewood Cliffs, NJ, 2nd edn., 1988. Barnett, V., and Lewis, T., Outliers in Statistical Data, Wiley, New York, 2nd edn., 1984. Bray, J. H., and Maxwell, S. E . , Multivariate Analysis of Variance, Sage University Papers Series on Quantitative Appli- cations in the Social Sciences 07-054, Beverly Hills, CA, 1986. Morrison, D. F., Multivariate Statistical Methods, McGraw-Hill, New York, 2nd edn., 1976. Harris, R. A . , A Primer of Multivariate Statistics, Academic Press, New York. 1975. Kshirsagar, A . M. , Multivariate Analysis, Marcel Dekker, New York, 1972. Reynolds, L. M., and Cooper, T . , Water Quality Parameters ASTM STP 573, American Society for Testing and Materials, Philadelphia, PA, 1975. Pimentel, R. A . , Morphometrics, Kendall-Hunt, Dubuque, IA, 1979. Gnanadesikan, R., and Kettenring, J . R., Biometrics, 1972,28, 81. Snedecor, G. W., and Cachran, W. G . , Statistical Methods, Iowa State University Press, Ames, IA, 7th edn., 1980. Steel, R. G. D., and Torrie, J. H . , Principles and Procedures of Statistics, McGraw-Hill, New York, 1960. Wilkinson, L., SYSTAT: The System for Statistics, SYSTAT, Evanston, IL, 1988. Paper 1104279C Received August 15, 1991 Accepted January 28, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701085
出版商:RSC
年代:1992
数据来源: RSC
|
7. |
Edible fats and oils reference materials for sterols analysis with particular attention to cholesterol. Part 1. Investigation of some analytical aspects by experienced laboratories |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1093-1097
Georges Lognay,
Preview
|
PDF (617KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1093 Edible Fats and Oils Reference Materials for Sterols Analysis With Particular Attention to Cholesterol Part 1 Investigation of Some Analytical Aspects by Experienced Laboratories Georges Lognay and Michel Severin Faculty of Agricultural Sciences, Department of General and Organic Chemistry, Belgium Achim Boenke and Peter J. Wagstaffe Community Bureau of Reference, DGXII, Commission of European Communities, Brussels, Be Ig iu m B-5030 Gembloux, 200 Rue de la Loi, B-1049 The work reported here is integrated into a programme organized by the Community Bureau of Reference with the aim of developing edible oil reference materials (RMs) certified for cholesterol content. One vegetable oil (RM 162, a blend of soya and maize oils) and two animal fats (RM 163, a blend of pig and beef fats, and RM 164, an anhydrous milk fat) possessing, respectively, low, medium and high cholesterol contents were chosen for this purpose.The present paper summarizes the analytical conclusions resulting from three interlaboratory trials carried out t o identify and correct for the major sources of random and systematic errors linked t o the protocol and t o the gas-liquid chromatographic analysis. Several improvements t o the methodology, and recommendations, have been proposed for the determination of individual sterols within the certification exercise. The latter will be reported elsewhere. Keywords : Cholesterol determination; edible oils; reference mate rial; collaborative study Edible vegetable oils and animal fats consist mainly of triglycerides, which are responsible for their nutritional and physico-chemical properties.Some other minor lipids, such as sterols and tocopherols, attract the interest of food analysts because they have particular nutritional properties and their composition provides a ‘fingerprint’ of a lipid material. The qualitative and quantitative analysis of sterols is essential for nutrition studies and for identification of lipid mixtures when adulteration is suspected. Reliable determina- tion of cholesterol is particularly important both for food labelling purposes and in dietary studies. A preliminary intercomparison of methods for cholesterol organized within the Community Bureau of Reference (BCR) programme had revealed poor agreement and it was concluded by the participating specialists that there was a need for a series of non-spiked lipid reference materials (RMs) with certified cholesterol contents.’ Several analytical methods, including gravimetric, enzymic and spectrophotometric techniques, have been developed to determine the total sterol content in many foods and biological samples.* Currently, high-performance liquid chroma- tography (HPLC) or gas-liquid chromatography (GLC) is used for the establishment of a sterol profile.The separation of A5 and A7 classes is easily achievable on the analytical and semi-preparative scales by normal-phase HPLC.3 However, reversed-phase columns lead to a more complete resolution of molecular species.4-6 In spite of its simplicity, the technique suffers from insufficient selectivity and sensitivity (the com- mon sterol possesses only isolated double bonds, which exhibit low molar absorptivity, their absorption maxima lying between 200 and 210 nm).Except for the determination of ergosterol, a molecule with a typical cis-diene structure (I,,, = 282 nm), the routine application of HPLC to sterol analysis is still limited.7-9 Capillary GLC is recognized as the method of choice for the qualitative and quantitative analysis of sterols in complex matrices such as fats and oils. The sterols are often present at low concentrations in such samples (from 80 to 1200 mg per 100 g of lipid). Extraction and purification prior to GLC analysis is therefore necessary. The procedure generally adopted can be summarized as follows: the sample is saponified with alcoholic potassium hydroxide, and the unsaponifiable matter (USM) is extracted into diethyl ether or another suitable solvent such as hexane.After washing of the crude extract, the solvent is evaporated and the USM is fractionated into its components by thin-layer chroma- tography (TLC). The sterols are then re-extracted from the gel and their constituents are finally separated by GLC, either underivatized or as their trimethylsilyl ethers. Standardized or official methods based on this principle have been adopted.10-12 On the other hand, a method for the rapid isolation of sterols, involving chromatography on aluminium oxide columns, was developed by Hornberg13 and has recently been standardized. 14 The time-consuming, multi-step proto- col of conventional methods has several potential sources of analytical error due, not only to sample handling, but also to the chromatographic resolution for final quantification.This paper summarizes the analytical conclusions resulting from the initial intercomparison of methods that were carried out to identify and correct for the major sources of random and systematic error. Details of the certification of the content of cholesterol and other sterols in three edible lipid RMs (RM 162, a blend of soya and maize oils; RM 163, a blend of pig and beef fats; and RM 164, an anhydrous milk fat) will be reported elsewhere. Experimental A general scheme of the method used by the participating laboratories for the GLC determination of sterols in fatty material is presented in Table 1.The recovery experiments and the improvement of chromatographic conditions were carried out on RM 162, because this RM possesses a more complex sterolic composition than the two other animal fats (RM 163 and RM 164) selected. In addition, the cholesterol content of RM 162 is relatively low and, therefore, well suited to spiking and recovery experiments.1094 ANALYST, JULY 1992, VOL. 117 Cholesterol was the only sterol for which a standard of high purity was available, and in view of its particular importance for nutritional purposes, the calibration and recovery experi- ments were focused on this molecule. High-purity cholesterol (Sigma; ref. (28667) and betulin (Sigma; B9757) samples, each from the same batch, were sent to the participants in the intercomparison studies to eliminate any variability arising from differences in purity of the compounds obtained from different sources.Intercomparisons First intercomparison The first intercomparison exercise undertaken by seven laboratories, using the method that they considered to be the most accurate, had led to the following observations: (1) almost regardless of the method used, the preliminary tests had highlighted poor agreement between laboratories with respect to the absolute values;' (2) the internal standards (ISTDs) were not added sufficiently early in the procedure to control losses at every stage; and (3) examination of RM 162 by gas chromatography-mass spectrometry (GC-MS) con- firmed that TLC clean-up was necessary to remove interfering compounds such as a-tocopherol (co-eluted with cholesterol) or triterpene alcohols and 4-methylsterols, which could interfere with several phytosterols.15 A TLC step was, therefore, strongly recommended for the analysis of vegetable oils.This conclusion is in line with Homberg and Bielefeld's observations. 16 Second intercomparison The second study, in which seven laboratories participated, was designed to eliminate the more serious of the suspected sources of errors appearing during the extraction and the Table 1 General scheme for the determination of sterols in fatty materials Saponification of the fat Extraction of unsaponifiable (USM) into diethyl ether Water washings of the extract Fractionation of USM (preparative TLC) Isolation of sterols from the gel Derivatization (TMS) Capillary GLC analysis J..1 1 1 .1 J. calibration of the cholesterol determination. The protocol to be followed consisted of three distinct, but related, stages: (1) the determination of response factors for cholesterol trimethylsilyl ether (betulin = 1 .OO) at three concentration ratios (0.5, 1 and 2); (2) a check on the entire sterol determination method by analysis of prepared solutions of cholesterol and ISTDs, taking these solutions into the whole procedure, i.e., 'saponification', extraction, washings, TLC and GLC; and (3) the determination of recovery of cholesterol added to RM 162 unsaponifiable. In the second and third steps of analysis, cholesterol concentrations were calculated from the response factors obtained in step 1. The choice of the methodology for the treatment of the samples was left to the participant.The conditions, summarized in Tables 2 and 3, varied widely. Joint evaluation of the results at a meeting of participants, summarized in Table 4, led to the following observations and conclusions. (1) All the participants used betulin as ISTD. (2) The study of response factors demonstrated adequate linearity over the range of interest for cholesterol in oils and fats. Most laboratories found results very close to unity [mean of means k 1 standard deviation (SD) = 1.028 k 0.056, p = 211, implying near-similarity in GLC and flame-ionization detec- tion behaviour of cholesterol and betulin trimethylsilyl ethers. The results of Laboratory 2 were around 1.1 and although no reason was found for this, it was a consistent value in that laboratory.(3) The recovery of cholesterol from simple solutions was quantitative (mean of means k 1 SD = 100 k 5%, p = 21). Extreme minimum and maximum values were 90% (Laboratory 3) and 108% (Laboratory 19). (4) In general, recoveries close to 100% (mean of means L 1 SD = 100 L 4%, p = 21) were achieved, implying good control of errors following the saponification-extraction process. The lower results from Laboratory 3 suggested a systematic bias; on the other hand, the higher values of Laboratory 19 were explained by a partial TLC separation of the betulin and cholesterol bands, which increased the risk of losses in the re-extraction process. The agreement for low cholesterol contents of unspiked RM 162 remained very poor.Indeed, the relative standard deviation (RSD) within laboratories (repeatability) varied from 2 to 31%, and the RSD between laboratories (reproduci- bility) was 60%. This was probably because the method was calibrated for high levels of sterols. The high results of Laboratory 2 (14 _+ 1 mg per 100 g) were excluded for the statistical evaluation. In view of the good recoveries of cholesterol after saponifi- cation (Table 4), it was concluded that a substantial part of the analytical errors arose during the initial saponification/extrac- tion stages. Further work was, therefore, concentrated on this point. Table 2 Second interlaboratory trials. Analytical conditions for saponification and extraction Laboratory Saponification conditions Extraction of the USM code for 5 g of sample into diethyl ether Eluent for TLC clean-up* 2 t 3 Not specified 1 mol dm-3 ethanolic KOH (50 cm?), 1 h at 80 "C 5 7 11 1st 19 50 cm3 ethanol + 10 cm3 KOH, 10% for 30 min 2 mol dm-3 methanolic KOH (20 cm3), 1 h at 80 "C 2 mol dm-3 methanolic KOH (100cm3), 1hat90"C 1 mol dm-3 ethanolic KOH (50 cm3) for 1 h 20 cm3 ethanol + 5 cm3 KOH (60%), 15 rnin at 80 "C 3 X 100 cm3 (15 rnin each) Chloroform-diethyl ether Not specified Hexane-diethyl ether-acetic acid 1 X 250 cm3 (2 min) 1 X 50 cm3 (30 min) 1 x 100 cm3, 2 x 50 cm3 3 X 100 cm3 (1 rnin each) Chloroform 1 X 100 cm3, 2 X 50 cm3 (90 + 10, v/v) (60 + 40 + 1, v/v/v) Chloroform Chloroform-diethyl ether- Hexane-diethyl ether ammonia (90 + 10 + 0.5, v/v/v) (70 + 30, v/v) (1 min each) Chloroform-diethyl ether- (1 min each) ammonia (90 + 10 + 0.5, v/v/v) Recovery of the sterols from the silica gel Hot chloroform (X 3) Diethyl ether (20 cm3 for 2 min) Diethyl ether Diethyl ether (x2) Chloroform ( x 1) Chloroform-diethyl ether Diethyl ether (7+3,v/v)(x5) Diethyl ether (5 cm3) * All TLC was performed on silica gel Si 60. t Standardized methods were used by Laboratory 2 (AFNOR) and Laboratory 15 (IUPAC).ANALYST, JULY 1992, VOL.117 1095 ~~~~~ Table 3 Second interlaboratory trials. Analytical conditions for GLC Laboratory Injector code Derivatization* Column Temperature OV-1701t (30m x 0.33 mm i.d., dfS = 0.3 pm) Isothermal (280°C) CPSIL-19$ (25 m x 0.2 mm i.d., df = 0.22 pm) Isothermal (250 "C) CPSIL-19 (25 m x 0.32 mm i.d., df = 0.22 pm) Isothermal (260 "C) 2 3 5 7 Pyridine-BSTFA + 1% SE-521(25 m x 0.32 mm i.d., df = 0.15 pm) 70-220°C (20°C min-1) On column BSTFA-TMCS (80 + 20), Trisil, 10 min at room Pyridine-HMDS-TMCS Moving needle Split Split 30 min at 60 "C temperature (9 + 3 + 1) TMCS (1 + 1) TMCS (1 + 1) 22@-300 "C (6.5 "C min-1) 11 Pyridine-BSTFA + 1% SE-52 (25 m x 0.32mm i.d., df = 0.15 pm) Isothermal (260°C) Split 15 Split 19 Pyridine-MSTFA (5 + l ) , OV-1711(25 m X 0.32 mm i.d., df = 0.2 pm) 110-260°C (30 "C min-1) On-column Pyridine-HMDS-TMCS CPSIL-19 (25 m x 0.32 mm i.d., df = 0.22 pm) Isothermal (280 "C) (9+3+1) 90 min at 80 "C * BSTFA = N , 0-Bis(trimethylsily1)trifluoroacetamide; TMCS = trimethylchlorosilane; HMDS = hexamethyldisilazane; and MSTFA = t OV-1701: 5% cyanopropyl-7% phenyl-88% methyl-siloxane.$ df = Film thickness. $ CPSIL-19: 85% dimethyl-7% cyanopropyl-7% phenyl-1% vinyl-siloxane. 1 SE-52: 5% phenyl-95% methyl-siloxane. 11 OV-17: 50% phenyl-50% methyl-siloxane. N-methyl-N-trimethylsilyltrifluoroacetamide. ~~ Table 4 Response factors and recovery check of cholesterol (second study) Laboratory 2 3 n t 5 5 Low cholesterol level, C : B = 5* Mean 1.16 0.975 SD 0.02 0.023 Medium cholesterol level, C : B = 1* Mean 1.16 0.994 SD 0.02 0.03 High cholesterol level, C : B = 2* Mean 1.15 0.983 SD 0.02 0.014 n t 5 4 Low cholesterol level, C : B = 0.5* Mean 102.6 91.3 SD 1 205 Medium cholesterol level, C : B = 1* Mean 104.7 89.7 SD 1.5 2 High cholesterol level, C : B = 2* Mean 100.3 92.1 SD 1.5 4.4 C. Recovery of cholesterol added to RM 162 unsaponifiable (results in YO)- A .Response factors- B. Procedure check; recovery from simple solutions (results in %)- Low level of cholesterol added (5200 mg per 100 g of oil) n = 3t Recovery 97.8 95.2 SD 4.6 4.2 Recovery 106 94.2 SD 1.5 2.9 Unspiked RM 162 cholesterol Mean 13.8$ 9.6 (mg per l00g) n = 3 SD 1 1.3 Medium level of cholesterol added (400 mg per 100 g of oil) n = 3t * C : B = cholesterol : betulin concentration ratio. t n = number of replicates in each laboratory. t. Result excluded for statistical evaluation. 5 3 1.01 0.03 0.997 0.03 1.05 0.02 3 97.3 1.5 97.3 5 97.7 1.5 98 3.3 100.4 1.1 1.9 0.2 7 5 1.01 0.01 1.02 0.02 0.98 0.02 5 101.3 1.5 98.5 1.2 99.6 2 99.8 1.8 98.9 4.1 4.8 0.2 11 3 0.995 0.012 1.006 0.016 0.998 0.014 3 95 2.4 99.8 0.8 101.3 3.3 101.2 1.5 100.2 1.1 4.6 0.1 15 5 1.04 0.01 1.05 0.02 1.01 0.02 5 107.5 8.3 101.5 2.1 99.7 0.5 96.6 2.2 101.5 2.9 5 1.3 19 4 1.012 0.004 1.01 0.004 0.992 0.006 4 108.9 5.2 108.1 4.9 106 7.4 104 1 106 1 1.5 0.4 Third and fourth intercomparisons Participants agreed to apply a common saponificatiodextrac- tion procedure for the third intercomparison.The procedure combined the conditions used in each laboratory and although not necessarily convenient for routine analysis, it was ex- pected to ensure fully quantitative saponification and extrac- tion as requested for eventual RM certification purposes. The procedure was as follows: saponification with 100 cm3 of 2 mol dm-3 methanolic KOH solution for 1 h at 75-80 "C; extraction of the USM with 3 x 100 cm3 of diethyl ether; and washing of the ethereal extract with 3 x 40 cm3 of water.Before use, the method was validated by means of radiolabelled sterols, [3H]cholesterol and [3H]cholesteryl oleate, and radiometric measurements,17 which demonstrated that: (1) the added cholesterol (free or as oleate) was quantitatively recovered regardless of the material tested (sunflower oil or butter oil); (2) cholesteryl oleate was totally saponified; (3) losses by washing did not exceed 1%; and (4) there was no detectable amount of sterol degradation products (TLC plus radiodensitometric scanning). In order to complete the study, six participants studied the recovery of cholesterol when the common protocol was applied. The response factors for silylated cholesterol and betulin standard mixtures were first measured directly without TLC.An aliquot was subjected to the saponification/TLC steps, and1096 ANALYST, JULY 1992, VOL. 117 ~ ~ ~~~ Table 5 Third interlaboratory trials. Response factors for cholesterol to betulin, with and without saponificatiodTLC steps Laboratory Response factors code Mean SD t-Test* 2At 2BT 3A 3B 5A 5B 6A 6B 7A 7B 11A 11B 1.020 1.060 1.030 1.010 1.039 1.019 1.045 1.098 0.980 0.990 0.993 0.970 1.020 1.030 1.044 1.040 1.046 0.944 1.041 1.074 0.990 0.990 0.999 0.973 1.040 1.030 1.046 1.060 1.085 0.828 1.066 1.081 0.990 1.010 0.996 0.967 1.020 1.060 1.023 1.030 - - 1.038 1.010 0.990 0.993 - - * Statistical evaluation: t-Test on paired values (95% confidence limits). t A; Without TLC; B; with TLC. $ NS: No significant difference of the means.0 S: Significant difference of the means. 1.010 1.020 1.080 1.050 - 1.036 - 1.038 - 1.057 - 0.930 1.058 1.049 - 1.084 1.OOO 0.994 1.010 0.998 - 0.995 - 0.970 0.010 NSS 0.020 0.010 NS 0.021 0.025 NS 0.096 0.012 S § 0.013 0.011 NS 0.011 0.003 S 0.030 2 0 15 30 Time/mi n Fig. 1 Chromato ram of the sterols of RM 162. Column, CPSIL- 19CB (Chrompackf, 25 m x 0.32 mm i.d., 0.2 pm film thickness. Cold 'on-column' injector. Temperature programme: 60-285 "C at 30 "C min-1, 40 kPa He, FID detector at 300 "C. 1, Cholesterol; 2, campesterol; 3, campestanol; 4, stigmasterol; 5 , stigmastanol + fucosterol; 6, 6-sitosterol; 7, A5-avenasterol; 8, A7-stigmasterol; 9, A7-avenasterol; and 10, betulin (ISTD) the response factors were measured again. A comparison of the two sets of results presented in Table 5 shows that the calculated response factors, with or without treatment, are virtually identical except for those of Laboratories 4 and 14 for which a significant difference (95% confidence limits) was detected.It was, therefore, necessary for the participants to confirm that the saponification/TLC step had no influence on the calibration procedure. If the results differed by more than +5%, the cause had to be identified and corrected before continuing the study. Conclusions The different interlaboratory trials led to some improvement of the methodology, and several recommendations were proposed for the determination of the individual sterol content in fats and oils within the certification exercise. Saponification. The rigorous saponification procedure dis- cussed above was mandatory.Internal standard. In spite of its particular structure (a lupane triterpenoi'd bearing two hydroxy groups), necessi- tating a careful assessment of the derivatization procedure, 9 8 7 - ._ 6 , " 5 0 4 0 3 s r 2 2 i 1 0) i E f 6 ' 3 ,I 2 - i . 0 2 3 4 5 6 7 8 9 1 1 1 5 1 9 Laboratory code Fig. 2 Comparison of the results obtained in three preliminary intercom arisons and the final certification of the cholesterol content of RM l&. (a) 1st intercomparison, 2.50 k 1.13 mg per 100 g; (b) 2nd intercomparison, 4.85 * 2.90 m er 100 g; (c) 3rd intercom arison, 2.78 k 0.91 mg per 100 g; and tJ certification, 2.49 _+ 0 . d m g per 100 gANALYST, JULY 1992, VOL. 117 1097 the betulin is virtually co-eluted with the sterols on the silica gel plates and does not interfere with any other molecule during the GLC run.On the other hand, cholestane, another potent ISTD, suffers two major drawbacks. ( a ) It is not eluted with the other sterols and, therefore, must be added after the TLC run. For that reason, it does not compensate for any loss of sterols during the extraction of USM and the re-extraction of the sterols from the gel. ( b ) The cholestane is not silylated. According to Homberg and Bielefeld,l6 cholestane is particularly useful for the analysis of animal fats where a TLC fractionation is not strictly obligatory. For sterol determination, it is recommended to add the ISTD not later than the saponification stage and at a level that is appropriate, i.e., producing a peak height similar to that of the compound to be determined.In order to guard against undetected drift during a working day, the calibration and sample solutions should be injected alternately. Derivatization . All the chromatograms furnished by the participating laboratories revealed that the different con- ditions of derivatization used throughout the studies led to complete silylation of both cholesterol and of betulin and that this was not a substantial source of error. GLC stationary phases. The use of chromatographic col- umns with high efficiency and selectivity is recommended in order to obtain the best achievable separation of sterols. Stationary phases of medium polarity, such as poly- cyanopropylphenyl(methy1)siloxane (OV-1701, CP-SIL19 CB, DB-1701, etc.) (see Table 3 ) , fulfil these requirements.p-Sitosterol and A5-avenasterol, unresolved on some apolar columns, were better separated on OV-1701 or equivalent phases. Under such conditions, additional peaks were distin- guishable in the tails of the peaks for campesterol and p-sitosterol (Fig. 1). As shown in Fig. 2, there was a sharp decrease in the standard deviation of the mean when these recommendations were followed and it was believed that the accuracy of the results for low cholesterol levels in soya-maize oil (RM 162) significantly (95% confidence limits) improved during the successive interlaboratory trials. The certification of the cholesterol level in the three BCR RMs will be reported elsewhere. This paper places on record the collaborative work of laboratories from several European countries and to whom the authors give full acknowledgement. The following special- ists contributed to the preliminary studies andor to the final certification exercise.G. Contarini and P. M. Toppino, Istituto Sperimentale Lattiero Caseario, Lodi, Milano, Italy. F. Mordret and F. Lacoste, Institut des Corps Gras, Pessac, France. W. D. Pocklington, J. Pearse and M. Burn, Laboratory of the Government Chemist, Teddington, UK. S. P. Kochar and R. Griffith, Leatherhead Food Research Association, Leatherhead, UK. V. Eckelmans, Ministerie van Economische Zaken, Brussels, I Belgium. S. L. Reynolds, Ministry of Agriculture, Fisheries and Food, London, UK. T. Leth, National Food Institute, Soborg, Denmark. B. Muuse and J. de Jong, Rijks-Kwaliteitsinstituut voor Land en S.Mannino, Universita degli Studi di Milano, Milano, Italy. Tuinbouwprodukten, Wageningen, The Netherlands. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 References Pocklington, W. D., Frezenius’ 2. Anal. Chem., 1988,332,674. Gorog, S. , Quantitative Analysis of Steroids, Elsevier, Amster- dam, 1983, pp. 247-289. Mordret, F., Ajana, H., and Gauchet, C., Rev. Fr. Corps Gras, 1985, 32, 308. Morisaki, M., and Ikekawa, N., Chem. Pharm. Bull., 1984,32, 865. Kesselmeier, J., Eichenberger, W. , and Urban, D., Physiol. Plantarum, 1987, 70, 610. Perrin, J. L., and Raoux, R., Rev. Fr. Corps Gras, 1988, 35, 328. Schwadorf, K., and Muller, H., J. Assoc. Off. Anal. Chem., 1989, 72,457. Cahagnier, B., Ind. Agroaliment., 1988, 1J2, 5. Osswald, W., Holl, W., and Elstner, E., Z . Naturforsch., Teif C, 1986,41, 542. IUPAC, Standard Methods for the Analysis of Oils, Fats and Derivatives, Method 2,403, Blackwell, Oxford, UK, 7th (revised and enlarged) edn., 1987. Determinazione del contenuto di steroli mediante gascromato- grafia con colonna capillare. Norme Grassi e derivati. Metodo NGD C72. Rev. Ital. Sost. Grasse, 1987, C4, 553. Corps Gras d’origine animale et vCgCtale. Dosage des faibles teneurs en CholestCrol. Norme Francaise, 1983, NF T 60-249. Homberg, E., Fat Sci. Technol., 1987,89,215. Arens, M., Fiebig, H., and Homberg, E., Fat Sci. Technol., 1990,92, 189. Homberg, E., and Bielefeld, B., Fat Sci. Technol., 1990, 92, 478. Homberg, E., and Bielefeld, B., Fat Sci. Technol., 1987, 89, 255. Lognay, G., Dreze, P., Wagstaffe, P. J., Marlier, M., and Severin, M., Analyst, 1989, 114, 1287. Paper 1 lO6454A Received December 30, 1991 Accepted March 6, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701093
出版商:RSC
年代:1992
数据来源: RSC
|
8. |
Focal plane charge detector for use in mass spectrometry |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1099-1104
Keith Birkinshaw,
Preview
|
PDF (854KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1099 Focal Plane Charge Detector for Use in Mass Spectrometry Keith Birkinshaw Department of Physics, University College of Wales, Aberystwyth, Dyfed SY23 3BZ, UK Since the advent of the microchannel plate (MCP) electron multiplier, the design of photon, electron and ion focal plane detectors (FPDs) has been an active but specialized research area. Recent rapid advances in support of integrated circuit design have brought powerful high-technology tools into the reach of research groups requiring custom FPDs. Focal plane detectors with an MCP front end are used in mass spectrometry, ultraviolet spectrophotometry, electron energy loss spectroscopy, etc., and are outlined. The design of a working charge detector on silicon is described and the proposed design of a new high-resolution FPD based on this detector is presented, with an explanation of the trade-offs made between the many design variables. Keywords: Focal plane charge detector; mass spectrometry; silicon sensor The term ‘focal plane detector’ (FPD) implies the simul- taneous detection of a spatially resolved spectrum over a section of the focal plane of an instrument which is much wider than that sampled by a single slit.In a magnetic sector mass spectrometer, for example, an FPD would cover several mass units of the mass spectrum instead of a fraction of a unit. A crude FPD would be a photographic plate, which, after exposure, would be scanned with a microdensitometer to extract the spectrum. This is, of course, extremely inconve- nient with no time resolution and has a low sensitivity and dynamic range and a non-electrical output, but has the advantages of small size, high collection efficiency and low cost.The need for a high-resolution FPD to replace the single detector of a high-resolution magnetic sector mass spec- trometer has been evident for a long time. Likewise, the need in ion and electron scattering experiments is equally felt. The primary aims of the FPD designer are to design a detector with a high collection efficiency, high resolution and low cost (production and running costs) while equalling or bettering other important aspects of performance, including noise immunity and low power consumption, and hence mass spectra can be accumulated more rapidly and very small samples can be analysed accurately. There has been a gradual evolution of detectors of various types over the past two decades which has followed the evolution in technology.Only those detectors which consist of a microchannel plate (MCP) electron multiplier followed by an anode or array of anodes (with emphasis on one-dimensional arrays) will be considered. The information on the spatial distribution of ions, electrons or photons falling on the MCP is amplified and transmitted as pulses of electrons to underlying anodes. It is the development of the MCP that has allowed the advance of this type of detector design. As technology has advanced there has been a gradual reduction in the size and cost of detector electronics, which has permitted the high level of integration proposed here.Anode configurations of high resolution have not been a limitation ( e . g . , resistive strip,l one-dimensional array of metal strips,2J wire grids4), but total integration of anodes and associated electronics on a silicon chip, although an obvious step,5.6 has awaited a reduction in circuitry size. Richter and H o ~ have reviewed FPDs (position-sensitive detectors) and their relevance to time-resolved electron energy loss spectroscopy. This paper summarizes develop- ment to date of focal plane charge detectors (FPDs) and describes an FPD design undertaken at Aberystwyth. In spite of their high development costs, FPDs are cost effective in expensive projects such as satellite and space shuttle experiments. Indeed, the latter was the stimulus for much investment in detector design.2.4.R11 Technology similar to that needed for the above FPD design is also needed for nuclear particle detection5.12 and this has contributed to the knowledge and momentum in this area.Early FPDs had a high collection efficiency and high resolution, but only at low count rates (up to about 106 counts s-1) as the logging of one event must be completed before the next occurs. Later FPDs, with the benefits of new technology, integrated discrete detector sites and more of the associated electronics on single chips so that the simultaneous detection of particles at many sites and a high collection efficiency at high total count rates are possible. Two important distinctions between an FPD and the traditional defining slit/single detector should be noted.First, the resolution of the single slit is not influenced by the multiplier or detector but only by the slit. This is not the case for an FPD incorporating an MCP, where resolution is lost at the interface. Where a stack of two MCPs is used to amplify the signal further, resolution is also lost at the MCP-MCP interface. The electro-optical FPD discussed below has several interfaces with loss of resolution at each one. Second, variations in sensitivity across an array could occur owing to variations in the MCP sensitivity andor sensor sensitivity. Again, this problem does not arise for a single detector. Sensor design and MCP quality should minimize this problem. Remaining variations could be removed by calibration and software correction or by sampling each part of the spectrum by each detector, i.e., by scanning the spectrum across the detector array.The direct implication of the potentially very high data collection rate for a detector array is that data must be accumulated on-chip as they cannot be shifted over the bus at a sufficiently high rate and it is not feasible to have an external output pin for each detector of a high-resolution detector array with, say, 400 detectors cm-1. Hence it is necessary to have a dedicated counter associated with each charge sensor, but the higher the number of counter bits, the greater is the area of the chip and the lower the yield. There is, therefore, a trade-off between number of counter bits and yield. The Aberystwyth design involved a careful trade-off between many such parameters, some of which are explained below.The market for FPDs is relatively small and, therefore, the driving force for their development on silicon has not come from volume sales, where most of the design expertise has been traditionally concentrated. The need for FPDs is felt most in areas where there is little silicon design expertise and the driving force is a large increase in efficiency of very expensive apparatus. Because of the rapidly falling costs and increasing accessibility of design tools, custom FPD design is now within the range of lower cost projects in industry and research. The major problem is the lack of know how of users,1100 ANALYST, JULY 1992, VOL. 117 but those developing skills in this area will be poised to take advantage of the rapid advances in silicon technology which will result in lower cost and higher performance.This paper outlines the development of FPDs over the last 20 years and explains the choice of the key features in the proposed design of a new FPD at Aberystwyth. Types of Focal Plane Detector Several types of FPD are given in Table 1, each of which incorporates an MCP. The performance of each FPD is summarized and references to original work are given. Figures in the table are taken from the references. The figure for the resolution for the discrete anodes is the array length divided by the number of anodes. Resistive Strip This consists of a thin layer of resistive material (e.g., carbon) of typical length 2 cm and resistance 20 kS2 deposited on an insulating substrate with metallized ends.Electrical connec- tions are made to the ends. A pulse of electrons from the MCP falls on the resistive strip and travels to both ends. The pulses arriving at each end are measured (arrival time or charge) and this gives the position of the pulse to about 50 pm.13 The count rate is limited by the difficulty of distinguishing signals for more than one e k n t at a time. Wire Grids A one-dimensional array of thin wires placed underneath an MCP will collect an electron pulse on several wires and the centre of gravity of the pulse can be interpolated.14 Crossed grids have been used to locate an electron pulse in two dimensions.4 For an n x rn array of wires only n + rn amplifiers are needed, but the maximum count rate is again limited by the difficulty of resolving more than one pulse at a time.Mainly one-dimensional grids are considered here. A one-dimensional array of wires is a discrete anode array but is distinguished from those listed under Discrete Anodes by its usage. The grid wires (anodes) are used to measure the profile of a single electron pulse and hence calculate accurately the position of the pulse, but only at low particle flux. To achieve both high resolution and high collection efficiency at high count rates (see under Discrete Anodes) necessitates the logging of the position of arrival of many particles at the same time. The implication for the design is that many independent detectors of high spatial resolution with associated electronics are needed. In addition, the MCP must be positioned close to the anode array so that resolution is not lost by spreading of the MCP output pulse.Metal Anodes on Substrate This is electrically equivalent to the wire grid method but the detector anodes are deposited on an insulating substrate. Metal Apodes on MCP Anodes can be deposited directly on the back face of an MCP and the collected chargepulses measured. Normally the faces of the MCP would be uniformly metallized leaving channels open, but in this instance strips of metal are deposited on the rear surface of the MCP, blocking the channel exits. Coded Anodes Charge falling on, e.g., 1024 detector anodes is capacitatively coupled to ten underlying perpendicular tracks. The track widths are coded so that each of the 1024 detector anodes produces a different pattern of coupled signals on the underlying tracks.Wedge and Strip This device consists of only three interdigitated electrodes, each with its own charge-sensitive amplifier. The width of the electrodes varies in the x and y directions so that the relative signals give the x and y positions of a pulse of charge. Discrete Anodes These arrays consist of many self-contained detectors. It is possible to have many configurations of anode, e.g., annular, radial, but each has its own charge sensor and counter. It is distinguished from the wire grid by the higher density of anodes and electronics which can be achieved by current Table 1 Survey of FPDs Event logging capacity Method Single Resistive strip Wire grids Metal on substrate Metal on MCP Coded anodes Wedge and strip Multiple Discrete anodes Electro-optical [charged-coupled detector (CCD>I (Photodiode array) Anode sizelmm 22 x 12 25 dia.0.025 dia. 0.1 dia. 25 x 0.4 6.5 x 0.025 3 x 0.225 25 x 3 13 x 0.015 Interdigitated 1.3 x 0.025 3.18 x 0.25 2 x 0.3 4 x 0.015 7.5 x 0.8 Array lengt h/mm 22 25 4 [6 x 6 20 30 18 (30 dia.) 26 35 (2D)I 3.2 50 5 50 - No. of anodes 1 1 80 30 x 30 40 1024 40 7 1024 3 64 96 16 50 - Resolution/ CLm 300 50 50 10 300 - - 3000 25 50 50 520 310 25 lo00 - 256 photosites - - - - - - Ref. 1 13 14 4 9 8 15 9 11 10 2 16 3 17 18 19 20ANALYST, JULY 1992, VOL. 117 / Detector anode 1101 Pulse Switch - Pulse I m 1 fabrication technology, and instead of observing a charge envelope, the MCP is placed closer to the detector anodes so that the charge is distributed over a narrower range and ideally a single pulse can be counted on a single anode.In order to obtain a comparison of performance with a 2 cm long resistive strip and wire grid FPDs, consider a 2 cm long discrete anode array. At 25 pm resolution there will be 800 independent detectors and hence the maximum count rate will be about 800 times that of the resistive strip and wire grid FPDs, with the restriction that the count rate at any individual detector cannot exceed about lo7 counts s-l. The maximum count rate at a single detector is limited by two other factors. First, in order to obtain an undistorted spectrum two particles should not arrive at the same detector within 100 ns of each other or only one event will be registered, i.e., the pulse pair resolution is about 100 ns.Considering the statistical distribution of arrival times, this limits the maximum count rate to about 106 counts s-l. Second, a detector anode of dimensions 15 x 4000 pm can be activated by about lo00 channels of an MCP whose channels are set on a 12 pm pitch. If each channel has a recovery time of 10-2 s, this means that the maximum count rate is about 105 counts s-1. Taking into account the statistical distribution of the particle arrival times, this means that the maximum count rate is limited by the MCP to be about 104 counts s-1 per anode. Hence the maximum count rate is dominated by the MCP and not the response time of the detector electronics. Higher performance MCPs are available with a lower channel resistance and hence lower recovery time, but which also have a greater power dissipation.. Electro-optical The basis of this technique is to place a phosphor screen after the MCP to convert the electron pulses to photons and then detect the photons. Two methods for photon detection are focusing onto a CCD19 or channelling through a fibre-optic bundle onto a light-sensitive diode array.*O The collection efficiency is high but resolution is lost at several interfaces and the device is bulky and expensive. Microchannel Plate Electron Multiplier The development of the MCP was a necessary precursor to the development of FPDs. However, the advances in silicon technology have been so rapid that it is the MCP which now limits the performance of FPDs. Dynamic Range Consider a single channel of an MCP.The capacitance of the channel is typically 10-16 F and its resistance is typically 1014 Q. A particle falling on the channel initiates the generation of a pulse of electrons which discharges the voltage across the capacitance, and this voltage must recover before the arrival of a second particle or the gain will be reduced. As a rough guide, this recovery time is related to RC, or 10-2 s in the present example. The lowering of the gain at increasing particle count rate has been observed experimentally.2~9~13 Clearly, reducing R will reduce the recovery time but it will also increase the heat dissipation owing to the greater standing current across the MCP. Lower resistance MCPs are currently available but dissipate more energy by ohmic heating.Resolution Microchannel plates are often used in pairs for extra amplifica- tion of the detected particles and to reduce ion feedback. The spacing between the MCPs is typically set to 50 pm and the output from a single channel of the first MCP will activate several channels of the second MCP, giving an output pulse of about 106 electrons within 1 ns. The separation between the second MCP and the FPD is a critical factor in determining the FPD resolution.2>6 The smaller the separation and the greater the attractive voltage applied to the FPD, the smaller is the spreading of the electron pulse. However, it has been observed and calculated2.6 that if the separation is much greater than the channel diameter, the beam spreading cannot be effectively reduced by reasonable values of the attractive field.With careful setting of the MCP-array separation to typically 50 pm, a single particle leads to an electron shower from the lower MCP which spreads to typically 70 pm. Discrete Anode FPD Design The following sections describe the proposed Aberystwyth FPD design. The sensor and the counterhnterface described have been fabricated and are functional.17 The array described is currently under development and the factors which have determined important array design parameters are explained. Full operational tests of the array require mounting it in a high-resolution mass spectrometer with an interface that will exercise it at its maximum rate of operation. This is also under development and results will be reported elsewhere.Sensor A block diagram of a single detector is shown in Fig. 1. The detector anode is a strip of aluminium on the top surface of the integrated circuit (IC). It forms the input to the charge-sensing circuit and its capacitance can be reduced and hence the sensitivity of the detector increased by either reducing the area of the anode or increasing the thickness of the insulator between it and the underlying substrate.3J1 Charge of sufficient magnitude falling on the anode is sensed by the charge sensor (described below) and a pulse is transmitted to the counter (described below). The voltage generated by electrons falling on the anode could be a small fraction of a volt to many volts. The sensor input stage is a switch driven by the detector anode. The equilibrium voltage on the anode is VDIS and this holds the switch open. A small negative pulse on the anode closes the switch and triggers a cycle which both generates a 5 V pulse and pulls the anode back to VDIS via the discharge means.If VDIS is moved closer to the switching voltage, then less negative charge is required to switch the circuit and hence the sensitivity of the sensor can be simply varied. Positive pulses can be detected by setting VDIS to hold the switch closed. The output base voltage level is then high and a small positive pulse opens the switch, triggers a discharge cycle and generates a negative-going 5 V output pulse. The dead time of this sensor is only its switching time. This is in contrast to detectors which require the clocking of a measurement cycle and a reset cycle.Here the sensing of a charge pulse triggers both the generation of a 5 V pulse and self-discharge. Charge III Fig. 1 Block diagram of the self-resetting charge sensor circuit1102 ANALYST, JULY 1992, VOL. 117 Charge 101 Ill /I Detector anode L{$F 107 o v 105 15v Po v Fig. 2 Schematic transistor level sensor circuit Voltage level on detector anode (Net 101 \ VDIS 1 Fig. 3 Model of the charge sensor operation in terms of a hysteresis diagram A transistor level circuit is shown in Fig. 2. Others which implement the functions shown in Fig. 1 are under investiga- tion. The following considers the transistors as switches. It should be remembered that transistor switching is not an instantaneous process. Initially transistor M1 is open (not passing current), M2 is closed and M3 is open.A small negative pulse on net 101 closes M l and lowers the gate-source voltage of M2, which opens. The 106 voltage rises as M2 opens, which feeds back to close M1 and open M2 further. The voltage on net 105 is pulled down and this drives the output inverter (M4/M5) to give a positive output pulse. M3 is closed and 101 is discharged back to VDIS. Any charge falling on the detector anode which is insufficient to trigger the latter cycle is discharged with a time constant of about 50 ns. This will reduce the effects of ‘blooming’ but will have a negligible effect on the voltage pulse height of a typical 1 ns charge pulse from the MCP. The circuit operation can be understood from a hysteresis diagram (Fig. 3). A negative pulse on 101 lowers the voltage below the lower hysteresis level, the circuit switches, gener- ates a 5 V pulse and discharge of 101 returns the circuit to its initial condition above the upper hysteresis level.It can also be seen that the lower the hysteresis gap, the greater is the sensitivity of the sensor until a limit of a zero gap. The sensor can be designed to measure positive or negative d.c. current by removing the discharge resistor Rdis. Consider- ing negative current, charge accumulates on the detector electrode until the induced voltage reaches the lower hystere- sis level. This initiates the cycle which generates an output pulse and discharges the accumulated charge. The frequency of the output pulse is proportional to the incoming current.Countedhterface A schematic diagram of the 8 bit counter and control logic is shown in Fig. 4. The circuit is relatively simple and self- explanatory. Shift register / 8 Bit counter R (in) -AAad R (out) Input I 1 pulse I - Fig. 4 Schematic diagram of the counter-bus interface circuit Main bus Detector electrodes and charge sensors Fig. 5 Detector array not to scale. Charge sensors are situated beneath detector electrodes and are connected to counters by metal tracks Pulses from the charge sensor are gated to the counter and counting is stopped when the counter reaches 252 counts (MAXCNT goes high), which prevents overflow and allows three extra counts to be accumulated during the stopping time. This allowance of three counts is entirely adequate to prevent overflow in both modes of operation mentioned below.Counting is also stopped when the counter is being read (EN is high) and when an external signal is asserted (STCIN is high). Detection Array The layout of an array of the detectors is shown in Fig. 5. Associated with each detector electrode is a charge sensor and a counter-bus interface. At a resolution of 25 pm there will be 400 electrodes cm-1. This means, of course, that 400 copies cm-1 of the associated circuitry must be placed on the chip. The detector electrodes are distributed along one side of the chip with the charge sensors underneath the electrodes. Pulses are routed to associated 8 bit counters arranged in banks. A local bus is associated with each counter bank and each local bus is buffered onto the main bus.Ideally the whole of a long detector array would be placed on a single chip, but yield will prevent this in the foreseeable future and, therefore, smallerANALYST, JULY 1992, VOL. 117 1103 chips will be butted on a substrate and stitched together to form a long detector. A separate circuit is being designed to interface between the detector array and an external com- puter. The array can operate in two modes. In Mode 1, when the external controller recognizes that one of the counters has reached 252 counts (STCOUT asserted), it inhibits all further counting (by asserting STCIN) and then reads all the accumulated counts. This gives a complete mass spectrum with all peaks relatively correct. In Mode 2 the counters are read cyclically and continuously. This mode will detect low intensity ions at maximum efficiency but the spectrum may contain flat-topped peaks.In both modes counters are automatically reset after they have been read. Optimum Design At Aberystwyth, the key requirements of a mass spectrometer were identified at the outset and all these requirements were incorporated in the FPD design. This involved many trade- offs, some of which are mentioned below. The key require- ments were identified as a high collection efficiency, high resolution, high sensitivity and low noise, low power con- sumption and low cost (high yield, high lifetime, ease of assembly, etc.). This shortlist has many implications for the design. The following explains the trade-offs made between design parameters to achieve the requirements listed above.Anode dimensions The model for the anode array is shown in Fig. 6. Consider a fixed anode width ( w ) and separation ( s ) from adjacent anodes. Cross-talk depends on the ratio of the inter-anode capacitance (C,,) and the anode-to-substrate capacitance (Gas): cross-talk = C,$(C,, + Gas), and for low C,,, cross-talk = Caa/Cas. Both fall at roughly the same rate and the cross-talk will remain roughly constant as the anode length is reduced. However, as the anode capacitance decreases, its sensitivity increases and, therefore, a small value of 1 should be chosen consistent with the required collection efficiency. Consider a detector anode of fixed length and inter-anode spacing. As the anode width increases (and hence the anode pitch), its capacitance increases but the inter-anode capaci- tance remains the same and hence the cross-talk is reduced.If the electron pulse from the MCP is much wider than the anode, then, although the anode capacitance increases with width, the collected electrons also increase at roughly the same rate and hence the sensitivity is not a strong function of width. The compromise chosen for the anode pitch of 25 pm is approaching half the expected MCP electron pulse width. Greater than this gives a reduction in sensitivity and resolu- tion. Less than this gives an increase in cross-talk, more circuitry and no great improvement in resolution or sensi- tivity. Insulator Ah77 Substrate Fig. 6 optimum design (see text for details) Model of the sensor array used in the calculation of the For a fixed anode length and width, a larger spacing between the anodes reduces the collection efficiency and increases the insulating surface area where charge build-up may occur.Smaller spacing increases cross-talk. Therefore, it should be as small as possible consistent with cross-talk limits. In summary, the optimum dimensions for a particular application can be found as follows. Choose the minimum value of 1 consistent with a good collection efficiency. Decide the allowable cross-talk and equate this with the detector parameters: C,, = lw/h c,, = ltls Let the allowable cross-talk be 7%. Therefore, cross-talk = C,$C,, = (ltls) (hllw) = hthw = 7/100 For h = 10 pm and t = 1 pm, sw = 143 (larger value gives lower cross-talk) . In this simple calculation it is assumed that the insulator is present between electrodes and hence the relative permittivity can be omitted.The minimum allowable value of s is fixed by the silicon process used. In the current design a value of 10 pm was chosen for s and 15 pm for w , giving an over-all resolution of 25 pm. A higher resolution implies both greater cross-talk and a larger detector array with a correspondingly lower yield, as there are more detectors and associated circuitry per unit length. Counter size Consider a 10 cm long detector containing 4000 detectors. Assume the read time per counter is about 0.4 X 10-6 s. For 4000 counters, the read time (t,) is about 1.6 X 10-3 s. Assume the maximum count rate per detector is determined by the MCP and is about 104 counts s-1. At this rate, an 8 bit counter will fill in about 256/104 s or about 25 ms (ts), and a 4 bit counter will fill in 16/104 s or about 1.6 ms (t4).In mode 1 (stop and read), the dead time for a 4 bit counter would be t,/(t, + t4) = (1.6 x 10-3)/(3.2 X 10-3) = 50%, and that for an 8 bit counter would be t,/(t, + ts) = (1.6 x 10-3)/(26.6 x 10-3) = 6%. In mode 2 (continuous read), t, = t4 and hence 4 bit counters should be adequate. Considering the much reduced circuitry and increased yield when using 4 bit instead of 8 bit counters, it may be best to use the former if possible. Higher performance devices with improved MCPs require more than 4 bits if maximum collection efficiency is to be achieved, but it should be realized that reducing the read time or the array length by a factor of two is equivalent to adding one counter bit.Mode 2 is the only mode necessary if sufficiently large counters are used. Sensor characteristics In order to collect ions efficiently at high ion intensity, the pulse pair resolution of the detectors should be high. The switching cycle of the charge sensor during the detection of an event is about 100 ns, giving a maximum pulse count rate of about 107 Hz. The electron pulse emerging from the MCP will fall on more than one anode. Charge from the periphery of such pulses must not build up on detector anodes and give spurious counts or there will be a ‘blooming’ effect and a loss of resolution. The sensor has been designed so that charge build-up on the anodes does not occur, and it is insensitive to noise below about 10 MHz.Both of these features can be achieved by exploiting the fact that the duration of a charge pulse from an MCP is very short (about 1 ns), so that charge can be drained through a resistor from the anode with a time constant of say 50 ns without significantly affecting the peak of the anode voltage pulse.1104 ANALYST, JULY 1992, VOL. 117 Fabrication technology Power consumption should be kept to a minimum. Circuit performance degrades at increased temperature and as cooling in a vacuum is inefficient it is important to select a fabrication technology that dissipates little energy. CMOS satisfies this requirement and is a leading technology undergo- ing constant development. Most power is dissipated in digital CMOS circuits during switching and the maximum dissipation of sensors (not CMOS) and counters is about 0.25 W cm-1.FPD Interface Hosticka22 considered the prospects of VLSI readout and signal processing integrated with detectors on a single chip. The FPD designed at Aberystwyth has basic intelligence, as discussed above, and could be controlled through a very simple interface by an external computer, but in many applications the constraints of an existing system to which the sensor device is to be interfaced will often necessitate an intermediate interface to mesh the FPDs facilities and demands into the existing system without extensive redesign. It is important to keep the cost of the new interfacing scheme to a minimum consistent with obtaining optimum system performance. The Aberystwyth interface accumulates data from the sensor array at the maximum rate of the latter, outputs it on a slow serial line and provides a physical interface between high- and low-voltage environments.The burden of the interfacing and the performance enhan- cement is best located off the detector array itself, for several reasons, including the following: to retain the modularity of the array design; yield (the array will be large and increasing the size by including the interface on the same IC as the array itself would reduce the yield); and the interface design is greatly simplified by using existing ICs. The design of the interface will be the subject of a later publication. This work is supported by the LINK Industrial Measurement Systems programme, Vacuum Generators Analytical (Fisons Instruments), and ICI (Wilton).The assistance of the Ruther- ford-Appleton laboratory and the Edinburgh Microfabrica- tion Facility is acknowledged. The support and discussions of T. M. McGinnity, D. P. Langstaff, M. W. LawtonandD. M. Forbes are also gratefully acknowledged. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 References Moak, C. D., Datz, S., Santibanez, F. G., and Carlson, T. A., J. Electron Spectrosc. Related Phenom., 1975, 6 , 151. Timothy, J. G., and Bybee, R. L., Appf. Opt., 1975, 14, 1632. Hatfield, J. V., York, T. A., Comer, J., and Hicks, P. J., IEEE J. Solid-state Instrum., 1989, 24, 704. Kellog, E., Henry, P., Murray, S., and van Speybroeck, L., Rev. Sci. Instrum., 1976,47,282. Heijne, E. H. M., and Jarron, P., Nucl. Instrum. Methods Phys. Res., 1984,226, 12. Asplund, L., Gelius, U., Tove, P. A., Eriksson, S. A., and Bingefors, N., Nucl. Instrum. Methods Phys. Res., 1984, 226, 204. Richter, L. J., and Ho, W., Rev. Sci. Instrum., 1986, 57, 1469. Timothy, J. G., and Bybee, R. L., SPIE, 1981,265, 93. Liptak, M., Sandie, W. G., Shelley, E. G., Simpson, D. A., and Rosenbauer, H., IEEE Trans. Nucl. Sci., 1984, NS31,780. Martin, C., Jelinsky, P., Lampton, M., Malina, R. F., and Anger, H. O., Rev. Sci. Instrum., 1981, 52, 1067. McClintock, W. E., Barth, C. A., Steele, R. E., Lawrence, G. M., and Timothy, J. G., Appl. Opt., 1981, 21, 3071. Walker, J. T., Parker, S., Hyams, B., and Shapiro, S. L., Nucl. Instrum. Methods Phys. Res., 1984, 226, 200. Firmani, C., Ruiz, E., Carlson, C. W., Lampton, M., and Paresce, F., Rev. Sci. Instrum., 1982, 53, 570. Gott, R., Parkes, W., and Pounds, K. A., IEEE Trans. Nucl. Sci., 1970, NS17, 367. Padmore, T. S., Roberts, K. M., Padmore, H. A., and Thornton, G., Nucl. Instrum. Methods Phys. Res., 1988, A270, 582. Gurney, B. A., Ho, W., Richter, L. J., and Villarubia, J. S., Rev. Sci. Instrum., 1988,59,22. Birkinshaw, K., McGinnity, M., Langstaff, D. P., Lawton, M. W., and Forbes, D. M., Sensors Technology, Systems and Applications, Adam Hilger, Bristol, 1991, pp. 421-426. Adams, N. G., and Smith, D., J. Phys. E, 1974,7, 759. Hicks, P. J., Daviel, S., Wallbank, B., and Comer, J., J. Phys. E , 1980,13,713. Cotrell, J. S., and Evans, S., Rapid Commun. Mass Spectrom., 1987, 1, 1. Dettmer, R., IEE Rev., 1988, 411. Hosticka, B. J., Nucl. Instrum. Methods Phys. Res., 1984, 226, 185. Paper 1 I050396 Received October 2, 1991 Accepted February 13, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701099
出版商:RSC
年代:1992
数据来源: RSC
|
9. |
Gas chromatographic–mass spectrometric characterization of flavanones in citrus and grape juices |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1105-1109
Colin S. Creaser,
Preview
|
PDF (447KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1105 Gas Chromatographic-Mass Spectrometric Characterization of Flavanones in Citrus and Grape Juices Colin S. Creaser, Mohammed R. Koupai-Abyazani" and G. Richard Stephenson School of Chemical Sciences, University of East Anglia, Norwich NR4 7TJ, UK A method for the characterization of flavanones in fruit juices, involving solvent extraction, hydrolysis to the corresponding aglycones, trimethylsilylation and combined gas chromatography-mass spectrometry, is reported. The application of the method is demonstrated for the analysis of orange, lemon, grapefruit and grape juices. Keywords: Gas chromatograph y-mass spectrometry; trimeth ylsilyl derivatives; flavanones; fruit juices The flavonoids of citrus fruits have been extensively investi- gated because of their pharmacological activity, flavour impact on citrus juices and value as by-products of the citrus industry.1 The principal citrus flavanones, naringenin (I), hesperetin (II) and eriodictyol (111), do not occur in juices as the free aglycones, but are combined through the C-7 hydroxy group with a sugar component ,1 either P-neohesperidose (IV) or p-rutinose (V). Bitterness in some citrus fruits is attributed to the flavanone neohesperidosides, while flavanone rutino- sides are tasteless.24 2' 3kR3 I: R1, R2, R4 = OH; II: R1, R2, R3 = OH; 111: R1, R*, R3, R4 = OH R3 = H R4 = OMe CH20H OH OH IV OH W The distribution of these flavanone glycosides, and of the flavanone aglycones derived from them, is characteristic of many citrus fruits and juices, and also affects the quality of processed citrus products.5 Analysis for flavanone content has, therefore, been proposed as a means of characterizing the authenticity of lemon juice,6 of measuring the adulteration of citrus juices7 and of identifying the presence of orange juice in fruit drinks.8 * Present address: Department of Botany, University of British Columbia, Vancouver, Canada V6T 1ZK. Several techniques have been used for the determination of flavanones in citrus juices.The most widely used method is based on the spectrophotometric measurement of the yellow colour produced by flavanone glycosides in alkaline solution .9 Although more selective methods have been developed for flavanone determination, the Davis method9 is still used for those analyses where its simplicity outweighs its suscepti- bility to interference.Methods involving combinations of extraction, adsorption chromatography, paper chroma- tography, thin-layer chromatography and spectrophotometry have all been reported for the separation and identification of flavonoids in citrus and other fruit juices.10-15 High-perfor- mance liquid chromatography (HPLC) has been applied to the determination of polymethoxylated flavonoids in citrus juices16717 and citrus peel.18 The adoption of HPLC for the analysis of mixtures of these compounds greatly improves the speed and accuracy of their identification and determination, but affords poor resolution for complex mixtures of flavo- noids. Capillary gas chromatography (GC) and gas chroma- tography-mass spectrometry (GC-MS) procedures have been demonstrated to yield improved separations for flavanone aglycone trimethylsilyl (TMS) derivatives,19-*1 but they have not been applied to fruit juices.In this paper, a sensitive and selective method is reported for the identification of flavanone glycosides in fruit juices, as the corresponding aglycones, which involves use of solvent extraction, hydrolysis, trimethylsilylation and combined GC- MS. The application of the method is demonstrated for the analysis of orange, lemon, grapefruit and grape juices. Experimental Reagents and Materials The solvents ethanol and diethyl ether (Aldrich, Gillingham, Dorset, UK) were of HPLC grade and ethyl acetate (Fisons, Loughborough, Leicestershire, UK) was of Distol grade. Anhydrous pyridine (Pierce, Rockford, IL, USA) was of silylation grade, and anhydrous sodium sulfate (Fisons) was obtained at 99.5% purity.The silylating reagents, hex- amethyldisilazane (HMDS) (98% ) and trimethylchlorosilane (TMCS) (98%), were obtained from Aldrich. Flavanone standards were purchased from Apin Chemicals (Abingdon, Oxfordshire, UK). Procedure A 100 cm3 sample of hand-squeezed, clarified orange juice was extracted with 2 x 100 cm3 of diethyl ether in a 500 cm3 separating funnel. The ether extracts were discarded. The aqueous phase was then extracted with 4 X 100 cm3 of ethyl acetate, and the organic phase was evaporated to dryness under reduced pressure at 35-40 "C. The residue was taken up1106 100 50 ANALYST, JULY 1992, VOL. 117 -(el - in 10 cm3 of ethanol; 5 cm3 of 5% HCl were added to a 3 cm3 portion of this solution, which was heated at 100 "C for 2 h in an oil-bath722 cooled, and extracted with 3 x 10 cm3 of ethyl acetate.The extracts were combined, dried through a plug of anhydrous sodium sulfate ( 3 3 x 2 cm i d . ) and evaporated to dryness. The residue was dissolved in 0.2 cm3 of pyridine, 0.2 cm3 of HMDS and 0.1 cm3 of TMCS, and the mixture was heated at 60 "C overnight.20.21 Samples (100 cm3) of hand-squeezed lemon, grapefruit and grape juices were extracted and pre-treated by the same procedure as that described for orange juice. Aliquots of each fruit juice were also spiked with standard naringenin, hespere- tin and eriodictyol at levels of 2 4 pg cm-3. GC-MS Analysis The derivatized samples were analysed by use of a Varian 3400 gas chromatograph (Palo Alto, CA, USA) directly interfaced with a Finnigan-MAT (San Jose, CA, USA) ion trap mass spectrometer operated via an AT/personal computer. The GC separation was carried out on a capillary column (50 m x 0.25 mm i.d.) of RSL 200 BP (0.2 pm film thickness), from Alltech (Carnforth, Lancashire, UK), under the following conditions: helium carrier gas pressure, 9 psi (62 kPa); injector and transfer line temperatures, 300 and 280 "C, respectively; oven temperature programme, 130 "C for 0.5 min, then heated to 235 "C at 30 "C min-1 and from 235 to 290 "C at 1 "C min-1.The MS conditions were electron ionization under automatic gain control at a trap temperature of 150 "C; scan speed, 1 scan S-1. Results and Discussion The polyhydroxylated flavanone glycosides present in fruit juices are insufficiently volatile for direct GC separation, and require hydrolysis to the corresponding aglycones and derivat- ization prior to analysis.Trimethylsilylation has been used successfully for this class of compounds by several groups of workers.19JO The flavanone aglycones, unlike the other flavonoids, yield a mixture of the TMS derivatives of the flavanone (e.g., VI for eriodictyol) and the corresponding chalcone (VII) in the presence of HMDS and TMCS under mild derivatizing conditions.20721 However, heating the reac- tion mixture at 60 "C overnight results in quantitative conversion into the TMS derivative of the chalcone. (CH3)3Si0 0 VI (CH3)3Si0 (CH3),Si0 0 VII Hydrolysis of the flavanone glycosides, followed by TMS derivatization of the aglycones, provides, therefore, a conve- nient method for the characterization of the glycosides, as the corresponding TMS chalcone derivatives, by GC-MS. The orange, grapefruit, lemon and grape juice extracts were prepared by a straightforward extraction procedure and were readily hydrolysed with HCl to yield the flavanone aglycones.The aglycones were converted into the TMS derivatives of the corresponding chalcones under the derivatization conditions used. The flavanones were then identified by comparison of the GC retention times and mass spectra of the TMS chalcone derivatives with those of standards prepared from the flavan- one aglycones after derivatization under equivalent condi- tions. Juice extracts were also spiked with standards prior to hydrolysis and derivatization to confirm GC retention times and mass spectral assignments.100 (=) 50 I 1 i 3 I 400 800 1200 1600 2000 Scan number Fig. 1 Selected ion chromatograms (m/z 545-633) for (a) a standard mixture of flavanone aglycones after TMS derivatization; ( b ) the TMS derivatization of oran e juice extract; (c) the TMS derivatization of lemon juice extract; 6) the TMS derivatization of grapefruit juice extract; and (e) the TMS derivatization of grape juice extract. 1, Naringenin; 2, hesperetin; and 3, eriodictyolANALYST, JULY 1992, VOL. 117 - 575 562 267 307 369 473 147 57 105 ] 179 1-1 - I . . . I I , 1 I 1 I I 1107 loo 50 Table 1 GC-MS retention times and limits of detection for TMS chalcone derivatives of flavanone aglycones Limit of Compound time/min pg cm-3 Retention detection*/ Naringenin 23.31 0.02 Hesperetin 26.47 0.02 Eriodict yo1 27.59 0.03 * Based on a signal-to-noise ratio of 2 : 1.( C) -73 - 73 267 353 532 575620 147 95 I 633 100 200 300 400 500 600 mlz Fig. 2 EI mass spectra of the TMS chalcone derivatives of (a) peak 1, naringenin; (b) peak 2, hesperetin; and (c) peak 3, eriodictyol from lemon juice extract [Fig. l(c)] The mass spectra of the TMS chalcone derivatives of naringenin, hesperetin and eriodictyol are characterized by a prominent [M - 151' ion at mlz 545, 575 and 633, respec- tively.23 These ions are the base peaks in each spectrum. The selected ion chromatogram for the sum of ion intensities in the range mlz 545-633, for the TMS chalcone derivatives obtained from a mixture of naringenin, hesperetin and eriodictyol, is shown in Fig.l(a). Table 1 gives data on retention times and limits of detection, obtained by GC-MS, for these com- pounds. Analysis for the TMS derivatives in hydrolysed orange juice extract , by GC-MS and co-chromatography with appropriate reference compounds, showed the presence of naringenin and hesperetin , but eriodictyol was not detected [Fig. l(b)]. Typical profiles for the TMS chalcone derivatives of flavanone aglycones derived from lemon and grapefruit juices are shown in Fig. l(c) and ( d ) , respectively. These ion Table 2 Concentrations and recoveries of flavanone aglycones in orange, lemon, grapefruit and grape juices Flavanone concentration*/pg ~ m - ~ Fruit juice Naringenin Hesperetin Eriodictyol Orange 0.46 (90)t 1.04 (90) NDS (86) Lemon 0.15 (88) 0.58 (87) 0.17 (86) Grapefruit 0.61 (90) ND (88) ND (85) Grape ND (87) ND (85) ND (83) * Repeatability, 9.4% in the range 0.4-1.0 pg ~ m - ~ .t Recoveries (Yo) in parentheses, for samples spiked at 2 4 $ ND = Not detected. pg cm-3. a C m -0 3 : 50 .- a > .- + - 01 n 0 100 200 300 400 500 600 mlz Fig. 3 EI mass spectrum of trihydroxymonomethoxyflavanone TMS derivative (peak 4) in the chromatogram obtained from lemon juice extract [Fig. l(c)] chromatograms establish that naringenin, hesperetin and eriodictyol are all present in lemon juice, while the predomi- nant flavanone component of grapefruit juice is naringenin. In contrast to orange, lemon and grapefruit juices, no flavanones were detected in grape juice at a level above the GC-MS limits of detection [Fig.l(e)]. Chalcone production from free flavanone aglycones was not detected in any of the juices after TMS derivatization of unhydrolysed extracts, confirming that the flavanones are present only as the glycosides in the fruit juice extracts examined. The mass spectra of the TMS derivatives of flavanones extracted from lemon juice are shown in Fig. 2. These match the spectra of the TMS chalcone derivatives of standard samples.23 Table 2 gives the flavanone concentrations found in orange, lemon, grapefruit and grape juices expressed in terms of the corresponding free flavanone aglycone. These data, based on peak area measurements, are the average of values obtained by using external standard calibration and standard additions procedures.Good agreement was obtained for both methods. Recoveries of naringenin, hesperetin and eriodictyol were in the range 83-90% for spiked fruit juices (Table 2). The chromatogram obtained after TMS derivatization of the lemon juice extract showed an unidentified peak (4), with a retention time of 27.4 min, in addition to peaks correspond- ing to naringenin, hesperetin and eriodictyol [Fig. l(c)]. The mass spectrum of this compound is shown in Fig. 3. The presence of an ion at mlz 575 in the mass spectrum of this peak (Fig. 3), tentatively assigned to the [M - 15]+ fragment ion, indicates that this is the TMS derivative of a tetrahydroxymo- nomethoxychalcone [Scheme 1, (b)] formed from a trihy- droxymonomethoxyflavanone [Scheme 1, (a)] .23 The [M - CO]+' ion at mlz 562 is a characteristic ion for chalcones.The ion at mlz 369 indicates that three of the hydroxy groups are in the A-ring [Scheme 1, (b)], of which one must be at the C-5 position of the original flavanone [Scheme 1, (a)].23 This fragment, which contains the intact A-ring, corresponds to the A2+ ion for underivatized chalcones,24.25 and is formed by the retro-Diels-Alder cleavage of the molecular ion. The absence1108 ANALYST, JULY 1992, VOL. 117 \ OH 0 Derivatization (CH3)3Si0 0 0 0’ ‘si’ C< ‘CH, [M - 15]+ rnlz 575 mlz 369 Scheme 1 OCH3 OSi(CH3)3 [M - CO]” mlz 562 of an [M - 88]+’ ion indicates that there are no adjacent TMS groups in the A-ring. On the basis of the observed ions, the original flavanone is assigned as the trihydroxymonomethoxy- flavanone [Scheme 1, (a)].23 Conclusion Analysis by GC-MS of the TMS derivatives of hydrolysed flavanone glycosides provides a sensitive and selective method for the characterization of these compounds in fruit juices, and for the identification of flavanones for which reference standards are not available.Detection limits and recoveries for the TMS derivatives of the aglycones are satisfactory for their determination in fruit juices. References 1 Horowitz, R. M., and Gentili, B., in Citrus Science and Technology, eds. Nagy, S., Shaw, P. E., and Veldhuis, M. K., AVI Publishing Co., Westport, CT, 1977, vol. 1. Horowitz, R. M., and Gentili, B., Tetrahedron, 1963, 19, 773. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Horowitz, R.M., in Biochemistry of Phenolic Compounds, ed. Harborne, J. B., Academic Press, New York, 1964. Horowitz, R. M., and Gentili, B., J. Agric. Food Chem., 1969, 17, 696. Kefford, J. F., and Chandler, B. V., The Chemical Constituents of Citrus Fruits, Academic Press, New York, 1970. Rolle, L. A., and Vandercook, C. E., J. Assoc. Off. Anal. Chem., 1963,46, 362. Wagner, D., and Monselise, J. J., Isr. J. Technol., 1963, 1, 33; Chem. Abstr., 1964, 60, 6140a. Koch, J., and Haase-Sajak, E., Dtsch. Lebensm.-Rundsch., 1965, 61, 199. Davis, W. B., Anal. Chem., 1947, 19, 476. Albach, R. F., and Redman, G. H., Phytochemistry, 1969, 8, 127. Nishiura, M., Esaki, S., and Kamiya, S., Agric. Biol. Chem. (Tokyo), 1969,33, 1109. Vandercook, C. E., and Rolle, L. A., J.Ass,i. Off. Agric. Chem., 1963,46, 359. Yokoyama, H., J. Assoc. Off. Agric. Chem., 1965,48, 530. Hagen, R. E., Dunlap, W. J., Mizelle, J. W., Wender, S. H., Lime, B. J., Albach, R. F., and Griffiths, F. P., Anal. Biochem., 1965, 12, 472. Dunlap, W. J., Hagen, R. E., and Wender, S. H., J. Food Sci., 1962, 27, 597.ANALYST, JULY 1992, VOL. 117 1109 16 Sendra, J. M., Navarro, J. L., and Izquierdo, L., J. Chroma- togr. Sci., 1988, 26, 443. 17 Heimhuber, B., Galensa, R., and Herrmann, K., J. Chroma- togr., 1988,439,481. 18 Park, G. L., Avery, S. M., Byers, J. L., and Nelson, D. B., Food Technol, 1983, 37, 98. 19 Greenaway, W., English, S., Wollenweber, E., and Whatley, F. R., J. Chromatogr., 1989,481, 352. 20 Creaser, C. S., Koupai-Abyazani, M. R., and Stephenson, G. R., J. Chromatogr., 1989, 478,415. 21 Creaser, C. S., Koupai-Abyazani, M. R., and Stephenson, G. R., J. Chromatogr., 1991,586, 323. 22 Coffin, D. E., and Dupont, J. E., J. Assoc. Off. Anal. Chem., 1971, 54, 1211. 23 Creaser, C. S., Koupai-Abyazani, M. R., and Stephenson, G. R., Org. Mass Spectrom., 1991, 26, 157. 24 Mabry, T. J., and Markham, K. R., in The Flavonoids, eds. Harborne, J. B., Mabry, T. J., and Mabry, H., Chapman and Hall, London, 1975. Mabry, T. J., and Ulubelen, A., in Biochemical Application of Mass Spectrometry, eds. Waller, G . R.. and Dermer, 0. C., Wiley-Interscience, New York, 1973. 25 Paper 210071 1 H Received February 11, 1992 Accepted March 6, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701105
出版商:RSC
年代:1992
数据来源: RSC
|
10. |
Detection and identification of volatile substances by headspace capillary gas chromatography to aid the diagnosis of acute poisoning |
|
Analyst,
Volume 117,
Issue 7,
1992,
Page 1111-1127
Peter J. Streete,
Preview
|
PDF (1760KB)
|
|
摘要:
ANALYST, JULY 1992, VOL. 117 1111 20 0 Detection and Identification of Volatile Substances by Headspace Capillary Gas Chromatography to Aid the Diagnosis of Acute Poisoning I - 1 1 I I I I I 1 Peter J. Streete and Manjit Ruprah National Poisons Unit, Guy's and Lewisham NHS Trust, Avonley Road, London SE14 5ER, UK John D. Ramsey Toxicology Unit, Department of Cardiological Sciences, St. George's Hospital Medical School, London SWl7ORE, UK Robert J. Flanagan" National Poisons Unit, Guy's and Lewisham NHS Trust, Avonley Road, London SE14 5ER, UK Headspace gas chromatography with split flame-ionization-electron-capture detection is a simple method of screening for a wide range of volatile substances in biological fluids. A 60 m x 0.53 mm i.d. thick-film (5 pm) f used-si I ica ca pi I la ry coated with SPB- 1 (Su pelc hem) with split f lame-ion ization-electron-ca ptu re detection provides a valuable alternative to packed columns in this work.Most commonly abused compounds, including many with very low boiling-points such as bromochlorodifluoromethane (BCF), butane, dimethyl ether, FC 11 , FC 12, isobutane and propane, can be retained and differentiated at an initial column temperature of 40 "C followed by programming to 200 "C. The total analysis time is 26 min. Retention and detector response data were generated for 244 compounds. Good peak shapes are obtained for polar analytes such as ethanol and injections of up to 0.30 cm3 of headspace can be performed with no discernable loss of efficiency. The sensitivity is thus at least as good as that attainable with packed columns.Of the commonly encountered compounds, only isobutane-methanol and paraldehyde-toluene are at all difficult to differentiate. Quantitative measurements can be performed either isothermally or by using the temperature programme. Keywords: Volatile substance abuse; biological samples; headspace gas chromatography; temperature programming; SPB- 1 capillary column There are now more than 100 deaths each year in the UK from volatile substance abuse (VSA, 'glue sniffing' , inhalant abuse, solvent abuse), mainly amongst adolescent males. 1 About 1.5-10% of UK secondary school children (boys and girls in approximately equal numbers) have experimented with VSA and about 1% are current users. Similar prevalence rates have been reported from many other countries within the last 10 years.In addition, there are a large number of reports not only of sudden death but also of long-term hazards associated with VSA.2 A wide range of compounds may be abused by inhalation (Table 1). However, the compounds encountered in UK VSA-related deaths are nowadays often very volatile substances such as those found in fuel gases, aerosol propel- lants and halon fire extinguishers (Fig. 1). The aerosol propellants used in the UK are at present mostly 'butane', although dimethyl ether either alone or as an azeotropic * To whom correspondence should be addressed. mixture with fluorocarbon (FC) 22 is used in other parts of Europe. A further compound, FC 134a, is under evaluation for use in medical aerosols.Analytical confirmation of the diagnosis of VSA is impor- tant , especially when investigating sudden death. Toxicologi- cal analyses may also be helpful if ingestion of solvents or solvent-based products is suspected or vapours may have been inhaled accidentally. Headspace sample preparation together with temperature-programmed gas chromatography (GC) and split flame-ionization-electron-capture detection (FID-ECD) provides a simple method of screening for a wide range of volatiles in biological specimens. Ramsey and Flanagan3 used a packed column (2 m x 2 mm i.d., 0.3% m/m Carbowax 20M on Carbopack C) programmed from 35 to 175 "C. On-column septum injections of up to 0.40 cm3 of headspace could be performed and thus good sensitivity (of the order of 0.1 mg dm-3 or better using 0.20 cm3 of sample) could be obtained.Moreover, most compounds of interest could be retained without resort to sub-ambient operation and the system could be used isothermally at an appropriate temperat- ure for quantitative analyses. Disadvantages included the poor resolution of some very volatile substances, the long total analysis time (40 min) and variation in the peak shape given by alcohols between different batches of column packing. Bonded-phase wide-bore capillary columns permit rela- tively large-volume septum injections and can offer advan- tages of improved efficiency, reproducibility and reliability. We have found that a 60 m x 0.53 mm i.d. fused-silica capillary coated with the dimethylpolysiloxane phase SPB-1(5 ym film thickness) offers many advantages over the packed column described above.In particular, improved resolution of very volatile compounds is obtained even at an initial temperature of 40 "C, the total analysis time can be reduced to 26 min and good peak shapes are obtained even for alcohols. Septum injections of up to 0.30 cm3 of headspace can be performed with no noticeable effect on column efficiency, hence the sensitivity is at least as good as that attainable with a packed column. The aim of this paper is to present retention1112 ANALYST, JULY 1992, VOL. 117 Table 1 Some volatile compounds which may be abused by inhalation (cfi, ref. 2) Aliphatic hydrocarbons- Acetylene (ethyne) Butane* Isobutane (2-methylpropane)* Hexanet Propane* Benzene (benzol) Toluene (methylbenzene, phenylmethane, toluol) Xylene (dimethylbenzene, xylol)$ Mixed hydrocarbons- Petrol (gasoline)$ Petroleum ethers (light petro1eums)fl Bromochlorodifluoromethane (BCF, FC 12B1) Carbon tetrachloride (tetrachloromethane) Chlorodifluoromethane (FC 22) Chloroform (trichloromethane) Dichlorodifluoromethane (FC 12) Dichloromethane (methylene chloride) 172-Dichloropropane (propylene dichloride) Enflurane (2-chloro-l,1 ,2-trifluoroethyl difluoromethyl ether) Fluorotrichloromethane (FC 11) Halothane (Zbromo-2-chloro-1 , I, 1-trifluoroethane) Isoflurane (l-chloro-2,2,2-trifluoroethyl difluoromethyl ether) Methoxyflurane (2,2-dichloro-1 ,l-difluoroethyl methyl ether) Monochloroethane (ethyl chloride) Tetrachloroethylene (perchloroethylene) 1,l71,2-Tetrafluoroethane (FC 134a) 1,1,l-Trichloroethane (methylchloroform, Genklene) Trichloroethylene (trike, Trilene) 1,1,2-Trichlorotrifluoroethane (FC 113) Acetone (dimethyl ketone, propanone) Butanone (butan-2-one7 methyl ethyl ketone, MEK) Diethyl ether (ethoxyethane) Dimethyl ether (DME, methoxymethane) Ethyl acetate Methyl acetate Methyl tert-butyl ether (MTBE) Methyl isobutyl ketone (MIBK, isopropylacetone, 4-methylpen- Nitrous oxide (dinitrogen monoxide , 'laughing gas') * Components of liquefied petroleum gas (LPG).t Commercial hexane is a mixture of hexane and heptane with small amounts of some higher aliphatic hydrocarbons. $ Mainly m-xylene (173-dimethylbenzene). $ Boiling-point range 40-200 "C, atmospheric pressure. fl Mixtures of pentanes, hexanes, etc., with specified boiling-point Aromatic hydrocarbons- Halogenated compounds- Oxygen compounh- tan-2-one) ranges (e.g., 40-60 "C).and reproducibility data obtained using the SPB-1 capillary column and to illustrate the use of this column in case work. Experimental Gas Chromatography The gas chromatograph (Hewlett-Packard Model 5890) was fitted with a Hewlett-Packard splitless capillary septum injector. The column was a 60 m x 0.53 mm i.d. fused-silica capillary coated with SPB-1 (5 pm film) (Supelchem UK, Saffron Walden, Essex, UK). The carrier gas was helium (flow rate 8.6 cm3 min-1). The injector and detector temperatures were 150 and 275 "C, respectively. The column oven, after a 6 min isothermal period, was programmed from 40 to 80 "C at 5 "C min-1 and then to 200 "C at 10 "C min-1 (total analysis time 26 min).Detection was by split FID-63Ni constant- current ECD (SGE outlet splitter system OSS-2, splitting ratio about 5 : 1). The hydrogen and air (FID) inlet pressures were Tabie 2 Preparation of the qualitative standard mixture (a) Qualitative standard mixture*- Compound Volume addedcm3t BCF Butane Dimethyl ether Isobutane FC 11 FC 12 FC 113 Propane 0.005 $ 1.0 $ 0.02 0.3 0.5 $ (b) Liquid components mixture§- Compound Volume addedcm3 Acetone Butanone Carbon tetrachloride Chloroform Ethanol Ethylbenzene Halothane Hexane Methyl isobutyl ketone Propan-2-01 Tetrachloroeth ylene Toluene 1 , 1 , 1-Trichloroethane 171,2-Trichloroethane Trichloroethylene 2,2,2-Trichloroethanol 7.5 5.0 0.05 0.5 5.0 2.5 0.1 5.0 2.5 5.0 0.025 2.5 0.25 1 .o 0.25 0.015 * Prepared in 125 cm3 gas sampling bulb.7 Volume of vapour phase in headspace vial. $2.0 cm3 of commercial 'butane' added (cf. , Table 1). § Add 0.01 cm3 to mixture of gaseous components in gas sampling bulb as in (a). 100 kPa (15 lb in-2) and 150 kPa (22 lb in-z), respectively. The ECD purge (nitrogen) flow rate was about 35 cm3 min-l. Data capture was by means of Hewlett-Packard 3396A recording integrators. The column was conditioned by programming from 30 to 260 "C with carrier flow at 2 "C min-1 and held for 16 h before use. Establishment of Retention Data Pure compounds were initially sampled in the vapour phase. However, direct liquid injection was employed for compounds that were not sufficiently volatile at 60 "C. Generally the amount of compound injected was sufficient to give full-scale deflection (FSD) or thereabouts at the detector sensitivities normally used in sample analyses (FID 80 pA FSD, ECD 2 kHz FSD).The ECD responses were coded as nil (0), poor (1) or good (2). Retention times were measured from the injection point and were also calculated relative to 1,1,2- trichloroethane. Kovdts retention indices were calculated from the data generated in the temperature programme by applying the following equation during the isothermal period or during the individual ramps of the programme:4*5 where RZ(x) = retention index of x , z = number of carbon atoms in alkane eluting before x , RT(x) = retention time of x , RT(z) = retention time of z, and RT(z + 1) = retention time of alkane with z + 1 carbon atoms eluting after x .Qualitative Standard Mixture A qualitative standard mixture was analysed prior to sample analyses. This mixture was prepared by adding the appro- priate volume of the gaseous components [Table 2(a)] toANALYST, JULY 1992, VOL. 117 1113 a clean, sealed, 125 cm3 gas-sampling bulb (Supelco 2-2146) and adding a portion (0.01 cm3) of the liquid components stock mixture [Table 2(b)]. This latter mixture was stable for at least 6 months when stored in a glass-stoppered vessel at -5 to -20 "C. Internal Standard Solution Ethylbenzene and 1 , l ,Ztrichloroethane were obtained from BDH (now Merck) (Poole, Dorset, UK) and tested by GC for the presence of contaminants, especially toluene and l , l , l - trichloroethane, before use. Approximately 50 mg of each compound were measured into 50 cm3 glass calibrated flasks containing outdated blood-bank whole blood.After thorough mixing, 1.0 cm3 of the 1,1,2-trichloroethane solution and 2.5 cm3 of the ethylbenzene solution were diluted to 100 cm3 using outdated blood-bank whole blood4e-ionized water (1 + 24) to give the working internal standard solution. The final ethylbenzene and 1,l ,2-trichloroethane concentrations were about 25 and 10 mg dm-3, respectively. This solution remained usable for not less than 2 years if stored in 5 cm3 portions at -5 to -20 "C in screw-topped glass bottles. Sample Preparation Blood and urine Internal standard solution (0.20 cm3) was added to a 7 cm3 glass septum vial (Schubert, Portsmouth, Hampshire, UK) using an air-displacement pipette (Eppendorf; BDH). The vial was then sealed with a crimped-on PTFE-lined silicone disc (Kontron, St.Albans, Hertfordshire, UK). The vial was incubated at 65 "C in a heating block and, after 15 min, a portion (0.10-0.30 cm3) of the headspace was taken using a gas-tight glass syringe (SGE), which had been warmed by being placed on the heating block (10 min), and injected onto the column. Subsequently, the sample (whole blood, plasma, serum or urine) (0.20 cm3) was added to the sealed vial using a 1.0 cm3 plastic disposable syringe fitted with a 1 in, 25 gauge Luer needle and, after at least 15 min, a second portion of the headspace was taken for analysis. After the injection the plunger was removed from the gas-tight syringe and the assembly placed on the heating block until the next injection to ensure evaporation of any remaining analytets).The syringe was rinsed occasionally with methanol to remove deposits and again allowed to dry before use. Tissues Samples of solid tissues were analysed as above after adding a proteolytic enzyme to the incubation mixture. Thus, 20-50 mg wet mass of tissue were dissected from the centre of the specimen, preferably whilst the specimen was frozen. Dupli- cate portions of the specimen were incubated (65 "C, 15 min) with internal standard solution (200 mm3) and about 1 mg of Subtilisin A (Novo, Windsor, Berkshire, UK) prior to the analysis of 0.10-0.30 cm3 of headspace as described above. The reagent blank analysis was performed in a separate vial. Products It is very important that all products sent for analysis are packaged and stored entirely separately from biological samples to prevent cross-contamination.Aerosols and fuel gases were analysed after releasing a portion of the product into a headspace vial. Adhesives and similar products were usually introduced into a glass vial. The vial was sealed and, after 1-15 min, a portion (0.05-0.10 cm3) of the headspace was taken for analysis. Liquids were analysed in the same way except that it was often possible to withdraw a portion (0.0054.05 cm3) of the headspace directly from the container. Quantitative Analyses Quantitative assays were performed in duplicate either isothermally or with a temperature programme and using the appropriate detector. If concentrations of ECD-responding compounds were very high it was sometimes more convenient to use FID for quantitative work.Assay calibration was by analysis of standard solutions prepared as described below; the same solutions were used in the analysis of blood and of tissue digests. Analyte concentrations in the range 0.1-10 or 0.5-50 mg dm-3 were usually adequate in cases of acute poisoning. Liquid analytes Calibration solutions were prepared by adding a known volume of the analyte to a calibrated flask containing 'blank' blood using a positive-displacement pipette and ascertaining the exact amount added by weighing. Appropriate volume to volume dilutions were then performed, taking care to mini- mize losses of analyte by handling reagents and glassware at 4 "C and storing samples and standards at 4 "C with minimal headspace.6 Small (2 cm3) glass vials with caps lined with aluminium foil are convenient for performing standard dilutions.Portions of the standards were transferred into headspace vials for analysis as described above and a calibration graph of peak height ratio of analyte to internal standard against analyte concentration was prepared. Often either 1,1,2-trichloroethane or ethylbenzene could be used as the internal standard. Carbon tetrachloride and l,l,l-tri- chloroethane were best determined isothermally at a column temperature of 120 "C, while tetrachloroethylene, toluene, 2,2,2-trichloroethanol and trichloroethylene were best deter- mined at 140 "C. Gaseous analytes Calibration mixtures were prepared directly into headspace vials.7 Septum bottles of about 125 cm3 capacity were calibrated by weighing the amount of de-ionized water each could contain.Each bottle was then dried and filled with nitrogen. A piece of aluminium foil (about 1 cm2) was added to aid mixing and the vial was sealed. After recording atmospheric pressure, the pure analyte, usually supplied in a small cylinder, was transferred into a 125 cm3 glass gas sampling bulb fitted with a septum port (Supelco 2-2146) at atmospheric pressure. An appropriate volume of vapour was then taken from the gas sampling bulb using a gas-tight syringe and added to a calibrated septum bottle. Care was taken to ensure that the contents of all vessels remained at atmospheric pressure. Thus, if the volume of gas transferred was greater than 0.1 cm3 a short vent needle was inserted through the septum well away from the point of the gas-tight syringe needle.After thorough mixing, further dilutions were pre- pared as required. Finally, known volumes of diluted analyte vapour were transferred using a gas-tight syringe into head- space vials containing the same volume of 'blank' blood as used in sample analyses. A constant volume of appropriately diluted internal standard (2,2-dimethylpropane for butane) vapour was also added to the sample and standard vials. Results and Discussion Retention and Relative Detector Response Even with sub-ambient operation and temperature program- ming, many capillary columns elute very volatile compounds too quickly if less volatile components of interest are to elute at reasonable retention times.However, the 60 m thick-film SPB-1 column retained and resolved many very volatile compounds at an initial temperature of 40 "C while allowing a total analysis time of only 26 min. The reductions in costs and in the time taken in recycling which arise directly from the use of this relatively high starting temperature are considerable,1114 ANALYST, JULY 1992, VOL. 117 t % 2 e 0 P LL t fn 0 P 2 n ow 21 I 24 1 16 17 15 18 0 5 10 15 20 25 Timelmin Fig. 2 a ) and ( b ) Analysis of the qualitative standard mixture (cf., Table 25 with detector sensitivities (FSD) (a) FID 3.2 nA and ( 6 ) ECD 64 kHz. Column, 60 m X 0.53 mm i.d. SPB-1 (5 pm film); oven temperature, 40 "C (6 min), then to 80 "C at 5 "C min-1, then to 200 "C at 10 "C min-1; and injection volume, =0.010 cm3.Peaks: 1 = propane, 2 = FC 12,3 = dimethyl ether, 4 = isobutane, 5 = butane, 6 = BCF, 7 = ethanol, 8 = acetone, 9 = propan-2-01,lO = FC 11,11 = FC 113, 12 = halothane, 13 = butanone, 14 = hexane, 15 = chloroform, 16 = 1,1,l-trichloroethane, 17 = carbon tetrachloride, 18 = trichloroethylene, 19 = meth 1 isobutyl ketone, 20 = 1,1,2- trichloroethane (internal standardr, 21 = toluene, 22 = tetrachlo- roethylene, 23 = 2,2,2-trichloroethanol, and 24 = ethylbenzene (internal standard) especially if liquid carbon dioxide cooling would otherwise have been necessary. It is of interest that Pekari et a1.8 used two linked 30 m x 0.53 mm i.d. capillaries (both 2.65 pm film thickness) coated with 100% dimethylpolysiloxane and 5% phenyl-95% dimethylpolysiloxane, respectively, in the deter- mination of benzene and toluene in blood headspace. They found that a programmed run to 200 "C was needed to obtain optimum sensitivity and selectivity, but that the use of the effectively 60 m 'medium film' column allowed a starting temperature of 50 "C to be used.The programme was run each day before undertaking sample analyses in order to remove any contaminants that had accumulated since the system was last used. The analysis of the qualitative standard mixture (Table 2) is illustrated in Fig. 2. Note especially the good peak shapes given by ethanol and propan-2-01 and the absence of a peak derived from the septum (cf., ref. 3). No deterioration in peak shape has been observed in routine use over a 1 year period. The SPB-1 column is operated well below its maximum recommended temperature (320 "C) so the column life should be long.Retention and detector response data for 244 compounds are given in Table 3. Compounds that were injected as liquids are identified by asterisks; these substances have been included in the database primarily to facilitate analysis of products and other non-biological specimens. Compounds that did not elute during the programme generally had boiling-points (atmo- spheric pressure) at 170 "C or above and retention indices (n-alkane) on SE-30, OV-1 or OV-101 packed columns of lo00 or more (cf., Table 3). Amongst the compounds found not to elute were camphor, 1-chlorooctane, cycloheptanone, decane, 1 ,2-dichlorobenzene, 1,6dichlorobenzene, 2,6- dimethylheptan-4-one, ethchlorvynol, 2-ethylhexan-1-01, 2-ethylhexyl acetate, hexachloroethane, 4-methylbenzal- dehyde, N-methylformamide, nonan-2-one, nonan-5-one, octan-1-01 and octylamine. The injection of hydrogen (retention time 2.49 min, relative retention with respect to 1 ,2,2-trichloroethane 0.134) pro- vided a measure of the void volume of the system (retention time of methane 2.52 min; cf., Table 3).There was a slight difference in the absolute retention time of compounds which responded at each detector because of the presence of the effluent splitter. Retention times were calculated relative to 1 , l ,2-trichloroethane as this compound gave a response on both detectors at the sensitivities normally employed. However, the retention data quoted in Table 3 were derived from the ECD for compounds responding on that detector.The classification of ECD response (Table 3) is empirical and is simply a guide to aid peak assignment. Compounds responding strongly on the ECD were primarily halogenated substances, but many compounds containing nitro or keto moieties also responded. Other substances such as ally1 isothiocyanate and dinitrogen monoxide also showed a good response. In contrast, the response to some halogenated compounds such as 1,1,1,2-tetrafluoroethane was relatively poor and certain ketones, e.g., the heptanones, showed no response. The retention data have proved highly reproducible in routine use over a 6 month period (Table 4) and should be applicable to other SPB-1 columns (and indeed to other dimethylpolysiloxane-coated capillaries) of similar dimen- sions and film thickness.However, the carrier gas flow rate might have to be adjusted to give retention data identical with those given in Table 3. We find that the retention time is more convenient than the retention index when assessing retention of unknowns. Franke et al.9 and others have emphasized the value of retention indices when transferring GC retention data between systems. However, Franke et al. themselves used retention indices based on alkan-1-01s (primary n-alkane alcohols) rather than on n-alkanes when using columns packed with Carbopack materials in the analysis of solvents and other volatiles. This was a strange choice as primary alcohols often give tailing peaks on such columns and thus give poorly reproducible retention data.The Kovhts retention indices (n-alkane) calculated for each compound with the temperature programme on the SPB-1 column are given in Table 3(a). Literature values for Kovhts retention indices on SE-30, OV-1 or OV-101 packed col- umns,3,*0 if available, are also given in this table; if two different packed column retention indices were reported, the mean was taken. It was found that only compounds with a retention index of <loo0 eluted from the SPB-1 column with the temperature programme used. Therefore, in order to calculate the retention indices for compounds eluting between 23.56 min (retention time of nonane) and 26.00 min, it was necessary to ascertain the retention time of decane. This was measured by continuing the final ramp for a further 2 min (retention time of decane 26.06 min).The retention indices of homologous series of acetates, formates, primary alcohols, aldehydes and alk-1-enes on the SPB-1 column are plotted inANALYST, JULY 1992, VOL. 117 1115 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig. 2 for chromatographic conditions)* (a) Alphabetical order- Compound Acetaldehyde Acetone Acetonitrile Acetonylacetone: see Hexane-2,5-dione Acetylacetone: see Pentane-2,Cdione Acetylene Acrylonitrile Allyl glycidyl ether Allyl isothiocyanate Amy1 . . .: see Pentyl . . . 3CF: see Bromochlorodifluoromethane 3enzene 3enzaldehyde 3enzonitrile 3icyclo[4.3 .O]nonane 3romoacetonitrile 3romobenzene 3romochlorodifluoromethane 3romochloromethane !-Bromo-2-chloro- 1 , l , 1-trifluoroethane: 3romodichloromethane l-Bromo-2,3-epoxypropane 3 romo form 3romomethane 1-Bromopropane 3romotrichloromethane 3romotrifluoromethane 3utanal 3utane-2,3-dione 3utane 3utan-1-01 3utan-2-01 ert-Butanol: see 2-Methylpropan-2-01 3utanone 3ut- 1-ene 3utyl acetate 3utyl chloride: see 1-Chlorobutane 3utyl formate 3utyl iodide: see I-Iodobutane 3utyl nitrite 3utyraldehyde: see Butanal see Halothane Capronaldehyde: see Hexanal Caprylene: see Oct-1-ene Carbon disulfide Carbon tetrachloride Chloral hydrate 1-Chlorobutane Chlorbutol: see Chlorobutanol Chlorobenzene Chlorobutanol 2-Chloro-1,l-difluoroethane 2-Chloro-1 ,I-difluoroethylene Chlorodifluoromethane l-Chloro-2,3-epoxypropane Chloroethane: see Monochloroethane 2-Chloroethanol Chloroform 1-Chloro-2-met h ylbenzene 1-Chloro-3-methylbenzene 1-Chloro-4-methyl benzene 2-Chlorophenol 1-Chloropropane 2-Chloro-l,l,l-trifluoroethane l-Chlor0-2,2,2-trifluoroethyl difluoromethyl ether: 2-Chloro-l,1,2-trifluoroethyl difluoromethyl ether: Cryofluorane: see 1,2-Dichlorotetrafluoroethane see Isoflurane see Enflurane RT/min 3.59 5.66 5.22 2.63 6.50 22.62 22.31 14.39 24.75 25.16 25.48 17.65 24.45 4.07 11.40 16.16 19.54 22.82 4.47 12.24 18.98 2.77 9.98 9.72 4.09 14.08 10.80 10.18 3.94 20.36 17.02 12.41 8.03 14.70 16.60 13.64 21.90 25.44 3.41 3.46 3.14 16.42 13.57 11.65 24.96 25.03 25.05 25.83 8.31 3.73 RRT 0.192 0.303 0.279 0.141 0.348 1.211 1.192 0.770 1.325 1.347 1.364 0.943 1.306 0.217 0.609 0.863 1.044 1.219 0.239 0.654 1.014 0.148 0.534 0.519 0.219 0.754 0.578 0.545 0.211 1.090 0.911 0.663 0.429 0.785 0.887 0.730 1.170 1.359 0.182 0.185 0.168 0.877 0.725 0.622 1.336 1.337 1.343 1.380 0.444 0.199 ECD 0 1 0 0 0 0 2 0 0 0 0 2 2 2 2 2 2 2 2 2 2 2 0 2 0 0 0 1 0 0 0 2 1 2 2 2 1 2 1 1 2 2 2 2 1 1 1 2 1 1 Calc.352 460 443 165 492 869 585 655 948 965 977 725 936 398 598 690 773 875 414 614 859 222 568 563 400 650 585 572 386 795 708 617 527 661 698 641 844 976 335 339 309 695 640 603 956 959 960 992 533 365 Lit. 372 469 455 - 590 880 660 956 - - - 945 405 660 715 805 911 - - 810 - - - 400 651 624 579 390 794 701 608 524 659 705 642 860 949 - - - 720 643 605 - - - - 570 375 Retention index Formula mass 44.1 58.1 41.1 26.0 53.1 114.2 99.2 78.1 106.1 103.1 124.2 120.0 157.0 165.4 129.4 163.8 137.0 252.8 94.9 123.0 198.3 148.9 72.1 86.1 58.1 74.1 74.1 72.1 56.1 116.2 102.1 103.1 76.1 153.8 165.4 92.6 112.6 177.5 100.5 98.5 86.5 92.5 80.5 119.4 126.6 126.6 126.6 128.6 78.5 118.5 B.p.PC 21 57 82 -81 77 154 151 80 179 191 167 627 156 -3 68 90 136 149 -94 71 103 -58 75 89 -1 117 100 80 -6 125 107 78 47 77 98 79 131 167 - 10 - 18 -41 118 129 61 159 162 162 175 47 7 CAS Registry No.75-07-0 67-64-1 75-05-8 74-86-2 107-13-1 106-92-3 57-06-7 71-43-2 100-52-7 100-47-0 496-10-6 590-17-0 108-86-1 353-59-3 74-97-5 75-27-4 3132-64-7 75-25-2 74-83-9 106-94-5 75-62-7 75-63-8 123-72-8 43 1-03-8 106-97-8 71-36-3 78-92-2 78-93-3 106-98-9 123-86-4 592-84-7 544-16-1 75-15-0 56-23-5 302-17-0 106-69-3 108-90-7 57-15-8 75-68-3 359-10-4 75-45-6 106-89-8 107-07-3 67-66-3 95-49-8 106-41-8 106-43-4 95-57-8 540-54-5 75-88-71116 ANALYST, JULY 1992, VOL.117 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig. 2 for chromatographic conditions) *-continued (a) Alphabetical order- Compound Cumene Cyanogen bromide Cyclohexane Cyclohexanol Cyclohexanone Cyclohexene C yclopropane Diacetone alcohol: see 4-Hydroxy-4-methylpentan-2-one Diacetyl: see Butane-2,3-dione Dibromodifluoromethane 1 ,ZDibromoethane Dibromomethane 2,2-Dichloro- 1,l-difluoroethyl methyl ether: see Methoxyflurane Dichlorodifluoromethane 1,l-Dichloroethane 1,2-Dichloroethane 1,l-Dichloroethylene 1,2-Dichloroethylene (both isomers) Dichloromethane 1,2-Dichloropropane 1,3-Dichloropropane 1,3-Dichloropropan-2-ol 1,2-Dichlorotetrafluoroethane Diethylamine Diethyl ether Diethyl ketone: see Pentan-3-one 1,l-Difluorotetrachloroethane 1,2-Difluorotetrachloroethane Diisopropyl ether Dimethoxymethane: see Methylal N, N-Dimeth ylacetamide Dimethyl disulfide: see Methyl disulfide Dimethyl ether N, N-Dimeth ylformamide 2,5-Dimethylfuran Dimethyl ketone: see Acetone 2,2-Dimethylpropane 2,6-Dimethylpyridine Dimethyl sulfide: see Methyl sulfide Dimethyl sulfoxide Dinitrogen monoxide: see Nitrous oxide Dipropyl ketone: See Heptan-4-one D i o x a n e 1,3-Dioxolane DME: see Dimethyl ether DMF: see N,N-Dimethylformamide DMSO: see Dimethyl sulfoxide Enflurane Epibromohydrin: see l-Bromo-2,3-epoxypropane Epichlorohydrin: see l-Chloro-2,3-epoxypropane 1,2-Epoxybutane Ethane Ethanol EthanolamineS 2-Ethoxyethanol 2-Ethoxyethyl acetate Ethyl acetate Ethylamine Ethylbenzene Ethyl Cellosolve: see 2-Ethoxyethanol Ethyl chloride: see Monochloroethane Ethylene Ethylene chlorohydrin: see 2-Chloroethanol Ethylene glycol$ Ethylene oxide Ethyl formate Ethyl iodide: see Iodoethane CEthylrnorpholine RT/min 24.19 6.25 14.91 22.77 22.90 15.69 8.29 6.12 20.20 15.80 3.18 9.57 13.09 7.28 9.19 7.45 15.87 19.20 22.74 3.59 9.53 6.69 16.77 16.90 11.43 21.61 3.34 18.47 16.53 4.32 22.62 19.97 16.16 11.13 6.14 10.63 2.69 4.80 13.75 16.38 22.83 11.42 4.30 22.38 2.63 15.15 4.22 6.82 22.68 RRT 1.295 0.334 0.798 1.219 1.226 0.840 0.444 0.327 1.079 0.844 0.170 0.511 0.699 0.389 0.491 0.398 0.848 1.026 1.215 0.192 0.510 0.358 0.896 0.903 0.612 1.157 0.179 0.989 0.885 0.231 1.211 1.069 0.865 0.596 0.328 0.569 0.144 0.257 0.736 0.877 1.222 0.611 0.230 1.198 0.141 0.811 0.226 0.365 1.214 ECD 0 2 0 0 1 0 0 2 2 2 2 2 2 2 2 2 1 2 2 2 0 0 2 2 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Calc. 925 482 666 874 878 681 533 477 790 683 313 559 630 512 552 515 684 765 873 352 559 499 702 704 598 835 328 746 697 409 869 784 690 592 478 582 200 427 643 694 876 598 408 861 165 670 405 52 1 871 Lit.905 664 899 875 703 - - - 823 733 305 563 63 1 556 515 - - - 885 361 515 785 730 594 - - - - 697 - - - 696 635 462 600 200 42 1 780 701 874 596 860 - - 772 400 545 - Retention index Formula mass 120.2 105.9 84.2 100.2 98.1 82.1 42.1 209.8 187.9 173.8 120.9 99.0 99.0 96.9 96.9 84.9 113.0 113.0 129.0 170.9 73.1 74.1 203.8 203.8 102.2 87.1 46.1 73.1 96.1 72.2 107.1 78.1 88.1 74.1 184.5 72.1 30.1 46.1 61.1 90.1 132.2 88.1 45.1 106.2 28.1 62.1 44.1 74.1 115.2 B .p.PC 152 62 81 161 156 83 -33 25 133 98 - 30 58 83 61 60 40 96 122 174 4 56 35 92 93 69 167 - 24 153 94 10 144 191 101 76 57 65 -88 79 70 135 156 77 17 136 -104 198 11 53 139 CAS Registry No. 98-82-8 506-68-3 110-82-7 108-93-0 108-94- 1 110-83-8 75-19-4 75-61-6 106-93-4 74-95-3 75-71-8 75-34-3 107-06-2 156-59-2 540-59-0 75-09-2 78-87-5 142-28-9 96-23-1 76- 14-2 109-89-7 60-29-7 76-11-9 76-12-0 108-20-3 127-19-5 115-10-6 68-12-2 625-86-5 463-82-1 108-48-5 67-68-5 123-91-1 646-06-0 13838-16-9 106-88-7 74-84-0 64- 17-5 141-43-5 110-80-5 11 1-15-9 141-78-6 75-04-7 100-41-4 74-85-1 107-21-1 75-21-8 109-94-4 100-74-3ANALYST, JULY 1992, VOL.117 1117 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig.2 for chromatographic conditions)*-continued (a) Alphabetical order- Retention index CAS RT/min RRT ECD Calc. Lit. mass B.p./”C No. Formula Registry 99 105-37-3 16.42 0.879 0 695 679 102.1 Compound Ethyl propionate Ethyl propyl ketone: see Hexan-3-one FC 11: see Fluorotrichloromethane FC 12: see Dichlorodifluoromethane FC 12B1: see Bromochlorodifluoromethane FC 22: see Chlorodifluoromethane FC 112: see 1,2-DifluorotetrachIoroethane FC 113: see 1,1,2-Trichlorotrifluoroethane FC 114: see 1,2-Dichlorotetrafluoroethane FC 134a: see 1,1,1,2-Tetrafluoroethane FC 142: see 2-Chloro-1 ,l-difluoroethane Fluorotrichloromethane Formaldehyde Formaldehyde dimethyl acetal: see Methylal Furfural 6.13 0.327 2.86 0.153 2 0 478 484 137.4 24 75-69-4 247 - 30.0 -20 50-00-0 806 825 96.1 162 98-01 - 1 20.74 1.108 1 Halothane 2,2,3,3,4,4,4-Heptafluorobutan-l-ol Heptanal Heptane Heptan- 1-01 Heptan-2-one Heptan-3-one Heptan-4-one Hept-l-ene Hexahydroindan: see Bicyclo[4.3.0]nonane Hexanal Hexane Hexane-2S-dione Hexan-1-01 Hexan-2-01 Hexan-2-one Hexan-3-one Hex-l-ene Hexyl acetate Hexyl formate 4-Hydroxy-4-methylpentan-2-one 8.76 0.468 9.25 0.494 23.05 1.234 16.70 0.894 24.88 1.332 22.79 1.220 22.62 1.211 22.27 1.192 16.18 0.866 543 553 883 700 953 874 869 857 690 533 883 700 955 880 857 - - - 197.4 200.1 114.2 100.2 116.2 114.2 114.2 144.2 98.2 50 103 153 98 176 152 148 144 94 151-67-7 375-01-9 11 1-71-7 142-82-5 11 1-70-6 110-43-0 106-35-4 123-19-3 592-76-7 19.82 1.061 11.51 0.616 23.25 1.242 22.12 1.184 19.95 1.068 19.45 1.041 19.33 1.035 10.89 0.583 25.83 1.383 23.84 1.276 21.17 1.131 0 0 1 0 0 0 0 0 0 0 1 78 1 600 890 852 784 77 1 768 587 992 911 820 - 600 894 860 786 787 78 1 - 100.2 86.2 114.1 102.2 102.2 100.2 100.2 84.2 144.2 130.2 116.2 130 69 188 157 140 127 124 64 170 155 168 66-25- 1 110-54-3 110-13-4 111-27-3 626-93-7 591-78-6 589-38-8 592-4 1-6 142-92-7 629-33-4 123-42-2 - 82 1 l-Iodobutane Iodoet hane Iodomethane l-Iodopropane Isoamyl... : see Isopentyl.. . Isobutane Isobutanol: see 2-Methylpropan-1-01 Isobutyl acetate Isobutylamine Isobutyl nitrite Isoflurane Isooctane Isopentyl acetate Isopentyl alcohol: see 3-Methylbutan-1-01 Isopentylamine Isopentyl nitrite (‘amyl nitrite’) Isoprene Isopropanol : see Propan-2-01 Isopropyl acetate Isopropylacetone: see Methyl isobutyl ketone Isopropyl formate Isopropyl nitrate Isovaleraldehyde: see 3-Methylbutanal 20.91 1.117 11.55 0.617 7.13 0.381 16.90 0.903 81 1 840 184.0 130 542-69-8 601 680 156.0 72 75-03-6 509 515 142.0 43 74-88-4 107-08-4 705 785 170.0 102 3.61 0.193 354 370 58.1 -12 75-28-5 0 19.00 1.017 10.95 0.586 10.37 0.555 5.52 0.295 16.31 0.873 22.34 1.196 759 754 116.2 118 110-19-0 588 - 73.1 69 78-81-9 576 - 103.1 67 542-56-3 454 - 184.5 49 26675-46-7 692 725 114.2 99 540-84- 1 884 130.2 142 123-92-2 859 16.42 0.879 15.87 0.848 6.91 0.370 0 2 0 695 - 87.2 95 107-85-7 684 680 117.2 98 110-46-3 504 - 68.1 34 78-79-5 14.10 0.755 0 650 648 102.1 89 108-21 -4 9.55 0.511 14.84 0.793 0 2 559 567 88.1 68 625-55-8 664 693 105.1 103 1712-64-7 Limonene 25.69 1.375 0 986 1053 136.2 177 138-86-3 MEK: see Butanone Meparfynol: see 3-Methylpent-l-yn-3-01 2-Mercaptoethanol Methane Methanol 2-Methoxyethanol 17.63 0.942 2.52 0.135 3.60 0.192 12.38 0.663 724 795 78.1 158 60-24-2 100 100 16.0 -161 74-82-8 353 49 1 32.0 65 67-56- 1 617 616 76.1 124 109-86-41118 ANALYST, JULY 1992, VOL.117 Table 3 Retention and relative detector resnonse data on the SPB-1 column system (see legend to Fig. 2 for chromatographic 1 conditions)*--continued (a) Alphabetical order- Compound Methoxyflurane Methyl acetate Methylal Methyl bromide: see Bromomethane 2-Methylbuta-l,3-diene: see Isoprene 3-Methylbutanal 2-Methylbutan-1-01 2-Methylbutan-2-01 3-Methylbutan-1-01 3-Methylbutan-Zone: see Methyl isopropyl ketone Methyl tert-butyl ether Methyl butyl ketone: see Hexan-Zone Methyl butyrate Methyl Cellosolve: see 2-Methoxyethanol Methylchloroform: see 1,l ,l-Trichloroethane Methyl cyanide: see Acetonitrile Meth ylcyclohexane Methylcyclopentane Methyl cyclopropyl ketone Methyl disulfide Methylene chloride: see Dichloromethane Methyl ethyl ketone: see Butanone Methyl formate 6-Methylhept-S-en-2-0ne 2-Methylhex-l-ene Methyl hexanoate Methyl iodide: see Iodomethane Methyl isobutyl ketone Methyl isopropyl ketone Methyl methacrylate 2-Methylpentane 3-Methylpentane 2-Methylpentan-2-01 4-Methylpentan-2-one: see Methyl isobutyl ketone Methylpentynol: see 3-Methylpent-l-yn-3-01 3-Methylpent-l-yn-3-01 2-Methylpropanal 2-Methylpropan-1-01 2-Methylpropan-2-01 Methyl propionate 2-Methylpropylamine: see Isobutylamine Methyl propyl ketone: see Pentan-2-one l-Meth ylpyrrole Methyl sulfide MIBK: see Methyl isobutyl ketone Monochloroethane Morpholine MTBE: see Methyl terr-butyl ether Neopentane: see 2,ZDimethylpropane Ni troethane Nitromethane l-Nitropropane 2-Nitropropane Nitrous oxide Nonane Octanal Octane Octan-2-01 Octan-Zone Octan-3-one Octan-4-one Oct-l-ene Oct-2-yne Paralde h yde Pentane-2,3-dione Pentane-2,Qdione Pentanal Pentane RT/min 17.04 7.30 7.08 13.43 17.63 12.96 17.45 9.60 16.94 17.82 13.11 16.85 18.07 4.28 25.24 16.05 25.50 17.60 13.71 16.48 9.90 10.61 17.45 16.42 8.46 12.22 7.14 12.33 17.56 7.12 4.77 19.77 12.47 7.98 17.00 15.27 2.66 23.56 25.70 20.57 25.65 25.39 25.29 24.99 20.17 22.57 19.28 15.37 19.17 15 .so 6.72 RRT 0.910 0.391 0.379 0.719 0.944 0.694 0.934 0.514 0.907 0.954 0.702 0.902 0.965 0.229 1.351 0.859 1.365 0.942 0.734 0.882 0.530 0.568 0.934 0.879 0.453 0.654 0.382 0.660 0.940 0.381 0.255 1.056 0.666 0.427 0.907 0.816 0.142 1.261 1.376 1.101 1.373 1.359 1.354 1.338 1.080 1.208 1.032 0.821 1.026 0.830 0.360 ECD 2 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 2 1 2 1 2 2 2 0 0 0 0 0 0 0 0 0 0 1 0 0 0 Retention index Formula Calc.Lit. 617 512 508 637 724 628 719 560 706 729 631 704 735 407 968 687 978 723 642 696 566 581 719 695 536 614 509 616 722 508 426 779 618 526 708 672 182 900 986 800 984 974 970 958 790 867 767 674 764 677 500 701 513 505 - - 636 720 - 723 748 650 730 - 499 725 - - 724 650 699 610 725 - 715 619 512 639 - 715 - 447 800 655 565 725 685 900 990 800 - - - - - 790 870 771 681 790 500 - mass 165.0 74.1 76.1 86.1 88.2 88.2 88.2 88.2 102.1 98.2 84.2 84.1 94.2 60.1 126.2 98.2 130.2 100.2 86.0 loo.1 86.2 86.2 102.2 98.1 72.1 74.1 74.1 88.1 81.1 62.1 64.5 87.1 75.1 61 .O 89.1 89.1 44.0 128.3 128.2 114.2 130.2 128.2 128.2 128.2 112.2 110.2 132.2 100. 1 100.1 86.1 72.2 B.p.PC 105 57 42 93 128 103 132 55 102 101 73 114 110 32 58 92 151 118 93 100 60 64 124 121 64 108 82 80 111 36 12 129 115 101 132 120 - 88 151 68 126 179 173 169 164 122 138 124 115 141 103 36 CAS Registry No. 7 6 - 3 8 - 0 79-20-9 109-87-5 590-86-3 137-32-6 75-85-4 123-5 1-3 1634-04-4 623-42-7 108-87-2 96-37-7 765-43-5 624-92-0 107-31-3 110-93-0 6094-02-6 106-70-7 108-10- 1 563-80-4 80-62-6 107 - 8 3 - 5 96-14-0 590-36-3 77-75-8 74-84-2 78-83-1 75-65-0 554-12-1 96-54-8 75-18-3 75-00-3 110-91-8 79-24-3 75-52-5 108-03-2 79-46-9 10024-97-2 11 1-84-2 124- 13-0 111-65-9 123-93-6 111-13-7 106-68-3 11 1-66-0 123-63-7 600-14-6 123-54-6 110-62-3 109-66-0ANALYST, JULY 1992, VOL.117 1119 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig. 2 for chromatographic conditions)*-continued (a) Alphabetical order- Compound Pentan-1-01 Pentan-2-one Pentan-3-one Pent-1-ene Pentyl acetate Pentyl formate tert-Pentyl alcohol: see 2-Methylbutan-2-01 Perchloroethylene: see Tetrachloroethylene Perfluoropropane a-Pinene Piperidine Propanal Propane Propane-172-diol$ Propane-173-diol$ Propan-1-01 Propan-2-01 Propanone: see Acetone Propionaldehyde: see Propanal Propionitrile Propyl acetate Propylamine Propyl bromide: see 1-Bromopropane Propyl chloride: see 1-Chloropropane Prop ylene Propyl formate Propyl iodide: see 1-Iodopropane Pyridine Pyrrole Pyrrolidine Styrene Sulfur hexafluoride 1,1,2,2-Tetrabromoethane 1,1,1,2-TetrachIoroethane 1,1,2,2-Tetrachloroethane Tetrachloroe thylene Tetrachloromethane: see Carbon tetrachloride 1,1,1,2-Tetrafluoroethane Tetrahydrofuran THF: see Tetrahydrofuran Toluene Tribromomethane: see Bromoform 1 , 1,l-Trichloroethane 1,l ,2-Trichloroethane 2,2.2-Trichloroethanol Trichloroethylene Trichloromethane: Gxhloroform RT/min 18.74 15.07 15.58 6.28 23.31 20.87 6.08 24.81 18.81 5.77 3.05 17.09 20.29 8.61 6.04 8.57 16.51 7.60 2.99 11.84 17.65 17.97 15.21 23.18 2.58 25.50 21.85 23.33 20.89 2.76 12.31 19.14 13.56 18.72 22.29 16.25 1,l,l-Trichloro-2-methyli>\ropan-2-ol: see Chlorobutanol l72,3-Trich1oropropane 23.53 1 , 1 , 1-Trichloropropan-2-01 23.85 1 , 1,l -Trichlorotrifluoroethane 7.90 1 , 1,2-Trichlorotrifluoroethane 8.01 Trieth ylamine 15.84 2,2,2-Trifluoroethanol 5.17 2,2,2-Trifluoroethyl chloride: see 2-Chloro- Trifluoromethyl bromide: see Bromotrifluoromethane Trimethylene: see Cyclopropane 2,2,4-Trimethylpentane: see Isooctane 1,l ,l-trifluoroethane Valeraldehyde: see Pentanal y-Valerolactone Vinyl chloride Vin ylidine chloride m-Xylene o-Xylene p-Xylene RRT 1.003 0.807 0.834 0.336 1.248 1.117 0.325 1.328 1.007 0.309 0.163 0.915 1.086 0.461 0.323 0.459 0.884 0.407 0.160 0.634 0.945 0.962 0.814 1.241 0.138 1.362 1.167 1.246 1.116 0.148 0.659 1.025 0.724 1 .ooo 1.191 0.868 1.257 1.274 0.422 0.428 0.848 0.276 ECD 0 1 1 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 2 2 2 1 0 0 2 2 2 2 2 2 2 2 0 1 Retention index Formula Calc .753 669 678 483 892 810 476 950 755 464 300 710 793 539 474 539 696 518 283 606 725 733 671 887 135 978 843 892 81 1 219 615 763 639 752 858 69 1 899 912 525 527 684 441 Lit. 763 680 683 - - 772 - 942 782 300 745 820 57 1 530 - 580 696 - 3 10 603 725 755 695 890 - - 870 905 807 - 638 768 634 727 859 710 910 920 530 555 580 - mass 88.2 86.1 86.1 70.1 130.2 116.1 188.0 136.2 85.2 58.1 44.1 76.1 76.1 60.1 60.1 55.1 102.1 59.1 42.1 88.1 79.1 67.1 71.1 104.1 146.1 345.7 167.9 167.9 165.9 102.0 72.1 92.1 133.4 133.4 149.4 131.4 147.4 163.4 187.4 187.4 101.2 100.0 B.p.PC 138 102 102 30 149 132 - 39 156 106 49 - 42 187 214 97 83 97 102 49 -48 81 115 130 89 145 -64 229 131 146 121 - 27 66 111 74 113 151 87 156 162 46 48 90 75 CAS Registry No.71-41-0 107-87-9 96-22-0 109-67-1 628-63-7 638-49-3 76-19-7 80-56-8 110-89-4 123-38-6 74-98-6 57-55-6 504-63-2 71-23-8 67-63-0 107-12-1 109-60-4 107-10-8 1 15-07- 1 110-74-7 110-86- 1 109-97-7 123-75-1 100-42-5 2551-62-4 79-27-6 630-20-6 79-34-5 127-18-4 811-97-2 109-99-9 108-88-3 71-55-6 79-00-5 115-20-8 79-01-6 96-18-4 76-00-6 354-58-5 76- 13- 1 12 1-44-8 75-89-8 23.98 1.281 1 917 921 100.1 218 108-29-2 6.10 0.326 2 476 440 62.5 -14 75-01-4 7.24 0.387 2 511 515 97.0 32 75-35-4 22.62 1.211 0 869 871 106.2 138 108-38-3 23.32 1.250 0 892 895 106.2 144 95-47-6 22.66 1.213 0 870 870 106.2 138 106-42-31120 ANALYST, JULY 1992, VOL. 117 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig.2 for chromatographic conditions)*-continued (b) Retention time order- Compound Methane Sulfur hexafluoride Acetylene Ethylene Nitrous oxide Ethane 1,1,1,2-Tetrafluoroethane Bromotrifluoromethane Formaldehyde Prop ylene Propane Chlorodifluoromethane Dichlorodifluoromethane Dimethyl ether 2-Chloro-l , 1-difluoroethane 2-Chloro-l , 1-difluoroethylene Acetaldehyde 1,2-Dichlorotetrafluoroethane Methanol Isobutane 2-Chloro-l,l,l-trifluoroethane But-1-ene Bromochlorodifluoromethane Butane Ethylene oxide Methyl formate Ethylamine 2,2-Dimethylpropane Bromomethane Monochloroethane Ethanol 2,2,2-Trifluoroethanol Acetonitrile Isoflurane Acetone Propanal Propan-2-01 Peffluoropropane Vinyl chloride Dibromodifluoromethane Fluorotrichloromethane Enflurane Cy anogen bromide Pent-1-ene Acrylonitrile Diethyl ether Pentane Ethyl formate Isoprene Methylal Methyl sulfide Iodomethane 2-Methylpropan-2-01 Vinylidine chloride 1,l-Dichloroethylene Methyl acetate Dichloromethane Prop ylamine l,l,l-Trichlorotrifluoroethane Nitromethane 1,1,2-Trichlorotrifluoroethane Carbon disulfide Cyclopropane 1-Chloropropane 2-Methylpropanal Propionitrile Propan-1-01 Halothane 1 ,ZDichloroethylene (both isomers) 2,2,3,3,4,4,4-Heptafluorobutan-l-o1 Diethylamine RT/ min 2.52 2.58 2.63 2.63 2.66 2.69 2.76 2.77 2.85 2.99 3.05 3.14 3.18 3.34 3.41 3.46 3.59 3.59 3.60 3.61 3.73 3.94 4.07 4.09 4.22 4.28 4.30 4.32 4.47 4.77 4.80 5.17 5.22 5.52 5.66 5.77 6.04 6.08 6.10 6.12 6.13 6.14 6.25 6.28 6.50 6.69 6.72 6.82 6.91 7.08 7.12 7.13 7.14 7.24 7.28 7.30 7.45 7.60 7.90 7.98 8.01 8.03 8.29 8.31 8.46 8.57 8.61 8.76 9.19 9.25 9.53 RRT 0.135 0.138 0.141 0.141 0.142 0.144 0.148 0.148 0.153 0.160 0.163 0.168 0.170 0.179 0.182 0.185 0.192 0.192 0.192 0.193 0.199 0.211 0.217 0.219 0.226 0.229 0.230 0.231 0.239 0.255 0.257 0.276 0.279 0.295 0.303 0.309 0.323 0.325 0.326 0.327 0.327 0.328 0.334 0.336 0.348 0.358 0.360 0.365 0.370 0.379 0.381 0.381 0.382 0.387 0.389 0.391 0.398 0.407 0.422 0.427 0.428 0.429 0.444 0.444 0.453 0.459 0.461 0.468 0.491 0.494 0.510 ECD 0 2 0 0 2 0 1 2 0 0 0 2 2 0 1 2 0 2 0 0 2 0 2 0 0 0 0 0 2 2 0 1 0 2 1 0 0 2 2 2 2 2 2 0 0 0 0 0 0 0 0 2 0 2 2 0 2 0 2 1 2 1 0 1 0 0 0 2 2 2 0ANALYST, JULY 1992, VOL.117 1121 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig.2 for chromatographic conditions)*-continued (b) Retention time order- Compound Isopropyl formate 1,l-Dichloroethane Methyl tert-butyl ether Butane-2,3-dione 2-Methylpentane B utanal Butanone Isobutyl nitrite 3-Methylpentane 1,2-Epoxybutane Butan-2-01 Hex-1-ene Isobutylamine 1,3-Dioxolane Bromochloromethane Ethyl acetate Diisopropyl ether Hexane Iodoethane Chloroform Propyl formate 2-Meth ylpropan-1-01 1-Bromopropane Te trahydrofuran Methyl propionate 2-Methoxyethanol Butyl nitrite Nitroet hane 2-Methylbutan-2-01 1,2-Dichloroethane Meth ylcyclopentane 3-Methylbutanal l,l,l-Trichloroethane 2-Chloroethanol 1-Chlorobutane Methyl isopropyl ketone EthanolamineS Butan- 1-01 Isopropyl acetate Benzene Carbon tetrachloride Isopropyl nitrate Cyclohexane pent an-2-one Ethylene glycol$ Pyrrolidine 2-Nitropropane Pentane-2,3-dione Pen tanal Pentan-3-one Cyclohexene Dibromomethane Trieth ylamine 1,2-Dichloropropane Isopentyl nitrite (‘amyl nitrite’) 2-Methylhex-1-ene Bromodichloromethane Dioxane Hept-1-ene Trichloroet h ylene Isooctane 2-Ethoxyethanol l-Chloro-2,3-epoxypropane Ethyl propionate Isopentylamine 3-Methylpent-1-yn-3-01 Methyl methacrylate Propyl acetate 2,5-Dimethylfuran Chloral hydrate Heptane RT/ min 9.55 9.57 9.60 9.72 9.90 9.98 10.18 10.37 10.61 10.63 10.80 10.89 10.95 11.13 11.40 11.42 11.43 11.51 11.55 11.65 11.84 12.22 12.24 12.31 12.33 12.38 12.41 12.47 12.96 13.09 13.11 13.43 13.56 13.57 13.64 13.71 13.75 14.08 14.10 14.39 14.70 14.84 14.91 15.07 15.15 15.21 15.27 15.37 15.50 15.58 15.69 15.80 15.84 15.87 15.87 16.05 16.16 16.16 16.18 16.25 16.31 16.38 16.42 16.42 16.42 16.42 16.48 16.51 16.53 16.60 16.70 RRT 0.511 0.511 0.514 0.519 0.530 0.534 0.545 0.555 0.568 0.569 0.578 0.583 0.586 0.596 0.609 0.611 0.612 0.616 0.617 0.622 0.634 0.654 0.654 0.659 0.660 0.663 0.663 0.666 0.694 0.699 0.702 0.719 0.724 0.725 0.730 0.734 0.736 0.754 0.755 0.770 0.785 0.793 0.798 0.807 0.811 0.814 0.816 0.821 0.830 0.834 0.840 0.844 0.848 0.848 0.848 0.859 0.863 0.865 0.866 0.868 0.873 0.877 0.877 0.879 0.879 0.879 0.882 0.884 0.885 0.887 0.894 ECD 0 2 0 2 0 0 1 2 0 0 0 0 0 0 2 0 0 0 2 2 0 0 2 0 0 0 2 2 0 2 0 0 2 2 2 1 0 0 0 0 2 2 0 1 0 0 2 1 0 1 0 2 0 1 2 0 2 0 0 2 0 0 2 0 0 0 0 0 1 2 01122 ANALYST, JULY 1992, VOL.117 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig. 2 for chromatographic conditions)*-continued (b) Retention time order- Compound 1,l-Difluorotetrachloroethane Methyl cyclopropyl ketone 1,2-DifluorotetrachIoroethane 1-Iodopropane Methyl butyrate I-Nitropropane Butyl formate Methoxyflurane Propane-1 ,2-diol$ 3-Meth ylbutan-l-ol 2-Methylpentan-2-01 1-Meth ylpyrrole Methyl isobutyl ketone 2-Mercaptoethanol 2-Methylbutan-1-01 Bromoacetonitrile Pyridine Methy lcyclohexane Pyrrole Methyl disulfide N, N-Dimethylformamide 1.1 ,2-Trichloroethane Pentan- 1-01 Piperidine Bromotrichloromethane Isobutyl acetate Toluene Pentane-2,4-dione 1,3-Dichloropropane Paraldehyde Hexan-3-one Hexan-2-one l-Bomo-2,3-epoxypropane Morpholine Hexanal Hexan-2-01 Dimethyl sulfoxide Oct-1-ene 1,2-Dibromoethane Propane-l,3-diol$ Butyl acetate Octane Furfural Pentyl formate Tetrachloroeth ylene 1-Iodobutane 4-Hydroxy-4-methylpentan-2-one N, N-Dimethylacetamide 1,1,1 ,ZTetrachloroethane Chlorobenzene Hexan- 1-01 Heptan-4-one 2,2 ,2-Trichloroethanol Allyl isothiocyanate Isopentyl acetate Ethylbenzene Oct-2-yne Allyl glycidyl ether 2,6-Dimethylpyridine Heptan-3-one rn-Xylene p-Xylene 4-Eth ylmorpholine 1,3-Dichloropropan-2-ol C yclohexanol Heptan-2-one Bromoform 2-Ethoxyethyl acetate Cyclohexanone Heptanal Styrene RTI min 16.77 16.85 16.90 16.90 16.94 17.00 17.02 17.04 17.09 17.45 17.45 17.56 17.60 17.63 17.63 17.65 17.65 17.82 17.97 18.07 18.47 18.72 18.74 18.81 18.98 19.00 19.14 19.17 19.20 19.28 19.33 19.45 19.54 19.77 19.82 19.95 19.97 20.17 20.20 20.29 20.36 20.57 20.74 20.87 20.89 20.91 21.17 21.61 21.85 21.90 22.12 22.27 22.29 22.31 22.34 22.38 22.57 22.62 22.62 22.62 22.62 22.66 22.68 22.74 22.77 22.79 22.82 22.83 22.90 23.05 23.18 RRT 0.896 0.902 0.903 0.903 0.907 0.907 0.911 0.910 0.915 0.934 0.934 0.940 0.942 0.942 0.944 0.943 0.945 0.954 0.962 0.965 0.989 1 .Ooo 1.003 1.007 1.014 1.017 1.025 1.026 1.026 1.032 1.035 1.041 1.044 1.056 1.061 1.068 1.069 1.080 1.079 1.086 1.090 1.101 1.108 1.117 1.116 1.117 1.131 1.157 1.167 1.170 1.184 1.192 1.191 1.192 1.196 1.198 1.208 1.211 1.211 1.211 1.211 1.213 1.214 1.215 1.219 1.220 1.219 1.222 1.226 1.234 1.241 ECD 2 0 2 2 0 2 0 2 0 0 0 0 1 1 0 2 0 0 0 1 0 2 0 1 2 0 0 0 2 0 0 0 2 1 0 0 0 0 2 0 0 0 1 0 2 2 1 0 2 1 0 0 2 2 0 0 0 0 0 0 0 0 0 2 0 0 2 0 1 0 0ANALYST, JULY 1992, VOL.117 1123 Table 3 Retention and relative detector response data on the SPB-1 column system (see legend to Fig. 2 for chromatographic conditions) *-continued (b) Retention time order- Compound Hexane-2,5-dione Pentyl acetate o-Xylene 1,1,2,2-Tetrachloroethane 1,2,3-Trichloropropane Nonane Hexyl formate 1,l ,l-Trichloropropan-2-ol y-Valerolactone Cumene Bromobenzene Benzaldehyde a-Pinene Heptan-1-01 1-Chloro-2-methylbenzene Octan-4-one 1-Chloro-3-methylbenzene 1-Chloro-4-methylbenzene Benzonitrile 6-Methylhept-5-ene-2-one Octan-3-one Octan-Zone Chlorobutanol Bicyclo[4.3.0]nonane Methyl hexanoate 1,1,2,2-Tetrabromoethane Octan-2-01 Limonene Octanal 2-Chlorophenol Hexyl acetate RT/ min 23.25 23.31 23.32 23.33 23.53 23.56 23.84 23.85 23.98 24.19 24.45 24.75 24.81 24.88 24.96 24.99 25.03 25.05 25.16 25.24 25.29 25.39 25.44 25.48 25.50 25.50 25.65 25.69 25.70 25.83 25.83 RRT 1.242 1.248 1.250 1.246 1.257 1.261 1.276 1.274 1.281 1.295 1.306 1.325 1.328 1.332 1.336 1.338 1.337 1.343 1.347 1.351 1.354 1.359 1.359 1.364 1.365 1.362 1.373 1.375 1.376 1.380 1.383 ECD 1 0 0 2 2 0 0 2 1 0 2 0 0 0 1 0 1 1 0 0 0 0 2 0 0 2 0 0 0 2 0 * RT = retention time; RRT = retention time relative to 1,1,2-trichloroethane (on the ECD channel for compounds responding on that channel); ECD = relative ECD response (0 = nil, 1 = poor, 2 = good); Retention index = KovAts retention index (Calc.= calculated on the SPB-1 system, Lit. = literature value on SE-30,OV-1 or OV-lOl;3JOB.p.= boiling-point at atmospheric pressure; CAS Registry No. = Chemical Abstracts Service Registry Number. t B.p. at 24 mmHg. $ Compounds injected as liquids. Fig. 3. All of the homologues gave straight-line plots with gradients identical with that given (by definition) by the n-alkanes. Homologues of secondary alcohols and methyl ketones gave results virtually identical with those of the aldehydes, chloride homologues to the primary alcohols and iodide homologues to the acetates studied, respectively. These results provide evidence of the reliability of the data generated with the temperature programme on the SPB-1 column. The retention indices for higher molecular mass compounds (retention indices in the range 1000-3300) have been com- pared on dimethylpolysiloxane-coated packed and capillary columns. Japp et aZ.11 studied 75 compounds and reported good correlations ( r = 0.995 or better) between packed column data and those obtained on capillary columns with inside diameters up to 0.53 mm, but did not comment on any systematic bias.In contrast, bra-Tamayo ef aZ.12 reported that in 88 out of 103 instances the retention index was longer on the 0.2 mm i.d. capillary column used. Others have discussed the reproducibiIity of retention data on dimethyl- polysiloxane-coated capillaries and have reported variations in retention time with amount of analyte injected,13 but this is probably attributable to overloading of the relatively narrow- bore, low film thickness capillaries used. The retention indices measured on the SPB-1 capillary are plotted against the packed column retention indices [Table 3(a)] in Fig.4. Although there was a good correlation between the two sets of data ( r = 0.983, n = 165), linear regression analysis gave an intercept of +42.2 retention index units on the ordinate (packed column data). The difference in the retention indices on the two systems (RIspB-l - RIpacked) is plotted against the retention indices on the SPB-1 column in Fig. 5. Clearly there is a tendency for the retention index to be longer on the packed column across the range of compounds studied. It is possible that either polar interactions with the support in the packed column or difficulties in measuring the retention of very volatile compounds on packed columns are responsible for this finding.Application to Sample Analyses The sample preparation procedure was that of Ramsey and Flanagan3 except that 0.20 rather than 0.10 cm3 of internal standard solution were used and it was not necessary to use nitrogen-filled vials. A typical ‘reagent blank’ analysis is illustrated in Fig. 6. Apart from the internal standards the only identifiable compounds were small amounts of methanol, which may have originated from the laboratory atmosphere, and chloroform, which probably arose from chlorination of the public water supply used to feed the laboratory de-ionizer. The analysis of a blood specimen from an adolescent who died after abusing ‘butane’ gas and vapour from a typewriter correcting fluid is shown in Fig. 7. The analysis of a blood specimen from a patient who died after inhaling vapour from an electrical component cleaner is illustrated in Fig.8. I t . is clear that the concentrations of the components of interest1124 Chloroform ANALYST, JULY 1992, VOL. 117 100 1 Table 4 Inter-assay reproducibility data (n = 30 in each instance) (GC conditions as in Fig. 2) Retention Relative Compound Propane FC 12 Dimethyl ether Isobutane BCF Butane Ethanol Acetone Propan-2-01 FC 11 FC 113 Halothane Butanone Hexane Chloroform l,l,l-Trichloroethane Carbon tetrachloride Trichloroethylene Methyl isobutyl ketone 1,1,2-Trichloroethane Toluene Tetrachloroethylene 2,2 ,2-Trichloroethanol Et h y lbenzene time/ min 3.050 3.179 3.338 3.613 4.065 4.091 4.795 5.664 6.036 6.128 8.013 8.763 10.173 11.512 11.651 13.559 14.703 16.245 17.602 18.720 19.143 20.892 22.292 22.377 retention RSD (%) time 0.27 0.67 0.25 0.24 0.90 0.23 0.23 0.20 0.23 0.71 0.51 0.19 0.14 0.11 0.19 0.14 0.12 0.09 0.06 0.06 0.05 0.05 0.07 0.04 0.163 0.170 0.179 0.193 0.217 0.219 0.257 0.303 0.323 0.327 0.428 0.468 0.545 0.616 0.622 0.724 0.785 0.868 0.942 1.OOO 1.025 1.116 1.191 1.198 RSD (Yo) 0.32 0.64 0.28 0.29 0.89 0.19 0.25 0.20 0.20 0.67 0.49 0.13 0.12 0.08 0.15 0.09 0.09 0.05 0.04 co.01 0.04 0.08 co.01 - 1000 800 X 7 600 .- C 0 $ 400 a .- +a a 200 0 1 2 3 4 5 6 7 8 9 No.of carbon atoms Fig. 3 Kovdts retention indices measured using the temperature programme on the SPB-1 column of homologous series: A, acetates; B, formates; C, alcohols; D, aldehydes; and E, alk-1-enes plotted against number of carbon atoms (cf., Table 3) 1200 1000 800 0 Y m P ct 600 400 8 I I 1 200 400 600 800 1000 RkP6-7 Fig.4 KovAts retention indices of compounds measured using the temperature programme on the SPB-1 column plotted against literature values on SE-30/OV-l/OV-101 oacked columns ( r = 0.983. I a 8 8 I -200 I I I I 200 400 600 800 1000 RkPB-1 Fig. 5 Plot of the difference in the Kovdts retention indices on the SPB-1 and packed columns against retention index on the SPB-1 column (cf., Fig. 4) t ln 0 P 2 0 L t ln 0 P 2 n u w Ethyl benzene (4 Methanol t I. P J I 1 1 I I 1,1,2-Trichloro. ethane I I 1 I I 0 5 10 15 20 25 Timelmi n Fig. 6 Analysis of the internal standard solution (GC conditions as in Fig. 2): sample volume, 0.20 cm3; injection volume, 0.30 cm3 headspace; and detector sensitivities (FSD): (a) FID 80 pA and (b) ECD 2 kHz were well above the limit of detection of the system.Indeed, although no formal studies have been performed, the sensitiv- ity attainable appears to be similar to that obtained using an FID-ECD splitting ratio of 10 : 1 with the modified Carbopack packed column system, i.e., of the order of 0.01 mg dm-3 for ECD-responding compounds and 0.1 mg dm-3 for the remainder.3 Hence sensitivity enhancement either by 'salting- out' or the use of purge-and-trap devices is unnecessary when working with clinical or forensic specimens. The compounds studied included those listed in Table 1 and other common halons, solvents and metabolites and products of putrefaction such as methyl sulfide. Of the commonly encountered compounds only isobutane-methanol and tolu- ene-paraldehyde are at all difficult to resolve.If paraldehyde is suspected the addition of 6 mol dm-3 sulfuric acid (0.20 cm3) to the vial followed by reincubation should remove the paraldehyde peak and lead to an increase in the acetaldehyde n = 165; cf., Table 3). Solid line: y = x ' peak. Measirement of released acetaldehyde has bkenANALYST, JULY 1992, VOL. 117 1125 t v) 0 P E 0 LL t v) 0 9. E n u w :a) sobutan ’ropane 4 3utane Ethylbenzene 1 I- 1,1,2-Trichloro- ethane 1,1,1 -Tric hloro- ethane 0 5 10 15 20 25 Ti me/m i n Fig. 7 Analysis of a whole blood specimen (0.20 cm3) from a patient who died after abusing cigarette lighter refills and a typewriter correcting fluid containing 1,l ,l-trichloroethane (GC conditions as in Fig.2): injection volume, 0.30 cm3 headspace; detector sensitivities (FSD): (a) FID 80 pA and (b) ECD 2 kHz; whole blood l,l,l- trichloroethane concentration 1.2 mg dm-3 advocated for metaldehyde assay in biological specimens,14 although this approach has not proved successful in our hands. Isobutane is unlikely to be found in the complete absence of butane and propane (cf., Table 1). Methanol is rapidly oxidized in aqueous solution on mixing with potassium dichromate (5% d v ) in dilute sulfuric acid (6 mol dm-3). Methanol cannot be directly removed from blood in this way but can be oxidized after headspace transfer to a second, warmed vial before adding the dichromate reagent. However, such transfer may be associated with a considerable decrease in sensitivity.Other workers using packed columns have emphasized the need to use retention data from two different columns before reporting results.9.15 However, as in any toxicological investi- gation the results must never be considered in isolation from any clinical or circumstantial evidence. In addition, the use of an efficient capillary column together with two different detectors confers a high degree of selectivity, particularly for low formula mass compounds where there are very few alternative structures. If more rigorous identification is required, GC combined with mass spectrometry (MS) or Fourier transform infrared (FTIR) spectrometry may be used. However, GC-MS can be difficult when the fragments produced are of less than mlz 40, particularly if the instrument is also used for purposes other than solvent analyses.In t v) 0 a 2 0 LL t v) 0 P 2 n E [a) Ethyl benzene b) FC 12 FC 11 I, 1,1.2-Trichloroethane ; 113 1- I I 1 1 0 5 10 15 20 25 Time/m i n Fig. 8 Analysis of a whole blood specimen (0.20 cm3) from a patient who died after abusing an aerosol designed for cleaning electrical components and which contained FCs 11,12, and 113 (GC conditions as in Fig. 2): injection volume 0.15 cm3 headspace; detector sensitivities (FSD): (a) FID 160 pA and (b) ECD 4 kHz; whole blood FC 11 and 113 concentrations 3.2 and 1.0 mg dm-3, respectively particular, the available sensitivity and spectra of the low molecular mass alkanes renders them very difficult to confirm by GC-MS. Gas chromatography-FTIR is more appropriate to the analysis of volatiles, but the sensitivity is relatively poor particularly when compared.with ECD.In addition, interfer- ences, particularly from water and carbon dioxide in biological specimens, can be troublesome. The likelihood of detecting exposure to volatile substances by headspace GC of blood is influenced by the nature of the compound(s) involved, the extent and duration of exposure, the time of sampling in relation to the time elapsed since exposure and the precautions taken when collecting and storing the sample.2 In one series of suspected abusers, volatile compounds or metabolites were detected in 79 out of 125 cases.16 In 69 (87%) of the positive cases the samples were obtained within 10 h of the suspected exposure. Nevertheless, exposure can be detected using later samples.Thus, in separate cases toluene was detected at 40 h and 2,2,2- trichloroethanol (from trichloroethylene) at 48 h. 16 Analysis of urinary metabolites may extend the time in which exposure may be detected but, of the compounds commonly abused, only toluene, the xylenes and some chlorinated solvents, notably trichloroethylene, have suitable metabolites.2 On the other hand, direct MS of expired air can detect many compounds several days post-exposure. However, the use of this technique is limited by the need to take breath directly from the patient. Chronic petrol ‘sniffing’ has been diagnosed by the measurement of blood lead concentrations17 or the detection of aromatic components such as toluene.18 However, with some petrols and with other complex mixtures such as light petroleums (Table 1) the blood concentrations of the indivi- dual components may be below the limit of detection of the method even after massive exposure.This is illustrated by the1126 ANALYST, JULY 1992, VOL. 117 analysis of blood after human exposure to the petroleum distillate white spirit (British Standard 245 : 1976) (boiling- point range 150-200 "C distributed around nonane) at a concentration of 577 mg m-3 for 4 h. Blood total hydrocarbons (initially about 1.5 mg dm-3) were measurable for only 0.3 h post-exposure (limit of detection 0.5 mg dm-3). However, after similar exposure (520 mg m-3, 3.3 h) to nonane alone, nonane excretion could be followed for at least 3 h.193 In both instances the headspace vial was incubated at 80 "C to maximize the amount of analyte volatilized.The lower sensitivity for white spirit was attributed to the distribution of the hydrocarbon load among many peaks rather than the single peak given by nonane. Most volatile compounds are relatively stable in blood if simple precautions are taken. The container should be glass, preferably with a cap lined with metal foil; greater losses may occur if plastic containers are used. The tube should be as full as possible and, ideally, should only be opened when required for analysis and then only when cold (4 OC).6 If the sample volume is limited it is advisable to select the container to match the volume of blood so that there is minimal headspace. An anticoagulant (lithium heparin or ethylenediaminetetra- acetic acid) should be used.Specimen storage between -5 and 4 "C is recommended and, for esters such as ethyl and methyl acetates, addition of 1% m/v of sodium fluoride is advisable to minimize esterase activity. However, many samples submitted in far from ideal circumstances still give useful qualitative results. It is vital that any products thought to have been abused are packed and stored separately from biological specimens to avoid cross-contamination. In a suspected VSA fatality, analysis of tissues (especially fatty tissues such as brain) may prove useful as high concentrations of volatile compounds may be found even if very little is detectable in blood. Tissue specimens should be stored before analysis in the same way as blood. Detection of a volatile compound in blood does not always indicate VSA or occupational/environmental exposure to solvent vapour.Acetone and some homologues may occur in high concentrations in ketotic patients. Large amounts of acetone and butanone may also occur in blood and urine from children with acetoacetyl CoA thiolase deficiency and may indicate the diagnosis.21 Acetone is also the major metabolite of exogenous propan-2-01 in man .223 Conversely, propan-2- 01 has recently been found in blood from ketotic ~atients.2~ Other ketones may give rise to alcohols in vivo. For example, cyclohexanol is the principal metabolite of cyclohexanone in man.25 A further complication is that contamination of the sample with ethanol or propan-2-01 may occur if an alcohol- soaked swab is used to cleanse skin prior to venepuncture.Other volatile compounds such as halothane and paraldehyde may be used in therapy, and chlorobutanol (chlorbutol), a sedative which is also used as a bacteriocide in some mucous heparin preparations, for example, may also occur. Small amounts of hexanal may arise from degradation of fatty acids in blood on long-term storage, even after storage at -5 to -20 "C.6 Hexanal is resolved from toluene using the temperature- programmed system described above but resolution may be lost if an isothermal quantitative analysis is performed. However, interference from hexanal is only likely to be important if very low concentrations of toluene (0.1 rng dm-3 or less) are to be measured. On the other hand, massive interference from ethylbenzene, rn-lp-xylene and o-xylene has been encountered in samples collected into Sarstedt Mono- vitte Serum Gel blood tubes.Information on the composition of commonly abused products has been given previously.2 When interpreting the results of qualitative analyses it is important to remember that some compounds usually occur in association one with another (Table 5). Blood toluene concentrations in samples from 132 VSA patients ranged from 0.2 to 70 mg dm-3 and were above 5 mg dm-3 in 22 of the 25 deaths.16 Blood Table 5 Associated compounds Compound Acetone BCF Butane Cyclohexanone Dimethyl ether Ethyl acetate* FC 11 FC 12 FC 22 Halothane Isobutane Methyl acetate* Propane Propan-2-01 1,l , 1-Trichloroethane 2,2.2-Trichloroethanol Trichloroe thylene Common associated compound(s) ketoacidosis, propan-2-01 (metabolite, rare) Butanone and higher ketones in FC 11 Butan-1-01, butan-2-01, butanone Cyclohexanol (metabolite) FC 22 Ethanol (metabolite) BCF, FC 12 FC 11 Dimethyl ether 2-Chloro-l,l-difluoroethylene, (metabolites, rare), isobutane, propane 2-chloro-1 ,l,l-trifluoroethane (metabolites, rare) Butane, propane Methanol (metabolite) Butane, isobutane Acetone (metabolite) Isopropyl nitrate (stabilizer, rare) Trichloroethylene (also metabolite of chloral hydrate , dichloralphenazone and trichlofos) 2,2,2-Trichloroethanol (metabolite), chloroform [possibly from thermal degradation of trichloroacetic acid (metabolite) in vitro] * Parent compounds not normally detected in blood.l,l,l-trichloroethane concentrations ranged from 0.1 to 60 mg dm-3 in samples from 66 VSA patients, 29 of whom died.16 However, the possibility of loss of analyte from the sample prior to the analysis must always be considered, especially with very volatile analytes such as butane.Indeed, the difficulty in ensuring that unacceptable losses do not occur during sample collection, transport and storage is the major reason why measurement of such compounds is not often justified. Conclusions The 60 m SPB-1 column has been found to be a valuable alternative to packed columns in the headspace GC analysis of specimens from patients suspected of VSA. Most commonly abused compounds, including many with very low boiling- points such as BCF, butane, dimethyl ether, FC 11, FC 12, isobutane and propane, can be retained and differentiated at an initial column temperature of 40 "C followed by program- ming to 200 "C.The total analysis time is only 26 min. Good peak shapes are obtained for polar analytes such as ethanol and on-column injections of up to 0.30 cm3 of headspace can be performed with no discernable loss of efficiency. The sensitivity is thus at least as good as that attainable with packed columns. Of the commonly occurring compounds only isobutane-methanol and paraldehyde-toluene are at all diffi- cult to differentiate. Quantitative analyses can be performed by using appropriate calibration standards. We thank Supelchem UK for the gift of the SPB-1 column, the British Aerosol Manufacturer's Association , Re-Solv and ICI Pharmaceuticals for financial support and Abbott Labora- tories, C-VET, the Health and Safety Executive Laboratories, Cricklewood, and Rh6ne-Poulenc for gifts of pure com- pounds. References 1 Wright, S. P., Pottier, A. C. W., Taylor, J. C., Norman, C. L., Anderson, H. R., and Ramsey, J. D., Trends in DeathsANALYST, JULY 1992, VOL. 117 1127 2 3 4 5 6 7 8 9 10 11 12 13 14 Associated with Abuse of Volatile Substances 1971-1989, St. George’s Hospital Medical School, London, 1990. Flanagan, R. J., Ruprah, M., Meredith, T. J., and Ramsey, J. D., Drug Safety, 1990, 5 , 359. Ramsey, J. D., and Flanagan, R. J., J. Chromatogr., 1982,240, 423. van den Dool, H., and Kratz, P. D., J. Chromatogr., 1963, 11, 463. Lee, J., and Taylor, D. R., Chromatographia, 1983, 16, 286. Gill, R., Hatchett, S. E., Osselton, M. D., Wilson, H. K., and Ramsey, J. D., J. Anal. Toxicol., 1988, 12, 141. Gill, R., Hatchett, S. E., Warner, H. E., Osselton, M. D., Wilson, H. K., Wilcox, A. H., and Ramsey, J. D., in Proceedings of the Meeting of the International Association of Forensic Toxicologists, Glasgow, August 1989, Aberdeen Uni- versity Press, Aberdeen, in the press. Pekari, K., Riekkola, M.-L., and Aitio, A., J. Chromatogr., 1989,491, 309. Franke, J. P., Wijsbeek, J., de Zeeuw, R. A., Moller, M. R., and Niermeyer, H., J. Anal. Toxicol., 1988, 12,20. Ardrey, R. E., de Zeeuw, R. A., Finkle, B. S., Franke, J. P., Moffat, A. C., Moller, M. R., and Muller, R. K., Gas- chromatographic Retention Indices of Toxicologically Relevant Substances on SE-30or OV-1, VCH, Weinheim, 2nd edn., 1985. Japp, M., Gill, R., and Osselton, M. D., J. Forensic Sci., 1987, 32, 1574. Lora-Tamayo, C., Rams, M. A,, and Chacon, J. M. R., J. Chromatogr., 1986, 374, 73. Bogusz, M., Wijsbeek, J., Franke, J. P., and de Zeeuw, R. A., J. Anal. Toxicol., 1983, 7, 188. Griffiths, C. J., J. Chromatogr., 1984, 295, 240. 15 16 17 18 19 20 21 22 23 24 25 Goebel, K.-J., J. Chromatogr., 1982, 235, 119. Meredith, T. J., Ruprah, M., Liddle, A., and Flanagan, R. J., H i m . Toxicol.. 1989. 8,277. Bruckner, J. V., and Peterson, R. G., in Review of Inhalants: Euphoria to Dysfunction (NIDA Research Monograph, 15), eds. Sharp, C. W., and Brehm, R. L., National Institute on Drug Abuse, Rockville, MD, 1977, p. 124. Nagata, T., Kageura, M., Hara, K., and Totoki, K., Nippon Hoigaku Zasshi, 1977,31, 136. Gill, R., Warner, H. E., Broster, C. G., Osselton, M. D., Ramsey, J. D., Wilson, H. K., and Wilcox, A. H., Med. Sci. Law, 1991,31, 201. Gill, R., Osselton, M. D., Broad, J. E., and Ramsey, J. D., Med. Sci. Law, 1991, 31, 214. Leonard, J. V., Middleton, B., and Seakins, J. W. T., Pediatr. Res., 1987, 21, 211. Daniel, D. R., McAnalley, B. H., and Garriott, J. C., J. Anal. Toxicol., 1981,5, 110. Kawai, T., Yasugi, T., Horiguchi, S., Uchida, Y., Iwami, O., Iguchi, H., Inoue, O., Watanabe, T., Nakatsuka, H., and Ikeda, M., Int. Arch. Occup. Environ. Health, 1990, 62, 409. Bailey, D. N., Clin. Toxicol., 1990, 28, 459. Sakata, M., Kikuchi, J., Haga, M., Ishiyama, N., Maeda, T., Ise, T., and Hikita, N., Clin. Toxicol., 1989, 27, 67. Paper 2/00450J Received January 28, 1992 Accepted February 14, 1992
ISSN:0003-2654
DOI:10.1039/AN9921701111
出版商:RSC
年代:1992
数据来源: RSC
|
|