ISSN:0003-2654    

DOI:10.1039/AN99217BP013

出版商:RSC    

年代:1992

数据来源: RSC

ISSN:0003-2654    

DOI:10.1039/AN99217FX015

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALAO 1 17(4) 695-830 (1 992)The AnalystApril 1992The Analytical Journal of The Royal Society of ChemistryCONTENTS695 EDITORIAL-Harp Minhas697 Urinary Cotinine Fluoroimmunoassay for Smoking Status Screening Adapted t o an Automated Analyser-Sergei A.701 Study of the Determination of the Adulteration of Petrol With Kerosene Using Fuel Cell Based Equipment-M. Shahru707 Calculation of Method Evaluation Functions for Inductively Coupled Plasma Atomic Emission Spectrometric Analysis713 Ion Chromatographic Study of the Effect of Ammonium Nitrate as a Modifier in Electrothermal Atomic Absorption717 Effects of Moisture on the Cold Vapour Determination of Mercury and Its Removal by Use of Membrane Dryer721 Analysis of Volatiles From Oranges in Good and Bad Condition by Gas Chromatography and Gas Chromatography-727 Fluorescent Chiral Derivatization Reagents for Carboxylic Acid Enantiomers in High-performance Liquid Chromato-735 High-performance Ion Chromatographic Analysis of Stainless Steels-Rajananda Saraswati, T.H. Rao741 Silica Gel Modified With Titan Yellow as a Sorbent for Separation and Preconcentration of Trace Amounts of Heavy745 Peroxides. Part 11. lodimetric Analysis of Dialkyl and Dicumenyl Peroxides-Leonard S. Silbert751 Determination of Trace Amounts of Carcinogenic Substances: Adsorptive Stripping Voltammetry of l-[4’-(Phenyl-azo)phenyl]-3,3-dimethyltriazene at a Hanging Mercury Drop Electrode-Jiii Barek, Arnold G. Fogg757 Study of the Faradaic Transfer of Ions in a 1,2-Dichloroethane Extraction System.Similarities and Differences inCharacter Between Cobalt(ii) and Nickel(ii) With 1,lO-Phenanthroline as the Complexing Agent-Wang Fang, LinSinru761 Effect of Flow Injection Parameters on the Selectivity of an Iodide-selective Electrode-David E. Davey, Dennis E.Mulcahy, Gregory R. O’Connell767 Flow Injection Determination of Triton X-100 With On-line Solid-phase Extraction-Charles Moeder, Nelu Grinberg,Holly J. Perpall, Gary Bicker, Patricia Tway ,773 Flow Injection Stopped-flow Kinetic Determination of the Anxiolytic Sedative Bromazepam in Dosage Forms-SalahM. Sultan777 Background-correction Methods for the Determination of Caffeine in Beverages, Coffee and Tea by UsingSecond-derivative Ultraviolet Spectrophotometry-Oi-Wah Lau, Shiu-Fai Luk, Oi-Ming Cheng, Teresa P.Y. Chiu785 Derivative Spectrophotometric Analysis of Two-component Mixtures Using a Compensation Technique-Abdel-AzizM. Wahbi, Fawzy A. El-Yazbi, Magda H. Barary, Suzy M. Sabri791 Spectrophotometric Determination of Magnesium(i1) With Emodin (1,3,8-Trihydroxy-6-methylanthraquinone)-Tarasankar Pal, Nikhil R. Jana, Pradip K. DasEremin, Ruth E. Coxon, David L. Colbert, John Landon, David S. SmithBahari, W. J. Criddle, J. D. R. Thomasfor Iron, Manganese and Titanium From Metal-spiked Filter Samples-lnge Lise Brink Olsen, Erik HolstSpectrometry-Muhammad Mansha Chaudhry, David LittlejohnTubes-Warren T. Corns, Les Ebdon, Steve J. Hill, Peter 6. StockwellMass Spectrometry-David V. McCalley, Juan F. Torres-Grifolgraphy-Toshimasa Toyo’oka, Mumio Ishibashi, Tadao TeraoMetals From Alkaline Earth or Alkali Metal Salts-Ryszard KocjanPAPERS PRESENTED AT THE XXVll CSI, BERGEN, NORWAY, JUNE 6-14,1991795 X-ray Spectral Microanalysis of the Phase Composition of High Temperature Superconductor Bismuth-Lead-Strontium-Calcium-Copper-Oxygen Ceramics Using Chemometric Approaches-lgor I.Bondarenko, Boris A.Treiger, Victor V. Rezvitskii, Lev N. Mazalov803 X-ray Spectral Microinvestigation of the Chemical States of Atoms in Cd,Hg, -,Te-GaAs Thin Films-Victor V.Rezvitskii, Boris A. Treiger, lgor I. Bondarenko, Lev N. Mazalov807 Spectroscopic Properties of Lanthanides Sorbed on Polymeric Matrices-Svetlana V. Beltyu kova, Gyorgy M.Balamtsarashvili, Tatyana B. Kravchenko813 Peculiarities of the Distribution of Man-made Radionuclides in Several European Seas-0.V. Stepanets, V. S. Karpov,V. M. Komarevsky, A. P. Borisov, I. T. Farrahov, G. Y. Solov‘eva, L. A. Pilipets, G. F. Batrakov, T. A. Chudinovskyhcontinued inside back coverTypeset and printed by Black Bear Press Limited, Cambridge, England00e3-2G54c199214-ANALYST, APRIL 1992, VOL. 117REPORT BY THE ANALYTICAL METHODS COMMITTEE817 Determination of Sulfadimidine in Medicated Animal Feeds-Analytical Methods CommitteeCOMMU NlCATlON823 Novel Methods for Derivatization of Mercury(a) and Methylmercury(ii) Compounds for Analysis-Peter J. Craig, Darren825 BOOK REVIEWS828 ERRATUM829 CUMULATIVE AUTHOR INDEXMennie, Naman Ostah, Olivier F. X. Donard, Fabienne MartinEIGHT PEAK INDEXOF MASS SPECTRA4th EditionThis qi ality compilation is recognised by many mass spectrometrists as the most useful index of massspectra in print today.Key features of the 4th Edition include:0 easy access to 8 1,123 mass spectra via unique indexing0 25% more compounds covered over previous edition0 spectra not currently available in any other commercial collectionsrapid identificatioa of unknowns by simple peak intensity matchingconvenient browsing of the dataProbably the best printed index of mass spectra in the world!For further information about the new edition, please contact:Sales and Promotion DepartmentRoyal Society of ChemistryThomas Graham HouseScience Park, Milton RoadCambridge CB4 4WF, United KingdomTel: +44 (0) 223 420066. Fax +44 (0) 223 423623. Telex: 818293 ROYALROYALSOCIETY OFCHEMISTRY~ 6 /InformationServicesCircle 007 for further informatio

ISSN:0003-2654    

DOI:10.1039/AN99217BX017

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, APRIL 1992, VOL. 117 695 Editorial The Publication Process 1. 2. 3. 4. 5. As the editorial staff of The Analyst attend various meetings throughout the year, in this and other countries, it often surprises us how few of our readers, authors and referees are actually aware of how the publication process works. In order to enlighten you all and to clarify how RSCjournals process manuscripts (MS) from receipt to final publication, we have put together a flow diagram (see overleaf) that explains the process in some detail. In addition I have added some average times to various stages in order to enable you to work out for yourselves where a manuscript resides the longest. On receipt of the manuscript an acknowledgement is sent to the author; our professional and academically trained staff then assign appropriate referees from our database containing over 3500 that are available for consultation (same day).Reviewers are allowed up to 14 days to assess and return the manuscript. (This can be lengthened if referees are away or indisposed, and we need to assign an alternative referee.) On receipt of the referee reports the paper is accepted/ rejectedheturned for revision (same day) after editorial consideration and the author notified. Author revision is the longest part of the process; the average revision time for 1991 was 55 days. This could be the result of many contributing factors: (i) other commit- ments; (ii) postal delays; (iii) holidays; (iv) author has no access to a personal computer hence MS has to be retyped in its entirety; (v) extensive revisions are necessary; (vi) further laboratory work is necessary.Postal delays are a problem, particularly in North and South America, Canada, Russia and Italy. We do realise that all these factors can play a part in the time taken for revision. However, this is a rate-determining step, so if you as authors can return your revised MSs as soon as possible we can reduce our times to publication (by at least a month) and hence ensure that your work is published in the shortest possible time. Once revised the MS is reassessed and either accepted or rejected and the author notified (same day). 6. The MS is edited and sent to the printers together with the rest of the MSs to be included in a particular issue to be typeset. Typesetting usually takes about 30 days.7. The typeset MS ( i e . , the proof) is checked and read by a member of the editorial staff (not the same person that edited it originally) to ensure that no errors have occurred on typesetting. The manuscript is then sent back to the author to ensure that we have not misinterpreted hidher ideas in our editing and to check for any errors. Authors are allowed up to 20 days for this stage. 8. Staff then incorporate the author’s corrections’ (the make-up stage) and then I as editor read all the papers in the issue (the buck stops here!) before assigning index terms, choosing the order of papers and giving final approval for publication. In the meantime covers, adverts and figures also have to be edited, checked and approved. The issue is then sent back to the printer for final correction (ours and author’s) before being printed, bound and then distributed (another 30 days).I have, of coursc, taken a fairly straight path through the system, but as one would expect many alternative permuta- tions are possible. Some of these are quicker and others much slower. You will observe that each manuscript is read at least three times by three different members of staff. We believe that as a result of this checking and double checking The Analyst is second to no other journal in the field, in terms of editorial quality. Finally, although we offer all our authors the benefit of the interpretational skills of our professional staff we hope our regular authors will learn from their previous papers, so when it comes to submitting further papers to us, we will have much less work to do. New authors should check the style of the journal beforc preparing their work for submission; this would again help to reduce our times to publication, to our mutual benefit . If you have any further enquiries about the publication of papers in The Analyst, please do not hesitate to contact the editorial office. Harp Minhas EditorANALYST, APRIL 1992, VOL. 117 696 RSC JOURNALS PUBLICATION PROCESS 1 Ms from author Same day 1 f Revision by author 22 days 55 days (av) I Editorial reassessment 1 14 days v 1 Referee’s reassessment m [Typesetting) Same day , opy-editing) I 30 days 1 1 Editorial proof-checking I 2-5 days 20 days I I Author’s proof-checking I 20 days p % z q 30 days Publication The times given are maximum allowed times, except for author revision which is an average.

ISSN:0003-2654    

DOI:10.1039/AN9921700695

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST. APRIL 1992, VOL. 117 Urinary Cotinine Fluoroimmunoassay for Adapted to an Automated Analyser Sergei A. Eremin Department of Chemistry, Moscow State University, Moscow 697 Smoking Status Screening 119899, Russia Ruth E. Coxon, David L. Colbert, John Landon and David S. Smith Department of Chemical Pathology, St. Bartholomew‘s Hospital, London ECI M 6BQ, UK A polarization fluoroimmunoassay for cotinine, a major metabolite of nicotine, has been adapted for fully automated screening of urine samples on the Abbott TDx analyser. The method has sensitivity and specificity suitable for the discrimination of active smokers from non-smokers (including passive smokers) by application of a cut-off at 0.5 m g 1-1 of total urinary cotinine. Most active smokers’ urine gave results over 1 m g 1-1, whereas apparent levels in non-smokers were 0.08 mg 1-1 or lower.A result for one sample can be obtained in about 5 min and a throughput of 80 samples h-1 can be maintained for large-scale screening applications. Keywords: Fluoroimmunoassa y; cotinine; urine screening; automation; smoking A simple test for smoking status is valuable in epidemiological studies,’.’ in confirming self-reported smoking behaviou1-3.~ and in clinical investigation and management of smoking- related conditions.5 There is also a demand for health- and insurance-screening purposes. Cotinine, a major metabolite of nicotine, is a useful marker of smoking as it is derived principally from tobacco.2 Cotinine levels have been reported to reach 0.5 mg I-’ or more (median 1.6 mg 1 - 1 ) in the urine of most active cigarette smokers, whereas passive smoking resulted in levels no higher than 0.04 mg 1-1, as determined by a radioimmunoassay.1 Urine is the most conveniently collected body fluid for screening tests. Urinary cotinine has been assayed by gas chromato- graphy,6,7 high-performance liquid chromatography8 and non-specific spectrophotometric methods;ZA however, immu- noassay is often the most practical method for large numbers of samples. Although the sensitivity of gas chromatography,7 radioimmunoassay1.S.7 or enzyme-linked immunosorbent assay9 may be required in the investigation of the exposure of the population to environmental tobacco smoke,7 a screen to discriminate smokers’ from non-smokers’ urine can involve simpler methods, including those non-isotopic immunoassays that can be performed without the need for a separation step.The Abbott TDx analyser, an automated instrument for the performance of polarization fluoroimmunoassay (PFIA) ,lo has become widely accepted for the assay of drugs and other small molecules. This paper describes the adaptation of previously developed cotinine PFIA reagents” for urine screening on the TDx instrument. Experimental PFIA Reagents Fluorescein-labelled cotinine tracer (trans-4’-carboxycotinine conjugated to fluoresceinthiocarbamylethylenediamine), coti- nine antiserum raised in sheep and ‘synthetic urine’ were as described previously.11 (-)-Cotinhe from Sigma, Poole, Dorset, UK, was used to prepare standards in synthetic urine. Abbott TDx dilution buffer (Abbott Diagnostics, Maiden- head, Berkshire, UK), consisting of 0.1 mol dm-3 sodium phosphate, pH 7.4, containing 0.1 g 1-1 of bovine y-globulins and 1 g 1 - 1 of sodium azide,12 was used as diluent throughout.PFIA Analysers The Abbott TDx analyser was used with disposable sample cartridges and 50 x 10 mm cylindrical glass cuvettes from Abbott or Sigma. Single-reagent PFIA11 was performed with use of a Model LS-20 polarization fluorimeter (Perkin-Elmer, Beaconsfield, Buckinghamshire, UK). Reagents for TDx Assay Abbott TDx reagents are provided in purpose-designed packs, which typically contain three pots with about 3 ml each of tracer, antiserum and ‘pre-treatment solution.’ The pre- treatment pot often contains a detergent formulation to eliminate non-specific binding effects, principally in the assay of serum samples.Reagent packs carry a bar code, identifying each assay by number, which is read when they are placed inside the analyser. The code consists of eight bars, each approximately 3 mm wide, with the two rightmost bars normally black and the other six coded in a simple binary format: ‘1’ for a black bar and ‘0’ for a white bar. The different assay numbers define various protocols (sample and reagent volumes, pipetting sequences, incubation times, and software parameters for standard curve fits). By the selection of an appropriate assay number to obtain suitable analyser perfor- mance characteristics, it is possible to adapt the TDx instrument for use with non-dedicated PFIA reagents.13-15 TDx Assay Procedure For the cotinine assay, use the cortisol-assay software of the TDx instrument (assay number 46).Take a used three-pot reagent pack and thoroughly wash out the pots. Fill the tracer pot with 3 ml of 1.1 x 10-7 mol dm-3 fluorescein-labelled cotinine. Fill the antiserum pot with 3 ml of cotinine antiserum diluted 1 + 29 with the dilution buffer. Fill the pre-treatment solution pot with 3 ml of dilution buffer only. These amounts are sufficient for 100 tests in the analyser. Either use an actual cortisol reagent pack or take any other three-pot pack and fabricate the appropriate bar code. Hence, a code, reading left to right, of white, black, black, black, white, black (101110 binary = 46 decimal) yields the desired pack identification for assay number 46 (nominally a cortisol assay).The existing software parameters of the TDx instrument for the cortisol assay, including the sample volume of 8 PI, were used, with the exception that standard concentrations were adjusted to values of 0, 0.2, 0.4, 1.2, 4.0 and 12.0 mg I-’. In order to perform assays, load aliquots (approximately 50 PI) of samples (or standards) into sample cartridges on a TDx carousel and place empty glass cuvettes into corresponding adjacent carousel positions. Place the reagent pack and the698 ANALYST, APRIL 1992, VOL. 117 carousel into the analyser and start the assay routine, which is performed entirely automatically by the analyser, using its on-board sampler-diluter.“) On the first revolution of the carousel, 8 pl of sample are aspirated and dispensed with dilution buffer into a vacant position on the sample cartridge.On the second revolution, aliquots of the diluted samples are pipetted, together with pre-treatment solution (25 PI) and dilution buffer (to bring the volume to 975 PI), into the glass cuvettes, and the blank signals (vertically and horizontally polarized components of the fluorescence) from the diluted samples are measured. On the third revolution, additional aliquots of the diluted samples, together with tracer (25 pl), antiserum (25 pl) and additional pre-treatment solution (25 pl) and dilution buffer (to bring the total volume to 1.95 ml), are dispensed into the glass cuvettes and, after an incubation period to allow immunoassay binding reactions to take place, fluorescence signals are again measured.Fluorescence polari- zation is calculated from the blank-corrected signal com- ponents and printed out in ‘millipolarization’ units (mP). Comparison Single-reagent PFIA Procedure This was performed as described previously.11 Mix 10 pl of urine sample or cotinine standard in synthetic urine with 1.5 ml of a pre-equilibrated mixture (‘single reagent’) of tracer and antiserum. After incubation at room temperature for 30 min, measure the fluorescence polarization. Urine Specimens Urine from patients attending drug-dependency clinics had been subjected to routine screening for drugs at Hackney Hospital, London; by examination of the nicotine peak on gas chromatograms they were classified as originating from smokers (urinary nicotine greater than 0.5 mg 1-1) or non-smokers.Specimens were also collected from both smoking and non-smoking staff members of St. Barth- olomew’s Hospital and Medical College. Results and Discussion Optimization of TDx Assay The software of the TDx instrument places limits on the acceptable values of various assay parameters such as the total tracer fluorescence intensity, the mP value for the zero-concentration standard, the mP differences between successive standards, and the over-all assay span. The required values are given in the Assay Parameters printout or are found in the Assay Specific Curve Characteristics table in the System Operation Manual. By trial and error, tracer concentration, antiserum dilution and standard concentrations were adjusted until conditions were found which would give a satisfactory calibration on the TDx instrument. Fig.1 shows a typical calibration graph as successfully fitted by the software of the TDx instrument, with the root mean square of residual errors being 0.20 mP unit. This calibration was stored in the analyser; subsequently, urine specimens could be assayed within about 5 min for one sample or 13 min for a full carousel of 20 samples. Sensitivity The intrinsic sensitivity of the PFIA was determined according to Rodbard16 by measuring the response for 20 replicates of the zero-concentration standard (synthetic urine without cotinine). The minimum detectable concentration at the 95% confidence level was 0.03 mg 1-1; this corresponds to 0.24 ng of cotinine in the 8 p1 sampled volume in the TDx instrument.220 - n -z 200 C 0 ! 180 .- .I- .- L m - 160 0 0.1 1 10 Cotinine concentration/mg I-’ Fig. 1 Calibration graph for cotinine obtained by polarization fluoroimmunoassay on the Abbott TDx analyser Cut-off Level for Smoker/Non-smoker Screening Urine samples from 34 known non-smoking staff members (including passive smokers) were assayed; the mean mP response was equivalent to 0.01 mg I-* of cotinine. By using the mean mP response less three standard deviations as a measure of a practically achievable cut-off for positive sample discrimination,lI a value of 0.13 mg 1-1 was found. Precision Three urine specimens from active smokers were analysed ten times within a measurement series, yielding mean results of 0.57, 2.91 and 4.37 mg I-* of cotinine with within-series relative standard deviations (RSDs) of 4.8, 3.4 and 7.1%, respectively. Analysis of the same samples on five different days, using the same previously stored calibration graph in the TDx analyser, afforded between-series RSDs of 5.2, 6.8 and 7.0%, respectively.Accuracy A urine sample from a nonsmoker with 0.04 mg I-* of apparent cotinine was spiked to levels of 1.33, 2, 2.4 and 4 mg 1-1. Recovery of the added cotinine ranged from 101 to 109% (mean 105%). Two smokers’ samples with cotinine levels of 3.3 and 5.3 mg 1-1 were serially diluted by 1 + 1, 1 + 3 , 1 + 7 and 1 + 15 with dilution buffer. Assay results ranged from 90 to 112% of expected values (mean 98%). Assay of Urine Samples and Correlation with Existing Methods The patients’ urine samples (45 smokers’ and nine non-smokers’ specimens) and those from the staff members (10 smokers’ and 34 non-smokers’ specimens) were assayed in the TDx analyser.The threshold of the TDx instrument for flagging of high-concentration results was set at 0.5 mg 1 - 1 . All urine samples were correctly classified by the TDx assay: all the smokers’ samples contained more than 1 mg 1-1 of apparent cotinine by TDx (with one exception at 0.63 mg I - I ) , the highest value being 9.45 mg 1-1, whereas the non-smokers’ samples gave values of 0.08 mg 1 - 1 or lower. All the specimens were also screened by the established single-reagent PFIA, with use of a 1 mg I-’ cut-off calibrator,ll and again all were correctly assigned. Stability of Reagents and TDx Analyser Calibration The stability of the cotinine PFIA reagents in their original format” has been confirmed, with the single reagent stored at 2-8 “C showing no deterioration in performance after more than 2 years.In the TDx analyser, diluted tracer andANALYST, APRIL 1992, VOL. 117 699 antiserum kept in reagent pots at 2 4 ° C resulted in indistinguishable calibration graphs over a period of at least 3 months. Specificity and Applicability of the Assay Although the cotinine antiserum was previously found to have negligible cross-reactivity with nicotine or a variety of its known metabolites, 1 1 it is possible that cotinine immunoassays could be detecting additional species, such as glucuronic acid conjugates.7 Accordingly, immunoassays should be regarded as measuring ‘cotinine equivalents’ and will not necessarily correlate exactly with techniques such as gas chromatography.7 Nevertheless, the results presented here show how existing reagents can be adapted to provide fully automated cotinine screening, with appropriate sensitivity and specificity for smokerhon-smoker discrimination and with a throughput capacity of 80 samples h-1.References I 2 3 Wald. N . J . , Boreham, J . . Bailey, A., Ritchie, C.. Haddow, J . E., and Knight, G.. Lancet, 1984, i, 230. Barlow. R. D., Stone. R . B., Wald, N. J . , and Puhakainen, E. V. J., Clin. Chim. Actu. 1987, 165, 45. Peach, H.. Ellard, G. A.. Jenner. P. J . , and Morris, R. W., Thorax. 1985. 40, 351. 4 5 6 7 8 9 10 11 12 13 14 15 16 Smith, C. L., and Cooke, M., Analyst. 1987. 112, 1515. Knight. G. J., Palomaki, G. E., Lea, D. H . , and Haddow, J. E., Clin. Chem. (Winston-Salem, N . C.), 1989, 35, 1036. Feyerabend. C., and Russell, M. A. H., J . Pharm. Pharmacol., 1990, 42, 450. Anderson, I . G. M., Proctor, C. J . , and Husager. L.. Analyst, 1991, 116, 691. Barlow, R. D., Thompson, P. A., and Stone, R. B., J . Chromatogr., 1987. 419, 375. Langone, J. J . , Cook. G., Bjercke, R. J., and Lifschitz, M. H., J . Immunol. Methods, 1988, 114, 73. Jolley, M. E., Stroupe, S . D., Schwenzer, K. S . , Wang, C. J . , Lu-Steffes, M., Hill, H . D., Popelka, S. R., Holen, J. T., and Kelso, D. M., Clin. Chem. (Winston-Salem, N . C . ) . 1981, 27. 1575. Hansel, M. C., Rowell, F. J., Landon, J . , and Sidki, A. M., Ann. Clin. Biochem., 1986, 23, 596. Tait. J. F., Franklin, R. W., Simpson, J . B . , and Ashwood, E. R., Clin. Chem. (Winston-Salem, N . C.). 1986. 32, 248. Colbert, D. L., and Coxon, R. E., Clin. Chem. ( Winston-Salem, N . C . ) , 1988.34. 1948. Eremin, S. A., Zaitchev, S. V., Egorov, A. M., Ermakov, A. N.. Smirnov, A. V., and Izotov, B. N., Anal. Chim. Acra, 1989, 227, 287. Mellor, G. W., and Gallacher, G., Clin. Chem. (Winston- Salem, N . C . ) , 1990, 36, 110. Rodbard, D., Anal. Biochem., 1978, 90, 1. Paper 1 I05067B Received October 4, 1991 Accepted December 19, 1991

ISSN:0003-2654    

DOI:10.1039/AN9921700697

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, APRIL 1992, VOL. 117 70 1 Study of the Determination of the Adulteration of Petrol With Kerosene Using Fuel Cell Based Equipment M. Shahru Bahari, W. J. Criddle" and J. D. R. Thomas School of Chemistry and Applied Chemistry, University of Wales College of Cardiff, P.O. Box 912, Cardiff CF13T5, UK Techniques based on the headspace vapour analysis of ethanol using a fuel cell sensor have been studied as the basis of a procedure t o determine the degree of adulteration of petrol with kerosene. The procedure, using commercially available instrumentation incorporating a fuel cell sensor, is described and was found to be experimentally facile (up t o five determinations per hour) while giving results of good precision (0.21 % relative standard deviation) and accuracy (+2%).Although petrols of different origin need different dilutions with hexane for optimum instrumental response, calibration slopes for different brands are similar, thus offering the prospect of a viable portable adulteration monitor. Keywords: Petrol adulteration; kerosene; fuel cell; headspace analysis The problem of the adulteration of highly priced petrol with low-priced kerosene has led to many studies on the develop- men t of procedures for the rapid determination of the degree of adulteration.1-17 Petrol is normally a mixture of hydrocar- bons having a boiling-point range between 30 and 180 "C, and usually carries a high level rate of tax duty, whereas kerosene consists of higher boiling-point range hydrocarbons (150- 250°C) and usually carries a lower level of duty.Therefore, there is an incentive for the two to be intentionally blended by criminally motivated syndicates, and the adulterated product sold to consumers for monetary gain. The seriousness of this problem, espccially in countries where the tax differential between the two products is large, has stimulated much research on this subject.1-17 However, none of the procedures resulting from such studies was designed for use as an 'in-the-field' approach for determining petrol adulteration. We have recently reported a phase-titration procedure for determining kerosene adulteration in petrol in the 0-2070 range16 and a spectrophotometric procedure has been success- fully incorporated for the end-point determination. l 7 The procedure, whilst providing a simple, inexpensive and rapid method of determining quantitatively the degree of adulter- ation of petrol samples, requires a laboratory environment in its present state of development.In the continuing investigation it was the objective to provide an alternative and more convenient method by using the responses of platinum-based fuel cells exposed to the adulterated petrol headspace. The principle of early fuel cells using hydrogen and oxygenis as fuel has been widely used in a wide range of applications including providing a source of energy and water for astronauts o n the Apollo missions.19 The application of fuel cells in measuring ethanol in beverages and during the fermentation process has been investigated by Criddle ut d.20.21 The resulting development of a range of ethanol analysers, using platinum-based fuel cells as sensors, has led to the manufacturing by Lion Laboratories (Barry, South Glamorgan, UK) of firstly the AE-D3 Alcolmeter and later the DA-1 Alcolmeter.The latter instrument is fully auto- mated and suitable for ficld and laboratory use in the alcoholic beverage industries. The chemical reactions involved in these fuel cells for ethanol analysis have been described by the following equations:" CH3CHzOH + CH3CHOH + H+ + e- CH,CHOH -+ CH,CHO + H+ + e - * To whom correspondence should be addressed. Although the primary reaction product is acetaldehyde, a build-up of acetic acid was observed. However, at the end, both these compounds accounted for 100% of the products, and acetic acid was shown not to react further.23324 Fuel cell inert products which can accumulate on the fuel cell surface were claimed to interfere with subsequent readings resulting in a lowering of the fuel cell signal output.This blanketing effect is reported to be a major cause of the fuel cell fatiguing phenomenon observed in ethanol analysis.2s With respect to hydrocarbon fuels, ethanol and other oxygenated compounds have been introduced into petrol in order to enhance its octane rating.26 This use in petrol is increasing with the reduction of lead additives to meet the increasing worldwide demand for unleaded petrol. Lists of oxygenates and their allowable concentrations are given in Table 1.27 The presence of alcohols and other oxygenated compounds points to the type of responses observed in headspace ethanol analysis being also obtained on exposure of the fuel cell to the petrol headspace. The most popular replacement for lead alkyls has been tert-butyl methyl ether (TBME).2SX.2Y Other than this, the use of methanol and ethanol as octane boosters in the production Table 1 Oxygenated supplemcnt allowed by the EEC D i r ~ c t i v e .' ~ The usc of these components other than those specificd as additives at a concentration below 0.5% v/v in total is not affected by the Directive A B" Compound (% v/v) (% v/v) Methanol, suitable stabilizing agent must be added? 3 3 Ethanol, stabilizing agent may be necessary 5 5 Isopropyl alcohol 5 10 zert-Butyl alcohol 7 7 Isobutyl alcohol 7 10 per moleculet I0 15 Other organic oxygenates 7 1 0 Ethers. containing five or more carbon atoms Mixture of any organic oxygenates4 2.5% m/m 7% m/m oxygen oxygen not exceeding the individual limits fixed above for each componcnt * The use of this maximum limit must be accompanied by proper identification of the petrol pump nozzles.-1 In accordance with the national specifications or, whcre thcsc do not exist, industrial specifications. $ Acetone is authorized up to 0.8% v/v when it is present as a by-product of the manufacture of certain organic oxygcnatc com- pounds.702 ANALYST, APRIL 1992, VOL. 117 of petrol is also increasing. In addition to the oxygenated compounds, alkenes, which comprise up to 2% v/v in petrol, are also fuel cell active.30 In this work, the fuel cell response of the petrol mixtures was initially investigated using the manually operated Lion AE-D3 Alcolmeter, and the findings (see under Results and Discussion) were incorporated in the data from analyses with the Lion DA-1 instrument.Experimental Reagents Hexane [General Purpose Reagent, BDH (now Merck), Poole, Dorset, UK], TBME (spectrophotometric grade, Aldrich, Gillingham, Dorset, UK), methanol and ethanol (Technical grade, BDH), and hex-1-ene (99+%, Aldrich) were used as received. Premium (4 Star, 97 octane) grade (Texaco, Esso and BP), unleaded petrol (Esso) and kerosene (Texaco, the only locally available source) were used directly without pre-treatment. Instruments Lion AE-D3 Alcolmeter The AE-D3 instrument is essentially a sensor head (Fig. 1) for headspace analysis, and consists of a manually operated gas sampling valve which allows a standard (fixed) volume of vapour to be delivered to a Lion WR-type fuel cell.The potential developed in the cell is amplified in a control unit and the output displayed either on a chart recorder (fuel cell output profile), or as a range of arbitrary units normally used in association with breath and blood ethanol analysis (LCD display and printer) and derived from the maximum potential achieved. The temperature of the sample solutions used for measure- ments with the AE-D3 instrument was measured with a temperature probe (Yellow Springs Instrument, Model 3020) associated with a Yellow Springs Instrument Model 32 conductance meter. Lion DA -I Alcolmeter The DA-1 instrument is a fully automated ethanol analyser having similar LCD display and printer facilities to the AE-D3 instrument and designed for the determination of ethanol in alcoholic beverages using dynamic headspace analysis, i.e., the analysis of a gas stream flowing through a solution of a volatile analyte, e .g . , ethanol in water.") The basis of the operation of this unit has been described elsewhere.?? The DA-1 instrument incorporates automatic temperature correc- tion between the standard and sample vessels. J M 1 Amplifier hiiq Fig. 1 Schematic diagram of the sensor head of the Lion AE-D3 instrument. A, Fuel cell disc; B, cell case; C, working electrode; D. sampling port; E. sampling needle; F, reaction products outlet; G. sample volume; H, diaphragm; J , set button; K, spring; L, sprung catch; M, read button; N.spring; P, magnet; Q. recd switch: R, magnet; S , reed switch; T. thermistor; and U, platinum contacts Procedure Data acquisition modes (DAMS) The Lion AE-D3 Alcolmeter (sample volume 1.2 cm3) amplified fuel cell output was connected to a chart recorder (Linear, Model 550) and used to study the behaviour of the fuel cell (signal, output profile and reproducibility) with respect to the various samples (DAM 1). In addition, values proportional to the output maxima of the AE-D3 and DA-1 instruments can be read either through a digital display or from a printout (DAM 2). This latter mode was used to allow direct comparison of data from the AE-D3 and DA-1 instruments (see under Instruments). Headspace sampling technique for the A E- 0 3 instrument Headspace samples were taken above liquid samples (in 100 cm3 flasks fitted with self-sealing septa) and the fuel cell output was measured with the AE-D3 Alcolmeter, where both DAM 1 and DAM 2 were used to record analytical data.A hypodermic ventilating needle (stainless steel, 15 cm) was pierced through the septum with its tip immersed in the solution to equilibrate the headspace pressure to atmospheric pressure during sampling (Fig. 2). The sampling needle (Microlance, 5 cm) attached to the sample head (see Fig. 1) was inserted through the septum and the headspace sample was then drawn into the fuel cell by pressing the READ (sampling) button. Successive sampling was carried out immediately after the appearance of the instrument ready indicator light (green). A red indicator light is observed when the instrument is in its initialization or operational mode.Immediately after the peak maximum had been noted, the cell was rapidly cleared of excess of sample by several activations of the SET and READ buttons of the sample head. Response characteristics of the A E-D3 instrument with respect to sample composition The signal profiles (DAM 1) and the contribution of fuel cell fatiguing effects of the hexane solution of petrol and some model compounds (DAM 2) were investigated using the above technique. Leaded and unleaded petrol (50% v/v), methanol (4 ppm), ethanol (4 ppm), TBME (2% v/v) and hex-1-ene (20% v/v) in hexane solution (50 cm3) were used for such tests. Each flask was placed in a thermostatic bath at 25 "C and the contents were agitated with a magnetic stirrer.Influence of petrol concentration on the output of the AE-D3 instrument The optimum concentration of the petrol-hexane mixture was investigated by measuring the headspace response of various *i B I D Fig. 2 Experimental set-up for headspace analysis using the Lion AE-D3 Alcolmeter. A , AE-D3 sensor head; B, temperature probe; C , ventilating needle; D, magnet; and E, septumANALYST, APRIL 1992, VOL. 117 703 concentrations of leaded petrol in hexane (50 cm3,4-28% v/v) at 25°C. Based on results obtained above (see under Response characteristics of the AE-D3 instrument with respect to sample composition), data were recorded only in the linear working range of the response of the AE-D3 Alcolmeter. Effect of adulteration on the response of the AE-D3 instrument The calibration studies of the response of the AE-D3 instrument to adulterated petrols were carried out by filling two vessels, respectively, with unadulterated (calibration) and adulterated petrol (leaded) samples (0-20% v/v) in hexane (10% v/v), and agitating at 25°C.The AE-D3 instrument (DAM 2) was calibrated using the unadulterated sample vessel and the 100% range setting. Note that the AE-D3 instrument was re-calibrated before each measurement of the adulterated sample. Effect of solution temperature on fuel cell response By using the procedure described under Influence of petrol concentration on the output of the AE-D3 instrument, the correlation between the headspace concentration and solution temperature was investigated by filling two vessels with an identical standard petrol mixture in hexane (10% v/v, 100 cm3), thermostating one vessel at 25 "C and varying the temperature of the other vessel in the range 10-40"C.The thermostated vessel was used for calibrating the AE-D3 instrument (DAM 2). Note that the AE-D3 instrument was re-calibrated with the 25 "C standard before each variable temperature sample. Optimization of petrol Concentration f o r operation of the Lion DA-1 Alcolmeter Various concentrations of premium-grade petrol (1O-50% v/v) in hexane (100 cm3) were prepared and investigated using the DA-1 instrument by filling both vessels (standard and sample) with the same solution. From the results obtained, the optimum concentration for each sample batch was determined such that the concentrations used were sufficiently high to give a sufficiently high fuel cell output to satisfy the analytical requirements of the DA-1 instrument.Calibration studies using the DA-l Alcolmeter Calibration studies were performed using the same procedure as above but varying the degree of adulteration by kerosene (0-20% v/v, 2% v/v increments) of the solution in the sample vessel. The concentration of the petrol samples in hexane for the different batches was based on the results obtained from the preceding studies (see under Optimization of petrol concentration for operation of the Lion DA-1 Alcolmeter). The first three response readings on the first sample of the calibration run were rejected. After this, the DA-1 instrument was re-set to its operational cycle, both reference and sample vessels being recharged with appropriate petrol samples, the subsequent data being used for the calibration graph and analytical measurements.Results and Discussion Response of Fuel Cell Based Equipment to Various Headspace Vapours At present, there is no realistic method available for the 'in-the-field' measurement of kerosene in petrol, and the aim of this work was to apply available equipment that is both portable and automated to this need. The equipment chosen incorporates a platinum-based fuel cell vapour sensor and it was, therefore, necessary to study the behaviour of the petrol-fuel cell system. The Lion AE-D3 and DA-1 Alcol- meter units are basically similar in operation, but differ in the level of automation. The AE-D3 instrument is essentially a manually operated unit equipped with a fuel cell sensor located in a hand-operated sampling head, capable of sampling 1.2 cm3 of a vapour.The DA-1 instrument is completely automatic in its operation, but incorporates the same fuel cell sensor, the reference and sample solutions used to generate a dynamic headspace being placed in glass vessels which are an integral part of the unit. Of the two instruments, the AE-D3 model is more convenient for fundamental studies, as it is a simple procedure to attach a chart recorder directly to the amplified fuel cell output. By using this equipment, the behaviour of petrol, kerosene and some petrol additives (oxygenates) and a model unsaturated organic compound was studied from the view- point of output (maximum peak potential) and output profile. In order to obtain convenient values for petrol responses, it was necessary to dilute the petrol with a fuel cell inactive solvent, and hexane was found to be ideal for this purpose, and was used throughout this study for dilution purposes. The composition of petrol has been the subject of several publications, detailed analyses by both ga~3l-~S and liquid36 chromatography revealing a wide range of oxygenates and unsaturated compounds, some of which occur naturally, others being added (oxygenates) to enhance octane ratings.In this work several model compounds were chosen and their fuel cell behaviour was studied in order to ascertain the behaviour of the petrol-fuel cell system. I t is clear that the output profile for petrol must be an accumulation of outputs from various substances, and attempts were made to ascertain what types of compound contribute to the petrol profile.Apart from petrols from three different refiners, several compounds were studied, viz., methanol, ethanol, TBME and hex-1-ene, and their profiles are shown in Fig. 3. I n order to permit a convenient comparison of these profiles, the concentration of these compounds in hexane was adjusted to give outputs of the same order of magnitude. The large concentration variations stem from the wide differences in air-solvent partition coefficients of these substances in hexane which in turn derive from the large variations in polarity of the solutes in the non-polar solvent. I t was deduced, from a qualitative examination of the profiles shown in Fig.3, that the petrol profile derives from a combination of the profiles from oxygenates and unsaturates. The profiles show a marked similarity between those of petrol and the alcohols up to the profile maximum, but the long tail resembles that of TBME and hex-1-ene. A B C D E , 0 1 2 3 4 Time/min Fig. 3 AE-D3 Alcolmeter output profiles of fuel headspace samples from the following hexane solutions (50.0cm3) at 25.0"C. A, Hex-1-ene (20% v h ) ; B , TBME (2% v/v); C, leaded petrol (Esso, 50% v/v); D, methanol (4 ppm); and E. ethanol (4 ppm)704 ANALYST, APRIL 1992, VOL. 117 It might be that future studies of such profiles could be used to ascertain the presence of such substances in petrol, but such a study would need to be more extensive than can be undertaken at present.However, when a fuel cell is used with the range of model compounds, successive sampling of the headspace generated yields marked differences in fuel cell performance. The data illustrated in Fig. 4 show that, particularly for petrol (leaded and unleaded) and hex-1-ene, the fuel cell output drops dramatically with each successive sample, whereas ethanol and TBME show a relatively constant output, methanol falling between petrol and ethanol. Again, as for the peak profile studies, petrol clearly shows profile charateristics that appear to reflect a composite from the model compounds. This observation should not be taken as suggesting that petrols contain a high concentration of hex-1-ene in particular but that unsaturates are likely to make a significant contribution to the output profile. The effect of lead alkyls, which could conceivably have contributed to this behaviour by catalyst poisoning, can be excluded as both leaded and unleaded petrols show similar behaviour (Fig.4). The relatively minor fatiguing phenomenon in ethanol analysis25 has been attributed to poisoning of the fuel cell by reaction products. With the ethanol system, acetic acid is a major product and is thought to be strongly hydrogen-bonded to the polar platinum surface. This surface layer is believed to restrict access by ethanol molecules to the active surface. In the same context, the same phenomenon can conceivably be attributed to reaction products and the higher boiling com- ponents in petrol being unable to escape from the electrode surface before the next sampling is performed. This was confirmed by allowing the fuel cell a rest period of 2 h after sampling, when the fuel cell returned to the original baseline value, thus supporting the blanketing theory.The hydrophobic nature of hexane does not appear to contribute to the fatiguing effect and it would be reasonable to 110 , I 5 100 ? 60 . L u a I I I - A I P P - U r l w - fi Q==&- rai E 50 ' I I I I I 0 2 4 6 8 10 12 Sampling order Fig. 4 AE-D3 Alcolmeter output behaviour of various fuel samples in hexane at 25°C. A, TBME; B, ethanol; C, methanol; D, petrol (Esso, unleaded); E, petrol (Esso, leaded); and F, hex-1-ene 600 I rn w .- 5 500 t e & 400 - g 300 C 0 2 200 L m u 4 100 a 0 4 8 12 16 20 24 28 Petrol in hexane (% v/v) Fig.5 the concentration of petrol (Esso. leaded) in hexane at 25 "C Relationship between the AE-D3 Alcolmeter response and conclude that the fatiguing effect does not result from the non-reactive compounds in petrol. As shown in Fig. 4, continuous sequential sampling of petrol results in an initial rapid fall-off of response, followed by a linear portion of the plot which can be used for analytical determinations owing to its relatively shallow slope. Therefore, all analytical data were taken in this latter range. The large drop in the fuel cell response in the initial sampling will not affect the true measurement of the sample if measurements are made in the linear portion of the plot, corresponding to the sixth and subsequent samplings (see Fig. 4).This means that the first five or six responses, treated as a fuel cell conditioning process to the petrol-hexane system, must be discarded. By taking analytical data in this sequential linear working region, the error arising from the decreasing response with sampling numbers is minimized. The fuel cell responses of the AE-D3 Alcolmeter at different petrol concentrations in hexane are shown in Fig. 5 , the response increasing linearly with increasing petrol concen- tration up to approximately 14% v/v. A 10% v/v petrol in hexane mixture (100 cm3) was, therefore, selected for subse- quent studies, this representing a percentage composition well within the linear range of the plot. Responses from Kerosene-adulterated Petrols For the studies on kerosene adulteration and the calibration studies using the AE-D3 Alcolmeter, five sets of responses were obtained from the linear working range for each of the samples and the means were plotted as a function of adulteration.The relationship between kerosene adulteration and the output of the AE-D3 instrument was found to be linear (Fig. 6), and although the data obtained showed modest precision, it indicated the potential of the procedure for field screening type measurements. As already mentioned, extensive studies have been carried out on the development of the DA-1 Alcolmeter for ethanol analysis.20 This instrument is equipped for fully automated operation once charged with standard and sample, and the operational mode and its portability make it particularly suitable for 'in-the-field' analysis.It is clear from the work described above that it can be used as the basis of an analytical method for the study of petrol adulteration, but some changes have to be made to suit the petrol type of system, most importantly the incorporation of a different temperature constant in the unit software. The fuel cell output and the solution temperature in ethanol analyses are exponentially related,'" the same being true for petrol (Fig. 7). The slopes of this exponential plot (tempera- ture coefficient) at different petrol concentrations (Table 2) are almost identical but the mean value (0.0118) is smaller than the comparable slope for ethanol (0.0281) which is incorporated in the DA-1 instrument software. With this 100 v) c .- 5 95 L .= 90 2 - $ 85 C 0 Q 80 m 2 75 a 70 ' I I I I 1 0 4 8 12 16 20 Kerosene in petrol (% v/v) Fig.6 Typical calibration plot for adulterated petrol (Esso. leaded) using the AE-D3 Alcolmeter at 25 "C. (Regression results: y = 95.17 - 0.87~; Y = 0.994)ANALYST, APRIL 1992, VOL. 117 705 I I 10 20 30 40 Tern pe rat u re/"C Fig. 7 Effect of solution temperature on the AE-D3 Alcolmeter output of petrol (Esso, leaded) in hexane solution (20% v/v, 100 cm3) difference in mind, the appropriate corrections were made to the results from the DA-1 Alcolmeter method. Concentrations suitable for use in the DA-1 instrument vary with petrol brands. Petrol from Texaco gave the highest electrical output and needed a petrol concentration of 20% v/v in hexane, compared with Esso (75% v/v) and BP (5o?i0 v/v) petrols in order to give suitable fuel cell signals in the operational range of the DA-1 instrument.This can be attributed to the difference in the concentration and types of the oxygenates and other fuel cell active materials added during the production of these petrols and is a parameter that must be established for each new brand of petrol. This could be carried out in association with the petrol refiners, for it is in their interest that the product dispensed from the pump does not reflect detrimentally on their reputation. Overall, however, the DA-1 Alcolmeter method gave substantially better results than those obtained with the AE-D3 instrument in terms of accuracy and precision. The similarity in the slope of the calibration graphs for different brands of petrol is shown in Fig.8, and despite the need for different dilutions of the various brands, this points favourably to the prospect of the application of the method to adulter- ation screening. Conclusion The procedures described here indicate that a platinum-based fuel cell sensor can be the basis of a rapid 'in-the-field' method for the determination of kerosene adulteration of petrol. The Lion DA-1 Alcolmeter was found to be particularly suitable for the analysis and is capable of modification to suit petrol analysis, essentially by changing the slope of the temperature correction factor. The authors thank the Public Services Department of Malay- sia, Kuala Lumpur, and the University of Technology of Malaysia for financial support to M.S. B., and Lion Labora- Table 2 Slopes of the logarithmic plots of the response of the AE-D3 instrument versus temperature Petrol concentration Correlation in hexane (% v/v) Slope coefficient ( r ) 5 0.01 15 0.991 10 0.0110 0.997 1s 0.0127 0.997 20 0.0114 0.998 25 0.0123 0.997 Mean: 0.01 18 & 0.006 .- U I I Kerosene in petrol (% v/v) Fig. 8 Calibration results for adulterated petrol (leaded) using the DA-1 Alcolmeter. [Regression results: A (BP), y = 5.004 - O.O49x, r = 0.998; B (Texaco), y = 5.058 - 0.0481, r = 0.996; and C (Esso), y = 5.073 - 0.0471, r = 0.9981 tories Ltd., Barry, South Glamorgan, UK, for the loan of the AE-D3 and DA-1 Alcolmeters. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 References Kryinitzky, J. A,, and Garret, D., Anal.Chem., 1956, 28, 967. Babitz, M., Bentur, S . , and Rocker, A., Bull. Res. Counc. Isr., 1960, 8C. Babitz, M., Bentur, S., and Rocker, A., Isr. J. Technol., 1965, 3, 220. Kanto Gikken Kogyo. K. K.. Jpn. Kokai Tokkyo Koho, JP 83 44 348, March 15, 1983. Goren, S., Isr. J. Chem., 1970,8, 809. Nagai, H., Jpn. Kokai Tokkyo Koho. JP 78 62 595, June 5,1978. DeAndre Bruening, I. M. R., and Branco, V. A. C . , Bol. Tec. Petrobras, 1980, 23, 117. Mitra, G. D., Fert. Technol., 1982, 19, 243. Kapoor, S. K., Kumar. P.. Malikk, V. P., Chhibber, S. K . , and Gupta, D. L., Res. Ind., 1978, 23. 94. Bhatti, S. S., and Singh. W., Acousr. Left., 1985, 8, 105. Verma. R. S., Kumar, K . , and Sharma, M. N., J. Indian Acad. Forensic Sci., 1985, 24, 22. Suri, S. K.. Prasad. K., Ahluwalia. J. C., and Rogers, D. W., Talanta, 1981, 28, 281. Sengupta, D., Analyst, 1987, 112, 615. Shoji. H., Jpn. Kokai Tokkyo Koho, JP 78 13 985, February 8, 1978. Husain, S., Alvi, S. N., and Nageswara Rao. R., Analyst, 1991, 116, 405. Bahari. M. S.. Criddle, W. J., and Thomas, J . D. R., Analyst, 1990, 115. 417. Bahari, M. S.. Criddle, W. J., and Thomas, J . D. R., Anal. Proc.. 1991, 28, 14. Grove. W. R.. Philos. Mag., 1839, 14, 127. Licbhafsky, H. A., and Grubb. W. T., J . Am. Rocket Soc.. 1961,31, 1183. Criddle, W. J., Parry, K. W., and Jones. T. P., Analyst. 1986. 111, 507. Criddle, W. J., Parry. K. W.. and Jones. T. P., Analyst, 1987. 112,615. Criddle, W. J., Parry, K. W., and Jones, T. P., Proc. Am. SOC. Brew. Chem.. 1989, 47, 1. Rightmire, R. A.. Rowland, R. L., Boos, D. L., and Beah. D. L., J . Electrochem. Soc., 1964, 111, 242.706 ANALYST. APRIL 1992, VOL. 117 24 25 26 Novak, M., Lantos, J., and Marta, F., Acta Phys. Chem.. 1972, 18, 151. Parry, K. W.. Ph.D. Thesis, University of Wales, 1988. Lang, G. J.. and Palmer, F. H., in Gasoline and Diesel Fuel Additives, ed. Owen, K., Wiley, Chichester, 1989, ch. 5, pp. Off. J. Eur. Comm., NO L334122. December 12,1985, Annex 1. Davies, T. A., Pet. Rev., 1989, May, 244. Baker, T., Pet. Times, 1990, April, 26. Otsuka, K., Shimizu, Y., Yamanaka, I., and Komatsu, T., Catal. Lett., 1989, 3, 365. 133-168. 27 28 29 30 31 32 33 34 35 36 Pauls, R. E., and McCoy, R. W., J. Chromatogr. Sci., 1981,19. 558. Pauls, R. E., J. Chromatogr. Sci.. 1985. 23, 437. DiSanzo, F. P.. J. Chromatogr. Sci.. 1990, 28, 73. Price, J. A., and Saunders, K. J., Analyst, 1984, 109. 829. Luke, L. A , , and Ray, J. E., Analyst, 1984, 109, 989. Agarwal, V. K.. Analyst. 1988, 113, 907. Paper I l04445A Received August 27, 1991 Accepted November 25, 1991

ISSN:0003-2654    

DOI:10.1039/AN9921700701

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, APRIL 1992, VOL. 117 707 Calculation of Method Evaluation Functions for Inductively Coupled Plasma Atomic Emission Spectrometric Analysis for Iron, Manganese and Titanium From Metal-spiked Filter Samples lnge Lise Brink Olsen* and Erik Holst Danish National institute of Occupational Health, Lersef Parkaile 705, DK-2 100 Copenhagen, Denmark The method evaluating function (MEF) is defined as the expected value of the analysis as a function of the 'true' content of the analyte in the samples. The MEF describes the complete analytical range and contains most of the relevant information on a chemical analytical method. The calculation of the MEF provides an easy way t o obtain stringent documentation on an analytical method. MEFs for the multi-element inductively coupled plasma atomic emission spectrometric (ICP-AES) determination of Fe, Mn and Ti have been established.Parameter estimators, based on the method of weighted least-squares regression, including confidence intervals for the zero-point error a and the slope of 1.3 of the regression curve, were determined. The overall analytical error, i.e., the day-to-day variation and the variation within a single run of the analysis were calculated. Certification limits were set, for use of the ICP-AES analysis in the development of interlaboratory control filters, based on the square root of the relative mean square error. Keywords: Inductively coupled plasma atomic emission spectrometry; multi-element analysis; method evaluating function; iron, manganese and titanium; quality control; relative mean square error The analysis of airborne particulates has long been the concern of the environmentalist, the industrial hygienist and the analytical chemist.The accuracy of the analytical results can only be verified by obtaining agreement with certified standard reference materials. Standard filter materials with a matrix composition and analyte concentration in the ranges of interest are normally not available commercially. Interlabora- tory comparisons of results obtained on real samples offer the simplest and most reliable means of evaluating accuracy. Interlaboratory control filters exposed to welding fume from an industrial working place are under development at our laboratory. Inductively coupled plasma atomic emission spectrometry (ICP-AES) appears to offer advantages of simultaneous multi-element analysis, low detection limits and wide linear dynamic range.Multi-element ICP-AES analysis for Fe, Mn and Ti is used in the verification of the interlaboratory control filters. The purpose of the present study was to validate the ICP-AES analysis by determining method evaluating func- tions (MEFs) for Ti, Mn and Fe from metal-spiked cellulose acetate filters. The following four parameters are usually determined to validate a recommended method:' (1) limit of detection (LOD) or criterion of detection; (2) limit of quantification, sometimes called the limit of determination; (3) the standard deviation of the unsystematic error (the precision); and (4) the systematic error (the accuracy or the trueness).When calculating the systematic error, only a single concentration is often taken into account,'-' and the repeat- ability and a possible concentration dependence is thereby neglected. Samples with known composition and amounts are normally used, e.g., National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1648 Urban Particulate Matter.'4 This paper is mainly concerned with the precision and the accuracy of the analytical method. A more realistic method is to evaluate the accuracy and the precision values as a function of the analyte concentration.5.6 An even more realistic estimate of the accuracy and the precision is obtained by also taking into account the variation between individual runs of the analytical procedure. This is the topic of the present study.The range of an analytical method can be defined in terms of the square root of the *' Present address: Skovgirdsvej 2. 3550 Slangcrup, Denmark. relative mean square error RMSE(p)i of the method. The RMSE(p)i is a function of the precision and the accuracy, each of which can depend on the analyte concentration.5." RMSE(p)l is calculated by using parameter estimators from curve fittings of an estimated MEF.7 The method evaluation experiment is designed so that RMSE(p)l combines the systematic error with the over-all analytical variation, i.e., the variation during a single run of the analytical method and variation between individual runs, including sample pre-treatment and recalibration of the instrument. The interval of definition for the analytical method can be determined by the demand to the RMSE(p)' of the method.The RMSE(p)i is a function of the precision and the accuracy, both of which can depend on the true analyte content. In order to reveal a concentration dependence, the MEF of the analytical method is calculated. The MEFs provide an easy way to obtain a stringent document on an analytical method and help to set rules for the quality control. The calculation of the MEF is, however, not without problems. The calculation of MEFs for the ICP-AES determination of Fe, Mn and Ti from metal-spiked filter samples will be presented in the following. Theory Any chemical analytical method is characterized by its MEF, which is the expected value as a function of the 'true' content of the analyte in the sample.Denoting the result of the chemical analysis by Y and the true content of the analyte by p, the MEF is given by: MEF(l4 = E(YIP) = &4 where MEF(p) is the method evaluation function, depending on p, i.e., the true analyte content, and E( Ylp) is the estimated value of Y , given p, i.e., the analytical result, Y , from a sample with known content p. It is tacitly assumed that the analytical method i s in statistical control, which means that the distribution of Y , given p, can be approximated by a normal distribution, i.e., (YIP) E "dd, (5J where 0,. is the precision, i.e., the standard deviation of the unsystematic error (the random error). Note that oY may be a function of p. It is very important to realize that (T,, is the total708 ANALYST.APRIL 1992, VOL. 117 standard deviation of the analytical method, i.e., a combina- tion of the standard deviation within a single run of the method and the standard deviation between independent runs of the method. The expected value of Y , given p, g(p), is the sum of the true content p and the accuracy, i.e., the systematic error 6, of the analytical method is 6, = g(p) - p. In general, 6, is a function of p. The linearity of the MEF, i.e., g(p) = a + pp, was tested by using the pure error lack of fit test;86, is then given by 6, = (a + pp) - p = a + (p - l ) p The systematic error is thus the sum of the so-called zero-point error a and the scale error (p - l)p, i.e., the systematic error has been resolved into a translational and a rotational bias.5-9 From an ideal point of view, an analytical method should be without any systematic error (i.e., a = 0 and p = l), which implies: The method is then in analytical control by definition.In practice, however, it is not possible to assure that 6, = 0 for all values of p in the range of definition for the analytical method. The aim of the quality control is, therefore, normally to secure that laYl < 6" for all values of p where is a pre-set constant. The variance oy of the unsystematic error should be simultaneously controlled, requiring that oy < 00 for all values of p, where oo is a pre-set constant. If the relative errors are to be controlled instead of the absolute errors, the requirements will be 16,/pI < 6dp = a. and oY/y < o& = ko, respectively, for all values of p in the definition range.The simultaneous control of 6, and o, is equivalent to control of the mean square error (MSE), which is defined as MSE(p) = E{[(YIp) - p12} = 6,* + oY2 = {[a + (p - 1)p]2 + oy2} Note that MSE is a function of the true content p. The control scheme is thus equivalent to MSE(p) < 602 + 0 0 ~ = MSEo Simultaneous control of 6,/p and oy/p is equivalent to control of the RMSE = MSE/p2 = { [ a l p + (p - 1)12 + oY2/p2! , which may be a more useful parameter. Note that RMSE(p) IS a decreasing function of p: RMSE(p) < (60/~)2 + (oo/p)' = RMSEo where RMSEo is a pre-set constant. The RMSE is calculated by inserting the estimators for a, p and a,. The limit of determination is then defined as the lowest concentration of p that produces a RMSE(p)g < (RMSE,,);.Experimental The calculation of MEFs requires standard solutions for calibration purposes and samples with a given true content. To eliminate any dependence between samples and the calibration graph, chemicals from different manufacturers were used to prepare standard solutions for calibration purposes and to prepare spiked 8.0 pm cellulose acetate filters used as samples with known contents (method evaluation samples). Spiked Fe, Mn and Ti method evaluation samples with true contents of 2.5, 12.5, 25, 50, 75, 100 and 125 pg per filter of each element were prepared by adding 50 pl of an appropriate standard solution to cellulose acetate filters. In the method evaluation study, 22 calibration graphs for each of the three elements were prepared.One, and only one, randomly selected method evaluation sample with known analyte concentrations was analysed by reference to each calibration graph. In order to test the linearity of the MEF, by a pure error test,8,10,11 several filters of each concentration were analysed except for the filter with 50 pg. All measurements were performed on an ICP-AES Plasma I1 spectrometer (Perkin Elmer). Inter-elemental spectral interference between the 21 elements, now analysed in one multi-element determination, were investigated by aspirating freshly prepared single-element solutions into the plasma, and optimum wavelengths regarding high sensitivity with less interference were selected. A detailed description of the automated parameter optimization is given in ref.12. The instrumental parameters used during all analyses for Fe, Mn and Ti are presented in Table 1. Reagents Solutions for calibration purposes were prepared from ferric nitrate, manganous nitrate and titanium nitrate BDH Spectro- sol standards. To prepare the method evaluation samples in the desired concentration range it was necessary to use standards from two different manufacturers. Spiked filters containing 2.5 and 12.5 pg per filter of each element were prepared from Fe, Mn and Ti Spectrascan element standards for atomic spectroscopy from Teknolab A / S (Drobak, Nor- way). The purity of these standards were 99.998, 99.98 and 99.9% for Fe, Mn and Ti, respectively. Merck Titrisols containing Fe, Mn and Ti (1.000 k 0.002 g) were used for filters containing 25, 75, 100 and 125 pg per filter of each element.To dilute calibration standards and to dissolve the filter samples, the following solvents were used: HN03, 65% (Merck, Suprapur); HCI, 30% (Merck, Suprapur); and de-ionized water (Millipore). Sampling Pre-treatment and Calibration Standards Before ICP-AES analysis the metal-spiked filter samples were dissolved. The filters were digested at 175 "C in 20 ml of mixed acid, consisting of 10 ml of 33% v/v HCI and 10 ml of 33% v/v HN03 in glass cylinders on a hot-plate for approximately 1 h until 5 ml of solution remained. The residual solution was then diluted to 50 ml with water for filters containing 2.5 and 12.5 pg of each metal, and otherwise to 25 ml, for measurement by ICP-AES. Samples were stored in plastic containers con- trolled for contamination.Only one method evaluation Table 1 ICP-AES instrumental parameters for determination of Fe, Mn and Ti Source Peak Survey Wavelength/ viewing window*/ window?/ Element nm height/mm nm nm Fe 238.204 13 0.030 0.045 Mn 260.569 8 0.030 0.080 Ti 336.121 12 0.020 0.100 Equilibration time 15 ms Sampling time 400 ms R.f. power 1260 W PMT gain 600 V Pump rate 1.0 ml min-' Background correction Automatic Argon flow rates Nebulizer flow 1 1 min-1 Plasma flow 15 1 min-1 Auxiliary flow 1 1 min-l * Designates the width surrounding the analytical line, which is used ? Sets the total wavelength range over which the spectral profile is to for curve fitting for intensity measurements. be taken.ANALYST, APRIL 1992, VOL.117 709 sample was digested at a time. To eliminate contamination, all glassware was stored overnight in the mixed acid solutions and rinsed thereafter five times with de-ionized water. The following Fe, Mn and Ti concentrations were used for the calibration graphs: 0.00,0.02,0.05,0.10,0.20,0.50, 1.00, 2.00, 5.00 pg ml-1. Multi-element calibration standards were diluted with a mixed acid (1.2 mol dm-3 HCl-1.8 mol dm-3 HN03) to match the matrix between samples and calibration standards. 13 All calculations were performed with the statis- tical data program Minitab.10 Results The following general equation has been accepted for defining the limit of detection:'.' LOD = kd . (sB where kd is an abitrary constant, and oB is the within-batch standard deviation of the blank.14 Table 2 ICP-AES analytical blank values for ten blank filters, the within-batch standard deviation oB and limits of detection (LOD) for Ti.Mn and Fe Blank No. 1 2 3 4 5 6 7 8 9 10 Blank response for Ti/yg per filter -0.250 -0.250 -0.225 -0.250 -0.250 -0.275 -0.250 -0.275 -0.200 -0.225 Blank response for filter 0.150 0.150 0.150 0.175 0.150 0.150 0.150 0.150 0.150 0.150 Mdyg Per Blank response for Fe/yg per filter 0.350 0.200 0.300 0.075 0.425 0.250 0.100 0.050 0.200 0.200 Mean response, dpg per filter -0.245 0.153 0.215 Overall batch standard deviation, (sB/yg per filter 0.0230 0.0079 0.1209 Limit of detection, 3oB/yg per filter 0.0689 0.0237 0.3626 Table 3 True contents and ICP-AES analytical results for Fe, Mn and Ti spiked filter samples 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Analytical results/pg per filter True content/ yg per filter Fe Mn Ti 2.5 2.5 12.5 12.5 12.5 25.0 25.0 25 .0 25.0 25.0 50.0 75 .0 75.0 100.0 100.0 100.0 100.0 125.0 125.0 125.0 125.0 125.0 * * 11.750 12.250 11.150 25.450 24.800 24.550 25.300 30.500 49.775 77.225 77.260 101.425 104.175 97.333 98.750 120.075 127.950 124.175 128.475 130.825 2.550 2.100 12.600 12.800 12.690 24.725 24.625 25.800 24.725 24.175 50.525 72.950 73.405 98.825 108.125 97.952 98.250 123.225 119.375 129.475 128.125 130.100 * Below LOD.1 Outlier-standard residual greater than 3. 2.750 2.900 13.450 13.314 12.800 25.575 22.550 26.000 25.875 26.500 49.975 77.138 79.558 100.750 t 101.848 102.625 126.625 126.600 133.350 129.250 131.200 The American Chemical Society's Subcommittee on Envi- ronmental Analytical Chemistry has recommended using a value of 3 for kd.14 To obtain estimates of the LOD values for Fe, Mn and Ti, ten blank filters were treated as samples and analysed.The estimators for oB and LOD for Ti, Fe and Mn are given in Table 2. Table 3 gives the analytical results and the true contents of the individual metal-spiked filter samples (method evaluation samples). MEFs for Ti, Mn and Fe, calculated from the results in Table 3, are shown in Fig. 1. The MEF data were first analysed by the linear least-squares regression .s Plots of the standardized residuals versus the true contents of Fe, Mn and Ti (Fig. 2) indicate that the standard deviation of the residuals depends on the concentration. The standard deviation of the residuals for the samples with true contents of 2.5 and 12.5 pg per filter deviate from the standard deviation of the residuals for the remaining samples.Weighted regres- sion analysis must be performed.11 It is then assumed that the variance of (YIP) is oy2/w, where w is the weight. Appropriate 40 C P 140 I (c) 100 I2O 1 0 20 40 60 80 100 120 140 True content/pg per filter Fig. 1 Method evaluation functions for ( a ) Ti, (b) Mn and (c) Fe, estimated from analytical results from Fe, Mn and Ti spiked cellulose acetate filters. ( a ) Zero-point error a (estimated) = 0.12 yg per filter; standard deviation for estimated a = 0.40 pg per filter; confidence interval for a = (-0.74; 0.97) pg per filter. Slope fl (estimated) = 1.03; standard deviation for estimated fl = 0.6 X 10-2; confidence interval for fl = (1.02; 1.04).( b ) Zero-point error a (estimated) = -0.15 yg per filter; standard deviation for estimated a = 0.65 pg per filter; confidence interval for a = (-1.51; 1.20) pg per filter. Slope fl (estimated) = 1.01; standard deviation for estimated (3 = 1.1 x 10-2; confidence interval for fl = (0.98; 1.03). (c) Zero-point error a (estimated) = -0.47 yg per filter; standard deviation for estimated a = 0.79 pg per filter; confidence interval for a = (-2.14; 1.20) pg per filter. Slope 0 (estimated) = 1.01; standard deviation for estimated p = 1.2 x 10-2; confidence interval for p = (0.99; 1.04)710 ANALYST, APRIL 1992, VOL. 117 1 C 0 O I - 1 O J 0.4 0.3 0.2 0.1 -2t O 0.6 0.5 2 1 '1 0 a -1 3 U -2 -3 -2 -'I0 0 0 0 0 -3 ' 20 40 60 80 100 120 140 True content/Fg per filter Fig.2 Mn and (c) Fe Standardized residuals versus the true content for (a) Ti, (b) weights should then be chosen, according to the differences in the dilution. By considering the three plots (Fig. 2) as a whole, the standard deviation of the MEF points, corresponding to the 2.5 and 12.5 pg per filter, diverge from the remaining MEF points by approximately half their standard deviation. The observations corresponding to these filters are, therefore, given the weight 1, while all the other observations are given the weight (%)* = 0.25. This is in accordance with the differences in the dilution. From pure error tests, the following significant levels (p) for linearity of the MEF were found: 93% < p < 96% for Mn, 57% < p < 59% for Ti, and 74% < p < 76% for Fe.There is, therefore, no reason to reject the hypothesis of E(YIp) = a + Estimators from the weighted least-squares curve fitting are summarized in Fig. 1, including confidence intervals for the regression coefficients. The 95% confidence intervals of the zero-point error (x = 0 tests were found to be 0.82, 0.78 and 0.56 for Mn, Ti and Fe, respectively. The p confidence intervals (Fig. 1) for Fe and Mn includes p = 1.00. The p = 1 hypothesis could not be rejected at the 0.05 significance level (significance levels 0.52 and 0.58, respectively). For Ti, (J = 1 is not included in the confidence interval (Fig. 1). The significance level for the hypothesis (J = 1 test was less than 0.001. Fig.3 shows the square root of the relative mean square error RMSE( p)i versus the true content for Ti, Mn and Fe. If a maximum analytical variation of 5% is accepted i.e. (RMSE,,)! (JP. I I I I I I I ;r, 0.4 5. w I 0.3 u 0.2 0.1 0 20 40 60 80 100 120 140 True contentlpg per filter Fig. 3 RMSE(p)I versus the true content for (a) Ti, ( b ) Mn and (c) Fe. (a) o , = 0.92 (weight = 1); (b) oy = 1.46 (weight = 1); and (c) oy = 1.01 (wdght = 1) = 0.05, the lowest certification limits (limits of determination) obtained from Fig. 3 are: 50 pg per filter for Ti, 50 pg per filter for Mn and 45 pg per filter for Fe. Discussion The MEFs can be considered as linear in the investigated range (Fig. l ) , according to the results from the pure error test. Parameter estimators for the MEFs were calculated by using weighted least-squares regression due to a concentration dependence on the standard deviation of the residuals (Fig.2). All the samples were digested independently but in the same manner; 2.5 pg and 12.5 pg per filter samples were then diluted to 50 ml with water instead of 25 ml as for the remaining samples. This dilution procedure was selected to achieve a concentration (pg ml-1) corresponding to the lowest point of the calibration graph. The intercept (x (the zero-point error) of the MEFs was found by tests not to be significantly different from 0 for the three elements investigated. The slope p of the MEFs for the Fe and Mn determinations was not significantly different from 1. Therefore, the systematic errors [a + (p - 1)p] can be assumed to be negligible for the ICP-AES determination of Mn and Fe.For the Ti determination, the scale error, ((J - 1)p, was significantly different from 0, and the results indicated an over-estimation of Ti. Analytical results for Ti, therefore, need to be corrected for the systematic error (recovery).ANALYST, APRIL 1992, VOL. 117 71 1 The RMSE(p)t values combine the systematic error with the over-all analytical variation, i.e., variation during a single run of the analytical method and variation caused by recalibration. RMSE(p)i could then be said to represent the total relative analytical error. Though the systematic error may be negligible according to the statistical calculations the estimated RMSE(p); displays a slight dependence of the concentration (Fig.3). This is, however, in accordance with the fact that RMSE(p) is a decreasing function of p. In the present study, the RMSE(p)i has been applied in an intuitive way. This is acceptable, owing to the relatively high number of samples. It must, however, be emphasized that the statistical characteristics of the estimator for RMSE( p)+ must be taken into consideration with a smaller number of samples. The certification limit, o r limit of determination, can be set, by demands to the relative mean square error. An over-all relative analytical error of 5% is assumed to be acceptable in the verification of filter samples in this concentration range of approximately 1/10 of the Danish threshold limit values. Conclusions The MEF contains most of the relevant information for a chemical analytical method.Therefore, it is important to obtain a reliable estimate of the MEF at an early stage in the development as it is possible to recognize systematic and random errors resulting from changes in the analytical method. In the present study with a pre-set maximum accepted relative analytical error at 5%, the lower certifica- tion limits (limits of determination) were found from the RMSE(p)i, calculated by using parameter estimators obtained from the curve fittings of the MEFs (using weighted least- squares regression). The fact that the analytical method for Fe and Mn is in analytical control could not be rejected. The analytical results for Ti have to be corrected in order to take account of the systematic error (the recovery). We gratefuly acknowledge D. Meincke and M. Toxvaerd- Larsen for technical assistance, and G. Frederiksen for writing this manuscript. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Pittwell, L. R., Anal. Proc., 1988, 25, 192. Shan, X.-Q., Wang, T.-B., and Ni, Z.-M., Fresenius’ 2. Anal. Chem.. 1987,326, 419. Wang. C.-F., Miau, T. T., Perng, J. Y., Yeh, S. J., Chiang, P. C., Tsai, H. T., and Yang, M. M., Analyst, 1989,114, 1067. Yamashige, T., Yamamoto, M., and Sunahara, H., Analyst, 1989, 114, 1071. Ramsey, M. H., Thompson, M.. and Banerjee, E. K . , Anal. Proc.. 1987.23. 260. Ripley, B. D., and Thompson, M.. Analyst, 1987, 112, 377. Hansen, A. M., Olsen, 1. B., Holst, E., and Poulsen, 0. M., Ann. Occup. Hyg., 1991, 35, 603. Armitage. P., Statistical Methods in Medical Research, Black- well, Oxford, UK, 1971. pp. 275-279. Ramsey. M. H., and Thompson, M., J . Anal. At. Spectrom., 1987. 2, 497. Minitab Reference Manual, Release 6.1, DWS, Duxbury Press, Boston, MA, 1988, pp. 341. Aarons, L., Toon, S . , and Rowland, M., J . Pharmacol. Methods, 1987, 17, 337. Ediger, R. D., At. Spectrosc., 1986, 7. Que Hee, S. S . , MacDonald, T. J., and Boyle, J . R., Anal. Chem., 198.5, 57, 1242. American Chemical Society Subcommittee on Environmental Analytical Chemistry, Anal. Chem., 1980. 52, 2242. Paper 1 I0041 I E Received January 29, 1991 Accepted October 24, 1991

ISSN:0003-2654    

DOI:10.1039/AN9921700707

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST. APRIL 1992, VOL. 117 713 Ion Chromatographic Study of the Effect of Ammonium Nitrate as a Modifier in Electrothermal Atomic Absorption Spectrometry Muhammad Mansha Chaudhry and David Littlejohn Department of Pure and Applied Chemistry, University of Strathclyde, Cathedral Street, Glasgow G7 7XL, UK Ammonium nitrate can be used as a chemical modifier in the determination of volatile elements by electrothermal atomic absorption spectrometry. It reacts with sodium chloride to produce ammonium chloride and ammonium nitrate, both of which are volatilized or decompose at temperatures below 400 "C. The effectiveness of the reaction has been studied by ion chromatography to analyse solutions of salt residues remaining on a pyrolytic graphite platform heated to different char temperatures.Determination of ammonium, chloride, nitrate, nitrite and sodium in the solutions allowed losses of the ions during the char step to be monitored. In the absence of ammonium nitrate, losses of sodium and chloride began about 700 "C and were complete by 1100°C. Addition of 300 pg of ammonium nitrate to 100 pg of sodium chloride caused an 8590% loss of chloride from the platform a t 200 "C. Complete removal of chloride was not achieved until a temperature of 1000°C was achieved. Removal of chloride was accompanied by almost total loss of ammonium and partial loss of nitrate at 200°C. From 200 to 600"C, a more gradual loss of the remaining nitrate occurred and the production of nitrite was observed in the range 300-700". In the presence of ammonium nitrate, loss of sodium was noted at 800"C, and was almost complete by 1100°C.Keywords: Ammonium nitrate modifier; electrothermal atomization; ion chromatography The use of chemical modifiers in electrothermal atomic absorption spectrometry (ETAAS) is now well established.' Many modifiers act by stabilizing the analyte to temperatures at which a significant fraction of the matrix components can be vaporized at the char or pyrolysis stage. Of these, palladium is widely recommended for elements such as Au, Bi, Pb and Se,*-i) which form inter-metallic compounds with the modifier. One of the first modifiers, ammonium nitrate, proposed by Ediger et al. ,7 has a different function to that of palladium and other analyte-stabilizing reagents. This modifier is added to allow the removal of sodium chloride, a major cause of interference in ETAAS, at a pyrolysis temperature at which the analyte is not yet volatile.Ediger et a1.7 proposed that, in the presence of an excess of ammonium nitrate, sodium chloride forms sodium nitrate and ammonium chloride at the drying stage by the following reaction: (boils at (decomposes at (decomposes at (sublimes 1410 "C) 210 "C) 380 "C) at 335 "C) This scheme implies that sodium nitrate, ammonium chloride and unreacted ammonium nitrate can be vaporized from the furnace at charring temperatures of about 400 "C or above. Consequently, ammonium nitrate has been used successfully by a number of workers as a modifier in methods for the determination of volatile elements, such as cadmium, in chloride-containing matrices. Pruszkowska et al.8 devised a procedure for the determina- tion of cadmium in sea-water and reported that 500 pg of ammonium nitrate reduced the background absorbance signal caused by 2 p1 of sea-water more effectively than did 200 pg of diammonium hydrogen phosphate.However, they suggested that addition of 8% v/v nitric acid with the phosphate would be as effective as ammonium nitrate in promoting the removal of chloride. A similar conclusion about the usefulness of nitric acid was also reached by Feitsma et al.9 They preferred 2% v/v nitric acid over a number of other modifiers for the determina- tion of cadmium in urine. In contrast, Yin et al.")reported that nitric acid was vaporized too early in the char stage to be effective in the removal of sodium chloride, whereas ammo- nium nitrate was retained long enough to form volatile ammonium chloride.They recommended the addition of 500 pg of palladium as a modifier for the determination of NaCl + NH4N03 --+ NaN03 + NH4Cl cadmium in biological materials. The inclusion of palladium allowed a char temperature of 900 "C to be used without loss of cadmium. Ammonium phosphate modifiers were rejected by Yin et ~ 1 . 1 0 as contamination with cadmium is frequently a problem. 1 1 In this study, the removal of chloride by reaction with ammonium nitrate during the char stage has been investigated by analysis of the residue left on a pyrolytic graphite platform after hzating to different char temperatures. The analysis was performed by ion chromatography, which allowed determina- tion of the amount of ammonium, chloride, nitrate, nitrite and sodium in the deposit after dissolution in distilled water.A comparison of the char temperature curves for each ion allowed confirmation of the above reaction and permitted an assessment of its efficiency in removing sodium chloride from the atomizer. Experimental Instruments A Philips (now Unicam) SP9 atomizer, fitted with a pyrolytic graphite-coated electrographite tube and a pyrolytic graphite platform, was used with oxygen-free nitrogen as the purge gas. Anion and cation determinations were performed with a Dionex 4000i ion chromatograph. The conditions used for the separation and detection of the ions are given in Table 1. Procedure A 20 p1 volume of the test solution was injected on to the platform in the graphite-tube atomizer, and dried at 150 "C with a ramp rate of 20 "C s-1 and a hold time of 30 s.The dried deposit was then heated to the required char temperature at 200°C s-1 for a hold time of 40 s. The atomizer was then cooled to ambient temperature, the platform was removed with tweezers, and the residue was dissolved in 10 ml of distilled water. The concentrations of the ions in the solutions injected into the Dionex chromatograph were generally less than 20 pg ml-* and were determined by calibration of the instrument with mixed anion or cation $tandard solutions prepared in distilled water. Example chromatograms are presented in Fig. 1714 3.5 3.0 - i - . 5 2.5 - i 0 - Lc c - al a 2.0 - ; ANALYST, APRIL 1092, VOL.117 1 .o 0.8 - 0.6 5. . Table 1 Ion chromatography with Dionex 4000i Anions Cl-, N03-. NO?- Anion column : HPIC-AS4A Anion eluent : 0.75 mmol dm-3 NaHN03 plus Suppressor column : Anion micro-membrane suppressor Regenerent solution : 5 mmol dm-3 H2SOJ Conductivity range : 10 ,US Cations : Na+,NH,+ Cation column : HPIC-Fast-Separation cation Cation eluent : 20 mmol dm-3 HCl plus Suppressor column : Cation micro-membrane suppressor Regenerant solution : 0.1 mol dm-3 tetrabutylammonium Conductivity range : 10 pS : 2.2 mmol dm-3 Na2C03 columns I and I1 0.3 mmol dm-3 DAP* HCI hydroxide * DAP = 2.3-diaminopropionic acid. Time - Fig. 1 Anion chromatogram of a solution of the residue of 100 pg of NaCl plus 300 pg of NH4N03 on a platform after heating to (a) 200 and (6) 400°C (for conditions see Table 1) Reagents All the reagents used were of analytical-reagent grade.Distilled water was used for the preparation of all solutions. Results and Discussion Ion chromatography was used to determine the masses of sodium and chloride left on a platform when separate additions of 100 pg of sodium chloride were heated to various char temperatures up to 1100°C. The data are presented in Figs. 2 and 3 as micromoles of sodium or chloride plotted against temperature. The char curves indicate that losses of sodium and chloride begin at temperatures above 700°C and are virtually complete by 1100 "C. This is in agreement with ETAAS analyses for sodium chloride carried out by other workers. 12.13 When 300 pg of ammonium nitrate was added to 100 pg of sodium chloride, 85 to 90% of the chloride was removed at 200°C (Fig.2). Complete removal of chloride was not achieved until the platform was heated to 1000°C. It is apparent, therefore, that conversion of sodium chloride into ammonium chloride is incomplete even when the number of moles of ammonium nitrate present is twice that of sodium chloride. The removal of the residual chloride at 1000°C is probably as a result of the vaporization of untreated sodium chloride. 0 400 800 1200 Char temperaturePC Fig. 2 Effect of char temperature on the amount of chloride, nitrate and nitrite remainingon a graphite platform (100 pg of NaCl; 300 pg of NH4N0,). A. Chloride (sodium chloride only); B. chloride in the presence of ammounium nitrate; C. nitrate in the presence of sodium chloride; and D, nitrite in the presence of sodium chloride 0 400 800 1200 Char temperaturePC Fig.3 Effect of char temperature on the amount of ammonium and sodium remaining on a graphite platform (100 pg of NaCl; 300 pg of NHlN03). A, Sodium (as sodium chloride); B, sodium in the presence of ammonium nitrate; and C, ammonium in the presence of sodium chloride The rapid loss of chloride at 200"C, in the presence of ammonium nitrate, was accompanied by the almost total removal of ammonium ions from the platform (Fig. 3), which confirms the formation and removal of volatile ammonium chloride. As an excess mass of ammonium nitrate was used, the complete removal of NH4+ confirms that the decomposi- tion of ammonium nitrate accompanies the formation and vaporization of ammonium chloride.This is supported by the observation (Fig. 2) that, at 200"C, there is a significant decrease in the mass of nitrate on the platform to a level equivalent in moles to the mass of sodium introduced as sodium chloride. The temperature at which 2 pmol of NO3- was vaporized corresponds to the reported decomposition temperature of ammonium nitrate.7 Figs. 2 and 3 indicate that, when ammonium nitrate is used as a modifier, the removal of most of the chloride and the decomposition of the excess of ammonium nitrate occur between 200 and 300"C, leaving a residue of sodium nitrate with about 10% of the entire mass of sodium chloride. Ediger et aZ.7 noted that ammonium chloride is reported to have a sublimation temperature of 335 "C, but the ion chromato- graphy results show that loss of chloride begins at a much lower temperature in the graphite furnace.Fig. 2 shows that between 300 and 600 "C there is a further gradual decrease in the mass of nitrate, which is probably dueANALYST, APRIL 1992, VOL. 117 715 to decomposition of sodium nitrate to sodium oxide. The melting point of sodium nitrate is reported to be 307 "C and the decomposition temperature is 380 "C.' Decomposition of sodium nitrate apparently occurs, at least in part, via the nitrite: 2NaN03(s)--+ 2NaN02(s) + 02(g) which is indicated by the appearance of nitrite in the ion chromatogram for char temperatures above 300 "C. Detect- able concentrations of nitrite were measured up to a char temperature of 600 "C. This allows the tentative suggestion that further stoichiometric decomposition of sodium nitrite to sodium oxide may involve release of equimolar amounts of NO and NO?.2NaNO?(s)-+ Na20(s) + NO(g) + NO?(g) Loss of sodium was noted at 800 "C and was mostly complete by 1100 "C (Fig. 3). The mechanism is likely to involve either carbon reduction of the oxide o r sublimation, which can occur above 700 to 800°C. Ion chromatography has proved to be a useful procedure for investigating the chemical reaction between micromolar amounts of sodium chloride and ammonium nitrate in a graphite furnace. The measurements reported here show that complete removal of chloride from sodium chloride is not achieved even with a 2-fold molar excess of ammonium nitrate. However, the fact that a substantial reduction in the mass of chloride occurs at 200°C explains why ammonium nitrate has proved to be such a useful modifier in the determination of cadmium and other volatile elements by ETAAS.4 5 6 7 8 9 10 11 12 13 References Tsalev, D. L., Slaveykova, V. I., and Mandjukov, P. B . , Spectrochim. Acta Rev., 1990, 13, 225. Shan, X.-Q., and Wang, D.-X., Anal. Chim. Acta, 1985. 173, 315. Teague-Nishirnura, J . E., Tominaga, T., Katsura, T., and Matsurnoto, K., Anal. Chem.. 1987, 59, 1647. Egila, J.. Littlejohn. D., Ottaway, J . M., and Shan, X.-Q., J. Anal. At. Spectrom.. 1987, 2, 293. Weiz, B., Schlemmer, G., and Mudakavi, J . R., J. Anal. At. Spectrom., 1988, 3 , 93. Beach, L. M., Spectroscopy, 1987, 2, 21. Ediger, R. D., Peterson, G. E., and Kerker, J . D., At. Absorpt. Newsl.. 1974, 13, 61. Pruszkowska, E., Carnrick, G. R., and Slavin, W . . Anal. Chem., 1983, 55, 182. Feitsma, K. G., Franke. J. P., and de Zeeuw, R. A., Analyst, 1984, 109, 789. Yin. X., Schlernmer, G., and Welz, B . , Anal. Chem.. 1987.59. 1462. Welz, B . , and Schlemmer, G., J. Anal. At. Spectrom., 1986, 1, 119. Pitchard, M. W., and Reeves, R. D., Anal. Chim. Acta. 2976, 82. 103. Welz, B . , Akman, S . , and Schlemmer, G., J. Anal. At. Spectrom., 1987. 2. 793. Paper I I04 772 H Received September 16, 1991 Accepted October 28, I991

ISSN:0003-2654    

DOI:10.1039/AN9921700713

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, APRIL 1992, VOL. 117 717 Effects of Moisture on the Cold Vapour Determination of Mercury and Its Removal by Use of Membrane Dryer Tubes Warren T. Corns, Les Ebdon and Steve J. Hill Plymouth Analytical Chemistry Research Unit, Department of Environmental Sciences, Polytechnic South West, Drake Circus, Plymouth, Devon P14 8AA, UK Peter B. Stockwell PS Analytical Ltd., Arthur House, 64 Chaucer Business Park, Watery lane, Kemsing, Sevenoaks, Kent TN75 6QY, UK Continuous-flow vapour generation is now a well-established sample-introduction technique for mercury and the hydride-forming elements. One of the problems associated with this technique is moisture carryover, which originates during the gas-liquid separation process. Excessive moisture carryover causes gradual loss in sensitivity and baseline drift for atomic absorption and atomic fluorescence spectrometry and inductively coupled plasma detection systems.Chemical desiccants and physical moisture traps can be used t o reduce moisture for short periods of time, although these may give rise t o contamination and analyte losses. A more effective way t o remove moisture carryover has been found by using a semi-permeable Nafion membrane dryer tube, which continuously desolvates the wet gaseous stream. This device improved the long-term stability and enhanced sensitivity. A relative standard deviation of 2% was achieved for 90 runs of a 1 pg I-' standard, obtained over a period of 3 h. Keywords: Vapour generation; atomic fluorescence spectrometry; semi-permeable membrane dryer tube; mercury determination; moisture removal The toxicological effects of mercury compounds on environ- mental systems has long been recognized.This has led to the development of many techniques for the determination of mercury in a wide variety of samples. Originally, colorimetric and spectrophotometric methods, based on the formation of coloured chelates between mercury ions and dithizone or dinaphthylthiocarbazone, were adopted. I However, these methods had poor sensitivity and entailed long and tedious manipulations, which often caused contamination and loss of analyte. Today, the most commonly used method for deter- mining mercury is cold vapour atomic absorption spec- trometry (CVAAS). This technique was first described by Poluektov and Vitkun as early as 19632 and later popularized by Hatch and Ott.3 Since that time, numerous modifications have been reported435 and several commercial systems are now available.Such vapour-generation systems are normally in the form of an accessory, which is coupled with an atomic spectrometer. Reducing agents such as tin(i1) chloride or sodium tetrahydro- borate(ii1) are commonly used to reduce mercury ions to gaseous elemental mercury. The mercury vapour is then stripped out of solution by using a gas-liquid separator, and then delivered to the detector. Although CVAAS has now become a widely accepted and utilized technique, there are several disadvantages associated with the use of AAS as the detector. These result from the limited linear calibration range and spectral interference arising from non-specific back- ground absorption of volatile organics, such as acetone and benzene .h West, in 1974,7 showed theoretically that cold vapour atomic fluorescence spectrometry (CVAFS) should be more sensitive and produce considerably less spectral interference from non-specific absorption compared with the correspond- ing AAS technique.This theoretical treatment was later verified practically by Thompson and Godden .g The appa- ratus they described was a dispersive system based on a modified AAS instrument, and detection limits of 0.02 ng were readily obtained. More recently, Stockwell et al.9 have shown that detection limits below 0.02 pg I-' can be obtained with ease by using continuous-flow vapour generation coupled with an AFS instrument.No AAS instrument is required, and the system is inexpensive and simple in both construction and operation. One of the problems associated with the determination of mercury by the cold vapour technique is that moisture occasionally condenses on the transfer tube walls and eventu- ally enters the detection system. Problems resulting from the moisture have been mentioned by numerous authors, who used various designs of separator. 1,5,10-*3 Excessive moisture carryover causes a gradual loss of sensitivity and baseline drift for AAS, AFS and inductively coupled plasma systems. The moisture, usually in the form of an aerosol, is a complicated mixture of reagents and sample matrix. In severe instances this mixture can cause fogging of optics and atom cell windows.The most commonly used method for the reduction of moisture carryover during the determination of mercury involves the use of desiccants, such as concentrated sulfuric acid,ll magnesium perchlorate,4 activated silica gel14 or anhydrous calcium chloride. 13 Although effective, these drying agents rapidly become saturated and, therefore, need constant replacement. They can also give rise to contamina- tion and loss of analyte at trace levels. This paper will address the problem of moisture carryover for continuous- flow vapour-generation AFS. The removal of moisture by using desiccants, physical water-traps and semi- permeable membrane drying tubes is discussed with reference to long-term stability and sensitivity. Experimental Reagents All reagents used were of analytical-reagent grade, unless stated otherwise [BDH (now Merck), Poole, Dorset, UK].De-ionized water was used throughout (Milli-R04, Millipore, Milford, MA, USA). A 2% m/v tin(I1) chloride solution was used as the reductant, and trace amounts of mercury were removed from this solution by purging with argon for approximately 30 min. Standard solutions were prepared by appropriate dilution of a stock 1000 mg 1-1 mercury(i1) chloride solution (SpectrosoL, BDH) with 2% nitric acid. All glassware was soaked in 10% nitric acid for at least 24 h before use, and then rinsed five times with de-ionized water.718 Dryer gas out Dryer gas in ANALYST, APRIL 1992, VOL. 117 I Wet gas from- ' -Detector separator Hygroscopic membrane Fig. 1 Semi-permeable membrane dryer tube 160 I I I I I 1 0 20 40 60 80 100 Run number Fig.2 Variation of peak height with run number Instrumentation An automated continuous-flow vapour generator (PSA 10.003, PS Analytical, Sevenoaks, Kent, UK) was used to generate gaseous mercury. The generated mercury gas was then detected by AFS utilizing a 254 nm interference filter for wavelength isolation and reduction of background scatter (Merlin, PSA 10.023, PS Analytical). This instrumentation has been described in detail elsewhere.9 Wet gas from the gas-liquid separator was continuously dried by using a semi-permeable Nafion membrane dryer tube (Perma Pure Products, Monmouth Airport, Farmingdale, NJ, USA). This basically consists of two concentric tubes 24 cm in length connected with variable bore T-piece connectors (Fig.1). The outer tube is made from poly(tetrafluoroethy1ene) (PTFE) and has dimensions 4 mm i.d. x 6 mm 0.d. The inner tube is a hygroscopic Nafion membrane with dimensions 2 mm i.d. x 3 mm 0.d. As the wet gas from the separator passes through the inner membrane, the moisture is removed and transferred to the outer tube. Meanwhile, a dryer gas flows in a direction opposite to that of the wet gas, removing the moisture on the outer surface of the membrane. The dryer can be any dry gas such as air, nitrogen or argon. Results and Discussion The problem of moisture carryover originates from the gas-liquid separator design. Most commercial separators make use of a purging procedure to liberate mercury gas from solution and, therefore, have associated moisture carryover.The separator supplied with the PSA vapour generator has a swan-necked design, which minimizes moisture carryover; however, this is not totally effective and moisture in the form of an aerosol can be transported along the transfer tubes to the detector. Recently, Hoult14 assessed the extent of moisture carryover for this design of separator. This was achieved by placing an activated silica gel trap between the separator and detector. By weighing the silica gel trap before and after a 2 h analysis period the amount of moisture carryover was determined. With a purge-gas flow rate of 1 1 min-1, approximately 0.8 g h-1 of moisture was flushed through the system. Decreasing the flow rate to 0.5 1 min-1 reduced the moisture carryover to 0.5 g h-1. Hoult also noticed that the removal of moisture increased the net signal intensity by at least 60%.This increase ~~~ ~ Table 1 RSDs (%) for the first, second, third and fourth batch of 25 runs Method of moisture removal No drying facility Physical moisture trap Membrane dryer tube before simplex optimization after simplex optimization Membrane dryer tube Run number 1-25 26-50 51-75 76-100 1.2 5.5 6.0 11.4 2.3 2.2 14.8 - 2.3 4.7 2.5 3.3 1.8 0.8 1.2 1.2 Wet gas from separator \Mo;sture trapped I Fig. 3 Physical moisture trap was attributed to a reduction in fluorescence quenching, resulting from the presence of moisture. The severity of the moisture problem can be seen to a greater extent by performing a long-term stability test. This involved repeated runs of a standard over a period of several hours without the use of a drying facility.Fig. 2 shows the variation in peak height over a 3 h period for a 1 pg 1-1 mercury standard solution stabilized in 2% nitric acid, at a carrier-gas flow rate of 0.8 1 min-1, and an average run time of 2 min. The relative standard deviation (RSD) for 25 runs was 1.2%, after which erratic spikes begin to appear. These were found to be caused by the formation of large moisture droplets condensing in the transfer tube, and eventually entering the detector. In severe instances, large droplets in the transfer tube can cause pulsation in the separator, owing to small back-pressure fluctuations. There is also a general decrease in the net analyte signal with time and this is probably due to an increase in quenching and/or the dissolution of mercury vapour in the aerosol mixture. The solubility of the mercury will be largely dependent on the aerosol composition. Table 1 shows the RSDs for the first, second, third and fourth batch of 25 runs.One arrangement, which can be used to improve the stability with time, involves the use of a physical moisture trap (Fig. 3). This was constructed from a 100 ml conical flask with gas inlet and exit ports. The increase in dead space betweenANALYST, APRIL 1992, VOL. 117 719 40 I I I I 0 20 40 60 80 Run number Variation of peak height with run number using the physical Fig, 4 moisture trap t 2 160 13, aJ c Y .- g 120 a 80 I I I I I I 0 20 40 60 80 100 Run number Variation of peak height with run number using a semi- Fig.5 permeable membrane dryer tube before simplex optimization the separator and detector with the inclusion of this device did not adversely affect the sensitivity of the system. Its role was to catch or collect large moisture droplets. A stability test was performed at a carrier-gas flow rate of 0.8 1 min-1 with use of this arrangement, which indicated a marked improvement (Fig. 4). The system was stable for approximately 2 h, and the RSDs for the first two batches of 25 runs were 2.3 and 2.2%, respectively (Table 1). After this period, erratic spikes were again observed. The use of chemical or physical moisture traps are effective, but only for short periods of time, typically of the order of several hours. Precision can be restored by replacing the desiccant or by cleaning out and drying the flask; this, however, is time consuming and is a major drawback in terms of automation.Such devices may also give rise to contamina- tion and analyte losses during trace analysis. In order to overcome moisture carryover for longer periods of time, the moisture has to be removed continuously rather than trapped, using a device such as the semi-permeable membrane dryer tube described above and shown in Fig. 1. The manufacturers of this device recommend an outer dryer gas flow twice that of the inner carrier flow to remove moisture effectively. Based on this recommendation, a carrier-gas flow rate of 0.5 1 min-1 and a dryer-gas flow rate of 1.0 1 min-l were chosen and a stability test was performed (Fig. 5). This shows that, after approximately 40 runs or 1.5 h, the peak height begins to deteriorate dramatically, hence the higher RSD in the second batch of 25 runs (Table 1).It is also interesting to note that there are no erratic spikes as observed previously. Although the moisture is being removed, the hygroscopic membrane becomes saturated and this causes the membrane to swell. This swelling creates a back-pressure, which results in the displacement of liquid in the U-tube of the separator, and this in turn reduces the amount of agitation occurring and, therefore, reduces sensitivity. The variable step size simplex algorithm's was used to optimize the gas flow rates of the system. The optimum conditions were found to be 0.32 1 min- 1 carrier-gas flow rate and 2.61 I min-1 dryer-gas flow rate.These conditions confirmed our expectations, as the lower carrier flow obviously reduced the amount of aerosol gener- ated, and the higher dryer gas flow removed the moisture more efficiently. 60 E 50 E 2 40 0, r" 30 .- Y m L? 20 10 0 0.2 0.4 0.6 Carrier gas flow rate/l min-1 Fig. 6 Variation of peak height with carrier gas flow rate 100 r I Y a 60 I I I I 0 1 2 3 4 Dryer gas flow rate/l min-1 Variation of peak height with dryer gas flow rate Fig. 7 160 I E 3 2 0 1 g 100 v- a 80 1 I I I I 0 20 40 60 80 Run number Fig. 8 permeable membrane dryer tube after simplex optimization Variation of peak height with run number using a semi- Figs. 6 and 7 show univariate searches for carrier- and dryer-gas flow rates, respectively, with the other gas flows held at the optimum.From Fig. 6 it is clear that the carrier flow rate has a dramatic effect on the net analyte signal, and this can be explained in terms of carrier gas dilution of the analyte and aerosol production inside the separator. The optimum carrier-gas flow rate is considerably less than the conditions normally used, but as the inner diameter of the membrane is much smaller, high gas velocities ensure rapid delivery of the analyte to the detector. Dryer-gas flow rate had little effect on the net analyte signal, presumably because its only purpose is to remove moisture from the outer surface of the membrane. Using these conditions a further stability test was performed and the results are shown in Fig. 8. Over a period of 3 h the signal remained extremely stable, with no drift or erratic spikes (Table 1). An RSD of 2% was obtained for 90 analytical runs over a period of 3 h.The Nafion membrane is chemically resistant to all gases and liquids; however, the drying capacity may be decreased if it becomes contaminated with non- volatile liquids or salts. Organic liquids can be removed with 1,1,l-trichloroethane. Sulfuric acid is removed with de-ion- ized water. Inorganic salts, which may be absorbed into the membrane, can be removed with 10% nitric acid at 50-60"C. The authors have used this device routinely for 6 months without observing any decrease in moisture removal effi- ciency. At the end of each working day it is recommended to leave the dryer gas on for a further 10 min to ensure that all moisture is removed from the membrane.720 ANALYST, APRIL 1992, VOL.117 In order to test whether or not mercury was being lost as a result of diffusion through the membrane, a gold sand trap was installed at the outlet of the dryer tube.16 The argon dryer gas had been previously cleaned of mercury using gold sand and charcoal traps. No mercury was detected in the dryer-gas flow outlet, indicating no detectable diffusion of mercury through the membrane. Conclusions Continuous-flow vapour generation AFS is an extremely sensitive technique with detection limits typically below 20 ng I-' and linear calibration ranges up to five orders of magnitude. Most gas-liquid separator designs involve purging the mercury from solution, which can cause inherent problems from moisture carryover. This problem is most pronounced after extended periods of time, when large droplets begin to condense in the transfer tubes used.The effects of moisture are long-term instability and reduction in sensitivity. Chemical and physical moisture traps can be used to remove moisture, but they eventually become saturated, and may give rise to contamination and analyte losses. The precision can be restored by replacing the trap, but this is time consuming and is a major drawback in terms of automation. The most effective way to remove the moisture is with the use of a semi-permeable membrane dryer tube, which continuously desolvates the wet gaseous stream. It was found that the operating conditions for this device were critical for consistent removal of moisture. The semi-permeable membrane dryer tube is also suitable for use with the hydride-forming elements, and this is currently being investigated in these laboratories. We are grateful to PS Analytical Ltd.and the Science and Engineering Research Council for financial support and the award of a CASE Studentship to one of us (W. T . C.). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 References Lindstedt, G., Analyst, 1970, 95, 264. Poluektov, N. S., and Vitkun, R. A., Zh. Anal. Khim., 1963, 18.33. Hatch, W. R., and Ott, W. L., Anal. Chem., 1968, 40, 2085. Hawley, J . E., and Ingle, J . D., Anal. Chem., 1975. 47. 719. Rooney, R. C., Analyst, 1976, 101, 678. Corcoran, F. L., Am. Lab., 1974, 6, 69. West, C. D., Anal. Chem., 1974, 46, 797. Thompson, K. C., and Godden. R. G., Analyst, 1975,100,544. Stockwell, P. B., Thompson, K . C., Henson, A., Temmerman, E., and Vandecasteele, C., Int. Labmate, 1989, 14. 45. Sturman, B. T., Appl. Spectrosc.. 1985,39, 48. Vijan, P. N., Rayner, A. C.. Sturgis, D., and Wood, G. R., Anal. Chim. Acta. 1976.88, 329. Welz, B.. Melcher, M., Sinemus. H. W., and Maier, D., At. Spectrosc., 1984,5, 37. Stuart, D. C., Anal. Chim. Acra, 1978, 101, 429. Hoult. G., Certificate in Advanced Analytical Chemistry Project Report, Sheffield City Polytechnic, UK, 1990. Ebdon, L., Cave, M. R.. and Mothorpe, D. J . , Anal. Chim. Acta, 1980, 115, 179. Ebdon, L., Corns, W. T., Stockwell, P. B., and Stockwell, P. M., J . Autom. Chem., 1989, 11, 247. Paper 1104577F Received September 4, 1991 Accepted November 8, 1991

ISSN:0003-2654    

DOI:10.1039/AN9921700717

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, APRIL 1992, VOL. 117 72 1 Analysis of Volatiles From Oranges in Good and Bad Condition by Gas Chromatography and Gas Chromatography-Mass Spectrometry David V. McCalley Faculty of Applied Sciences, Bristol Polytechnic, Frenchay, Bristol BS16 IQK UK Juan F. Torres-Grifol Food Machinery Espariola, S.A., Jesus Morante Borras 24, 46012 Valencia, Spain The vapour from whole Navel oranges was examined by using headspace techniques followed by capillary gas chromatography and gas chromatography-mass spectrometry with the aim of distinguishing good, damaged and diseased fruit by chemical means. Low concentrations of limonene were detected in the headspace of undamaged fruit, but mechanical damage caused considerable increases in the concentration of limonene and volatile terpene peel-oil constituents.Diseased oranges affected by common Penicilliurn infections produced different headspace profiles, containing relatively high concentrations of acetaldehyde, simple alcohols and esters. The method was reproducible (relative standard deviation I-2%) and 23 potential compounds of interest could be separated in a single run by using a column containing a thick film of polar stationary phase. Keywords: Oranges; capillary gas Chromatography; headspace analysis The emission of volatile compounds from whole citrus fruit provides a potentially useful indicator of fruit condition and quality. Temperature, humidity, composition of the storage atmosphere, mechanical damage and infection can affect the emission of volatile compounds, thereby providing a possible means of diagnosis of the condition and history of the fruit.Conditions must be carefully controlled during storage and transportation of oranges and other citrus fruit in order to maintain optimum quality. Much work has been performed on the composition and emission of volatiles from orange essential oils. Gas chromatography (GC), based on various types of column and pre-separation techniques, has been the method of choice for analysis; in a review, Shawl has collated quantitative data for about 40 compounds found in orange essential oil. Other studies have involved analysis of orange juice, using either solvent extraction or headspace techniques, followed by GC, typically with methylphenylsilicone station- ary phases.24 These studies have aimed to show changes in the composition of the juice on storage.Results have been variable, although decreases in some desirable flavour con- stituents and increases in undesirable components have been shown. However, very little work has been performed on the analysis of volatiles from whole oranges. Attaway and Oberbacher,s who used packed-column GC and gas chromatography-mass spectrometry (GC-MS) , iden- tified the following compounds in the aroma of intact Hamlin oranges: ethyl acetate, ethyl butyrate, ethyl hexanoate, ethyl octanoate, limonene and ethanol. Typically, 70 fruits were placed in glass jars at 21 "C, and the volatiles were collected over a period of 12 d on activated charcoal. The trapped material was subsequently extracted with diethyl ether, and the extract was analysed by GC, with use of a thermal- conductivity detector.No quantitative data were reported and individual oranges were not studied, presumably because of the lack of sensitivity of the detection techniques employed. Norman et al.,h who used headspace analysis and packed- column GC, compared the emission of volatiles from injured and uninjured Valencia oranges at temperatures ranging from 20 to 38 "C. At 20 "C, small amounts of limonene and octanal were detected, whereas at 38°C the amount of volatiles increased approximately 75 times, with acetaldehyde, ethyl acetate, ethanol and the terpenes, a-pinene and myrcene, also found. Absolute quantitative data were not reported. Later, the same group investigated the production of ethanol, acetaldehyde and methanol by intact oranges during and after storage under nitrogen, by using direct headspace analysis and packed-column GC with a porous polymer column .7 3 During storage under nitrogen, the production of these volatiles increased substantially, with ethanol levels as high as 700 pg h-1, per 100 g of orange, being reported. Mass spectrometry was not used in either study, although in the latter investigation, retention data on several different packed columns, together with qualitative functional group tests, were used as confirmatory techniques for peak identity. The aim of the present work was to ascertain whether GC and GC-MS analysis of volatile compounds emitted from oranges at room temperature could be used to distinguish between fresh, damaged and diseased fruit.The long-term aim of this work is the production of miniaturized sensors capable of continuous monitoring of the volatiles produced by oranges during storage and transport. For this purpose, it was desirable to determine the actual concentrations of the volatiles in the headspace above oranges, rather than obtain- ing the relative composition of the vapour components as has been performed in some other studies. Experimental Gas chromatography was performed with a Model 5890A instrument (Hewlett-Packard, Waldbronn, Germany), equipped with a single flame-ionization detector. The injector and detector were maintained at 275 "C. The column used was of 15 m x 0.53 mm i.d. fused silica, containing an immobilized film of polyethylene glycol (DB-Wax) of thickness 1.0 pm (J&W Scientific, Folsom, CA, USA).Stationary-phase bleed during temperature programming was corrected by using an electronically stored blank chromatogram. The carrier gas was helium, at an average linear velocity of 35 cm s-1. The column temperature was maintained at 35 "C for 5 min, followed by programming to 200 "C at 10 "C min-1. A 500 p1 sample of the headspace was introduced into the column with a gas-tight syringe, by using the split-injection technique and a split ratio of 1 : 3. Gaseous standard samples were prepared by injection of small amounts of liquid into a standard solvent bottle with a cap modified with a GC septum holder. In order to ensure complete vaporization, the amount of liquids injected was always considerably below the amount necessary to produce saturation.9 Determination was based on comparison of peak areas obtained from a chromatographic data station (Trivec- tor, Sandy, Bedfordshire, UK), by use of the external- standard technique.722 ANALYST, APRIL 1992, VOL.117 Gas chromatography-mass spectrometry was performed with a Model 5995C instrument (Hewlett-Packard) under the following MS conditions: direct inlet, transfer line 250 "C, ion source 240 "C, and quadrupole separator 240 "C. The column used was of 25 m x 0.32 mm i.d. fused silica, containing a 1.0pm film of BP-20 (SGE, Milton Keynes, Buckingham- shire, UK), a phase similar to DB-wax. Other GC conditions used were similar to those described above, although a split ratio of 1 : 10 was used with this smaller diameter column.Unwaxed fresh Navel oranges of low to medium ripeness and of approximately 200g mass were used in all the experiments; these oranges had received no post-harvest fungicide treatment. Two different infections that arose naturally on the fruit were studied. In the first, the oranges became covered with olive-green spores, with the fruit gradually becoming shrunken and mummified. The infection was sub-cultured on potato dex- trose agar amended with 0.2% yeast extract and 0.2% peptone (all from Unipath, Basingstoke, Hampshire, UK). Optical- microscopic examination of preparations stained with a solution of Cotton Blue in lactophenol [CH,CH(OH)- COOC6H5] showed short spore-bearing structures with smooth spores of variable size, ellipsoidalkylindrical, typically 4 X 6.5 pm.This fungus was identified as Penicillium digitatum Sacc. The second infection was characterized by blue spores with a white margin. Sub-culturing and microscopic examination of the stained material showed long spore-bearing structures, with spores spherical to ellipsoidal, typically 3.5-4 pm. The second infection was identified as P. italicum Wehmer. Infection of fresh oranges was performed from these cultures by making a puncture incision in the skin of the fruit and introducing a small number of spores. The oranges were placed in a beaker containing wet tissue paper to maintain an adequate level of moisture for the growth of the infection. The beakers were covered with paper to prevent cross-infection yet allow gas exchange. The oranges were placed in a glass reaction vessel of 1.25 1 capacity, which incorporated a GC septum holder to allow removal of headspace.Standard grey septa (Hewlett-Packard) were used. All sampling was performed after 1 h, with the vessel maintained at room temperature (20 "C) unless stated otherwise. After examination, the oranges were removed from the vessel and stored in air at room temperature. Results and Discussion Although trapping of volatiles on sorbents such as porous polymers and activated charcoal, followed by thermal desorp- tion and cryofocusing techniques, yields improved sensitivity over headspace techniques, the latter technique was used in this investigation. Headspace analysis is much simpler, and does not require complex and expensive additions to a standard GC system.Although analysis of the headspace over whole oranges at elevated temperature increases the concentration of volatiles and thereby facilitates the analysis, it was decided to maintain the sampling vessel at 20°C for all the experiments on the assumption that this temperature would more closely resem- ble actual temperatures experienced in storage and transport. Further, oranges subjected to temperatures of 50 "C or above, as used previously for orange juice,3 could become thermally shocked, leading to changes in the volatile compounds emitted. This would prevent the continued study of the same orange over periods of several days, as was the aim here. Nevertheless, a potential problem to be considered for work at room temperature is the possible retention of higher-boiling components by the syringe during sampling, which could lead to carryover of material from one sample to the next.No such effects were observed with volatile analytes such as ethanol and ethyl acetate; however, for less volatile substances such as limonene, this effect could be significant. Pre-heating ot the gas syringe at 60°C before sample abstraction and injection was found to alleviate this problem. Blanks were always run prior to the injection of samples to check for syringe contamination. If shown to be necessary, the syringe was dismantled, thoroughly cleaned by heating or by use of an aqueous detergent solution and re-checked before use. Previous workers have used porous-polymer packed GC columns for the analysis of headspace samples of whole fruit.Their high retention is ideal for very volatile compounds, although temperature programming in order to elute less volatile species can be problematic owing to baseline instabil- ity caused by column bleed.8 The injection of large volumes of gas into capillary columns could be detrimental to perfor- mance owing to the long times necessary for the sample to enter the column. Further, the retention of very volatile substances, such as acetaldehyde and methanol, is too low on thin-film capillary columns to achieve an adequate separation. However, the optimum volumetric flow rate of wide-bore capillary columns is considerably higher than that of narrow- bore columns, allowing faster transfer of sample to the column, and the retention of very volatile polar substances, such as acetaldehyde and methanol, is considerably increased on a thick film of polar phase.t 3 2 0 Q L c 0 0, c 0" 1 1 2 3 10 9 7 1 13 19 12 14 16 17 22 21 I 0 5 10 15 20 tlmin Fig. 1 Chromatogram of gaseous standard sample on a 15 m wide-bore column coated with DB-wax. 1, Acetaldehyde; 2, ethyl formate; 3, ethyl acetate; 4. methanol; 5 , ethanol; 6, ethyl propan- oate; 7, a-pinene; 8, ethyl butyrate: 9, hexanal; 10, p-pinene; I 1. ethyl valerate; 12, myrcene; 13, limonene; 14, ethyl hexanoate; 15, octanal; 16, ethyl heptanoate; 17, ethyl octanoate; 18, decanal; 19, linalool; 20 and 21, citral (cis and truns isomers); 22. geranyl acetate: and 23, geraniol. Column temperature programmed from 35 to 200°C at 10°C min-1 with 5 min initial hold.For other conditions, see under Experimental Table 1 Precision* of injection of gaseous standard samples using split injection Compound Concentratiodmg 1-1 RSD (%) Ethyl formate Ethyl acetate Methanol Ethanol Acetaldehyde Limonene 0.60 0.91 1.78 0.87 0.52 1.26 1.65 1.07 1.08 2.01 0.29 1.28 * The precision was calculated from six injections of a mixture of the standard compounds.ANALYST. APRIL 1992, VOL. 117 723 Fig. 1 shows the separation of 23 volatile compounds mentioned in the literature as being present in orange oil o r in the headspace above whole oranges. In addition, it is possible to determine ethene with this system, although this compound 13 12 * 115 1 1. 0 5 10 15 20 tlmin Fig. 2 Chromatogram of headspace (500 PI) from an orange damaged by making shallow cuts in the skin.For chromatographic conditions and peak identification, see Fig. 1. Peaks marked with an asterisk are thought to be minor terpenes Table 2 Quantitative analysis and mass spectral identification of volatile compounds from a mechanically damaged orange Con- Peak timdmin Significant ions ( r n l z ) Identity pg 1 - 1 Retention centration/ 1 5 7 12 13 15 0.95 3.13 5.02 7.91 9.05 9.72 10.65 11.48 29% 42 43 44 3 1 f 32 43 44 45 46 69777991 93T 121 69777991 931- 121 41t69777991 93 67 68i 77 79 93 107 65 67 77 79 80 931- 10s --f I36 136 121 136 121 136 121 Acetaldehyde 10 Ethanol 40 a-Pinene 150 Terpene 98 M yrcene 220 Limonenc 5300 Terpene 20 Octanal 2 * Peaks gave typical terpene mass spectra. i. Base peak. $ The amount of octanal present was insufficient for mass spectral identification.elutes near to the dead time of the column and hence may suffer co-elution with other very volatile constituents. The column afforded 18400 theoretical plates for ethyl octanoate, measured in an isothermal injection of 500 p1 of the mixed gaseous standard at 90°C; this value is similar to that found for liquid injections when using a column of the given dimensions. Injection of smaller volumes did not produce significant improvement in column efficiency, and injection of 1 ml resulted in a noticeable decrease in efficiency, especially for the early eluting peaks. Obviously, there are no solvent effects acting in headspace analysis to re-concentrate sample bands;'() solvent effects could result in improved efficiency for peaks not 'cold trapped' when liquids are injected.The precision for selected standard compounds, shown in Table 1, is very good. The poor precision of split injection is generally attributed to the non-linear splitting of droplets of liquid within the injector." The high precision shown here is probably due to the absence of such detrimental effects in headspace sampling. Gas standards were freshly prepared immediately before analysis of samples; however, they could be preserved without significant losses from the septum-sealed bottle for up to 12 h. The instrument was calibrated by injection of standard gas mixtures of the same six compounds. The calibration graphs were linear, and correlation coefficients for ethyl formate (range 0-0.60 mg I - l ) , ethyl acetate (0-1.78 mg I - I ) , methanol (0-0.52 mg I - I ) , ethanol (0-1.65 mg I - I ) , acetalde- hyde (0-0.29 mg 1-1) and limonene (0-1.08 mg I-*) were at least 0.999.Detection limits were calculated, with reference to standard procedures,l2 to be approximately 1 pg 1 - 1 for each compound. The fresh, whole oranges used had little discernible smell when intact. In order to increase the concentration of volatiles, two oranges were placed in the collection vessel and the headspace was monitored after 1 h at 20°C. From duplicate experiments, only limonene, at an average concen- tration of 12 pg I-', could be detected. In order to simulate mechanical damage of oranges, a shallow cut (1 cm long) was made in the skin of the oranges. A large increase in the concentration of limonene (monitored after 1 h in the collection vessel) was obtained (937 pg I - l ) , together with small amounts of ethanol (8 pg 1-1) and myrcene (22 pg I-').Peaks were identified by comparison of retention times with those of standard compounds. Subjection of oranges to such minor mechanical damage, as might occur during harvesting of the fruit, did not inevitably give rise to infection. Monitoring of such oranges, with storage in air between measurements, showed that the levels of the above volatiles dropped gradually to those shown by undamaged fruit over a period of several days; this confirms that small injuries may dry out and 'heal'. In order to increase the amount of volatiles released so that minor components could be observed and mass spectral Table 3 Variation of profile of volatiles released from an orange infected by P.digifaturn with time Concentration/pg I-' ~ ~~ Time/d Diameterkm CH3CH0 HCOOEt MeCOOEt MeOH EtOH Limonene 4 2 4 --I - 188 185 14 - 372 709 41 5 4 30 - 6 6 64 14 - 372 87 1 268 7 8 57 23 19 385 86 1 106 8 10 46 305 452 328 853 105 1 1 total 28 1300 4160 I14 600 242 (spor* 0.5) (spar 2) (spot. 6 ) infection (spor 70%) * 5por = diameter of 5porulating portion ot infection. -1 = Not detected.724 ANALYST, APRIL 1992, VOL. 117 confirmation of peak identity obtained, ten shallow cuts (each 10 cm long) were made in a similar fruit. The chromatogram shown in Fig. 2 was obtained. Analysis by GC-MS generated the data shown in Table 2. The composition is in general accord with that of orange essential oils as quoted by Shaw;' naturally, only constituents that are sufficiently volatile will partition significantly into the gas phase.The peaks marked with an asterisk gave typical terpene mass spectra, although because of the similar nature of such spectra, unequivocal identification was not possible. It is possible that these minor terpenes are sabinene and valencene. Clearly, elevated levels t al C x 2 L Y 0 0, Y 0" I i 5 0 5 10 15 tlrnin Fig. 3 Chromatogram of headspace (500 PI) from an orange infected with sporulating Penicillium digitaturn. For conditions and peak identification, see Fig. 1 t v) 0 P 2 L 4-4 0 aJ Y 0" of terpenes, especially limonene, indicate mechanical damage of oranges, and the concentration of limonene in the gas phase is considerable even when the damage is relatively minor.Fresh oranges were then infected with confirmed cultures of P. digitaturn Sacc. and P. italicum Wehmer (see under Experimental); these organisms are the two most common post-harvest pathogens of oranges. The progress of each infection was studied daily, with the oranges placed in the sampling vessel for 1 h prior to analysis, before being returned to normal air storage at room temperature. Each infection was repeated twice, and a typical series of data for P. digitatum is shown in Table 3. In each instance, the initial stages of infection were characterized by a gradual increase in the production of acetaldehyde, methanol and ethanol, which reached a plateau before subsequently decreasing. The onset of sporulation of the fungus coincided exactly with the production of two further compounds, which were identified as ethyl formate and ethyl acetate by comparison of retention times with those of standards.The production of these esters increased considerably as the orange became covered with sporulating organisms. Fig. 3 shows a chromatogram of the headspace from an orange infected with sporulating P. digitatum. Further confirmation of all peak identities was obtained by GC-MS, which for methanol gave characteristic ions at mlz 29,30,31 (base peak) and 32; for ethyl formate at mlz 29,31, 43 (base peak), 44 and 74; and for ethyl acetate at mlz 29,43 (base peak), 45, 61, 70 and 88. Characteristic ions for acetaldehyde, ethanol and limonene were as shown in Table 2.The production of limonene and eventually of minor peel oil constituents, such as myrcene and octanal, which was ob- served in more advanced stages of decay (data not shown), is probably as a result of cell damage caused by the infection. It was possible to detect the presence of characteristic volatiles from this fungus at a much earlier stage than could be obtained by identification with conventional optical-microscopic tech- 5 13 0 1 2 3 4 Time in vessel/h Fig. 4 Effect of sampling time on headspace concentration of A. acetaldehyde; B, methanol; C, ethanol; and D, limonene in an orange infected with Penicillium digitarum (before sporulation) 0 5 10 15 20 tlmin Fig. 5 Chromatogram of headspace (500 p1) of an orange infected with sporulating Penicilfiurn iralicurn.For conditions and peak identification, see Fig. 1 Table 4 Variation of profile of volatiles released from an orange infected by P. iralicurn with time Concentrationlyg 1-1 Time/d Diametedcm MeCHO HCOOEt MeCOOEt MeOH EtOH Limonene - 48 17 17 4 1 -* - - 218 148 28 5 2 6 5 38 - - 368 656 106 7 5 48 - 352 663 78 - - - * Not detected.ANALYST. APRIL 1992, VOL. 117 725 niques. Further, analysis of the volatiles from large numbers of oranges could reveal the presence of infection without unpacking and visual inspection of the fruit. A second orange infected 5 d previously with P. digitaturn was studied, with sample abstraction after 2 and 3 h in addition to the normal 1 h period. The results are displayed graphically in Fig. 4. It is evident that, in this complex living biological system, there is not a simple equilibrium such as that which exists between a liquid sample and its associated headspace.The level of volatiles does not stabilize com- pletely, even after considerable periods, presumably because of the continuing metabolism of the organisms. It was decided to take samples after 1 h, because this gave adequate concentrations of volatiles, while minimizing the possibility of producing anaerobic conditions at the end of the sampling period. Respiration of oranges produces carbon dioxide , which may affect the production of volatiles in the manner shown previously.7.8 The profile of volatiles from oranges infected with P. italicurn in the initial stages of infection was similar, with acetaldehyde, methanol, ethanol and limonene again released. This fungus appeared to be slower growing than the previous infection and it was not possible to achieve total coverage of the orange with fungus.Table 4 shows the variation in the concentration of the volatiles with time. However, even after extensive sporulation of the fungus, ethyl formate and ethyl acetate were not detected in the headspace of the orange. Hence, the profile of volatiles could be used to distinguish between the two infections. Fig. 5 shows a chromatogram of the headspace of an orange infected with P. italicurn after sporulation of the organism. Conclusion The work shows that clear differences exist in the profile of volatile compounds from whole oranges when in good condition, when mechanically damaged, and when infected. Further, it could even be possible to distinguish between types of common infection. The study has revealed candidate compounds and their likely concentrations, which will be used in further studies involving sensors to monitor the quality and condition of fruit. The authors acknowledge financial assistance from the European Community under the ESPRIT programme, project ‘Fruit’, project number 5379. The authors thank P. Spencer-Phillips for assistance with the microbiological iden- tification of the organisms. 1 2 3 4 5 6 7 8 9 10 11 12 References Shaw, P. E., J . Agric. Food Chem., 1979, 27, 246. Marsili, R., Kilmer, G., and Miller, N., LG-GC, 1989, 7, 778. Paik, J . S., and Venables, A. C., J . Chromatogr., 1991, 540, 456. Moshonas, M. G., and Shaw, P. E., J . Agric. Food Chem., 1989,37, 157. Attaway. J . A., and Oberbacher, M. F . , J. Food Sci., 1968,33, 287. Norman, S . , Craft, C. C.. and Davis, P. L., J. Food Sci., 1967, 32, 656. Norman, S. M., and Craft, C. C., J . Am. SOC. Hort. Sci., 1971, 96,464. Norman, S., J. A m . Soc. Nort. Sci., 1970, 95, 777. Sigma-Aldrich Material Safety Data Sheets (CD ROM), Sigma- Aldrich, Milwaukee, 1989. Grob. K., and Grob, G., J . Chromatogr. Sci., 1969, 7 , 584. Grob, K . , and Neukom, H. P., J. Chromatogr., 1982,236,297. Analytical Methods Committee, Analyst, 1987, 112, 199. Paper I f05093A Received October 7, 1991 Accepted December 2, I991

ISSN:0003-2654    

DOI:10.1039/AN9921700721

出版商:RSC    

年代:1992

数据来源: RSC