摘要:

THE ANALYSTTHE ANALYTICAL JOURNAL OF THE CHEMICAL SOCIETYEDITORIAL ADVISORY BOARD"Chairman: H. J. Cluley (Wembley)"L. S. Bark (Salford)R. Belcher (Birmingham)L. J. Bellamy, C.B.E. (Waltham Abbey)L. S. Birks (U.S.A.)E. Bishop (Exeter)L. R. P. Butler (South Africa)E. A. M. F. Dahmen (The Netherlands)A. C. Docherty (Billingham)D. Dyrssen (Sweden)J. Hoste (Belgium)H. M. N. H. Irving (feeds)H. Kaiser (Germany)M. T. Kelley (U.S.A.)W. Kemula (Poland)"W. T. Elwell (Birmingham)"J. A. Hunter (Edinburgh)"G. F. Kirkbright (London)G. W. C. Milner (Harwell)G. H. Morrison (U.S.A.)"J. M. Ottaway (Glasgow)"G. E. Penketh (Billingham)"T. B. Pierce (Harwell)E. Pungor (Hungary)D. I. Rees (London)"R. Sawyer (London)P. H. Scholes (Sheffield)"W. H.C. Shaw (Greenford)S. Siggia (U.S.A.)A. A. Smales, O.B.E. (Harwell)A. Walsh (Australia)T. S. West (Aberdeen)A. L. Wilson (Medmenham)P. Zuman (U.S.A.)"A. Townshend (Birmingham)"Members of the Board serving on The Analyst Publications CommitteeREGIONAL ADVISORY EDITORSDr. J. Aggett. Department of Chemistry, University of Auckland, Private Bag, Auckland, NEWProfessor G. Ghersini, Laboratori CISE, Casella Postale 3986, 201 00 Milano, ITALY.Professor L. Gierst, Universit6 Libre de Bruxelles, Facult6 des Sciences, Avenue F.-D. Roosevelt 50,Professor R. Herrmann, Abteilung fur Med. Physik., 63 Giessen, Schlangenzahl 29, GERMANY.Professor W. E. A. McBryde, Dean of Faculty of Science, University of Waterloo,Waterloo, Ontario,Dr.W. Wayne Meinke, KMS Fusion Inc., 3941 Research Park Drive, P.O. Box 1567, Ann Arbor,Dr. 1. Rubeika, Geological Survey of Czechoslovakia, Kostelni 26, Praha 7, CZECHOSLOVAKIA.Dr. J. Rbiieka, Chemistry Department A, Technical University of Denmark, 2800 Lyngby, DENMARK.Professor K. Saito, Department of Chemistry, Tohoku University, Sendai, JAPAN.Dr. A. Strasheim, National Physical Research Laboratory, P.O. Box 395, Pretoria, SOUTH AFRICA.ZEALAND.Bruxelles, BELGIUM.CANADA.Mich. 48106, U.S.A.Published by The Chemical SocietyEditorial: The Director of Publications, The Chemical Society, Burlington House,London, W1 V OBN. Telephone 01 -734 9864. Telex No. 268001.Advertisements: J. Arthur Cook, 9 Lloyd Square, London, WC1 X 9BA. Telephone 01 -837 631 5.Subscriptions (non-members): The Chemical Society Publications Sales Office, Blackhorse Road,Letchworth, Herts., SG6 1 HN.Volume 101 No 1206@ The Chemical Society 1976September 197

ISSN:0003-2654    

DOI:10.1039/AN97601FX033

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

ANALAO I01 (1 206) 689-760 ( I 976)ISSN 0003-2654September 1976THE ANALYSTTHE ANALYTICAL JOURNAL OF THE CHEMICAL SOCIETYCONTENTS689 EDITORIAL: Short PapersORIGINAL PAPERS690 Duplicate Analysis in Geochemical Practice. Part 1. Theoretical Approach andEst i mat i o n o f An a I y t i ca I Rep rod u c i b i I i ty- M i c h a e I T h om pso n and R i c h a rdJ. Howarth699 Duplicate Analysis in Geochemical Practice. Part 11. Examination o f ProposedMethod and Examples o f Its Use-Richard J. Howarth and Michael Thompson710 Determination o f Organic and Inorganic Carbon in Soils by Potentiometry-L. Th. Begheijn71 7 Colorimetric Determination o f Phenylephrine Hydrochloride in PharmaceuticalPreparations-Yehia M. Dessouky and Laila N. Gad El Rub720 Chromatographic Determination of Promethazine Hydrochloride in AqueousSolution-B. J.Meakin, D. J. G. Davies, Norma Cox and John Stephens728 Determination of Polyoxyethylene in Small Amounts o f Non-ionic Detergents byHydrogen Bromide Fission Followed by Gas Chromatography-I. I. Kadujiand J. B. Stead732 Determination o f Pesticides by Derivative Formation. Part IV. A SensitiveGas-chromatographic Method for the Determination o f MCPA and MCPBHerbicides after Esterification with 1 -Bromomethyl-2,3,4,5,6-pentafluoro-benzene-Haig Agemian and A. S. Y. Chau738 Assay o f Amprolium in Poultry Feedingstuffs by High-performance LiquidChromatography-G. B. Cox and K. Sugden742 Determination o f a Non-volatile Nitrosamine by Using Denitrosation and aChemiluminescence Analyser-M. J. Downes, M. W. Edwards, T. S. Elsey andC. L. Walters749753 Application o f an Oscillating-mirror Rapid-scanning Spectrometer t o Simul-taneous Multi-element Microwave Plasma Emission Spectrometry-OliverRose, Jr., Daryl W. Mincey, Alexander M. Yacynych, William R. Heineman andJoseph A. CarusoDetermination o f Bismuth in Blood and Urine-R. C. RooneyCO M M U N I CAT10 NSulphur Response o f the Alkali Flame-ionisation Detector-R. A. Hoodless, M.Sargent and R. D. Treble757759 Book ReviewsSummaries o f Papers in this lssue-Pages iv, v, viii, i xPrinted by Heffers Printers Ltd, Cambridge, EnglandEntered as Second Class at New York, USA, Post Offic

ISSN:0003-2654    

DOI:10.1039/AN97601BX035

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

iv SUMMARIES OF PAPERS I N THIS ISSUE September, 1976Summaries of Papers in this IssueDuplicate Analyses in Geochemical PracticeReproducibilityDuplicate analytical results can be used to give rapid and realistic estimatesof precision in analytical systems. In particular the effects of varying con-centration of analyte on the variance of the measurement can be taken intoaccount. When only small numbers of duplicate observations are available,the precision can be rapidly tested against an empirical standard level by useof a special control chart. Some common data-recording practices, however,can lead to erroneous estimates of detection limit, irrespective of the esti-mation procedure employed.MICHAEL THOMPSON and RICHARD J. HOWARTHApplied Geochemistry Research Group, Department of Geology, Imperial Collcge01 Science and Technology, London, SW7 2BP.Analyst, 1976, 101, 690-698.Part I.Theoretical Approach and Estimation of AnalyticalDuplicate Analysis in Geochemical PracticePart 11. Examination of Proposed Method and Examples of Its UseMonte Carlo simulation has been appliccl to test the robustness of a methodfor estimating precision as a function of concentration. The effect of dcvi-ations from the basic assumptions underlying the method are shown to begenerally fairly small. The causes of such departures can be identified whenthey occur with actual laboratory results. Methods of recording laboratoryobservations can cause an over-optimistic bias of precision estimates in somecircumstances.RICHARD J.HOWARTH and MICHAEL THOMPSONApplied Geochemistry Research Group, Department of Geology, Imperial Coliegeof Science and Technology, London, SW7 2RP.Analyst, 1976, 101, 699-709.Determination of Organic and Inorganic Carbon inSoils by PotentiometryA method is described for tlic direct, rapid and precise determination of0-6 mg of organic and/or inorgmic carbon in soils. The equivalent amountof carbon dioxide, liberated by improved combustion techniques, is absorbedand precipitated in a solution containing 2.50 minol of sodium hydroxideand 0.96 mmol of barium chloride. Kext, the resultant barium carbonateis dissolved by the addition of 0.99 mmol of EDTA (disodium salt). ThepH of the final solution is related to the amount of carbon dioxide presentby means of a calibration graph.By use of a pH meter with an expandedrange of 1 pH unit, sensitivity to 10 pg of carbon is achieved. Analyses ofpure organic compounds, such ;IS benzoic acid and hydroquinone, showrecoveries accurate to within 0.1 mg of carbon. Results for calcareous soilsare in close agreement with vsliies for calcium oxide obtained by use ofX-ray fluorescence spectrometry.Interference from chloride is effectively eliminated by a preliminaryevaporation step.L. Th. BEGHEIJNDepartment of Soil Szienic ; ~ i i d Geology-, Agricultural University, Wageningen,The Netherlands.,4naZyst, 1976, 101, 710-716Sqbtembev, 1976 SUMMARIES OF PAPERS I N THIS ISSUEColorimetric Determination of Phenylephrine Hydrochloridein Pharmaceutical PreparationsAn accurate, selective and simple method for the determination of phenyl-ephrine hydrochloride is proposed.The method is based on the formationof the nitroso derivative of phenylephrine in the presence of copper ions.The method has been applied successfully to pharmaceutical preparationswithout prior separation of phenylephrine. Other substances commonlypresent in such formulations do not interfere in the determination.YEHIA M. DESSOUKYDepartment of Pharmaceutical Chemistrv, Faculty of Pharmacy, Cairo University,Cairo, Egypt.and LAILA N. GAD EL RUBResearch Department, Soci6t6 Misr pour l’lndustrie Pharmaceutique, 92 El MatariaStreet, Post El-Zeitoun, Cairo, Egypt.Analyst, 1976, 101, 717-719.VChromatographic Determination of Promethazine Hydrochloridein Aqueous SolutionThin-layer and gas - liquid chromatographic techniques that are suitable forthe determination of promethazine hydrochloride in aqueous solutionscontaining breakdown products are described.In both assay processes thedegraded drug solutions are directly applied to the chromatograph withoutpre-extraction. The validity of the assays was confirmed in a kinetic study,which showed that the first-order rate constants obtained for the thermalor photolytic degradation of promethazine hydrochloride were not significantlydifferent .B. J. MEAKIN, D. J. @. DAVIES, NORMA COX and JOHN STEVENSPharmaceutics Group, School of Pharmacy and Pharmacology, University of Bath,Bath, BA2 7AY.Analyst, 1976, 101, 720-727.Determination of Polyoxyethylene in Small Amounts of Non-ionicDetergents by Hydrogen Bromide Fission Followed byGas ChromatographyA procedure is described for determining the polyoxyethylene content ofsamples of ethoxylated materials in the range 50-500pg, such as thoseextracted from biodegradability tests on non-ionic detergents, surface waters,effluents or other media containing these species at very low concentrations.The method, which is an adaptation of the hydrogen bromide fission - gas-chromatographic method for determining the oxyalkylene ratio in copolymers,is applicable to materials that have short polyoxyethylene chains in additionto those in the typical non-ionic detergent range.I. 1. KADUJI and J. B. STEADAnalytical Research Department, Imikro Chemicals Limited, P.O. Box 1, Eccles,Manchester, M30 OBH.ANalvst, 1976, 101, 728-731

ISSN:0003-2654    

DOI:10.1039/AN97601FP069

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

...Vlll SUMMARIES OF PAPERS I N THIS ISSUEDetermination of Pesticides by Derivative FormationDetermination of MCPA and MCPB Herbicides after Esterificationwith 1 - Bromomethyl-2,3,4,5,6- pentafluorobenzeneA sensitive quantitative method for the gas-chromatographic determinationof 4-chloro-2-methylphenoxyacetic acid (MCPA) and 4-(4-chloro-Z-methyl-phenoxy)butyric acid (MCPB) in natural water is presented. The 2,3,4,5,6-pentafluorobenzyl (PFB) esters of MCI'A and MCPB have longer retentiontimes and provide increased sensitivity over the conventional methyl and2-chloroethyl esters of these acids. Concentrations as low as 0.1 p g 1-1of MCPA and 0.2 p g 1-1 of MCPB (for a. 1-1 water sample) can be determinedby this method.HAIG AGEMIAN and A. S. Y . CHAIJInland Waters Directorate, Water Quality Branch, Ontario Region, Canada Centrefor Inland Waters, P.O.Box 5050, 867 Lakeshore Road, Burlington, Ontario,L7R 4A6, Canada.Axalyst, 1976, 101, 732-737.September, 1976Part IV. A Sensitive Gas-chromatographic Method for theAssay of Amprolium in Poultry Feedingstuffs by High-performanceLiquid chromatographyA simple, rapid method, based on high-performance liquid chromatography,is described for the assay of amprolium in poultry feedingstuffs. The drug isextracted from the feed by use of an ethanol -water (3 $- 1 V j V ) mixtureand aliquots of this solution are chroinatographed on a silica gel column,using a methanolic mobile phase containing ammonium nitrate and ammoniasolution. The method is applicable to feedingstuffs containing at least5 mg kg-l of amprolium.The constituents of grass (loo/, nz/m) and fish(10% m/m) meals and the presence of pyrimetliamine, ethopabate, sulpha-quinoxaline, arsanilic acid, nitrovin, iiifursol, dimetridazole, virginiamycinand vitamins A, D, and E do not interfere in the analysis.G. B. COX and K. SUGDENDepartment of Industry, Laboratory of the Government Chemist, Cornwall House,Stamford Street, London, SE1 9NQ.Axalyst, 1976, 101, 738-741.Determination of a Non-volatile Nitrosamine by UsingDenitrosation and a Cheniiluminescence AnalyserA method has been devised for the determination of N-nitrososarcosine,representative of the group of non-volat ile nitrosamines of physiological origin,whereby the nitrosamine is denitrosated with hydrogen bromide to form thevolatile nitrosyl bromide. After dehakogenation in a heated converter, thenitrosamine can be determined as nitric oxide by using a chemiluminescenceanalyser, the response being linear up to a t least 200,ug.The limit of de-tection is below 6 ng, and is therefore much lower than that for the deter-mination of N-nitrososarcosine as inorganic nitrite in solution after denitro-sation. A response is also obtained fi-om inorganic nitrite but this can bedifferentiated from that of N-nitrososarcosine by the use of acetic acid alone,prior to denitrosation with hydrogen bromide in the same solvent.M. J. DOWNES, M. W. EDWARDS, T. S. ELSEY and C. L. WALTERSBritish Food Manufacturing Industries Research Association, Randalls Road,Leatherhead, Surrey, KT22 7 K Y .Analyst, 1976, 101, 142-748September, 1976 SUMMARIES OF PAPERS I N THIS ISSUEDetermination of Bismuth in Blood and UrineBismuth is determined in blood and urine by means of atomic-absorptionspectrophotometry following generation of bismuth hydride from a wet-oxidised sample.Recoveries are shown to be satisfactory, and the detectionlimit using analytical-reagent grade reagents is less than 0.01 pg ml-l.The direct determination of bismuth in urine by means of hydride generationhas been investigated, and it is shown that there is severe suppression andpoor recovery.R. C. ROONEYRooney 13 Ward Ltd., Rlackwater Station Estate, Camberley, Surrey.APzalyst, 1976, 101, 749-752.Application of an Oscillating -mirror Rapid-scanning Spectrometerto Simultaneous Multi- element Microwave Plasma EmissionSpectrometryThe potential value of an oscillating-mirror rapid-scanning spectrometer asa detector for simultaneous multi-element emission spectrometry is demon-strated.A microwave argon plasma was used as the emission source witha carbon cup sampling device for analyte introduction. Analytical graphsfor bismuth, cadmium, manganese and magnesium prepared simultaneouslyare shown. The sensitivity of this system for cadmium, manganese andmagnesium is a t the nanogram level. The spectrometer has a wide opticalrange of 200-930 nm that can be scanned in a single sweep and a large dynamicrange as radiation is detected by a photomultiplier tube.OLIVER ROSE, Jr., DARYL W.MINCEY, ALEXANDER M. YACYNYCH,WILLIAM R. HEINEMAN and JOSEPH A. CARUSODepartment of Chemistry, University of Cincinnati, Cincinnati, Ohio 4522 1, USA.Analyst, 1976, 101, 753-756.Sulphur Response of the Alkali Flame -ionisation DetectorCounununicationR. A. HOODLESS, M. SARGENT and R. D. TREBLEDepartment of Industry, Laboratory of the Government Chemist, Cornwall House,Stamford Street, London, SE1 9NQ.Analyst, 1976, 101, 757-758.iX THE ANALYST Se$tember, 1976Reprints of Review PapersReprints of the following Review Papers published in The Analyst since 1967 are available fromthe Publications Sales Officer, The Chemical Society, Blackhorse Road, Letchworth, Herts., SG61HN (not through Trade Agents).The price per reprint is jtIl; orders for six or rnore reprints of the same or different Reviewsare subject t o a discount of 26%.The appropriate remittance, made out to The ChemicalSociety, should accompany any order.“Activation Analysis,” by R. F. Coleman and T. B. Pierce (January, 1967).“Techniques in Gas Chromatography. Part I. Choice of Solid Supports,” by F. J . Palfranian“Heterocyclic Azo Dyestuffs in Analytical Chemistry,” by R. G. Anderson and G. Nickless“Determination of Residues of Organophosphorus Pesticides in Food,” by D. C. Abbott and“Radioactive Tracer Methods in Inorganic Trace Analysis : Recent Advances, ’’ by J. VV.“Gamma-activation Analysis,” by C. A. Baker (October, 1967).“Precipitation from Homogeneous Solution,” by P .F. S. Cartwright, E. J. Newman and“Industrial Gas Analysis,” by (the late) H. N. Wilson and G. Ill. S. Duff (December, 1967).“The Application of Atomic-absorption Spectrophotometry to the Analysis of Iron and“Inorganic Ion Exchange in Organic and Aqueous - Organic Solvents,” by G. J. Moody and“Radiometric Methods for the Determination of Fluorine,” by J. I<. Foreman (June, 1969).“Techniques in Gas Chromatography. Part 11. Developments in the van Deemter RateTheory of Column Performance,” by E. A. Walker and J. F. Palframan (August, 1969).“Techniques in Gas Chromatography. Part 111. Choice of Detectors,” by T. A. Gough andE. A. Walker (January, 1970).“Laser Raman Spectroscopy,” by P. J. Hendra and C. J. Vear (April, 1970).“Ion-selective Membrane Electrodes, ” by El-no Pungor and KlAra T6th (July, 1970).“X-ray Fluorescence Analysis,” by K.G. Carr-Brion and K. W. Payne (December, 1970).“Mass Spectrometry for the Analysis of Organic Compounds,” by A. E. Williams and H. E“The Application of Non-flame Atom Cells in Atomic-absorption and Atomic-fluorescence“Liquid Scintillation Counting as an Analytical Tool,” by J. A. B. Gibson and A. E. Lally“The Determination of Some 1,4 Benzodiazepines and Their Metabolites in Body Fluids, ’’“Atomic-fluorescence Spectrometry as an Analytical Technique,” by R. F. Browner(October, 1974).“The Use of Precipitate Based Silicone Rubber Ion-selective Electrodes and Silicone RubberBased Graphite Voltammetric Electrodes in Continuous Analysis,” by 2s.FCher,G . Nagy, K. Tbth and E. Pungor (November, 1974).“The Examination of Meat Products with Special Reference to the Assessment of the MeatContent,” by D. Pearson (February, 1975).“Chemiluminescence in Gas Analysis and Flame-emission Spectrophotometry,” by J . H.Glover (July, 1975).“The Analytical Role of Ion-selective and Gas-sensing Electrodes in Enzymology, ” byG. J. Moody and J. D. R. Thomas (September, 1975).“Thiazolylazo Dyes and Their Applications in Analytical Chemistry,” by H%vard R. Hovind(November, 1975).“Sample Preparation in the Micro-determinat ion of Organic Compounds in Plasma or Urine,”by Eric Reid (January, 1976).“Recent Advances in the Ring Oven Technique,” by Herbert Weisz (March, 1976).“The Radioimmunoassay of Drugs,” by J. Landon and A. C. Moffat (April, 1976).“Analysis and Assay of Polyene Antifungal Antibiotics,” by A. H. Thomas (May, 1976).“Properties and Uses of the Colorimetric Reagents 2-Nitroso-5-dimethy1aminophenol and2-Nitroso-5-diethylaminophenol for Cobalt,” by Kyoji TBei and Shoj i Motomizu (July,1976).and E. A. Walker (February, 1967).(April, 1967).H. Egan (August, 1967).McMillan (September, 1967).D. W. Wilson (November, 1967).Steel,” by P. H. Scholes (April, 1968).J. D. R. Thomas (September, 1968).Stagg (January, 1971).Spectroscopy,” by G. F. Kirkbright (Seiptember, 197 1).(October, 1971).by J. M. Clifford and W. Franklin Smyth (May, 1974).“Fibre Identification and Analysis of Fibre Blends,” by H. M. Appleyard (August, 1976)

ISSN:0003-2654    

DOI:10.1039/AN97601BP073

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

SEPTEMBER 1976 The Analyst Vol. 101 No. 1206 Editorial Short Papers A category of contributions designated “Short Papers” will be introduced into The Analyst in 1977. It was formerly the practice of the journal to publish, in addition to “Original Papers,’’ other contributions of limited length and scope which were designated “Notes,” and later “Short Papers.” This separate category of papers was discontinued after 1966, in the belief that such contributions might be interpreted as representing work of a lower scientific standard than that described in the full papers which comprised the major part of the journal. (Such an interpretation could have been fostered by the practice of using a smaller size of print for the Notes and Short Papers.) The thought at that stage was that if work was of sufficient merit to justify publication, then it should appear as a paper in its own right, regardless of brevity and narrowness of interest, and be of equal standing with longer papers.The question of categories of papers in the journal has recently been reconsidered by The Analyst Publications Committee. While recognising the laudable objectives which prompted the earlier decision to abandon Short Papers, the Committee believes that the absence of such a category may inhibit potential authors from submitting to the journal papers of limited length and scope, and that provision for the publication of such contributions could be of advan- tage to authors and readers alike. The decision has therefore been taken to re-institute the category of Short Papers from the beginning of 1977.I t is emphasised that such papers will differ from full papers only in size and in breadth of subject matter, not in quality. The Short Papers will be subjected to the same rigorous scrutiny by the duplicate refereeing system as is used for full papers; the Short Papers will also be printed in the same type. However, the brevity and narrower scope of the Short Papers should facilitate refereeing and editorial processing, and in this way it is aimed to offer more rapid publication for the Short Papers. Authors are therefore invited now to provide contributions in the form of Short Papers, on topics of suitably limited scope, for publication at the beginning of 1977. We are not, of course, seeking to encourage authors merely to abbreviate their contributions; where the subject matter is of conventional magnitude, authors are asked to prepare their manuscripts for publication in the normal, full paper form. Finally, it should be made clear that the re-introduction of Short Papers will not mean the discontinuance of “Communications.” These are also of limited length but are aimed at providing the maximum speed of publication of urgent matters of scientific and analytical importance, and are not formally refereed. Hence the Communications fulfil a need some- what different from that of Short Papers, and their publication will be continued. H. J. CLULEY Chairman, The Analyst Publications Committee 689

ISSN:0003-2654    

DOI:10.1039/AN9760100689

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

690 Awalyst, September, 1976, Vol. 101, p p . 690-698 Duplicate Analysis in Geochemical Practice Part 1. Theoretical Approach and Estimation of Analytical Reproducibility Michael Thompson and Richard J. Howarth Applied Geochemistry Reseavch G Y O U ~ , Department of Geology, Imperial College of Science and Technology, London, SW7 2BP Duplicate analytical results can be used to give rapid and realistic estimates of precision in analytical systems. In particular the effects of varying con- centration of analyte on the variance of the measurement can be taken into account. When only small numbers of duplicate observations are available, the precision can be rapidly tested against an empirical standard level by use of a special control chart. Some common data-recording practices, however, can lead to erroneous estimates of detection limit, irrespective of the esti- mation procedure employed.The estimation of analytical reproducibility is an integral part of geochemical analysis and has been the subject of much discussion in the technical literature.1,2 This discussion arose because one requires a separation of the variation in a data set into that due to laboratory and sampling causes, and the residuum of real interest, the geochemical variation. The formal requirements for a valid analysis of variance have been discussed by Mie~ch.~ Even when these conditions are not met, or a complete analysis of variance is not required, both the analyst and the geochemist usually need to know, at the time of the analysis, whether the laboratory variance is suitably low for subsequent interpretation.Despite this requirement, methods currently in use for the estimation of analytical repro- ducibility are usually unsatisfactory in one or more respects, as follows: (i) The materials used in the estimation procedure are not usually representative of the samples being analysed. Ideally, the samples themselves should be used for the estimation, as in Garrett’s method.2 (ii) The methods used take no account of possible changes in absolute and relative repro- ducibility over the concentration range of the analyte, but derive some sort of an average value for the range. (iii) Some data recording practices, described below, which are common among analysts, can lead to grossly incorrect estimates of variance. (iv) There is a lack of clarity and consistency in the definition and application of terms such as “precision” and “detection limit” used to characterise an analytical procedure, especially in connection with (ii).We have described methods by which duplicate analyses can be used to estimate variation, and indicated their advantage^.^ Following experience in applying these and other methods in geochemical practice, we are attempting in this paper to achieve two objects: firstly, to establish a unified and comprehensive approach to the discussion of reproducibility in analysis that would be of general applicability in the geochemical field; and secondly, to demonstrate simple and effective methods, utilising duplicate analyses, which can be used for the estimation of the analytical reproducibility, or for testing it against an empirical standard level.These two objects, when fulfilled, jointly overcome the unsatisfactory aspects of the methods currently in use. In Part I1 we present studies of the performance of the advocated method, based on data derived from both geochemical practice and from computer-simulations de- signed to explore the outcome of deviations from the basic assumptions underlying this approach. Gaussian Distribution in Analytical Measurement The assumption that analytical variation closely follows the Gaussian (normal) distribution underlies most of our discussion, and it is necessary to examine this assumption carefully and to be aware of situations where it may not apply. The mathematical conception of theTHOMPSON AND HOWARTH 691 Gaussian curve is that it results from the combination of a large number of small independent errors.In analytical practice, these small fundamental errors can be regarded as resulting from variations in manipulative operations, such as sub-sampling, weighing, dissolution, dilution and presentation to an instrument of the working solution, and from electronic noise from the instrument. Despite the fact that some of these contributions may not be srriall, they are usually independent and the resulting distribution is not usually distinguishable from a random sample drawn from a Gaussian population. Non-Gaussian forms are apparent in practice when two particular conditions are met : one or more of the contributing factors is no longer relatively small ( i e ., it makes a substantial contribution to the total variation) ; and its particular frequency distribution is non-Gaussian. There are a number of well recognised situations where these conditions can arise, some of which are discussed by Ecksch1ager.j (i) The sample is heterogeneous, the analyte being largely or completely concentrated in a small proportion of the particles constituting the sample, e.g., tin as cassiterite in sediments. This situation leads to a large sub-sampling variance and a skewed concentration distribution. This topic has been discussed extensively by Ingamells and Switzer.6 (ii) The precision of the method is poor, and the calibration is intrinsically non-linear, e.g., in the region of the detection limit of spectrographic methods, where the calibration is logarithmic.This situation has been discussed by Ahrens’ among others. (iii) The concentration levels are within an order of magnitude of the digital resolution of the instrument. For example, lead concentrations determined by atomic-absorption spectroscopy are commonly recorded as integer multiples of 0.1 pg ml-1, with no intermediate values. The final values, referring to the original samples (after multiplying the instrumental value by a factor) take only discrete values, such as 0, 5, 10, 15, . , . p.p.m. This custom produces a discontinuous frequency distribution of error. (iv) The concentration levels are near the detection limit, and sub-zero readings are set to zero. Alternatively, readings below the detection limit are set to the detection limit or recorded as “less than.” This practice produces a truncated distribution and is commonly found in association with (iii).In this connection, it is worth emphasising that, while the idea of negative (or even zero) concentration has no physical significance, a negative measure- ment of concentration is feasible and, when considered statistically (i.e., as an estimate with confidence limits), meaningful. These values can be distinguished conceptually from ordinary random variations as arising from mistakes or gross errors in procedure. In short, they really belong to a different population of results. Methods for detecting and eliminating “fliers” have been discussed by Harvey8 and others. The uncritical use of statistics calculated on observations drawn from such non-Gaussian forms is likely to lead to erroneous conclusions, especially where the variability of measure- ment is high relative to the actual level being measured. In particular, it is liable to affect estimates of the detection limit, which is, in trace analysis, a crucial characteristic of an analytical system.In practice, apart from the circumstances outlined above, error distributions noticeably divergent from the Gaussian are rarely encountered. Even at very poor levels of repro- ducibility, an impracticable number of observations are required to reject the usual Gaussian hypothesis for a sample drawn from a log-normal population (e.g., over 1000 observations for coefficients of variation less than 20%). However, as markedly non-Gaussian distributions of analytical variation may be present unexpectedly in the data, it is good practice, when estimating precision, to test for compatibility with the Gaussian assumption.(v) The data set contains “wild” results or “fliers.” Advantages of Duplicate Determinations in the Estimation of Precision In normal practice, in order to estimate the within-batch precision, one or more samples are analysed repeatedly (usually contiguously) and the standard deviation calculated in the normal way. Alternatively, a statistical series interspersed with the actual sampless can be analysed. (i) The materials selected for the repeated analysis or the statistical series may not ade- quately represent the behaviour of the samples. Even if they are carefully selected to match the chemical composition of the samples, they will almost certainly receive extra preparation These practices have limitations, outlined below.692 THOMPSON AND HOWARTH: DUPLICATE ANALYSIS Analyst, Vol.I01 in the way of grinding, which is necessary to ensure homogeneity over long periods. Moreover, the selection and preparation of these materials may involve a great expenditure of time and care. (ii) If the control samples are analysed contiguously and, especially if they are recognisable as such to the analyst, precision estimates are likely to be unrealistically low. (iii) Estimates of standard deviation have relatively large standard errors [a/4(2n) for the Gaussian curve] unless large numbers of observations are made, which then becomes an inefficient use of analytical resources.(iv) No conclusive information regarding the variation of standard deviation at different concentration levels can be obtained. The methods described below, which are based on duplicate analysis of all, or a random selection, of the actual samples remove or alleviate these drawbacks. The results truly represent the samples, no labour is expended in the preparation of standard materials, no data are wasted, as the duplicates can be averaged to improve the precision of the results, and conclusive information can be obtained regarding the precision at various levels of the analyte. However, it must be emphasised that. in order to obtain an accurate estimate of within-batch precision, each of the two duplicates must be taken through the whole analytical procedure as if it were a separate sample, and the position of the second duplicate in the analysis sequence must not be systematically related to the first.Repeat analyses in different batches are likely to have a greater systematic (between-batch) variation than a random component, unless the entire sequence of samples, plus duplicates, is analysed in a random order. A strong systematic difference between duplicate pairs will invalidate the procedures to be described, but can easily be detected, as shown in Part 11. The Concept of an Analytical System This concept is fundamental to any consistent discussion of analytical precision, as is shown by Kaiser,lo who used the term “complete analytical procedure.” We have extended his concept slightly, and define an analytical system as comprising the following items, each of which makes an independent contribution to the total variance: (i) the set of samples, with the analyte in a specific matrix; (ii) the exactly defined analytical procedure; and (iii) the particular instruments used.Only with each of these factors defined can one measure precision. In short, one cannot meaningfully ask “what is the precision of a method?” because it will be determined not only by the procedure, but by the nature of the samples, and may also depend, inter alia, on the manufacture or the type of instrument used. A procedure applied to limestone samples may well have a different detection limit when applied to sandstones. A corollary of this concept is that the samples for a system should be drawn from a homogeneous type, otherwise one may be attempting to measure a meaningless average precision between two distinct systems.From a practical point of view, it is the precision of a system that is actually of interest to the user of the data, as it corresponds to the over-all laboratory variance that he is likely to encounter. Variation of Precision with Concentration The standard deviation of a determination within a system tends to vary with concentration of the analyte, but this situation does not appear to have been dealt with adequately in the literature. We have found the following treatment of the subject to be helpful. The variation can be expressed generally by the expression u, = f ( c ) where uc is the standard deviation at concentration c.Without any loss of generality, this expression can be replaced by a polynomial expression : u, = k, + klc + k2c2 + . . . + k,cn where each k is a constant. The first term (KO) can be replaced by u,, the standard deviation at zero concentration. It appears in practice that in many instances, the non-linear terms containing powers greaterSeptember, 1976 I N GEOCHEMICAL PRACTICE. PART I 693 than unity can be ignored, but this should not be accepted in particular situations without being scrutinised (see Part 11). The expression then becomes .. . - ( 1 ) I+, =ao + klc . . .. .. Analytical variation is often specified in terms of precision, which in geochemical practice is usually taken to be twice the coefficient of variation.Following this convention .. .. ' * (2) .. .. (3) p , = 2 4 . . .. where p , is the precision at concentration c. Combination with equation ( 1 ) gives p , = 2a0/c + k . . .. .. where k = 2k1. In this equation, the precision approaches the value k asymptotically at high concentrations, and rises increasingly fast as c tends to zero, as illustrated in Fig. 1. Definitions of detection limit vary,1°-13 but all concur that it should be related to the standard deviation of the system measured at or near zero concentration of the analyte. Usually it is regarded as the concentration equal to a constant multiplied by this standard deviation, and values of 1.0, 2.0, 2 z / 2 and 3.0 have been suggested for this constant. A definition which follows from our treatment is that the detection limit is the concentration at which the precision is unity.This point is illustrated in Fig. 1, where cd is the detection limit. It is evaluated by the substitution in equations ( 2 ) and (3) of the values c = cd and p , = 1.0, giving As k is usually much less than unity, this definition corresponds to the selection of a factor approximating to 2.0, but varying with the value of k . C d = 2 0 , = 2 a o / ( l - K) . . .. .. ' (4) I cd Concentration Fig. 1. Variation of precision as a function of concentration according to equation (1). It is commonly held that the precision (9,) of determination of an analysis is relatively stable if the measurements are well above the detection limit, whereas the standard deviation steadily rises. The idea that relative error is comparatively constant is one of the theoretical reasons for the log-transformation of data before the analysis of ~ariance.~ It is possible to draw some general conclusions on the basis of the linear model of standard deviation.If reduced concentration (c,) is defined as the concentration relative to the detection limit, c/cd, equation (4) becomes and combination of this expression with equation (3) gives c = C,Cd = 2a,c,/(l - K) .. * . (5) p c = (1 - k)/cr + K .. ..694 THOMPSON AND HOWARTH : DUPLICATE ANALYSIS Analyst, VOl. 101 Fig. 2 shows equation (3) plotted with values of the parameter k ranging from 0.3 (30% precision) to 0.001 (0.1% precision). It can be men that the precision approaches to within 10% of its asymptotic value only above a reduced concentration given by c,. w 10/k.Within the range of k levels found in geochemical practice (approximately 0.3 to 0.001) p , will approach k only when c,. exceeds 30 or 1000, respectively. It is, therefore, not safe to assume that the precision will attain a steady value at concentrations less than two orders of magnitude greater than the detection limit. 1 .o 0.1 2 Y c 0 V Q .- (0 .- e 0.01 0.001 1 I I 10 100 1000 Reduced concentration (c,) Fig. 2. Variation of precision with “reduced concentration” a t various levels of the parameter k . Estimating o0 and k, from Duplicate Analyses If n pairs of values are drawn at random from a Gaussian distribution with mean p and variance, 02, i.e., N(p,02), the absolute difference, x’, between the pairs of values is distributed as the positive half of the Gaussian distribution N(pd,a2,), where cr2d = 2a2 and pa = 0.The median of the half-normal distribution corresponds with the quartile of the Gaussian, so that sample values of the median of x will tend to the mlean value 0.674 5ad with a standard error14 of 1.362 6ad/4n. In terms of the parent standard. deviation, writing crd = 2/20, the median tends to 0.953 90 with a standard error of 1.927 Oa/dn. This can be related to analytical practice as follows. If duplicate analyses are obtained for a number of samples within a narrow concentration range, the median of the absolute differences between corresponding values multiplied by 1.048 3(1/0.953 9) is an estimate of the standard deviation within that range.The arithmetic mean of all the results is an estimate of concentration within the range. If this procediure is repeated for a number of successive narrow concentration ranges, a set of corresponding concentration and standard deviation estimates is obtained. The relationship between them can be found by regression. If simple linear regression is used, the intercept (c = 0) is a, and the coefficient is k,, as illustrated in Fig. 3. The usual limitations of regression techniques15 apply here and careful note must be taken of the standard errors of the regression estimates. It is common to find that either oo or k, does not differ significantly from zero, and may be negaiive. An assumption implicit in this treatment is that each pair of values in a narrow concen- tration range is drawn from a population with the same standard deviation as the others in that range.More rigorously, if n samples are analysed in duplicate two values are obtained from each of n populations: N(pi,oi2), i = 1,n. The assumption is that, as the pi fall in a narrow range, the ui are identical. While this is unlikely to be exact, even in an analytical system (i.e., in a set of similar samples with the analyte in a constant matrix), the method has been found to be tolerant of considerable variations (see Part IT). However, it would produce meaningless results if, for example, the sample set consisted of a mixture of roughly equal numbers of two distinct sample types with different ui values. The detection limit can also be calculated from these statistics.September, 1976 IN GEOCHEMICAL PRACTICE.PART I 695 Fig. 3. Regression of median absolute differences against concentration, with the concentration The diagonal lines show the regression line with range divided into equal-frequency intervals. confidence limits of one standard error. Simple linear regression requires that the variance of the dependent variable is constant over different ranges of the independent variable. That condition is clearly not met in this procedure and, strictly speaking, weighted regression should be used. However, no bias is produced by the use of simple regression, although the solution obtained is not the minimum variance solution. Additionally, simple regression is based on the here unfulfilled condition that the independent variable is error-free.Again this deviation is unlikely to cause any practical problems because of the grouping procedure. Bartlett’s method16 may be a suitable alternative to regression in this situation. The optimum number of samples to use for each narrow concentration range is difficult to determine, because the final regression parameters are little affected by variations over a wide range. We have tried, for n samples, groups of between three and 4 3 , with no real difference in the final result. We normally use groups of 11 differences, as a convenient compromise between these extremes. The minimum total number of results required is also difficult to specify, and will depend, inter alia, on the concentration range spanned by the sample set. Generally, a minimum of 50 pairs is necessary, although smaller data sets can be treated by an alternative method described below.The use of the median as the estimator has some advantages, despite the fact that it is not the most efficient statistic. Firstly, it is almost equal to the standard deviation for the range, and in many instances no serious problems would result from assuming equality. Secondly, it is unaffected by a small proportion of “wild” results in the data set. This procedure causes no difficulty in practice. Brief Procedure for the Estimation of Precision (i) From the list of duplicate analyses (Xt, Y,, i = 1,n) obtain lists of the means of the pairs and the corresponding absolute difference, i.e., ( X i + Yi)/2 and IXi - Yi 1. (ii) Arrange the lists in increasing order of concentration means.(iii) From the first 11 results obtain the mean concentration and the median difference. (iv) Repeat this for each successive group of 11 results, ignoring any remainder less than 11. (v) Complete the linear regression of the medians on the means, and multiply the intercept, the coefficient and their standard errors by 1.048.696 THOMPSON AND HOWARTH : DUPLICATE ANALYSIS Art@1!J6t, VOl. 101 (vi) This procedure can be carried out either numerically or graphically on a scatter plot In the approximate graphical procedure, the factor of means zleysuus absolute differences. 1.048 can be ignored. Testing Precision by Small Numbers of Duplicates For a specified relationship between standard deviation and concentration, it is possible to predict the frequency distribution of absolute differences at any given concentration, assuming that the Gaussian error curve is followed.Observation of the actual distribution of a small number of experimentally determined vidues makes it possible to determine whether the experimental precision exceeds the specification. As shown above, the absolute differences are (distributed as the positive part of the dis- tribution N(0,202). The percentile points for this distribution can be calculated simply from standard tables for the Gaussian distribution,15' and some values are given in Table I. Thus the median (50th percentile) difference between duplicates at concentration c is 0.674 5ad = 0.674 5 x 2/20, = 0.953 9 o,. Likewise, the 90th percentile is given by 1.644 9 x 2/20, = 2.326 30,.As (T, is related to c, it is possible to obtain these percentiles through a specified form of equation (1). TABLE I PERCENTILE POINTS OF THE HALF-NORMAL DISTRIBUTION Percentile Normal variate Percentile Normal variate 0 10 20 30 40 50 60 0.000 0 0.125 7 0.253 3 0.385 3 0.524 4 0.674 5 0.841 6 70 1.036 4 80 1.281 6 90 1.644 9 95 1.960 0 99 2.575 8 99.9 3.290 6 Fig. 4 illustrates the derived percentiles for the system u, = 1.0 + 0.05~. If the experi- mental means and differences are plotted on this chart, it can be seen whether the distribution conforms to the specification or not by the number (m) of points falling above the given percentiles. It is possible to quantify the conclusions by calculation of the probability that m or more points fall above these limits. 30 I I 0 20 40 60 80 100 Mean' of duplicate results Fig.4. Percentiles of the absolute differences between duplicates for the system uC = 1.0 + 0.05c, with some experimental points. The meaning of the 12 experimental points in Fig. 4 can be assessed as follows. (i) A count is made of the number of points lying above the percentiles for the system (Table 11).September, 1976 IN GEOCHEMICAL PRACTICE. PART I 697 TABLE I1 PROBABILITIES ASSOCIATED WITH DATA OF FIG. 4 Points above Single event Combined event Percentile percentile (m) probability ( p ) probability (r) * 50 5 0.5 0.806 90 2 0.1 0.341 99 1 0.01 0.114 99.9 1 0.001 0.012 * Y = chance of m or more points falling above a given percentile, when (ii) The chances of observing five or more points above the median (0.806) or two or more above the 90th percentile (0.341) are high and there are no strong grounds for suspecting that the precision is worse than the specification for these data.However, as a solitary point falls above the 99.9th percentile and therefore constitutes a rare event, it is possible to regard this as a wild result, i.e., not really part of the system. the chance of a single point doing so is p . (iii) We can conclude that the analyses are within the control limits set. Mean of duplicate results Fig. 5. Control chart used for rapid check on analytical precision. Percentile lines for differences between duplicate results at 10% precision (coeffi- cient of variation = 0.05). For routine geochemical analysis, we have adopted a precision chart that is based on this system, as shown in Fig.5. It consists of the 90th and 99th percentiles of the function 0, = 0 . 0 5 ~ (i.e., 10% precision), plotted on logarithmic axes. This gives a valuable and immediate check on batches of analysis. The analyst can see at a glance whether his duplicates indicate a precision seriously worse than 10%. We therefore conclude that this method is a simple procedure for routine laboratory control, making optimum use of the results of duplicate analyses. 1. 2. 3. 4. 6. 6. 7. 8. References Craven, C. A. U., Trans. Instn Min. Metall., 1954, 63, 551. Garrett, R. G., Econ. Geol., 1969, 64, 568. Miesch, A. T., Prof. Pap. U.S. Geol. Sur., 1967, 574A. Thompson, M., and Howarth, R. J., Analyst, 1973, 98, 153. Eckschlager, K., translated by Chalmers, R. A., “Errors, Measurements and Results in Chemical Ingamells, C. O., and Switzer, P., Talanta, 1973, 20, 547. Ahrens, L. H., “Quantitative Spectrochemical Analysis of Silicates,” Pergamon Press, London, Harvey, P. K., Geochim. Cosmochim. Acta, 1974, 38, 435. Analysis,” Van Nostrand Rheinhold Co. Ltd., London, 1969, p. 103. 1955, pp. 39-46.698 9. 10. THOMPSON AND HOWARTH Stanton, R. E., “Rapid Methods of Trace Analysis,” Edward Arnold, London, 1966, p. 10. Kaiser, H., translated by Menzies, A. C., “The Limit of Detection of a Complete Analytical Procedure,” ROOS, J. B., Analyst, 1962, 87, 832. Wilson, A. L., Analyst, 1961, 86, 72. Svoboda, V., and Gorbatsch, R., 2. Analyt. Chem., 1968, 242, 1. Kendall, M. G., and Stuart, A., “The Advanced Theory of Statistics,’’ Volume 1, Third Edition, Griffin, London, 1969, p. 243. Draper, N. R., and Smith, H., “Applied Regression Analysis,” John Wiley and Sons, New York, 1966, p. 80. Bartlett, M. S., Biometrics, 1949, 5, 207. Lindley, D. V., and Miller, J . C. P., “Cambridge Elementary Statistical Tables,” Cambridge University Press, Cambridge, 1962, Table 2, p. 5. Received January 28th, 1976 Accepted A p d 21st, 1976 Adam Hilger, London, 1968. 11. 12. 13. 14. 15. 16. 17.

ISSN:0003-2654    

DOI:10.1039/AN9760100690

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

Analyst, September, 1976, Yol. 101, pp. 699-709 699 Duplicate Analysis in Geochemical Practice Part 11." Examination of Proposed Method and Examples of Its Use Richard J. Howarth and Michael Thompson Applied Geochemistry Research Group, Department of Geology, Imperial College of Science and Technology, London, SW7 2BP Monte Carlo simulation has been applied to test the robustness of a method for estimating precision as a function of concentration. The effect of devi- ations from the basic assumptions underlying the method are shown to be generally fairly small. The causes of such departures can be identified when they occur with actual laboratory results. Methods of recording laboratory observations can cause an over-optimistic bias of precision estimates in some circumstances. The use of a method based on the analysis of duplicate samples to estimate precision for an analytical system as a function of concentration was discussed in Part 1.l The basic assumptions were that the analytical error behaves as a normally distributed (Gaussian) variable and that the standard deviation of the observed concentration values is a linear function of the form a, = a.+ kc where 0, = standard deviation at concentration c, a,, = standard deviation at zero concen- tration and k = a constant. If the concentration relative to the detection limit (cd) is c,. = c/c,, we showed that the precision at concentration c,. can be expressed as P c = (1 - w,. + k We use Monte Carlo simulation methods in this Part to investigate the robustness of the proposed method, both in the Gaussian case and in non-ideal situations where the errors do not conforrn to the Gaussian model, and we have extended this work to investigate the effect of methods of recording the original observations on the determination of the parameters for the analytical system.Finally, we show some examples of the use of this procedure in actual laboratory quality control. Simulation Results Method The investigation of all aspects of the behaviour of the proposed method using duplicate analyses would be extremely lengthy to carry out in practice. It would require the analysis of large numbers of samples in duplicate (a costly and time-consuming task), as well as necessitating knowledge of the underlying system parameters which one is trying to estimate.However, by using Monte Carlo simulation techniques on a digital computer, we can obtain realistic estimates of the probable performance of the method with a variety of data sets each containing a distinct controlled deviation from the basic assumptions. These simulated results can be illustrated by actual results obtained in laboratory trials to obtain a realistic assessment of the performance of the method in practice. The computer program written for the simulation experiments used samples of 1 000 dupli- cate observations (x1,x2) drawn at random from specified populations belonging to a variety of analytical systems : (a) Gaussian: F(x) = N(c,o,), a, = 0.5 + 0.05~ (b) Log-normal: F(y) = N(O,a,), o, = 0.13 + 0.01~ (c) Gaussian: as system (a), but observation x2 has a constant bias added to it.* For Part I of this series, see p. 690. x = ( 1 0 V - 1) + c700 HOWARTH AND THOMPSON : DUPLICATE ANALYSIS Analyst, VoZ. 101 (d) Mixture distribution: observations drawn from two systems (I and 11) UCI = CTQZ + KIC { OCII = OOII + KIIC in the proportion P(1) and P(I1) such that P(1) + P(I1) = 1.0. In all calculations, a certain proportion, P(O), of zero concentrations was included and the concentrations above the zero point were drawn from a uniform distribution in the range 1.0 < c < 10.0. Such a model would imitate reasonably well the situation encountered in practice with geochemical analyses where a large proportion of the samples lie near the detection limit. In order to account for various methods of recording the final observed reading (equivalent to x,,~,), we rounded up or truncated the observation to a predefined number of decimal places.For example, a true instrumental value of 1.372 96 . . . when rounded to three places is R3 = 1.373, and truncated is T, = 1.372. Results were obtained for &,Ti where i = 0,3. The basic algorithm is as follows for the Gaussian case: (i) obtain a uniform random number U , 0.01 < U < 1.0, (ii) if U < P(0) then F(x) = N(O,u0), else (iii) obtain c uniform in the interval 1 < c *< 10 then F(x) = N(c,a,), CT, = oo + kc, (iv) obtain (x1,x2) EF(x), (v) add bias, x2 = x2 + 8, 6 > 0. (vi) round or truncate {x1,x2) to n decimal places: {x1*,x2*), (vii) “observations” {x1*,x2*) used to form (+(x,* + x2*), [XI* - x2* I}, Repeat steps (i) to (v) for all (1 000) duplicate pairs, then (viii) estimate a, and k from {&(x,* + x2*), [A:,* - x2* I}.Repeat steps (vi) to (viii) for n = 0, 3. These calculations were carried out in single-precision (60-bit word length) arithmetic on a CDC Cyber 73 computer. The standard CDC pseudo-random number generator RANF was used as the core of the method, a different seed being used for each set of 1 000 duplicates. (Steps (vi)-(viii), above, were obviously based on the same set of true duplicate values to obtain the observed values in each experiment.:) Generation of the (uncorrelated) {x1,x2} pairs from the normal (or log-normal) distributions N(O,o,), N(c,o,) used the standard Muller method., The data analysis was carried out using groups of 11 duplicates for the median by the regression technique described previous1,yl using our subroutine DUPAN2.The Gaussian Model The underlying system has the parameters uc == 0.5 + 0.05~ in terms of the concentration. The most accurate and unbiased recording of the observations will be to round-off the observa- ions, retaining any negative measurements as negattive in sign, as these are statistical estimates of a zero true value of concentration (this is not the same as stating the concentration is negative). We would expect a slight loss in accuracy if the observations are truncated on recording, rather than rounded. Rounding and Truncation Effects ; Recording of Negative Readings Fig. 1A shows the estimate of the standard deviation at zero concentration (Go) plotted as a function of the proportion of zero Observations P(0) in the data set, for results rounded to R,, R, and R, decimal places.(R, results do not differ significantly from those with more decimal places.) In all instances, negative observations have been correctly recorded as such. Fig. 1B shows generally similar results when the observations are truncated to To, T , and T , places. The results for To are, as one might expect, somewhat lower in general than R,, while T , and T , are in good agreement with R, and R,. In both series, agreement of the estimated value with the actual value of a, is good [particularly when the standard error of the regression to obtain h0 is allowed for. For example, it drops from 0.06 to 0.01 on the R, results, as P(0) increases].Poor estimation is obtained with extreme curtailment of the results, R, and To. In subsequent discussions, we will refer throughout to the best-case rounded results (R3) unless otherwise stated. In contrast, Fig. 1C shows the results for the data when values of negative measurements have been systematically recorded as zero. A large bias in the estimate of a0 results whenSeptember, 1976 I N GEOCHEMICAL PRACTICE. PART I1 701 0.5 (t? .. 1 A I ~r 8 -A B 0 a p A - A A A A A A L I I I I I I I I I I A 1.0 I- A A A A 0.10 1 8 t i A 8 0 - 0.01- I ' I I I I I ' - A ' I 1.0 - - - A - A A :a I k 0.10 a A 8 ' I A 0.05 ; 11 A 0.01 " l t t ' l l l ' Fig. 1. A, Effect of round-off error for R, (e), Rl (m) and I?, (A) ; and B, effect of truncation for T, (O), TI (0) and To ( A ) ; in both instances negative readings were left unchanged.C, As A, but all negative readings were .set to zero. the proportion of zero values is greater than 0.5. The bias lowers the value of &-,, which has the effect of making the method appear to be more precise than it actually is and under- estimating the detection limit. In the estimation of the constant term (k), the effect of the method of recording is the opposite (Fig. 2A and B), the magnitude of ,& having a positive bias on recording negative values as zero, when P(0) is a significant proportion of the data. As expected, estimates of k become increasingly less reliable when high proportions of zero observations are included. Gaussian error; estimates of a, for the system uc = 0.5 + 0 .0 5 ~ as a function of P(0). B P(0) P(0) Fig. 2. Gaussian error; effect of round-off error [R3 (a), R, (.I. I?, (A)] on estimates of k for thesystem a, = 0.5 + 0 . 0 5 ~ . A, Negative values recorded; B, negative values set to zero as a function of P(0).702 HOWARTH AND THOMPSON : DUPLICATE ANALYSIS Aaalyst, VoZ. 101 Effect of a Systematic Bias on the Second Observation Suppose that during the course of an analytical run bias occurs in the method and, in consequence, the second of the pair of Observations (x,) forming the duplicate has a systematic shift in value (of magnitude +6) from the first (xJ. Fig. 3 shows the effect of this bias on the correct estimation of the parameters for the same underlying system a, = 0.5 + O.O5c, for a proportion P(0) = 0.10 zero values.It is clear that until the bias becomes relatively large (6 > 0.25 in this example) the estimates Go and k are reasonably good. As the bias becomes larger, Go approaches 6 asymptotically, and k falls effectively to zero, as might be expected. The behaviour of oo is entirely predictable from the properties of the folded normal di~tribution,~ the point of the folding being greater than 0, with respect to the observations (x1,%! + 8). 0.2 0.01 0.1 1 5 Bias (6) 0.1 c 1 0.001 -- 0.01 0.1 1 5 Bias (6) Fig. 3. Gaussian error; effect of increasing bias (8) on the system 0, = 0.5 + 0 . 0 5 ~ with P(0) = 0.10, expressed as estimates A, a,,; B, R . The Log-normal Model The model for this trial was taken to be aLc = 0.13 + 0 . 0 1 ~ with F(y) = N(O,a,,) and x = (10” - 1) + c as this will yield a distribution comparable in range with the Gaussian model investigated above.It should be noted that the behaviour can be approximated by (T, = 0.3 + 0.03~ for the greater part of the probability distribution for a given con- centration, c . Figs. 4 and 5 show that the effect of a log-normal system error distribution is to have the parameters of the approximating linear function reasonably well estimated. The dampen- ing effect of the median, used for the regression. to fit the system parameters, could be ex- pected to be operating here to suppress the wild high differences produced in some duplicates by the log-normal error distribution. Finally, it will be noted that the effect of bias caused by recording negative observations as 0 is exactly the same as observed in the Gaussian model: h0 is lowered and ,$ raised.September, 1976 0 1.0 a 0.10 r 0 O 0 -a?- 4 ," 0 0.01 ' I i " ' ' IN GEOCHEMICAL PRACTICE.PART I1 703 Fig. 4. Log-normal error; effect of setting negative readings to zero, for the system UL = 0.13 + O.Olc, in estimate of oo as a function of P(0). 8, Negative readings retained; 0, negative readings set to 0. Analysis of a Mixture of Samples of Two Matrix Types Since a group of samples with differing matrices could be expected to have different parameters, a. and k , we might expect that, in an analytical run involving samples of two rather different matrix types, one would observe from the mixed duplicates properties ranging between the extremes of the two sets of samples.The response of the estimated over-all system parameters was investigated for samples characterised by a, = 0.5 + 0 . 0 5 ~ with various proportions of samples with a relationship given bya, = 0.75 + 0.05~; a, = 0.75 + 0.075~; CT, = 1.00 + 0.10~; anda, = 2.00+0.20c in four successive experiments. Again, 1000 duplicates were simulated for each trial; 10% zero determinations and Gaussian error (as above) was assumed throughout. The results for b0 and (Figs. 6 and 7) show that as the proportions of the second population rise, Go has an approximately linear relationship between aoI and uon, while that for ,$ may be non-linear. Estimation will obviously be best when one population is dominant. Fig. 5. Log-normal error; estimates of k for system uL = 0.13 + 0 .0 1 ~ as a function of P(0). Symbols as for Fig. 4. Parameter Estimation and Numbers of Duplicates Estimation of the Gaussian system u, = 0.5 + 0.05~ with numbers of duplicates between 50 and 1000 (Fig. 8) shows that the true system parameters uo and k generally lie within one standard error of the estimated values as obtained from the regression of lxl* - x2* I on +(xl* + x2*). As expected, the size of the standard error of regression falls as the sample size increases and there is a general inverse relationship between the errors in t?o and fl.704 HOWARTH AND THOMPSON : DUPLICATE ANALYSIS Analyst, VoZ. I01 Obviously the correlation between &, and the detection limit and & and the asymptotic precision means that the behaviour of ii0 and & predicts that of the other derived parameters.2.5 2.0 i (g 1.5 1 .o 0 P (11) Fig. 6. Effect of mixed populations (I and 11) on estimate of uo for various systems as a function of P(I1). oc = a + fiC a m 2 A 0.76 0.075 A 0.76 0.05 0.50 0.05 >I 0.20 I I 8 8 0 . 0 1 Fig. 7. Effect of mixed populations on estimate of k for various systems as a function of P(I1). Symbols as for Fig. 6 . Implications of These Trials The simulation experiments have shown that the proposed method of duplicate analysis is reasonably robust for a linear system with Gaussian error, or log-nomal error that can be reasonably approximated by a linear function for the analytical system. We show later how a non-linear system can be identified and how such a situation can be dealt with once it has been recognised.However, the most important implication of these trials is that the normal laboratory practice of discarding meaningless parts of the measurements (insignificant digits and negative measurements) may prevent correct identification of the size of the random errors inherent in the analytical system. Moreover, neglect of these terms will introduce an optimistic bias into the subsequent interpretation that will suggest that the precision of the analyses is better than it is in reality. Perhaps where a value is regarded as geochemically significant only t o the nearest whole part per million, at least one decimal digit should be recorded in order to estimate correctly the magnitude of the system errors. Once this has been done, the data can subsequently be treated in whole parts per- million, should this seem desirable in the light of the statistical appraisal.It is, perhaps, worth pointing out that if the duplicates were independent samples collected in the field and analysed once each, the method will assess the total sampling plus labora- tory error.September, 1976 IN GEOCHEMICAL PRACTICE. PART I1 705 I- I 0 1 oob Number of duplicates (n) 0 1000 Number of duplicates (n) Fig. 8. Variation of estimates of a, and k for the system a,, = 0.50 + 0 . 0 5 ~ with P(0) = 0.10 as a Error bars correspond to standard error of regression with a median function of number of duplicates, n. based on 11 pairs. Detection of Abnormal Behaviour in the Data Set In this section, we illustrate a simple method for evaluating the goodness of fit of the linear model and the effects of various types of departure from it, so that the causes of such departures can be identified in practice.Contingency Table If the linear model holds good for a particular analytical system, then we would expect the duplicate pair differences to conform to the half-normal distribution. Then, for an arbitrary concentration distribution, we would expect an equal number of observations to fall within ten-percentile intervals of the half-normal for any ten-percentile interval of the concentration distribution (Fig. 9). The two-way table of observed frequencies for the ten- percentile absolute difference and concentration intervals, referred to here as the contingency table, is generated automatically by the DUPAN~ computer program.Departures from the expected distribution are shown by lack of an even spread of frequency values about the expected frequency over the whole of the table. Simulated Examples The examples that follow illustrate the types of pattern that occur in the contingency table for various departures from the ideal model. In each of these simulated cases, one effect has been isolated to show the response in the contingency table, but in actual examples (some of which are discussed in the following section) these effects could be expected to act together in some instances. Gaussian error would yield cells randomly scattered through the table with frequencies around that of the expected frequency. Examples of departures from this situation are illustrated in Fig.10, the ratios of the observed to expected frequency being shown symbolically for each absolute difference and concentration decile of the table. Substantially the same pattern, with a diagonal of high frequencies and large patches of zero frequency values would be obtained with excessive rounding or truncation, whether or not measurements below zero are recorded as negative or set to zero. The effect of setting measurements below zero to zero [Fig. lO(ii)] is again a characteristic pattern. The high frequency of zero concentration values and the diagonal of moderate positive departure from the expected frequency below those cells with zero frequency is distinctive. In general, one could expect that if there is no systematic bias present in the analytical system, then the number of positive differences between the duplicates would be close to Fig.l O ( i ) illustrates the effect of excessive truncation.706 HOWARTH AND THOMPSON : DUPLICATE ANALYSIS Analyst, VoZ. 101 F \ IX, -x, I Fig. 9. Schematic illustration of the derivation of the contingency table for the frequency (F) of duplicates falling into a particular absolute difference ([XI - X,l) decile (ADD) and concentration (C) decile (CD). It is assumed that the ADD and CD intervals correspond to half-normal and arbitrary probability density functions, respectively. The latter is shown here as a Gaussian function for con- venience. half the total number of duplicates. However, if bias is present, this will not be true and significant departure from the expected number of positive differences will occur.The D U P A N ~ program tests the observed number of positive differences using binomial expecta- tions and it is a simple and sensitive test for systematic bias. Bias will also manifest itself in the contingency table by giving a broad band of high frequencies in the centre of the table, with low frequencies in the margins [Fig. 10(7;ii)] ; in these circumstances, the fitted model will be a best over-all approximation (Fig. 3) and the contingency table is reflecting the departures from it. A log-normally distributed analytical error is mainly shown by the behaviour of the low normal [Fig. lO(iv)]. Mixed distributions are fitted by a best over-all approximation and are indicated [Fig. lO(v)] mainly by the unusually high proportion of wild values above the 90th percentile of the half-normal. Fitting to a non-linear system gives a variety oE patterns in the contingency table, depend- ing on the degree of non-linearity of the true analytical system.A pattern such as that in Fig. lO(vi) is fairly characteristic, with strong departures from even frequency values, caused by the fitted (linear) model lying below the true curve at low and high concentrations and above it in the intermediate concentration ranges. Examples from Laboratory Results We have examined large numbers of duplicates during routine laboratory analyses in the Applied Geochemistry Research Group over the last few years. Hundreds of samples of broadly similar matrix (e.g. , rocks, soils or stream sediments) can be analysed for any parti- cular project by rapid colorimetric or atomic-absorption or emission spectrometric techniques as required.Analysis of duplicates from these prosjects has, in general, not shown any peculiar features, nor have we seen any evidence that either log-normal error distribution or non-linear analytical systems are operating.September, 1976 IN GEOCHEMICAL PRACTICE. PART I1 707 4 I 1 0 0 0 0 0 0 0 0 0 0 o o o o o o @ o ~ o o o o o @ o o o o o - 0 0 o @ o @ o 0 0 0 - 0 0 0 0 0 0 0 0 0 0 - n o 0 0 o o 0 o @ @ o - 4 @ @ @ @ @ 0 @ 0 0 @ - 0 0 0 0 0 0 0 0 0 0 - 0 0 0 0 0 0 0 0 0 0 - 0 0 0 0 0 0 0 0 0 0 - 0 I f I 1 I I I I I I a3 a3708 HOWARTH AND THOMPSON: DUPLICATE ANALYSIS Analjst, VoZ. 101 It is of interest, however, that in all instances in which samples were submitted for analysis without prior randomisation of the duplicates (and other samples) in the sequence (e.g., dupli- cates analysed successively), a moderate degree of bias was detected by the number of positive differences between duplicates.It was not strong enough, however, to be obviously reflected n - 0 - 0 0 - 0 0 -0 @ Q - 0 0 0 * - 0 0 0 8 - a 0 0 -0 0 0 -0 0 O O 0 - 0 @ @ o @ o @ o o I l l l I I I 1 1 1 Fig. 11. Contingency tables obtained with actual data; symbols as for Fig. 10. (i) Lead determined by direct-reading emission spectrograph with randomised samples, E = 3.15 ; (ii) arsenic determined colori- metrically, E = 6.66; (iii) zinc determined by atomic absorption, E = 7.20. Three examples that have been encountered in practice are given in Fig.11. Fig. ll(i) is typical of the usual results, in which no significant departure from the normal system is evident. It shows the results for lead in 315 soils, analysed in duplicate by a direct-reading emission spectrograph with samples submitted in a randomised sequence. Contrasting results are shown by 666 stream sediments analysed in duplicate colorimetrically for arsenic [Fig. ll(ii)], where the data were rounded to the nearest 4 p.p.m. The similarity to Fig. lo(;) is evident. Finally, Fig. ll(iii) shows results for 720 stream sediments analysed in duplicate for zinc by atomic-absorption spectroscopy. Again, the effect of the method of data recording is evident. The enhanced number of wild values in the 90th percentile for the model suggests the possibility of a mixed system effect [see Fig.lO(v)].Se+tembev, 1976 I N GEOCHEMICAL PRACTICE. PART I1 709 Conclusions Detailed study of the factors that can operate in an analytical system can be carried out using Monte Carlo simulation methods. The results lead to a practical guide for the analysis of the behaviour of such a system based on duplicate chemical analyses, the linear model of increasing variance with concentration being fitted in the general case. Experimental results have shown that in most instances, such a model is adequate and departures from it can be recognised. Although we have found circumstances in which the estimated slope of the regression is not significantly different from zero (i.e., constant precision over the concentration range), in general the linear model explains the variance in precision better than the hitherto accepted intrinsic hypothesis that there is no systematic change of precision with concentration. While a non-linear function may be preferable, as it may give a better fit in some instances, the curvature would have to be pronounced to be statistically recognisable and would require much more data than the linear model to fit accurately. We have not yet encountered any convincing evidence for such a system in practice. Comparison with simulated results shows that the contingency table representation of the fitted linear system allows adequate identification of probable causes for departure from the ideal linear system. The effect of data recording methods on the estimation of system behaviour has been shown to be critical, over-optimistic estimates of precision being obtained when data values are subject to severe truncation or round-off error. Computations for both parts of this paper were carried out on the Imperial College Computer Centre CDC 6400/7314 facility. References 1. Thompson, M., and Howarth, R. J., Analyst. 1976, 101, 690. 2. Jansson, B., “Random Number Generators,” Victor Petterson, Stockholm, 1966. 3. Leone, F. C . , Nelson, L. S., and Nottingham, R. B., Technometrics, 1961, 3, 543. NOTE-Reference 1 is to Part 1 of this series. Received January 28th, 1976 Accepted April 21st, 1976

ISSN:0003-2654    

DOI:10.1039/AN9760100699

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

710 Analyst, September, 1976, Vol. 101, pp. 710-716 Determination of Organic and Inorganic Carbon s i n Soils by Potentiometry L. Th. Begheijn De$artment of Soil Science and Geology, Agricultural University, Wageningen, The Netherlands A method is described for the direct, rapid and precise determination of 0-6 mg of organic and/or inorganic carbon in soils. The equivalent amount of carbon dioxide, liberated by improved combustion techniques, is absorbed and precipitated in a solution containing 2.50 mmol of sodium hydroxide and 0.96 mmol of barium chloride. Next, the resultant barium carbonate is dissolved by the addition of 0.99 mmol of EDTA (disodium salt). The pH of the final solution is related to the amount of carbon dioxide present by means of a calibration graph. By use of a pH meter with an expanded range of 1 pH unit, sensitivity to 1Opg of carbon is achieved.Analyses of pure organic compounds, such as benzoic acid arid hydroquinone, show recoveries accurate to within 0.1 mg of carbon. Results for calcareous soils are in close agreement with values for calcium oxide obtained by use of X-ray fluorescence spectrometry. Interference from chloride is effectively eliminated by a preliminary evaporation step. Carbon occurs in soils in both inorganic and organic forms, chiefly as calcite, dolomite or organic matter. The last material can be destroyed by wet or dry combustion techniques. Ames and Gaitherl first reported the complete wet oxidation of organic carbon in soils with a chromic acid - sulphuric acid solution. Because interfering sulphur dioxide might be co- evolved with the carbon dioxide, Clark and Ogg2 replaced the sulphuric acid with a mixture of sulphuric and orthophcsphork acids (3 + 2).Allison3 applied this improved sulphuric acid - orthophosphoric acid solution and reduced the us8ually bulky purifying train. The carbon dioxide was retained in absorption towers and weighed. He reported a good recovery com- pared with dry combustion procedures. Although several workers*-' have proposed alterna- tive digestion mixtures, Allison's method has been generally accepted as being the most reliable. A dry combustion technique for the determination of up to 10 mg of total carbon in soils using the Leco Automatic 70-Seconds Carbon Analyser (Tabatabai and Bremnerg) has given results that are in close agreement with those by all is on,'^ method.Non-aqueous absorption and titration techniques have been described in recent years.9-11 Read12 describes a sensitive method for the determination of carbon dioxide in silicate rocks by use of an absorption solution of 5% monoethariolamine in dimethylformamide and direct titration of the carbon dioxide with a solution of tetrabutylammonium hydroxide in toluene. This paper describes a rapid and simple routine method for the determination of carbon in the semi-micro range (0-6 mg of carbon), which is an alternative to non-aqueous titrimetry or the Leco carbon analyser, and which embodies an improvement of the wet-combustion step applied by Allison. Experimental Apparatus The distillation and absorption apparatus is shown in Fig.1, the parts being as follows: A, absorption tower filled with ascarite (sodium hydroxide-coated asbestos), 8-20 mesh; B, gas washing bottle filled with water (bubble counter) ; C, microscale gas burner, preferably shielded; D, 50-ml reaction vessel (Quickfit FP 50/1/1A); E, gas delivery tube fitted into F; F, cone - screw thread adaptor (Quickfit ST 51/13)1; G, expansion adaptor (Quickfit XA/21) ; H, Allihn condenser (Quickfit CX 7/02), with additional socket joint, size 14/23; I, adaptor (Quickfit M F 17/1) ; J, 50-ml separating funnel (Quickfit D2/11) ; K, three-way stopcock, plane (Vestale) ; L, absorption unit consisting of a 100-ml glass tube (about 4 cm in diameter, 13 cm long), closed by a perforated rubber stopper fitted with a fritted-glass gas-dispersion tube in the centre (diameter 25 mm, filter porosity 100 pm), a capillary glass vent tube and a glass delivery tube slightly bent towards the central tube and with an inner lining of plasticBEGHEI JN 71 1 tubing (2 mm i.d.) ; M, magnetic stirrer (Minimag KMO, IKA Werk) with PTFE-coated magnetic stirring rod (15 mm long) in the absorption tube.Connections A-B, B-E, I-K and I<-L are made with plastic tubing; the tubing carrying cooling water is made of rubber. For routine analyses in our laboratory two pairs of the apparatus were constructed, each pair mounted on a standard laboratory frame 1 x 1 m in size. Nitrogen+ A 6 n Fig. 1. Distillation and absorption apparatus. For key see text. Fig. 2. Dosed injec- tion system. For key see text.The injection system is illustrated in Fig. 2 and the parts are as follows: A, Cornwall continuous pipetting outfit (Becton, Dickinson and Co., New Jersey, USA), consisting of a 10-ml calibrated spout pipette with sinker and cannula (2 mni 0.d.); B, 1-1 polythene storage bottle (neck diameter 6 cm); C, perforated rubber stopper fitted with (1) a capillary glass vent tube or ascarite tube for sodium hydroxide, (2) a glass storage tube for the pipetting outfit, (3) a delivery tube, 15 mm in diameter, and (4) a 15-ml glass storage tube for the sinker. The injection system is constructed in triplicate, for the sodium hydroxide, barium chloride and EDTA solutions. In order to prevent damage to the pipetting outfit by the solutions during storage, the spout pipette is always stored filled with water and the sinker is placed in the 15-ml glass tube.The delivery tube is closed by a rubber stopper during st orage. Additional equipment valve (see Note 1). Nitrogen (99.98%, carbon dioxide free) is drawn from a cylinder that is fitted with reducing NOTE 1. The reducing valve should be set to a maximum pressure of 15 x lo3 Pa so that blow-outs of A pH/mV meter (Corning, Model 12), a double-junction reference electrode (Orion 900200) and a pH (glass) electrode (Ingold, Type 205) with a standard connector (Corning 476115, 36 in) are also used. Both electrodes are fitted in a single perforated rubber stopper and immersed in water in a 50-ml glass tube during storage. A magnetic stirrer is required. the apparatus are avoided even when the outlets are closed.712 Analyst, Vol.101 Reagents In addition, all solutions should be made up with freshly boiled distilled water. The bulk solutions are stored in 10-1 polythene bottles fitted with a stopcock near the bottom. An ascarite vent tube is mounted on the sodium hydroxide bottle. Reagents for the digestion (see Allison3) are as follows. Digestion acid for carbonates. Dissolve 57 ml of sulphuric acid (sp. gr. 1.84) and 92 g of iron(I1) sulphate heptahydrate in 600 ml of water. Cool the solution and dilute it to 11 with water. Digestion acid for organic matter. Mix three volumes of sulphuric acid (sp. gr. 1.84) with two volumes of orthophosphoric acid (sp. gr. 1.70). Potassium dichromate, powdered. Reagents for adsorption are as follows.Sodiunz hydroxide solzdion, 0.5% m/V. Dilute an analytical concentrate from an ampoule (Baker) containing 0.5 mol of sodium hydroxide to 41 according to the directions provided with the ampoules. Barium chloride solution, 1.0% m/V. Dissolve 23.46 g of barium chloride dihydrate in water and dilute to 21. Ethylenediaminetetraacetic acid, disodium salt, solution, 3.7% m/V. Dissolve 37.00 g of EDTA, disodium salt, in water and dilute to 11. BEGHEI JN : DETERMINATION OF ORGANIC AND Analytical-reagent grade chemicals should be used. Procedure Initial moistening of the soil Alperl3 reported low results for organic carbon when directly analysing air-dried soil, caused by protective gelatinous silicic coatings around the soil particles. An insufficient impregnation capacity on the part of the viscous digestion acid can also cause negative errors.Allison3 and Clark and Ogg2 wetted their samples (0.5-3 g) with 3 ml of water prior to analysis. An initial moistening step is included in the present procedure. Removal of Carbonates for determination of organic carbon Carbonates can be removed by use of dilute mineral acids prior to the determination of organic carbon, and subsequently analysed. As Roberts et al.14 found that up to 44% of organic carbon, from modern carbonate sediments, passes into solution or suspension as a result of the action of dilute sulphui-ic acid, freeze-drying was advised rather than discarding the acid filtrate. In the method described here, carbonates are removed by using a dilute sulphuric acid - iron(I1) sulphate solution mixture, as was reported by Allison,3 and organic carbon is subsequently determined in the residue obtained after evaporating off the excess of solution on a boiling water bath.Iron(I1) sulphate prevents the possible oxidation of organic matter by any manganese(1V) oxide present. Because Anderson and Harris6 reported the incomplete removal of carbonates, owing to the presence of sulphate coatings on the carbonate particles, powdering of the sample is advised. Moreover, the use of a 2 M hydrochloric acid - 5% iron(I1) chloride solution is advocated for dolomitic soils. Removal of chlorides Chloride interferes in the determination of organic carbon because of the formation of chromyl chloride and chlorine, which may partly neutralise the absorption solution.This interference is eliminated by carrying out a preliminary treatment as for the removal of carbonates. (Chloride does not interfere in the determination of inorganic carbon.) Outline of the determination Weigh a suitable amount of the powdered soil sample, containing up to 6 mg of organic and/or inorganic carbon, into the reaction vessel. Moisten it with a few drops of water. [If a separate preliminary test has indicated the presence of carbonates, inorganic carbon is first removed or determined (as for organic carbon) by heating with 5 ml of sulphuric acid - iron(I1) sulphate solution.] Evaporate the residue to dryness on a boiling water bath and add 250 mg of potassium dichromate, then fit the vessel to the apparatus and connect theSeptemlrcif, 1976 INORGANIC CARBON IN SOILS BY POTENTIOMETRY 713 carrier-gas delivery tube.Insert a magnetic stirring rod into the glass absorption tube and pass nitrogen through the tube at a rate of 2 bubbles per second for at least 10 min. Inject 20.0 ml each of sodium hydroxide and barium chloride solutions into the absorption tube. Next add 5 ml of the digestion acid to the separating funnel. Switch the three-way stopcock to a position open to the air and pass the acid through the condenser into the reaction flask. Immediately close the tap on the funnel and switch the three-way stopcock back to its original position. Then start the flow of cooling water and decrease the flow-rate of the carrier gas to 1 bubble per 2 seconds. Bring the solution to boiling in about 5 min by use of a flame 1 cm in height, held 1 cm below the reaction vessel.Boil gently for another 5 min in such a way that 110 visible white acid fumes rise above the second bulb section of the condenser. Flush the remaining carbon dioxide for another 10 min at a flow-rate of 2 bubbles per second. Next, turn the three-way stopcock to close all connections (see Note 2). Inject P0.0ml of EDTA solution into the absorption tube ; now turn the stopcock to admit air to the absorption section of the apparatus and after 1 min start the magnetic stirrer. After 10 min (see Note 3) slowly turn the stopcock again in order to force the solution out of the fritted-glass filter-tube. Separate the absorption vessel from the equipment and wash the delivery tube with a few drops (about 0.5ml) of water into the contents of the absorption vessel.Place the vessel over a magnetic stirrer and insert the electrodes. Finally, stop stirring and read the pH to two decimal places. Before every series of measurements the pH meter is calibrated (see Note 4). Two blanks should be run during each series of determina- tions. NOTES- the passage of acid fumes beyond the condenser. 2. By closing the stopcock to all connections full decompression of the system is avoided, preventing 3. During the 10 min following the admission of air the parallel set can be started up. 4. Use pH 7.00 buffer (Titrisol, Merck) and an EDTA, disodium salt - barium chloride - sodium hydroxide mixture (see below) at pH 12.00. Fluctuations about the latter pI-f occur, normally within 0.02 pH, and are compensated for by the manual temperature control knob (this fluctuation hardly affects the calibration at pH 7).The calibration a t pH 12.00 should be tested about every five readings. Preparation of calibration graph Add successive 0.25-ml portions of aqueous sodium carbonate (primary standard) solution (0.883%, m/V) and hydrochloric acid (0.167 M) from two 5-ml microburettes to a solution 0.20% in sodium hydroxide, 0.40% in barium chloride and 0.74% in EDTA, disodium salt (this last solution prepared by adding to a suitable vessel successively 10.0 ml of 3.7% EDTA, disodium salt solution, 20.0 ml of 1.0% barium chloride solution and 20.0 ml of 0.5% sodium hydroxide solution). Each pair of 0.25-ml increments represents an addition of 0.25 mg of carbon (as carbon dioxide).Mix the solution well by magnetically stirring it and note the pH after each pair of additions. Plot the pH values Z J ~ Y S U S milligrams of carbon added (Fig. 3). Reactions chloride. The reactions are as follows. The absorption solution contains 2.50 mmol of sodium hydroxide and 0.96 mmol of barium A bsorption. CO, + Ba2+ + 20H- -+ BaCO, + H20 gas precipitate NezttraZisatio.uc and dissolution. By addition of 0.99 mmol of EDTA, disodium salt, 1.98 mino1 of OH- are neutralised, and the barium carbonate is dissolved. H2EDTA2- + Ba2+ + 20H- -+ BaEDTA2- + 2H,O H2EDTA2- + BaCO, + 20H- -+ BaEDTA2- + CO,2- + 2H20 precipitate CO," + H,O + HCO, + OH-714 BEGHEI JN DETERMINATION OF ORGANIC AND Analyst, Vol. 101 8.0 0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Amount of carbodmg Fig.3. Relationship between pH and amount of carbon absorbed as carbon dioxide in a solution that is 0.20% m/V in sodium hydroxide, 0.40% m/V in barium chloride and 0.74% ?n/V in disodium EDTA (pH 12.00, calibrated). Thus, the calibration graph can be regarded as a titration graph of 0.52 mmol of sodium hydroxide against up to 0.50 mmol of carbon dioxide in an original solution that is 0.20% in sodium hydroxide, 0.40% in barium chloride and 0.74% in EDTA, disodiurn salt. The possible escape of free carbon dioxide at pH values below 8.5 limits the usefulness of the calibration graph beyond 7 mg of carbon. Sensitivity and accuracy The calibration graph is based on 28 successive additions of 0.25-ml aliquots of sodium carbonate solution and hydrochloric acid to 50 ml of the sodium hydroxide - barium chloride - EDTA, disodium salt, mixture, whereas only gaseous carbon dioxide is passed through the absorption solution during an actual analysis.In order to check whether dilution during calibration has any appreciable influence, five additions of 2 ml of distilled water were made to the final solutions with 1, 3 and 5 rng of absorbed carbon. Positive errors of 22 and 6 pg of carbon per 2-ml addition were found at the 1 and 3 mg of carbon levels, respectively, but the error was negligible at the 5-mg level. Therefore, to minimise the marked effect of dilution at low carbon contents, 2.00-ml aliquots of 0.883% aqueous sodium carbonate solution and 0.167 M hydrochloric acid are added to the final solution and 2.00 mg of carbon are subtracted from the amount of carbon found.This procedure has the additional advantage of avoiding measurement in the range 0-2 mg of carbon, where the sensitivity of the method is relatively low; ApH/AC (mg) is 0.165 at the l-mg level against 0.830 at the 3 mg of carbon level (Fig. 3). Results and Discussion The method was tested on reagent-grade organic compounds, such as benzoic acid and hydroquinone, which proved to be completely oxidised, in contrast to the findings of Anderson and Harris.6 Because graphite and charcoal may cause incomplete recovery of total carbon in some soils, active charcoal was analysed as well. The results are shown in Table I. In order to test the inorganic carbon determination, standard samples (25-50 mg) of calcium carbonate (analytical-reagent grade) and calcareous soil samples were analysed.Dilute hydrochloric acid was used for the dissolution of inorganic carbon. The recoverySeptember, 1976 INORGANIC CARBON I N SOILS BY POTENTIOMETRY TABLE I RESULTS OF CARBON ANALYSIS OF ORGANIC COMPOUNDS 715 Amount Compound takenlmg GluCCJ33 . . .. .. . . 6.84, 16.28 Hydroquinorie . . .. . . 5.78, 3.83 Ascorbic acid . . .. . . 7 2u, 15.70 Citric acid .. .. .. 12 05, 0.58 Henzoic acid . . .. . . 1.28, 6.94 Active charcoal" . . .. . . 3.48, 4.93 Carbon content/mg F- Theoretical Found 2.73, 6.51 2.70, 6.45 3.78, 2.50 3.85, 2.45 2.94, 6.42 2.90, 6.50 4.52, 0.22 4.45, 0.157 0.88, 4.77 0.80,t 4.80 3.15, 4.40 -* * Carbon content not specified; dried a t 105 "C.No 2-ml portions of sodium carbonate (0.8830/,) and hydrochloric acid (0.167 M) were added. of pure calcium carbonate (without soil) was l00.0%, with a standard deviation of 2.27(, (n = 8). The results for the soil samples are compared with results for calcium oxide, obtained by X-ray fluorescence spectrometry,15 in Table 11. (Amounts of calcium and carbonate species other than calcium carbonate were negligible in these samples.) TABLE I1 RESULTS OF INORGANIC CARBON ANA4LYSIS OF CALCAREOUS SOILS BY POTENTIOMETRY COMPARED WITH THOSE FOR CALCIUM OXIDE BY X-RAY FLUORESCENCE SPECTROMETRY l'otentioinetric results per 100 mg of dry, powdered soil samples; analyses on five batches on rive dii~ercnt days. san1p:t number 731230 23 1 232 233 234 235 236 Carbon by potentiometry/mg* 6.45, 5.55, 5.45 5.20, 5.35, 5.20, 5.25 5.80, 5.95, 5.80, 5.95 5.90, 6.20, 5.86, 6.10, 6.25 5.35, 5.55, 5.35 3.60, 3.75, 3.50, 3.70, 3.80 4.55, 4.70, 4.45 Mean/mg ( 5.48 5.25 5.88 6.06 5.42 3.67 4.57 CaC) content if all :arbon is in CaCO,, 7; 25.6 24.5 27.4 28.3 25.3 17.1 21.3 CaO content by X-ray fluorescence spectrometry, o/; 26.0 25.0 26.9 27.9 25.6 18.5 22.5 * Rcadings from graph are accurate to 0.05 mg of carbon.The reproducibility of the method is good (the standard deviation of individual determina- tions is 0.12 mg of carbon, n = 20), as is shown by results of five batches run on different days (Table 11). Conclusions In the direct and sensitive method described for the sequential determination of inorganic and organic carbon in soils, interferences are negligible.The method utilises improved com- bustion techniques and is five times more sensitive than the conventional gravimetric and titrimetric procedures. The apparatus is built up from low-cost, readily available, standard laboratory equipment and is simple to construct and operate. Up to 25 analyses can be carried out in 1 d by a skilled analyst using the method, which appears to be a promising alternative t o the non-aqueous titrimetric procedures and which can distinguish between organic and inorganic carbon, in contrast to dry combustion by use of the keco carbon analyser. It is expected that the application of the method will be extended from soil and rock samples to soil extracts and polluted waters, and possibly also to steel and air. The author thanks 0. D. Jeronimus for preparing the figures, and N. van Ereemen and R. Brinkman for critical comments on a draft of this paper. References 1. 2 . Ames, J. W., and Gaither, E. W., J . I H ~ . Engig Chem., 1914, 6, 661. Clark, N. A., and Ogg, C. I;. A., Soil Sci., 1942, 53, 27.716 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. BEGHEI JN Allison, L. E., Proc. Soil Sci. Soc. Am., 1960, 24, 36. McCready, R. M., and Hissid, W. Z., Ind. Engng Cheliw. Analyt. Edn, 1942, 14, 525. van Siyke, 33. D., and Folch, J., J . Biol. Chem., 1940, 136, 509. Anderson, J. U., and Harris, W., Proc. Soil Sci. SOC. Am., 1967, 31, 341. Nommick, H., Soil Sci., 1971, 111, 5. Tabatabai, M. A., and Bremner, J . M., Proc. Soil Sci. S’oc. Am., 1970, 34, 608. Grant, J. A., Hunter, J. A., and Massie, W. H. S., Analyst, 1963, 88, 134. Jones, R. F., Gale, P., Hopkins, P., and Powell, L. N., Analyst, 1965, 90, 623. Sen Gupta, J . G., Analytica Chim. Acta, 1970, 51, 437. Read, J . I., Analyst, 1972, 97, 134. Alper, P., J . Agric. Sci., Camb., 1938, 28, 187. Roberts, A. A., Palacas, J . G., and Frost, I. C., J . Sed’im. Petrol., 1973, 43, 1157. Halma, G., Acta Colloq. Spectrosc. Internationale X V I I , 1973, 11, 626. Received January 12th, 1976 Accepted April 21st, 1976

ISSN:0003-2654    

DOI:10.1039/AN9760100710

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

Analyst, September, 1976, Vol. 101, pp, 717-719 717 Colorimetric Determination of Phenylephrine Hydrochloride in Pharmaceutical Preparations* Yehia M. Dessoukyt Department of Pharmaceuticai Chemistry, Faculty of Pharmacy, Cairo University, Cairo, Egypt and Laila N. Gad El Rub Research Department, Socibtb Misr pour 1 'Industrie Pharmaceutique, 92 El Mataria Street, Post El-Zeitoun, Cairo, Egy@ An accurate, selective and simple method for the determination of phenyl- ephrine hydrochloride is proposed. The method is based on the formation of the nitroso derivative of phenylephrine in the presence of copper ions. The method has been applied successfully to pharmaceutical preparations without prior separation of phenylephrine. Other substances commonly present in such formulations do not interfere in the determination.Formation of the o-nitrosophenol derivative of benzene by the action of hydrogen peroxide, hydroxylammonium chloride and copper(I1) sulphate has been reported in the literaturelJ and the colorimetric determination of some benzyl derivatives by a method based on this reaction was described by bar to^.^ In the proposed method, nitrosation of phenylephrine is carried out by sodium nitrite in the presence of 0.2 iu hydrochloric acid and copper(I1) acetate. The red copper(I1) complex formed shows maximum absorption at 520 nm. A characteristic property of o-nitrosophenol is its ability to form highly chelated compounds with heavy metals. The presence of a copper salt is essential both to stabilise the nitrosyl radical and to ensure that o- and not p-nitroso- phenol is formed4; in the absence of a copper salt an unstable yellow colour that has alow absorbance is produced.Experimental A Carl Zeiss, type M4QI1, instrument was used. Apparatus Spectrophotometer. Reagents All reagents used should be of analytical-reagent grade. Sodium nitrite solution. Dissolve 5 g of sodium nitrite in 100 ml of water. CopPer(I1) acetate solution. Dissolve 2.5 g of copper(I1) acetate in 100 ml of water. Urea solution. Dissolve 1 g of urea in 100 ml of water. Hydrochloric acid, 0.2 M. Procedure Pipette a l-ml aliquot of sample solution containing the equivalent of 0.5 mgml-l ol phenylephrine hydrochloride into a 10-ml calibrated flask and add successively 1 ml of sodiun nitrite solution, 1 ml of copper(I1) acetate solution, 3.75 ml of 0.2 M hydrochloric acid anc 0.5 ml of urea solution, mixing the solution after each addition.Immerse the flask in a boilini water bath for 15 min, cool it in ice for 5 min, then add 1 ml of 0.2 M hydrochloric acid, make up to volume with water and thoroughly mix the contents of the flask. Measure the absorb. ance of this solution at 520 nm, using a l-cm cell, against a similarly prepared blank solution Calculate the concentration of phenylephrine hydrochloride by reference to a Cali bration graph prepared by applying the same procedure to solutions of phenylephrine hydro chloride reference standard of concentrations in the range 0.25-1.0 mg ml-1. * Paper presented at the 35th International Congress of Pharmaceutical Sciences, Dublin, Ireland t Present address: Faculty of Pharmacy, University of Tripoli, Tripoli, P.O.Box 4022, Libya. September lst-5th, 1975.718 DESSOUKY AND GAD EL RUB: COLORIMETRIC DETERMINATION OF Analyst, VoZ. 101 Results and Discussion The red complex formed under the specific experimental conditions described was found to be stable for several hours at room temperature when exposed to indirect sunlight; the absorbance values decreased very slightly after 24 1.1 (Table I). The colour obeyed Beer’s law over the range 0.25-1 mg ml-l. TABLE I EFFECT OF TIME ON THE STABILITY OF THE COLOUR Absorbance at 620 nm corresponding to 0.6 mg of phenylephrine hydrochloride; each value is the mean of three determinations. Time/h . . .. .. .. 0 :t 6 10 24 Absorbance . . .. . . 0.335 0.335 0.336 0.330 0.326 Several experiments in which the concentrations of nitrite and copper(I1) salt were inde- pendently varied were carried out in order to establish the optimum conditions of the reaction.Higher concentrations than those specified in the procedure produced unstable orange or brown colours having higher absorbance values. 0.3.- 0.2 - 0.1 - Fig. 1. Effect of pH on the absorb- ance of the colour. The effect of variation in pH on the development of the colour is shown in Fig. 1. Maximum colour intensity was obtained at pH 2. An additional 1 ml of 0.2 M hydrochloric acid, added TABLE I1 FORMULATIONS OF SAMPLES Sample Constituent Phenylephrine hydrochloride/mg . . . . Chlorpheniramine maleatelmg . . .. Ephedrine hydrochloride/mg . . . . Codeine phosphatelmg .. . . .. Sulphacetamide sodium/mg . . .. Phenyltoloxamine dihydrogen citrate/mg Naphazoline nitratelmg . . .. .. Chlorbutol/mg . . .. .. .. Additives, antioxidants, preservatives, flavours, etc. . . . . . . . . Vehicle (syrup, glycerin, propylene glycol, alcohol, buffer, distilled water, etc.) . . As required To 10 ml 11” 6 2 6 8 - As required To 10 ml IIIt 50 - - 0.26 10 - As required To 10 ml 1 IVS 10 6 - - 1.6 25 As required To 10 ml * Samples I and I1 (phenylephrine eye drops and Tussivan cough linctus) supplied by Soci6td Misr pour t Sample I11 (Toloxan nasal drops) supplied by Cid, Chemical Industries Development, Egypt. 1 Sample IV (Rhinoline infant nasal drops) supplied by AIDCO, The Arab Drug Company, Cairo, Egypt. 1’Industrie Pharmaceutique, Cairo, Egypt.Sf?$hWZbtT, 1976 PHENYLEPHRINE HYDROCHLORIDE I N PHARMACEUTICAL PREPARATIONS 719 after heating, was necessary for colour development with low concentrations of phenylephrine hydrochloride.The method was applied to cough linctuses, syrups, nasal drops and eye drops containing phenylephrine hydrochloride without prior separation. None of the ingredients included in the samples investigated (Table 11) showed any interference with the results. The proposed method was compared with the BPC 1968 method for phenylephrine eye drops.5 The results obtained were highly reproducible with a maximum error of -&2% compared with the moderately high results obtained by the official method (Table 111). The standard deviation of the method varied between 3 0 .5 and 1.2%. TABLE I11 RECOVERY OF PHENYLEPHRINE FROM PHARMACEUTICAL PREPARATIONS Sample Statedlmg Addedlmg I Phenylephrine eye drops . . 1 000 - 500 - I1 Tussivan syrup . . . . . . 5 I11 Toloxan nasal drops . . . . 50 - I V Rhinoline infant nasal drops . . 10 - 2.5 25 6 Recovery, % - 7 Proposed method* BPC method 100.3 (k0.7) 103 99.2 ( & 0.5) t 99.5 (k0.8) 108 98.4 ( & 0.5) t 100.6 (+0.8)? 101.1 ( k 0 . l ) 110 102 (f1.2) 112 99 (20.6)t * Mean of three determinations ; values in parentheses are standard deviations of individual results. 7 Recovery by the proposed method when a known amount of phenylephrine is added to the sample. Other phenolic compounds of pharmaceutical interest were also investigated. The red copper( 11) complex formed with phenylephrine under the specified experimental conditions did not develop with any of the compounds in Table IV, but unstable yellow, orange or brown colours having different absorption maxima were obtained. TABLE IV COLOURS PRODUCED BY REACTION WITH PHENOLIC COMPOUNDS Compound Adrenaline Isoprenaline Hexoprenaline Tetracycline Resorcinol Methyl salicylate Thymol Menthol Developed colour Yellow Yellow Yellow Brown Yellow Orange Brown precipitate Negative Absorption maxirnum/nm 420 410 410 Undefined 400 400 References 1. Baudisch, 8.’ Science, N.Y., 1940, 92, 336. 2. 3. 4. 5 . Konecny, J. O., J . Am. Chenz. SOG., 1955, 79, 5748. Bartos, J., Annls Pharm. Fr., 1969, 27, 759. Finar, I. L., “Organic Chemistry, The Fundamental Principles,” Third Edition, Volume I, Longmans, “British Pharmaceutical Codex, 1968,” The Pharmaceutical Press, London, 1968, p. 1056. Green and Co., London, 1959, p. 604. Received October Gth, 1975 Accepted January 20th, 1976

ISSN:0003-2654    

DOI:10.1039/AN9760100717

出版商:RSC    

年代:1976

数据来源: RSC

摘要:

720 Analyst, September, 1976, Vol. 101, $9. 720-727 Chromatographic Determination of Promethazine Hydrochloride in Aqueous Sollution B. J. Meakin, D. J. G. Davies, Norma Cox aind John Stevens Pharmaceutics Group, School of Pharmacy and Pharmacology, University of Bath, Bath, B A 2 7A Y Thin-layer and gas - liquid chromatographic techniques that are suitable for the determination of promethazine hydrochloride j n aqueous solutions containing breakdown products are described. In both assay processes the degraded drug solutions are directly applied to the cloiromatograph without pre-extraction. The validity of the assays was confirmed in a kinetic study, which showed that the first-order rate constants obtained for the thermal or photolytic degradation of promethazine hydrochloride were not significantly different.Promethazine hydrochloride is a member of the phenothiazine group of drugs that are widely used as antihistamines, hypnotics and tranquilisers. These drugs may be formulated as solid dose forrns or as aqueous-based medicines for oral, parenteral and topical administration. As with other phenothiazines, promethazine in aqueous solution is subject to thermal and photolytic degradation processes that are essentially oxidative in character, involving complex pathways that give rise to a number of degradation products.lS2 In order to assess the stability of promethazine in pharmaceutical formulations, it is necessary to establish assay procedures that are specific for the drug in the presence of breakdown products. Direct assays by standard ultraviolet or visible-light spectroscopy that have been used in stability studies3-7 fail to meet this criterion and it appears necessary to introduce a separation step prior to the assay of the parent drug.Although it is known that phenothiazine derivatives can be successfully separated from each other, from formulated medicines and from their breakdown products by both gas - liquid and thin-layer chromatography,lS2 there are no reports of these techniques being used quantitatively for high-stress accelerated stability sltudies, where the breakdown products constitute a major proportion of the system. This paper describes the comparative evaluation of thin-layer and gas - liquid chromatographic tecliniques, which are satisfactory for the determination of residual promethazine in both heat- and light-degraded solutions.Experiment:al Materials Promethazine hydrochloride BP was a gift from May and Baker Ltd. and was used as received. Orange I1 was recrystallised three times from 957(0V/V ethanol. All buffer salts were of analytical-reagent grade and other reagents viere of at least labor atory-reagent grade. Solvents were of analytical-reagent or spectroscopic grade. Water was freshly distilled from an all-glass still. The silica gel G (Merck) was extracted with diethyl ether for 24 h before use. Instruments Spectrophotometers. Pye Unicam SP500 Series 2 and SP1800 spectrophotometers were used with 1-cm cells. pH meters. Either a Pye Unicam 291 pH meter or a Radiometer Type 27 pH meter fitted with a pH A 630 P scale expander were used in conjunction with Pye-Ingold 405 combined glass - silver chloride electrodes.All pH measurements were made on solutions equilibrated to 25 & 0.1 "C; meters were standardised with two appropriate standard buffers.8 Gas - l i p i d chromatograph. A Pye Unicam Series 104 gas - liquid chromatograph with dual flame-ionisation detectors linked to a Vitatron UR406 linear integrating recorder was used. Injection directly into the column via a silicone rubber diaphragm was made using a Scientific Glass Engineering 10-pl syringe.MEAKIN, D-&VIES, COX AND STEVENS 721 Methods Thin-layer chromatographic assay Using the method of T a n ~ e y , ~ 5 pl of an aqueous solution containing up to 0.6% mlV of promethazine hydrochloride were spotted from an Agla micrometer syringe, fitted with a flat-ground needle, on to 0.25 mm thick silica gel G plates previously activated at 105-1 10 “C for 30 min.The plates were divided into areas and seven test spots applied. One of the central areas was retained for the blank. After drying the plates in air, development was carried out at room temperature, in the dark to prevent photodegradation, with a solvent mixture of acetone, methanol and ammonia (140 + 60 + 2). The running time for 10 cm was about 45 min and the XF value of the drug varied from 0.42 to 0.46 according to room tenipera- ture. Following development, the plates were dried in air in the dark for 1 h and the positions of tlie spots located by spraying 2-cni strips along each edge with Dragendorff’s reagent.Areas of silica gel (2.5 x 2.0 cm) containing the drug and the blank area were removed from the plates and transferred into glass-stoppered centrifuge tubes; 5 ml of aqueous 2.5 x 10-” M orange 11, adjusted to pH 2.0 with hydrochloric acid, and 4 ml of chloroform were then added and the tubes shaken for 1 h in the dark. They were then centrifuged at 4 000 rev min-l for 10 min and the absorbance of the aqueous layer was measured a t 486 nm. Three replicate determinations were carried out at each drug. level and the mean absorbance was subtracted from that of the blanks to give the decrease in absorbance of the aqueous layer. This value was used to calculate the concentration of promethazine hydrochloride in the sample using equation (l), which was obtained by linear regression analysis of replicate calibration plots: A =0.5585Y .... .. .. * (1) where -4 is the difference between tlie absorbance of blank and test solutions, and I’ is the percentage concentration of promethazine hydrochloride. Gas - liquid cltronaatogra$4ic assay Preparation of columns. A 30-g amount of Chroniosorb V? was silanised by reactim with 100 in1 of a 5% w/V solution of dimethyldiclilorosilane in toluene; the support was filtered off, washed with toluene, then methanol, and dried at 80 “C. A slurry of this silanisecl support was prepared in sufficient of a 5y0 nzlV solution of SE-30 in toluene to give a 576 m/V loading, evaporated to dryness on a rotary evaporator and finally oven dried at $0 “C. Silanisd glass columns (5 ft x Q in i.d.) were packed with this support and aged by heating at 250 “C for 48 11 with a flow-rate of argon carrier gas of 15 rnl min-l.Operating conditions. Column temperature, 220 “C; carrier gas, argon, flow-rate 45 rnl n1in-I ; flame gases, hydrogen 20 lb in-2, air 20 lb in-2; attenuator range, 2 000; recorder settings, selector 1, range 6, chart speed 2 cm min-l. Gas - liquid chromafog~a$hic assay techniqzce. Suitable dilutions of aqueous promethazine hydrochloride were prepared such that 5-rnl samples added to 4 nil of a 0.257, nzlV solution of clibutyl phthalate in absolute ethanol and adjusted to 10 ml with the latter gave solutions containing between 0.09 and 0.30% m/V drug and 0.1% wz/V internal standard in 50% VjV aqueous ethanol. Amounts of 3 pl of these solutions were injected into the column and the peak area ratios (R) of drug to internal standard determined in triplicate.The drug concen- tration was calculated from the mean value of K using equation (2) and the appropriate column constants (a and b, Table 111) determined by linear regression analysis of replicate calibration plots : R = b P - - n .. . . m . .. . . (2) Thermal and Photolytic Degradation of Promethazine Solutions The detailed procedure used for degradation studies is described elsewhere.lS2 For com- paring the two assay techniques, aqueous solutions containing either 0.3 or 0.5% nzlV of promethazine hydrochloride and 0.1% mlV of EDTA (disodmm salt) were prepared in Sorensen’s citrate buffer a t a pH of 4.0 and adjusted to an ionic strength of 0.5 M with potassium chloride.Immediately prior to degradation, the solutions were equilibrated with oxygen and an oxygen flow-rate of 5 or 10 ml min-l was maintained via a sintered-glass diffuser throughout the degradation experiments. Thermal degradation wits effected by722 MEAKIN et d. : CHROMATOGRAPHIC DETERMINATION OF Analyst, VOl. 101 heating the solution in a blackened vessel in a water-bath at 90 & 0.1 "C. Photolytic degrada- tion was carried out in a fan-cooled light box fitted with Atlas 15-W Northlight Colour- matching fluorescent tubes. The temperature of the irradiated solutions was 28 1.0 "C. Samples were withdrawn at appropriate time intervals and assayed by the thin-layer and gas - liquid chromatographic techniques described above. Replicate experiments were carried out for both degradation procedures and the results are shown in Tables IV and V.Results and Discussion Thin-layer Chromatographic Assay Preliminary experiments showed that a satisfactory solvent system for separating the drug from both heat- and light-induced breakdown products consisted of acetone, methanol and ammonia (140 + 60 + 2). Using the drug forinulation described above, five breakdown products resulted from thermal degradation and six from photolytic degradation. These breakdown products included N-alkylphenothiazines, phenothiazine dimers, sulphoxides and other oxidation products that were identified by mass spectrometry. Changes in stress conditions or drug formulation induced changes in the number and nature of these compounds and their detailed structure together with proposed degradation pathways will be published subsequently.Solvent extraction and measurement of the absorbance at 250 nm of the separated drug spots proved unsatisfactory owing to unacceptably large variations in the blanks. The blank variation could be reduced by filtration through sintered-glass or Millipore membranes but this procedure raised an additional problem, as significant adsorption of the drug on to the negatively charged filter media occurred. The problem of high ultraviolet blanks arising from trace impurities in solvents and ad- sorbents can often be obviated by forming coloured drug complexes, which may also give a marked improvement in assay sensitivity. However, unlike many of the other pheno- thiazine drugs, promethazine is not a good subject in this respect and none of the colour reactions that have been proposed for its identification and determination'J0-12 proved sufficiently sensitive for use with the material eluted from the plates.Attention was therefore turned to the ion-pair complexation technique, w:hich has been used for the determination of prometha~ine.1~9~~ Preliminary studies indicated that the ion-pair interaction of prome- thazine with methyl orange, bromothymol blue or bromophenol blue was insufficiently sensi- tive for the amount of drug applied to the plates but that the interaction with orange I1 was adequate. The formation and extraction of orange I1 ion-pairs into chloroform has been shown by several workers to be pH dependent,15-17 maximum extraction occurring at low pH.For promethazine, extraction of the complex was found to reach a maximum below pH 3.0 and a pH of 2.0 was therefore chosen for assay puirposes. An extraction time of 60 min was chosen for the assay following initial work that showed that equilibrium was achieved after 30 min. Extraction was carried out in the dark so as to prevent fading of the colour of the dye. In the ion-pair technique, it is possible to measure either the increase in absorbance of the organic layer or the decrease in absorbance from the aqueous layer. The latter procedure was chosen in this instance as it was necessary to centrifuge the extract in order to effect separation of the two liquid layers from silica gel particles, and the aqueous layer could be sampled directly from the centrifuge tubes.The validity of using the mean plate blank from any particular run in calculating drug concentration was checked by taking five blank areas from each of five plates after solvent development. The over-all coefficient of variation for the 25 blanks was 1.27% and an analysis of variance showed no significant difference between and within plates (Fo.05 = 1.42; F t a b = 5.75; fi = 0.05). The linearity and repro- ducibility of the assay process over the concentration range 0.1-0.6% m/V (5-30 pg of drug) were assessed by carrying out replicate calibration plots. Linear regression analysis gave the data shown in Table I. Comparison of the slopes and intercepts by a t-test showed that they were not significantly different.Consideration of the intercept data suggests that the calibration plots can be considered to pass through the origin and the drug concentration was therefore calculated from the combined regression slope of 0.558 5 using equation (1). The precision of the assay was evaluated for a 0.4% m/V solution of drug by carrying out six replicate thin-layer chromatographic determinations on three separate occasions, giving recoveries of 102.1, 103.7September, 1976 PROMETHAZINE HYDROCHLORIDE IN AQUEOUS SOLUTION 723 and 101.9yo, the corresponding coefficients of variation being 3.88, 3.60 and 3.95%, which were considered to be satisfactory. TABLE I REGRESSION DATA FROM CALIBRATION PLOTS FOR THE THIN-LAYER CHROMATOGRAPHIC ASSAY OF PROMETHAZINE HYDROCHLORIDE S tanuard Standard deviation of deviation of Correlation Determination Slope slope I iitercep t intercept coefficient I 0.55G 8 0.026 5 -0.001 2 0.009 6 0.997 I1 0.560 3 0.010 1 -0.008 5 0.003 7 0.999 Combined regression 0.558 5 0.013 8 -0.004 8 0.005 0 0.998 tslope = 0.12; tintercept ;= 0.71; ttab = 2.45;F = 0.05; 111 = 7211 = 5.Gas - Liquid Chromatographic Assay Preliminary experiments showed a satisfactory combination of drug, internal standard and co-solvent to be 0.09-0.30y0 m/V, 0.1% wz/V and 500/: VjV, respectivcly. Fig. 1 show a typical chromatogram with retention times of 3.5 min for dibutyl phthalate and 9.0 rnin for promethazine hydrochloride. Table I1 shows the reproducibility of the peak-arcla ratios for different drug concentrations based on five replicates.The coefficients of variation at con- centrations greater than 0.090/, were considered to be satisfactory but below this concentration they wcrc large enough to affect the assay results significantly. Drug concentrations ; i h o \ ~ 0.3;/, were suitably diluted to prevent alterations in attcnuator and recorder settings. 4 , ! , I 10 9 8 7 6 5 4 3 2 1 0 Ti me/m i 11 Fig. 1. Gas -liquid chromatogram of a solution containing A, promethazine hydrochloride and B, dibutyl phthalate; C is the solvent peak. Conditions are as described in text. As these assays were aimed at the determination of residual promethazine in pharmaceutical formulations subjected to high stress, it was necessary to determine the effect of some conimon stabilising agents, such as buffers and antioxidants, on the gas - liquid chromatographic TABLE I1 REPRODUCIBILITY OF PEAK-AREA RATIOS FOR THE GAS - LIQUID CHROMATOGRAPHIC ASSAY OF PROMETHAZINB HYDROCHLORIDE Promethazine hydrochloride &Tea11 peak-area Standard Coefficient concentration, yo m/ V ratio (H) deviation of R of variation, "/o 0.24 2.266 0.010 0.44 0.12 1.101 0.026 2.37 0.09 0 870 0.022 2.57 0.02 0.137 0.010 7.30724 MEAKIN et aZ.: CHROMATOGRAPHIC DETERMINATION OF Autalyst, VoZ. 101 calibration graphs. Calibration graphs involving various vehicles were theref ore plotted and the linear regression data given in Table I11 for column A showed that these vehicles had no significant effect on the regression constants. In addition, it was noted that after repeated use, a brown deposit was formed at the injection site, presumably due to breakdown of such materials as citrate buffer.Although this deposit ;dso appeared not to influence the column calibration equations significantly (Table 111, columns A and B), it was felt advisable to empty and re-pack the initial 10 cm and re-calibrate the column when the deposit became obvious. TABLE IIlE COLUMN CALIBRATION CONSTANTS FOR THE GAS - LIQUID CHROMATOGRAPHIC ASSAY OF PROMETHAZINE HYDROCHLORIDE Number of points on Column Drug calibration Slope Column condition vehicle A A A A A B B B B C C Freshly packed Water Freshly packed Citrate buffer, pH 4.0 Freshly packed Citrate buffer pH 4.0; O.;% EDTA: 3 months old; Citrate buffer, visible deposits pH 4.0; O.lO;, EDTA 3 months old; Citrate buffer, top 10 ern re- packed EDTA pH 4.0; 0.1% Freshly packed Water 7 months old.Water 7 months old; Water visible depdsit top 10 cm re- packed top 10 cm re- packed 7 months old; Water Freshly packed Water Freshly packed Water 10 9.790 10 9.i78 10 9.714 10 9.76 4 9 8.907 8 8.822 6 0.550 6 9.394 11 9.715 10 9.876 Standard deviation of slope 0.087 0.134 0.081 0,085 0.oeo 0.071 0.155 0.163 0.2512 0.199 0.185 Intercept (4 - 0.079 - 0.082 - 0.078 - 0.071 -0.077 -0.052 -0.082 - 0.047 - 0.036 - 0.047 -0.033 Standard Derived deviation Correlation statistical of intercept coefficient parameters (p=O.O5) 0.013 0.021 0.016 0.017 0.016 0.011 0.022 0.025 0.039 0.028 0.027 0.999 6 x~~~~~ = 0.52 0.999 2 x~ntese,t= 0.15 0.999 9 XLb = 9.49 0.999 8 - 0.999 7 - 0.999 8 tslope = 0.60 0.999 1 tintercept = 1.22 ttab = 2.16 0.999 4 t,lope = 0.52 0.998 1 tslope - 0.59 0.998 6 tiotemept = 0.36 ttab = 2.11 The individual columns A, B and C were prepared, packed and used by different operators, which accounts for the slight variations in regression constants and their precision.The mean regression constants relevant to a particular column were therefore used in equation (2) for calculating residual drug concentration. Table I11 shows that for the most precise work (column A), the negative intercept was highly significant and consequently intercept values were used in all gas - liquid chromatographic determinations, even when their use was statis- tically doubtful. The negative intercept that cornmonly occurs in column calibrations can be attributed to such factors as degradation on the column or the presence of a small number of irreversible binding sites.Slight variations in support preparation and column packing would alter the latter and could account in part for the differences in intercepts. This effect highlights the necessity for calibrating every column individually and the danger of using a single standard equation for the assay of a drug on nominally identical columns. Such differences could account for the three unsatisfactory results reported in an inter-laboratory evaluation of a gas - liquid chromatographic assay for chloramphenicoP8 using standardised columns and a single-point column calibration. In order to test the validity of the gas - liquid chromatographic assay for the drug in the presence of its breakdown products, a 0.5% m/V solution of promethazine hydrochloride was thermally degraded in the dark.The five breakdown products were separated by thin- layer chromatography, extracted with chloroform. and the solutions evaporated to dryness at room temperature. The residues were separately dissolved in 50% V/V ethanol and injected into the column under standard conditions, after increasing the instrumental sensi-Septenzber, 1976 PROMETHAZINE HYDROCHLORIDE IN AQUEOUS SOLUTION 725 tivity %-fold. The result is shown in Fig. 2, which indicates that these breakdown products would be unlikely to interfere with the drug and internal standard peaks. This conclusion was checked by applying 10 ml of a 0.5% m/V drug solution, thermally degraded to below 30% residual concentration, to thin-layer chromatographic plates and, after development, separation and extraction, bulking the breakdown products and evaporating them to dryness ; to the residue were added 5 ml of a 0.37; mjV solution of promethazine hydrochloride in water and 4ml of a 0.25% m/V solution of dibutyl phthalate in ethanol.On dilution to 10 ml with ethanol, this approximated to the assay situation when a l”/d m/V drug solution was degraded to about 30% residual concentration. A control without the breakdown products was prepared and each solution assayed five times, giving mean R values and standard errors of 1.52 & 0.039 and 1.46 & 0.027 for the degraded mixture and control, respectively, which values were not significantly different (to.o5 = 1.26; ttab = 2.31; 9 = 0.05).C B ii !I 1.2 Time/rnin Fig. 2. Position on the gas - liquid chromato- gram of the five breakdown products (peaks 1-5) from the thermal degradation of promethazine hydrochloride. A, Promethazine hydrochloride ; B, dibutyl phthalate; and C, solvent peak. (Peaks A and B are not to scale.) A similar evaluation with the photolytically induced breakdown products proved more difficult. Photolytic degradation of promethazine is a much slower process, yielding relatively small amounts of breakdown products and only three of the six photolytic products gave visible peaks at a sensitivity that was not seriously affected by “noise” or trace amounts of impurity from the thin-layer chromatographic procedure. An alternative kinetic check T.4BLE Iv THERMAL DEGRADATION OF PROMETHAZINE HYDROCHLORIDE Ccmctmtrations of promethazine hydrochloride in citrate buffer, pH 4.0, containing 0.1 yo EDTA, ionic strength 0 .A M, degraded at 90 “C. Time/h 0.00 1.00 2.00 3.00 4.75 5.50 6.25 7.00 8.00 0.00 1.00 2.00 4.00 5.00 6.00 8.00 Concentration found, % $7.21 V (----.--p---h-----l___l 7 0.311 0.275 0.260 0.253 0.228 0.240 0.198 0.188 0.162 0.165 0.146 0.151 0.129 0.122 0.122 0.119 0.10s 0.103 Gas - liquid chromatography Thin-layer chromatography 0.308 0.263 0.217 0.181 0.157 0.134 0.110 0.296 0.267 0.254 0.167 0.156 0.140 0.101726 MEAKIN et al. : CHROMATOGRAPHIC DETERMINATION OF Analyst, VoZ. 101 was therefore adopted, which involved assaying promethazine solutions subjected separately to light and thermal stress by thin-layer and gas ..liquid chromatographic techniques simul- taneously. The results for replicate kinetic experiments are shown in Tables IV and V. TABLE V PHOTOLYTIC DEGRADATION OF PROMETHAZINE HYDROCHLORIDE Concentrations of promethazine hydrochloride in citrate buffer, pH 4.0, containing 0.1 % EDTA, ionic strength 0.5 M. Concentration found, % m/ V Time/h Gas - 0.00 23.58 44.58 68.83 92.00 116.00 142.75 165.16 189.08 0.00 22.08 48.42 69.17 93.12 117.16 liquid chromatography 0.481 0.442 0.416 0.383 0.362 0.357 0.332 0.304 0.266 0.510 0.479 0.444 0.423 0.404 0.375 1 Thin-layer chromatography 0.466 0.443 0.409 0.370 0.351 0.342 0.32 1 0.291 0.275 0.504 0.485 0.441 0.410 0.405 0.377 Figs. 3 and 4 show that first-order rate plots from these data are linear, leading to the rate constants shown in Table VI.Comparison of both the raw data and the derived rate constants shows that both assay techniques give similar results, which is evidence that neither heat- nor light-induced breakdown products interfere significantly with the gas - liquid chromato- graphic assay of promethazine, at least to the degradation levels studied. 1.5 cv + \ - 1.4 4 F &? 6 1.3 0 .- CI L CI a 2 1.2 8 0" 1.1 m rn 2 - J 0 1 2 3 4 5 6, 7 8 0 40 80 120 160 200 Time/h Time/h I .o Fig. 3. First-order rate plots for the Fig. 4. First-order rate plots for the photolytic degradation of promethazine hydrochloride in citrate buffer solution, pH 4.0, ionic strength 0.5 M, containing 0.1% m/ V EDTA: symbols as given for Fig. 3. thermal degradation a t 90 "C of promethazine hydrochloride in citrate buffer solution, pH 4.0, ionic strength 0.5 M, containing 0.1 yo m/ I, EDTA: 0 and A, gas - liquid chromato- graphic assay; a and A, thin-layer chromato- graphic assay.Stp&%VZbCY, 1976 PRObIETHAZINE HYDXOCHLORIDE; I N AQUEOVTS SOLWTIO;?; TABLE VI DEGRADATION OF PROMETHAZINE HYDROCHLORIDE 72 7 First-order rate constants for the thermal (90 T) and photolytic degradation of promethazine hydrochloride in citrate buffer, pflI 4.0, containing 0.1% EDTA, ionic strength 0.5 M.Experi- Assay 1 ment Stress method A Heat GLC Heat TLC I3 Heat GLC Heat TLC C* Light GLC Light TLC D* Light GLC Light TLC First-order :ate constant (h) is-‘ 3.65 x 10-5 3.50 x 10-5 3.79 x 10-5 7-82 x 10-7 7.66 x 10-7 7.08 x 10-7 6.98 x 10-7 3.54 x 10-6 Standard deviation CorreIa tiori of 12 n coefficient t ( p = 0.05) x2 fP = 0.05) 0.996 to.,, 0.99 7.85 x 10-7 9 0,998 0.74 1.80 x 10-6 9 0.991 ttab 2.15 1.45 x 10-6 7 4.77 x 10-8 9 0.987 I,.,, 0.28 2.46 x 10-8 6 0.998 to.o5 0.16 2.10 x 10-6 7 0.993 ttab 2.23 3.27 x 9 0.994 ttab 2.15 6.00 x 10-8 6 0,986 ttab 2.31 * Experiments C and D were carried out a t slightly different light intensities and could not, therefore, be compared by the Barlett test.It can therefore be concluded that the gas - liquid chromatographic process described pro- vides a direct and easily manipulated assay for stability studies on aqueous formulations of prornethazine, which does not require prior extraction by an organic solvent. The thin- layer chromatographic technique, although more tedious, would be useful for the assay of the drug in situations such as those found in some hospital laboratories, where more sophisticated equipment is not readily available. This work was supported by studentships awarded by the SRC (J.S.) and the Pharma- ceutical Society of Great Britain (N.C.). 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. References Stevens, J., Ph.D. Thesis, University of Bath, 1973. Cox, IS. I., M.Sc. Thesis, University of Bath, 1975. Maghoub, A., M,Phil. Thesis, University of London, 1968. Felemeister, A., and Discher, C. A., J . Pharm. Sci., 1964, 53, 756. Felemeister, A., Schaubman, R., and I-Iowe, H., J . Phavm. ScZ‘., 1965, 54, 1589. Yamamoto, R., and Fujisawa, S., Yakugaku Zasslii, 1962, 82, 1396. Ryan, J. A., .J. Am. Pharm. Ass. Scient. Edn, 1959, 48, 240. Bates, R. G., “Determination of pH,” Second Edition, John Wiley & Sons, New Ytrrk and London, Tansey, I. P., ,W.Sc. Thesis, University of Bath, 1960. Rusiecki, W., and Henneberg, M., Acta Pol. Pharwz., 1964, 21, 23. Bulenkov, T. I., Medskaya Prom. S.S.S.R., 1973, 17, 26. Albert, F. M., Camboli, D., and Cimpu, V., Rul. I m t . Politeh. “Glzecarghe Gheorghui-Dej,” Buc., 1971, Nyberg, Id., J . Phavm. Pharmac., 1970, 22, 500. Matsui, F., and French, W. N., J . Pharm. Sci., 1971, 60, 287. Few, A. V.. and Ottewill, R. H., J . Colloid Sci., 1956, 11, 34. Zographi, G., Patel, P. R., and Weiner, N. D., J . Pharm. Sci., 1964, 53, 544. Scott, G. V., Analyt. Chem., 1968, 40, 768. Margosis, M., .J. Phavn?. Sci., 1974, 63, 435. 1973. 33, 41. Received Februaqi 23vd, 1976 Accepted April 14th, 1976

ISSN:0003-2654    

DOI:10.1039/AN9760100720

出版商:RSC    

年代:1976

数据来源: RSC