摘要:

THE ANALYSTTHE ANALYTICAL JOURNAL OF THE CHEMICAL SOCIETYEDITORIAL ADVISORY BOARD*Chairman: H. J. Cluley (Wembley)'L. S. Bark (Salford)R. Belc her (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. Doc herty (Billingham)D. Dyrssen (Sweden)'W. T. Elwell (Birmingham)J. Hoste (Belgium)'J. A. Hunter (Edinburgh)H. M. N. H. Irving (Leeds)M. T. Kelley (U.S.A.)W. Kemula (Poland)'G. F. Kirkbright (London)G . W. C. Milner (Harwell)G. H. Morrison (U.S.A.)H. W. Nurnberg (W. Germany)"J. M. Ottaway (Glasgow)"G. E. Penketh (Wilton)'T. B. Pierce (Harwell)E , P u ng or (Hungary)D. 1. 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. Watsh (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 CommitteeREG I0 NAL ADVl SORY ED I T 0 RSDr. J. Aggett. Department of Chemistry, Univlersity of Auckland, Private Bag, Auckland, NEWDr. G. Ghersini, Laboratori CISE, Casella Postale :3986, 20100 Milano, ITALY.Professor L. Gierst, Universitd Libre de Bruxelle:;, Facult6 des Sciences, Avenue F.-D. Roosevelt 50,Professor R. Herrmann. Abteilung fur Med. Physik., 63 Giessen, Schtangentahl 29, W. GERMANY.Professor W. A. E. 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. I. Rubeska, Geological Survey of Czechoslovakia, Kostelni 26, Praha 7, CZECHOSLOVAKIA.Dr. J, Rtkicka, 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.ZEALAN D.B ruxel les. BE LG I U M.CANADA.Mich. 481 06, U.S.A.Published by The Chemical SocietyEditorial: The Director of Publications, The Chemical Society, Burlington House,London, W1V OBN. Telephons 01 -734 9864. Telex No. 268001Advertisements: Advertisement Department. The Chemical Society, Burlington House, Piccadilly,London, WIV OBN. Telephone 01 -734 9864Subscriptions (non-members): The Chemical Society, Distribution Centre, Blackhorse Road,Letchworth, Herts., SG6 1 HNVolume 103 No 1226@ The Chemical Society 1978May 197

ISSN:0003-2654    

DOI:10.1039/AN97803FX017

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

ANALAO 103 (I 226) 41 7-528 (1 978)ISSN 0003-2654May 1978THE ANALYSTTHE ANALYTICAL JOURNAL OF THE CHEMICAL SOCIETYCONTENTS41 743844745246947548249249750951 3521525528REVIEW. Simultaneous Techniques in Thermal Analysis-F. Paulik and J.PaulikSolid-state Mercury(1) Chloride EIlectrode for Determining 0.1-1 .O pg mi-'Levels of Chloride i n Boiler Water and Other High-purity Waters-G. B.Marshall and D. MidgleyInfluence o f Ascorbic Acid on the Matrix Interferences Observed During theCarbon Furnace Atomic-absorption Spectrophotometric Determina :ionof Lead in Some Drinking Waters-J. G. T. Regan and J. WarrenQualification o f Estimates for Total Trace Elements in Foodstuffs UsingMeasurement by Atomic-absorption Spectrophotometry-W. H.EvansAnalysis of Metals Using a Glow-discharge Source with a Fluorescent AtomicVapour as Spectral-line lsolator-H. G. C. Human, N. P. Ferreira, R. A. Krugerand L. 3. P. ButlerRadiochemical Neutron-activation Analysis of Sulphide Ores Using ZincDiethyldithiocarbamate as Extraction Reagent-E. Pernicka, P. A. Schubigerand 0. MutterPyrolysis - Mass Spectrometry of Textile Fibres-J. C. Hughes, B. B. Whl:als andM. J. WhitehouseDetermination o f Probenecid in Serum by High-performance Liquid Ch omato-graphy-R. K. Harle and T. CowenPolarographic Method for the Identification of 1,4-Benzodiazep nes-W.Franklin Smyth, M. R. Smyth, J. A. [Groves and S. B. TanREPORTS BY THE ANALYTICAL METHODS COMMITTEESpectrophotometric Determination of Ronidazole in Animal FeedsIdentification of Prophylactic and Growth-promoting Drugs in Animal Feeding-stuffsGeneral Method for the Determination of Iron with 4,7-Diphenyl-I ,IO-pken-anthroline (Bathophenanthroline)Book ReviewsErrataSummaries of Papers in this fssue-Pages iv, viii, ix, x iPrinted by Heffers Printers Ltd Cambridge EnglandEntered as Second Class at New York, USA, Post OfficMay) 1978 SUMMARIES OF PAPERS IN THIS ISSUEPolarographic Method for the Identification of 1, 4-BenzodiazepinesThe polarographic behaviour of 12 therapeutically important 1,4-benzo-diazepines in Britton - Robinson universal buffers, pH 4.0 and pH 12.0,has been investigated.Differences in the polarographic peak potentials ofthese compounds in these niedia are explained. The rates of hydrolysis ofcertain benzodiazepines in acidic solution were investigated.Brornazepamand flunitrazepam, both of which possess a strongly electron-withdrawingsubstituent on the 5-o-phenyl group, were found to undergo rapid acidhydrolysis. On the basis of these findings, and taking into account theextraction profile of some of the compounds over a pH range, a scheme isdevised for the identification of any one or tnorc of 12 1,4-benzodiazepines.It is suggested that this procedure would be applicable to the analysis ofunknown formulations or body fluids in forensic cases where the parentcompound exists in relatively high concentrations compared with its meta-bolites.xiKeywovds : 1,4-Benzodiazepine identification ; polnrogvapbiyW. FRANKLIN SMYTH, M.R. SMYTH, J. A. GROVES and S. B. TANDepartment of Chemistry, Chelsea College, University of London, nlanresa Road,London, SW3 6LX.Analyst, 1978, 103, 497-508.Spectrophotometric Determination of Ronidazole in Animal FeedsReport prepared by the Medicinal Additives in Animal Feeds Sub-Committee“A.”Keywords : Ronidazole determistation ; animal feeds ; .spect?jophotowaetryANALYTICAL METHODS COMMITTEEThe Chemical Society, Ihrlington House, London, W1V OBX.Analyst, 1978, 103, 509-512.Identification of Prophylactic and Growth-promoting Drugs inAnimal FeedingstuffsReport prepared by the Medicinal Additives in Animal Feeds Sub-Committee“B.”Keywords : PvoPJzylactic drugs ; growth-promoting drugs ; feedingstufls analysis ;tba in-layer c JwomatogvaphyANALYTICAL METHODS COMMITTEEThe Chemical Society, Burlington House, London, W1V OBN.Analyst, 1978, 103, 513-520.General Method for the Determination of Iron with4,7- Diphenyl- 1,lO-phenanthroline (Bathophenanthroline)Report prepared by the Iron Sub-Committee.Keywords : Iron detevmination ; 4,7-diphenyl-l ,lO-phenathroZine ; batlzo-phenanthroline ; spectrophotometvyANALYTICAL METHODS COMMITTEEThe Chemical Society, Burlington House, London, W1V OBN.Analyst, 1978, 103, 521-524

ISSN:0003-2654    

DOI:10.1039/AN97803BX019

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

... May, 1978 THE ANALYST 111Titles from Ellis Horwood Ltd., Chichester, distributed b.y John Wiley & Sons LimitedVol. 1: Column Backings, GPC, GF and Gradient Elution and Vol. 2: Hydrophobic, Ion Exchangeand Affinity Methods, edited and compiled by R. Epton, Department of Physical Sciences, ThePolytechnic, Wolveuhampton.The first volume deals in parallel with the molecular sieve chromatography of synthetic polymers(GPC) and of biological polymers (GF). This dichotomy of treatment integrates a broad surveyof macromolecular fractionation which introduces recent advances, innovative technology, up-to-date industrial practice, and applications both preparative and analytical.The second volume describes important recent advances in the chromatography of biologicalmolecules.The contributions are based on lectures presented at an international symposiumorganized by the Macromolecular Group of the British Chemical Society. Publication was re-garded as a primary organisational objective of this meeting and authors were able to cooperatefully in producing a balanced text. The end result is a synthesis of many intellects, ideas andviewpoints.085312 069 2 378 pages March 1978 &18.00/$38.15 (vol. 1)085312 073 0 approx. 360 pages due April 1978 approx. &18.50~$37.00 (vol. 2)by J. GaspariE, Charles University, Hiradec KraIovk, Czechoslovakia, and J. ChurSCek, TechnicalUniversity, Pardubice, Czechoslovakia; (Translation Editor : R. A. Chalmers, University of Aberdeen)This handbook on the uses of thin-layer and paper chromatography shows briefly the theoreticalprinciples? but is mainly concerned with thc practical applications to organic compounds.Des-criptions of commercially available laboratory equipment introduce experimental proceduresand uses of the techniques in qualitative and quantitative analysis.(Ellis Horwood Series in Analytical Chemistry: Editors, Dr. R. A. Chalmers and Dr. M. Masson,University of Aberdeen).085312 041 7 362 pages April 1973 &1 8.00/$38.15Chromatography of Synthetic and Biological PolymersLaboratory Handbook of Paper and Thin Layer ChromatographyMetal Ions in Solutionby J. Burgess, Department of Chemistry, University of LeicesterThis important and needed treatise unifies the numerous varied aspects of the chemistry of metalions in solution found across the dual areas of inorganic and physical chemistry.It deals withmetal cations from all areas of the Periodic Table, from alkali and transition metal ions to lan-thanide and actinide ions, andt o less familiar species such as the transient T12+ aq and the newlycharacterized Pt2+ aq cations. Aqueous and non-aqueous media are covered, with discussion alsoof binary aqueous solvent systems.085312 027 7 approx. 560 pages In Press approx. &25.00/$47.50New Processes of Waste Water Treatment and Recoveryedited by G. Mattock, Director, Bostock Hill and Rigby Ltd., Birmingham, England.This book offers a planned coverage of recent advances in waste water technology, every chapterbeing a contribution of original research by authorities from the UK, USA, Germany, SouthAfrica, Canada, The Netherlands and France.It is thus representative of most recent thinkingin water pollution control, and conservation of protein and water resources.085312 096 X approx. 500 pages In Press approx. &25.00/ $50.00Available from all good booksellers or from Wiley. I f you wish to use American Express, Diners ClubBarclaycard or Access, please quote your card and numberiv SUMMARIES OF PAPERS I N THIS ISSUE May, 1978Summaries of Papers in this IssueSimultaneous Techniques in Thermal AnalysisA ReviewSummary of ContentsIntroductionDevelopment of simultaneous niethodsDifficulties of interpretationDerivative methodsStandardisation of experimental conditionsEGA and EGDExamination of decomposition of organic compoundsTD and ETAEC methodsManipulation of experimental conditionsConclusionTG - DTAKeywords : Review ; thermal anaiysis ; derivative and diflerential waeth,ods ;simultaneous methodsF. PAULIK and J.PAULIKInstitute for General and Analytical Chemistry, Technical University, Gellert ter 4,1502 Budapest XI, Hungary.Analyst, 1978, 103, 417437.Solid -state Mercury(1) Chloride Electrode for Determining0.1-1.0 pug ml-1 Levels of Chloride in Boiler Water and OtherHigh-purity WatersA solid-state electrode based on mercury(1) chloride and mercury(I1) sulphidehas been developed for determining chloride concentrations of 0.01-1 .O pg 1-1in boiler water. The greater sensitivity of the electrode compared with thatof silver - silver chloride electrodes enables concentrations as low as 0.01pg ml-l to be determined by a simple manual technique. The total standarddeviations at chloride concentrations of 0.5, 0.1 and 0.01 pg ml-l were 0.025,0.005 and 0.015 pgml-l, respectively. The electrode can be preparedeasily in the laboratory from comme:rcially available materials and a RBiiEkaSelectrode. The only significant interference is from iron(II1) ions and thisinterference can be eliminated by adding fluoride ions to the sample.Keywords : Chloride determination ; wuter analysis ; potentiometry ; chloride-selective electrode ; mercury ( I ) chloride electrodeG. B. MARSHALL and D. MIDGLEKCentral Electricity Research Laboratories, Kelvin Avenue, Leatherhead, Surrey,KT22 7SE.Analyst, 1978, 103, 438-446

ISSN:0003-2654    

DOI:10.1039/AN97803FP037

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

...VIIl THE ANALYST -%fay, 1978Influence of Ascorbic Acid on the Matrix Interferences ObservedDuring the Carbon Furnace Atomic-absorption SpectrophotometricDetermination of Lead in Some Drinking WatersNine samples of drinking water taken from a range of locations in Englandand Scotland have been analysed for lead by using carbon furnace atomic-absorption spectrophotometry. Spiking experiments have been carried outin order to determine the severity of the matrix interference. The suppressionof the lead signals ranged from 22 to 84%. No relationship was found toexist between the hardness of a water sample and its suppression effect.Further spiking experiments carried out in the presence of lo/, m/V ofascorbic acid showed that the suppression effect of eight of the water sampleswas reduced to a level of less than 5'x.The remaining water sample gavea suppression of 18%. This water was not the hardest examined, nor did itgive the highest suppression in the previous experiment.The natural lead contents of the nine waters were determined both bycarbon furnace atomic-absorption spectrophotometry in the presence ofascorbic acid and by a method that involves solvent extraction - ffarne atomicabsorption. Statistical analysis, using a t-test, indicated that there was nosignificant difference (at the 95% confidence level) in the results obtained byusing the two techniques.Keywords : Lead determination ; drinking water; cavbo.12 fuvnace atomic-absorfition spectrophotometry ; matrix interferences ; ascorbic acidJ.G. T. REGAN and:J. WARRENDepartment of Industry, Laboratory of the Government Chemist, Cornwall House,Stamford Street, London, SE1 ONQ.Analyst, 1978, 103, 447-451.Qualification of Estimates for Total Trace Elements in FoodstuffsUsing Measurement by Atomic -absorption SpectrophotometryThe qualification of results for small concentrations of elements in focdstuffsimplies a knowledge of the accuracy of a method when applied to focdstuffsand an assessment of the variation in results that exists in the applicationof that method.An attempt is made to describe the problems inherent in obtaining suchqualifications, and to suggest a standard procedure for accomplishing theseaims. From data obtained for a particdar method, a statistical appreciationwill give confidence limits and detecticln limits that can be applied to subse-quent results obtained, depending upon the nature of the exercise involved.Keywords : Accuracy and variation c f resulfs ; limit of detection ; statisticalappreciation ; foodstuffs analysis ; atomic-absorption spectvophotoynetryW. H.EVANSDepartment of Industry, Laboratory of the Government Chemist, Cornwall House,Stamford Street, London, SEl 9NQ.Analyst, 1978, 103, 452-468.Analysis of Metals Using a Glow-discharge Source with aFluorescent Atomic Vapoui- as Spectral-line IsolatorA compact and rugged spectral-line isolator based on atomic fluorescencefrom atoms generated by a low-pressure gas discharge has been constructed.The device is bolted to a standard glow-discharge source.A pulsed electricalcurrent generates the atoms and the fluorescence is measured by ineans of agated integrator. Several types of metals have been analysed, e.g., steel,cast iron, aluminium and gold. Good precisions and accuracies have beenobtained.H. G. C. HUMAN, N. P. FERREIRA, R. A. KRUGER and L. R. P. BUTLERNational Physical Research Laboratory, CSTR. P.O. Box 395, Pretoria 0001,Republic of South Africa.Analyst, 1978, 103, 469-47-2.Keywords : MetaE analysis ; atomic-JEuorescence spectroynetvk f a y , 1978 SUMMARIES OF PAPERS IN THIS ISSUERadiochemical Neutron - activation Analysis of Sulphide Ores UsingZinc Diethyldithiocarbamate as Extraction ReagentA procedure for the analysis of lead sulphide and mixed sulphide ores forsilver, arsenic, gold, cadmium, copper, manganese, antimony and zinc wasdeveloped with emphasis on the determination of the low gold to silverand arsenic to antimony ratios.Radiochemical neutron-activation analysiswas necessary and a solvent-extraction technique has been developed. Inthe first separation step arsenic(II1) chloride was extracted from the oresolution with benzene. The results are compared with the values obtainedafter separation of arsenic by distillation.Gold(III), silver(I), copper(II), cadmium(I1) and several other traceelements were extracted with zinc diethyldithiocarbamate in chloroform,whereas antimony(V) remained in the aqueous phase. The activities ofthe samples were counted on a germanium(1ithium) well-type detector andcompared with those of known volumes of standard solutions.Chemicalyields were determined by re-activation.The combination of conventional arsenic separation and this newlydeveloped diethyldithiocarbamate extraction technique proved to be a veryefficient and reliable method for the analysis of sulphide ores.Keywords : Sulphide ore analysis ; neutron-activation analysis ; radiochemicalseparation ; zinc diethyldithiocarbamate ; gavnma-ray spectrometryE. PERNICKA, P. A. SCHUBIGER and 0. MULLERMax-Planck-Institut fur Kernphysik, Postfach 103980, D-6900 Heidelberg, WestGermany.Analyst, 1978, 103, 475-481.ixPyrolysis - Mass Spectrometry of Textile FibresA procedure for pyrolysis - mass spectrometry is described and the spectra(mass pyrograms) of various textile fibres are presented.The method iscompared with infrared spectroscopy for the forensic characterisation ofsynthetic fibres. Samples of less than 5 pg can be analysed.Keywords ; Textile fibre cliavacterisation ; pyrolysis - ma.ss spectrometry ;infrared spectroscopyJ. C. HUGHES, B. B. WHEALS and M. J. WHITEHOUSEMetropolitan Police Forensic Science Laboratory, 109 Lambeth Road, London,SE1 7LP.Analyst, 1978, 103, 482-491.Determination of Probenecid in Serum by High-performanceLiquid ChromatographyThe determination of probenecid in serum samples by using high-performanceliquid chromatography is described. The method gives satisfactory resultsover the normal therapeutic range, namely up to 150 pgml-l of probenecidin serum, and is not affected by metabolites of the drug.The method doesnot require derivatisation of the drug, as in gas - liquid chromatographicprocedures, and is less subject t o interferences than spectrophotometricprocedures. It has been used in analysis of several hundred serum samplesand has given a satisfactory performance in respect of precision and accuracy.Keywords : Probenecid determination ; serum ; high-performance liquidclz YO mat ograph yR. K. HARLE and T. COWENInternational Development Laboratories, E. R. Squibb and Sons Limited, Moreton,Merseyside, L46 1QW.Analyst, 1978, 103, 492-49X THE ANALYST May, 1978NewE U R 0 -STAN D A R Dnow availableE. s 377 - 1Furnace Dust (LD Converter)Certified for the following elements :Fe, Si, Ca, Mg, At, Ti, Mn, P, S, Na,K, F, V, Cr, Ni, C, Zn, Pb, Cu and AsFull details obtainable from :Bureau of Analysed SamplesLtd.Newham Hall, Newby,Middlesbrough, Cleveland TS8 8EATelephone: Middlesbrough 31 721 6ANALYSTProfessionaIly quaIified analyst is required for our up-to-date anal y-tical laboratory which is situated in pleasant rural surroundings.This is a challenging new post which results in part from our recentpurchase of a scanning electron microscope with X-ray analysisfacilities.Postgraduate experience of energy dispersive X-ray analysisis necessary, together with some experience in, or a willingness i:olearn chemical, thermal (TGA, DSC) or spectroscopic (UV, IR, NMII,AA) techniques of analysis.Alternatively we might consider a graduate with considerable ex-perience of some of these latter techniques, and willing to learn X-rayanalysis.Applicants should possess drive, versatility, the ability to develop newanalytical methods with a minimum of supervision, and the abilityto work on a number of different topics simultaneously.A salary commensurate with experience will be offered, together withparticipation in the FSSU superannuation scheme.Please write in confidence to: Director of Research (quoting ref.BKT)The Malaysian Rubber Producers’ Research Association, Tun AbdulRazak Laboratory, Brickendonbury, Hertford SG13 8NL.CLASSIFIED ADVERTISEMENTSThe Rate for Classified Advertisements is g2.30 per singlecolumn centimetre (minimum L4.60).Box Numbers are charged a n extra sop.Deadline for classified copy is 20th of the mc’zlh precedingmonth of issue.All space orders, copy instructions and enyzriries should beaddressed to The Advertisement Department,The Chemical Society, Burlington House, Piccadilly,London WIV oBNTelephone 01-734 9864 Telex 268001THE QUEEN’S UNIVERSITYOF BELFASTMSc COURSE inANALYTICAL CHEMISTRYApplications are invited for admission to thisestablished 12 month full-time MSc coursewhich provides a comprehensive training inthe theory and practice of modern chemicaland instrumental methods of analysis.Appli-cants should normally possess an honoursdegree (or equivalent) in chemistry or cognatesubjects. Part-time courses are available.The Science Research Council has recognisedthe course for tenure of its Advanced CourseStudentships.A description booklet and application formscan be obtained from Professor D. ThorburnBurns, Dept. of Chemistry, Queen’s Universityof Belfast, Belfast,BT7 lNN,Northern Ireland.PLASMA ANALYSISScientific Conference Services Limited willbe holding a three day workshop forusers and potential users of InductivelyCoupled Plasmas as an analytical spectro-scopic source at Imperial College from3rd to 5th July 1978.Topics will include a brief theoreticalintroduction on plasma emission spectro-scopy, instrumentation, the application ofthe technique to various analytical fieldsand practical demonstrations. These willbe presented by internationally recognisedexperts in the ICP field.For further information contact :Dr. M. D. SilvesterScientific Conference Services Limited14 Trading Estate RoadGreat Western Trading EstateLondon NWlO 7L

ISSN:0003-2654    

DOI:10.1039/AN97803BP041

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

MAY 1978 The Analyst Vol. 103 No. 1226 Simultaneous Techniques in Thermal Analysis* A Review F. Paulik and J. Paulik Institute for General and Analytical Chemistry, Technical University, Gellert ter 4, 1502 Budapest XI, Hungary Summary of Contents Introduction Development of simultaneous methods Difficulties of interpretation Derivative methods Standardisation of experimental conditions EGA and EGD Examination of decomposition of organic compounds TD and ETA EC methods Manipulation of experimental conditions Conclusion TG - DTA Keywords : Review ; thermal analysis ; simultaneous methods derivative and Introduction The development of simultaneous thermoanalvtical differential methods ; methods has taken dace during the past two deLades and this development has man; aspects. However, withi; the scope 2 this review we wish first to analyse the causes, motives, aims and trends that led finally to the development of simultaneous methods.In the 1950s thermal analysis entered a new phase of development. The accuracy obtainable with classical methods had not met more stringent requirements and could not be increased by improving the measuring devices. In those days absolute temperature values could be measured by use of thermocouples with an accuracy higher by orders of magnitude than, for example, that with which decomposition temperatures could be determined by means of differential thermal analysis (DTA). For instance, two literature values for the decomposition temperature of calcium carbonate obtained by DTA are 800 and 900 "C. These values were determined with an error of only $1 "C, but in spite of this, and on the basis of these data, we cannot state with a greater certainty than 5 5 0 "C that the decomposition temperature of calcium carbonate is 850 "C.Also, mass change could be determined with an accuracy higher by orders of magnitude than that with which the amounts of two components could be determined if the decompo- sition processes of the components overlapped. Thus, if the sample mass became constant between the two decomposition processes, the amount of each component could easily be determined with an error of only *0.2%. However, if the two processes overlapped, then, as shown in Fig. 1, the error could increase to $20% or more. Accordingly, it became evident that a further development in thermal analysis could only be expected if new principles of examination were introduced and new measuring techniques developed.One of the most significant results of this activity was the birth and rapid development of simultaneous met hods. * Plenary Lecture presented at the First European Symposium on Thermal Analysis, Salford University, 417 As a consequence, intensive research activity began. September, 1976.418 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES Analyst, Vd. 103 Development of Simultaneous Methods The aims that were characteristic of this development were, in addition to multiplying the information obtainable, to increase resolution, to standardise the experimental conditions and to increase selectivity. Experimental conditions were manipulaked in order to realise these aims.The idea of coupling single classical thermoanalytical methods is straightforward, as it is well known that extension of an examination by use of an additional method increases both the certainty of interpretation and the accuracy of evaluation of the thermoanalytical curves not in a proportional, but in a multiple, way. This idea is demonstrated by the following example. Knowledge of the mineral composition of bauxites is important from the point of view of alumina production. However, as Fig. 2 shows, for this extreme example (selected for the purposes of demonstration), the composition of a bauxite sample cannot be determined at all if a thermobalance alone is used. The possible components are as follows (with decomposition temperatures in parentheses) : boehmite, a-A1,0,.H20 (520 "C) ; diaspore, y-A120,.H20 (540 "C) ; kaolinite, A120,.2Si0,.2H,0 (570 "C) ; alunite, K,SO,.- 3A1,03.3S0,.6H20 (570 "C) and K2S0,.3A1,0,.3S0, (790 "C) ; and calcite, Ca0.C02 (800 "C).The thermogravimetry (TG) curve alone cannot indicate which components are present and the decomposition of which components was responsible for the two-step mass-loss process. 50 It 10 mg 50 k 10 mg Temperature .-+ Tem pera tu re/"C Fig. 2. TG investigation of a bauxite. Fig. 1. Accuracy of mass measurement by means of thermogravimetry : (a), when the decomposition pro- cesses of the components of a sample do not overlap; and ( b ) , when overlap occurs. As Fig. 3 shows, the composition of the sample cannot be stated even if the derivative thermogravimetry (DTG) curve is also recorded, though the situation undoubtedly becomes clearer.The DTG peaks indicate the single temperature values characteristic of the different mineral components more accurately than does the TG curve. However, the presence of four bauxite minerals or, more precisely, three combinations of these four, is still possible. These combinations are : boehmite and alunite; boehmite, alunite and kaolinite ; and boehmite, kaolinite and calcite. Although it is not sufficient for the complete solution of the problem, it reduces the number of possible combinations to two, as can be seen from Fig. 4. A small exothermic DTA peak can be seen at 960 O C , which is characteristic of kaolinite only and clearly proves the presence of this mineral.The two possible combinations are, therefore : boehmite, alunite and kaolinite ; and boehmite, kaolinite The DTA curve offers further important information.May, 1978 I N THERMAL ANALYSIS. REVIEW 41 9 \Boehmite In v) 5 0 200 400 600 800 1000 Temperature/"C Fig. 3. Simultaneous TG and DTG inves- 40 tigation of the same bauxite as in Fig. 2. 0 200 400 600 800 1000 Temperature/"C Fig. 4. Simultaneous TG, DTG and DTA investigation of the same bauxite as in Figs. 2 and 3. and calcite. Accordingly, in this instance the simultaneous TG, DTG and DTA examina- tions must be completed by use of one of the evolved gas analytical (EGA) methods in order to determine whether sulphur(V1) oxide or carbon dioxide is liberated, i.e., whether the sample contains alunite or calcite, respectively, and if possible to determine the amount of the gaseous decomposition product.As Fig. 5 shows, by using EGA in addition to the other methods, not only can the qualita- tive composition of the sample be identified with certainty, but also the amount of each identified component can be determined with satisfactory accuracy. Uitticulties of lnterpretation Despite the fact that the combination of different methods is connected with the above- mentioned advantages, up to the 1950s, apparently for no reason, thermoanalysts used only single thermoanalytical methods for their investigations. The question of the reason for this choice of method arises. One of the causes undoubtedly lies in the difficulty of the common interpretation and evaluation of curves recorded by means of different thermoanalytical devices.This diffi- culty is shown for a TG and a DTA curve in Fig. 6. The curves demonstrate the thermal decomposition of the same dolomite sample, investigated by using a thermobalance to obtain the TG curve, and a classical steel block type DTA apparatus. On comparing the courses of the two curves, immediately two contradictions can be seen. Firstly, the TG curve represents the thermal decomposition as a one-step process while the DTA curve shows it as a two-step process. The explanation of this phenomenon lies in the different character of the two curves and the poor resolution of the TG curve. As will be shown below, this problem could be solved successfully by the derivation of the TG curve.At least as serious is the other contradiction. According to the two curves, the thermal decomposition of the dolomite did not take place in the same temperature range. According to the TG curve, the sample had totally decomposed below 900 O C , while according to the DTA curve, only the first part of the decomposition had taken place below this temperature and the decomposition was complete only at a temperature 100 "C higher. This phenomenon420 PAULIK AND PAULIK SIMULTANEOUS TECHNIQUES Analyst, Vol. 103 ~~ 9 DTA I - .-. -.-. '*\. W w o \ I * 18 rng of H 2 0 \5? I I 120 rng of boehrnite 1 I I I \ . \ ' I > 10 mg of H20 72 mg of kaolinite 0 200 400 600 800 1000 Ternperatu re/"C Fig. 5. Simultaneous TG, DTG, DTA and EGA investigation of the same bauxite as in Figs.2, 3 and 4. can be explained by the fact that different experimental conditions lead inevitably to different results. In order to eliminate this difficulty, thermoanalysts made efforts to standardise the experimental conditions. According to the thermoanalytical method applied the thermal changes examined are described in the form of either integral or differential functions. The two different kinds of curve illustrate the transformations in different ways, because of the difference in their resolution. We can detect a difference between two thermal effects described by two curves of different character, even if these curves demonstrate thermal processes, such as mass and enthalpy changes, that occur simultaneously. For instance, even the most experienced thermoanalyst could not decide with certainty at first glance whether the TG and DTA curves in Fig.7 illustrate the decomposition of the same bauxite sample or that of two different bauxites of similar but not identical composition. Derivative Methods This statement is supported by Fig. 8, in which, in addition to the simultaneously recorded DTA With the development of derivative methods this problem was fully eliminated.May, 1978 IN THERMAL ANALYSIS. REVIEW 421 I DTA L 0 200 400 600 800 l ( 600 703 800 900 1000 Temperature/" C Fig. 6. dolomite. Parallel TG and DTA examinations of TG 0 0 200 400 600 800 1000 Temperature/"C Fig. 7. Parallel DTA and TG examinations of bauxite. f I I I I I I 4 0 200 400 600 800 1000 Temperature/"C Fig. 8.Simultaneous TG, DTG and DTA investigations of the same bauxite as in Fig. 7.422 PAULIK AND PAULIK: SIMULTANEOUS TECHNIQUES Analyst, vol. 103 and TG curves of the above-mentioned bauxite sample, the DTG curve is shown. It can be seen that the courses of the DTG and DTA curves are similar, which occurs because owing to their mathematical relationship their characters are identical. Thus the DTG curve creates a basis for the comparison of TG and DTA curves. However, the recording of the DTG curve has further, even more significant, advantages than that just mentioned. The DTG curve, owing to its high resolution, is of significant help in selecting the characteristic temperatures as well as in describing the whole course of the transformation, and it therefore makes possible the reliable determination of the qualita- tive composition of multi-component samples.Thus, in Fig. 8, while reactions that follow one another overlap in the TG curve, these reactions can clearly be distinguished in the DTG curve. So, from knowledge of the TG curve alone, we could not decide whether the step on the TG curve in the temperature interval between 500 and 700°C is due to the decomposition of boehmite, diaspore or kaolinite, or to all of them together, but the DTG curve shows without any doubt that the sample contains boehmite and kaolinite. Further, by projecting the 560 "C minimum of the DTG curve on to the TG cu.rve, even the amount of these two minerals can be determined with a limited accuracy, which is an important additional advantage of recording the DTG curve.These advantages were suspected by all those thermoanalysts who were making efforts to develop the various differential and derivative methods in thermal analysis. Table I is a list of the pioneers in the field. Dejean first suggested the derivation of thermoanalytical functions in 1905. His measuring technique was theoretically correct but owing to practical difficulties his method was not used. The method of de Keyser, published in 1953, also has more theoretical than practical significance. The first derivative method that found practical application was worked out in 1954. The technique, based on the principle of induction, is suitable for recording a DTG curve. These initiatives were soon followed up and in the course of a single decade numerous measuring techniques were worked out for the derivation of thermoanalytical curves (DTA, TG), thermal analysis (TA), thermodilatometry (TD), evolved gas detection (EGD), evolved gas analysis (EGA), electrical conductivity analysis (EC) and thermomagnetic analysis (TM).Although the thermoanalytical curves can easily be derived by use of computers, for various reasons conventional methods are used in preference even now. As these computer tech- niques can technically be classed among computer rather than thermoanalytical methods, they are not listed in Table I. On theoretical con- siderations derivative methods are favoured as differential methods, owing to the principle In Table I differential and derivative methods are distinguished.Derivative TA TG TG TG TG DTA } EGD TD TD TD TG TA EGA EGD TG EC TM TABLE I DERIVATIVE AND DIFFERENTIAL METHODS Differential Date Workers 1905 Dejean 1954 Paulik and co-workers 1955 Paulik and co-workers 1956 Waters TG 1953 de Keyser TG 1955 Lambert 1959 Campbell et al. DTA 1959 196 1 1961 1963 DTA 1963 1965 1965 1966 1966 1969 1972 1974 1975 TG Freeman and Edelman Paulik et al. Paulik et al. Wilburn and Hesford Knofel Garn Rupert Paulik and co-workers Pannktier Byrd Price Forster, et al. Moskalewicz Reference 1 2 3, 4 5, 6 7 8 9 10 11, 6 12, 6 13 14 15 16 17, 18 19 20 21 22 23May, 1978 I N THERMAL ANALYSIS. REVIEW 423 of difference formation, are liable to smaller or greater errors, while derivation can be regarded as the ideal limiting case in difference formation.Standardisation of Experimental Conditions The 1950s saw the introduction of another innovation, the basic idea of which is as follows. If examinations are carried out in such a way that different thermal variables are measured in a single sample, then for each variable the experimental conditions will be identical and there will not be any shift in the course of the curves obtained. This idea is the basic principle of simultaneous methods. Fig. 9 demonstrates the thermal decomposition of dolomite. The curves in Fig. 9(a) were obtained by using a DTA apparatus and a deriving thermobalance separately, while the curves in Fig. 9(b) were recorded by applying the simultaneous TG, DTG and DTA technique. It can be seen that in the latter instance the DTA and DTG curves are congruent.Curves recorded by using two different devices differ in phase and in form. By observing the two pairs of curves we can form an idea of the difficulties that thermo- analysts had in the common interpretation and evaluation of thermal curves recorded by means of separate devices. Also, we can understand why the literature contains data for transformation temperatures of the same material that often differ by 50-100 "C. DTA :---To I 1 I 700 800 900 1 LO 700 800 900 1000 Fig. 9. (a), Parallel and ( b ) , simultaneous TG, DTG and DTA examinations of the same dolomite as in Fip. 6. The explanation for this phenomenon is as follows. Most of the decomposition reactions In a closed system a t of inorganic compounds are processes that lead to an equilibrium.decomposition pressure and temperature. For example, the decomposition pressure of calcium carbonate as a function of temperature is shown in curve 1 in Fig. 10. However, in general, in thermoanalytical investigations the sample is not examined in a closed space but in an open sample holder at atmospheric pressure and in the presence of air or of an inert ugc Therefnre the rnrnnncitinn nf tho ~ f a c in r n n f a r t 1xr;t-h the c n l i A nhgco ic rhanuincr Ior a gven pracricai example. correspond to the points given on the TG curve. ine single values 01 me parriai pressure or carDon aioxiae These values were obtained by extra- temperature. However, it should be noted that the curve was constructed by making certain suppositions and neglecting certain factors, so that the picture obtained is only qualitative. In spite of this the figure characterises the exceedingly complex process taking place in the sample.It is well known that under conditions of thermoanalytical investiga-424 loo 80 ae v; 60 % 2 4 0 - 20 0 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES Analyst, VoZ. 103 Ti CaCO3 - - 1 I I I - - q!S-- - - _,_ - -1,- -- r 1 Fig. 10. 1, Decomposition pressure of CaCO, veYsus temperature; and 2, TG curve of CaCO, obtained by using a low-walled crucible. tions reactions leading to an equilibrium are composed of many chemical and physical partial processes, some of which impede the transformation while others promote it. The path of the transformation is determined by the infinite succession of alternations taking place continuously in the formation and resolution of micro-equilibrium states.For example, when the dissociation of calcium carbonate begins, carbon dioxide will appear within the sample, in the space between the grains of the substance. If the partial pressure of this carbon dioxide approaches the value of the decomposition pressure that corresponds to the actual temperature, then the decomposition will slow down. It will even stop if the values of the theoretical and actual pressures become equal. The fact that the reaction leads to an equilibrium produces an impeding effect on the progress of the decomposition, i.e., it sets an upper limit to the decomposition rate. However, the momentary quasi-equilibrium will always overbalance and the decomposition will start again because, owing to the increased temperature, only a higher concentration of carbon dioxide will cause a new quasi-equilibrium to be established.In addition, another factor hastens the decomposition. The mixture composed of carbon dioxide and air diffuses continuously outwards from the sample, while air diffuses towards the centre of it. The carbon dioxide released is then replaced by amounts of newly formed carbon dioxide. However, the increase in sample temperature that actually promotes the heating programme and in an atrn&phere of con.May, 1978 I N THERMAL ANALYSIS. REVIEW 425 process occurs without hindrance only until the sample has taken up an amount of heat that corresponds to its heat capacity. In general, the thermal conductivity of the substances investigated is poor.Therefore, the sample is not able instantaneously to absorb from its surroundings the amount of heat necessary for the transformation to take place as this amount is greater by orders of magnitude than its heat capacity. For example, for calcium carbonate the molar heat capacity, Cp, is 26 cal mol-l "C-l at 890 "C and AHOdle8. is 42 600 cal mol-l. This is the reason why transformations generally take place slowly, as is demonstrated by curves 1 and 2 in Fig. 11. Curve 1 illustrates the ideal course of decomposition of calcium carbonate. In constructing this theoretical curve isothermal conditions and the presence of a carbon dioxide atmosphere were assumed. Curve 2 was recorded with a 10 "C min-l dynamic heating programme and in an atmosphere of carbon dioxide.Owing to the atmosphere of carbon dioxide the decomposition process was made independent of the gas transport, as we have supposed in constructing curve 1. Accordingly, the difference between the courses of the two curves can be attributed solely to the effect of the slow heat transfer (see also Fig. 18). Thus the progress of the decomposition is governed by the rate of temperature increase and of gas diffusion, which will be such that the momentary equilibrium of the system always corresponds to the correlation between decomposition pressure and temperature. Fig. 12 demonstrates the different gas-diffusion conditions obtaining in two sample holders of different shape. Comparison of the partial pressure values of carbon dioxide, indicated at the corresponding points of the TG curve, shows that the concentration of carbon dioxide within the sample changed in the two experiments in different ways.For example, when, in the low-walled sample holder, the decomposition of the sample was complete, the tempera- ture was at 850 "C, while the partial pressure of carbon dioxide was at 370 Torr. In contrast, a t the same temperature and pressure values, only 40% of the sample was decomposed in the high-walled sample holder and the decomposition was complete only at 890 "C. However, the partial pressure of carbon dioxide in the sample had reached 680 Torr a t this temperature. 100 80 60 s 2 40 1 v) 20 0 0 200 I- 2 400 & 0 --. v) 0" 600 760 600 700 800 850 900 Temperature/'C Fig.12. 1, Decomposition pressure of CaCO, vemus temperature; 2 , TG curve obtained with a low-walled crucible; and 3, TG curve obtained with a high-walled crucible. It follows from the above that all of the experimental conditions that may influence the rate of gas diffusion and also the rate of heat transfer exert a significant influence upon the course of thermal decomposition. Such conditions are, for example, the amount of sample, its layer thickness and compactness, size of the grains, heating rate, composition and pressure of the gaseous atmosphere and shape and size of the sample holder. However, for the sake of completeness it should be noted that in some instances, owing to the occurrence of even slower partial processes than those mentioned above (nucleus forma-426 PAULIK AND PAULIK: SIMULTANEOUS TECHNIQUES Analyst, vol.103 tion, nucleus growth, gas diffusion through the compact new phase, etc.) the transformation process will be even more complicated. On the other hand, for endothermic reactions not leading to an equilibrium, the situation is simpler, as the course of such transformations is independent of the concentration of the gaseous decomposition products and in most instances it is disadvantageously influenced only by the slow heat transfer. In the 1950s, recognition of these factors led to research activity to find appropriate means for the standardisation of experimental conditions, but a real breakthrough in this field could successfully be achieved only by the introduction of simultaneous techniques.The solution to the problem, which was to use a single sample, was exceedingly simple and the results obtained were ideal. Despite the rapid and wide application of simultaneous techniques, the parallel application of individual methods remained current and they are still employed nowadays. Great efforts were made to standardise the experimental conditions in the course of the develop- ment of these latter methods too. At the present stage of development both types of measurement have advantages and disadvantages. In the construction of equipment for simultaneous techniques, just in the interest of coupling different methods we are often compelled to select less precise techniques of measurement, while methods applied in parallel compensate for the lack of standard results with greater accuracy.Although the principle of examination precludes the possibility of using identical experi- mental conditions when applying complementary methods, these methods do not play a subordinate role among combined techniques. Let us remember, for example, the importance of complementary X-ray spectroscopy, infrared spectroscopy and electron microscopy. In spite of the importance of all the combined methods mentioned, in what follows we shall deal only with simultaneous measuring techniques. TG - DTA Among thermoanalytical methods TG and DTA are the two that yield the greatest and the most valuable information about the substance investigated. Consequently, it is under- standable that the coupling of these two methods was the first to be attempted.Researchers who made efforts in this direction are listed in Table 11. It can be seen that the first equipment of this type was constructed in 1955, and was suitable for the simultaneous recording of TG, DTG and DTA curves. This initiative was soon followed by several others. TABLE I1 SIMULTANEOUS DTA, TG AND DTG Method DTA - TG - DTG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG DTA - TG - DTG DTA - TG DTA - TG - DTG DTA - TG DTA - TG Date 1955 1957 1959 1960 1960 1961 1962 1962 1962 1963 1963 1964 1964 1964 1965 1968 Workers Reference Paulik et al. 24, 25 Powell 26 Papailhau 27 Blazek and Cisar 28 Reismann 29 Piece 30 Torkar et al. 31 Formanek and Dykast 32 Kissinger and Newman 33 McAdie 34 Kriiger and Bryden 35 Wiedemann 36 Khristianov and Korovyatnikov 37 Saito et al.38 Patai et al. 40 Charsley and Redfern 39 EGA and EGD Apart from TG and DTA methods, the examination of the gaseous decomposition products that evolve on heating may perhaps furnish the greatest and the most useful informationMay, 1978 I N THERMAL ANALYSIS. REVIEW 427 regarding both the composition of the substance investigated and the nature of the reactions taking place in the material. For this purpose a number of different methods have been developed, which are usually divided into two groups, EGD and EGA. According to the measurement technique, the character of the information obtained and the field of applica- tion each of these groups can be further divided into two sub-groups. A number of the EGD methods (Table 111) are based on the detection of gas evolution by means of a thermal conductivity cell or a gas density detector (Group I).The method of examination is simple and the equipment is cheap, but the information obtained is modest. The equipment clearly indicates the evolution of gas and conclusions can be reached based upon the amount of gas evolved. However, these methods are not adequate for the examina- tion of the quality of the evolved gas. In spite of this the combination of these methods with othcr thermoanalytical methods, for example with DTA or TD, or with mass spectrometry (hIS), may be very useful, as they can contribute to the better interpretation of the results obtained and enable physical transformations and decomposition reactions to be clearly distinguished.TABLE I11 EGD METHODS Method combined Method with Date Workers Reference Group I- Thermal conductivity and DTA 1960 Lodding and Hammel 41 densitome try DTA 1961 Ayres and Bens 42 DTA l9Gl Garn and Kessler 43 DTA 1962 Wendlandt 44 Group I I- Volumetric Barometric DTA 1953 Teitelbauin and Berg 45 DTA 1957 Gordon and Campbell 46 DTA 1964 Charles 47 TG 1965 Bancroft and Gesser 48 DTA 1967 Guenot et al. 49 TG 1968 Maicock and Pai Verneker 50 DTA 1969 Eousquet et al. 51 An example of the use of these techniques of examination is given in Fig. 13. The DTA curve of barium perchlorate hydrate is shown, and the interpretation of this curve alone would be difficult, but if the EGD curve is recorded simultaneously, as shown, then the common interpretation of the two curves permits us without risk of major error to state the following.The first three maxima on the EGD curve indicate the release of water, while the fourth is due to the escape of oxygen. The first maximum on the DTA curve can probably be explained by the melting of the hydrate, the fifth and sixth maxima by changes in crystal modification, and the seventh may be due to the melting of the anhydride. With the help of gas volumetric and gas barometric examinations (Table 111, Group 11) the amount of gas evolved can also be determined, but obtaining these curves is only an indirect means for the determination of the quality of the evolved gases. In addition to recording the curves of the gaseous decomposition products, it is also necessary to record the TG curve.Owing to the difference in the relative molecular masses of gases, in a f avourable instance we shall be able to select from possible gaseous decomposition products the most probable one. Here the EGD and TG curves of a calcium oxalate sample can be seen. The former indicates that in the three periods of the decomposition the volume of evolved gases was equal, but owing to the difference in the relative molecular masses of water, carbon monoxide and carbon dioxide, the sizes of the three steps in the TG curve are different. In more compli- cated instances this numerical difference forms the basis of calculations. The essential difference between the methods of EGD and EGA lies in the fact that with EGA both the quality and the amount of the gaseous decomposition products can be determined in a selective way.However, in addition to problems of standardisation, the The principle of this examination is demonstrated for a simple example in Fig. 14.PAULIK AND PAULIK: SIMULTANEOUS TECHNIQUES Analyst, VoZ. 103 Melting v Melting I Dehydration Decomposition I 1 1 I I 100 300 500 Temperature/"C 0 100 - E 2 200 - 5 2 300 400 0 200 400 600 800 1000 T e m p era t u r el * C Fig. 14. Simultaneous TG and EGD Fig. 13. Simultaneous DTA (1) and examinations Of CaC204.xH20. EGD (2) examinations of Ba(ClO,),.xH,O. separation of overlapping transformations, or in other words increasing the selectivity of the examination, causes the greatest trouble to thermoanalysts. We have already shown, for the determination of boehmite, alunite and kaolinite, the usefulness of EGA methods in separating overlapping reactions.A similarly convincing example is shown in Fig. 15. During the preparation of aluminium sulphate, if the optimum I I I 200 400 0 8 20 vi B - ln .s 40 60 '/3 H 2 S 04.24 H 2 0 \ -9 21 H2 0 I I I 200 400 Tern pera t u re/' C Fig. 15. Simultaneous DTA, TG, DTG and EGA examinations of (a), A1,(S04),.xH,0 of unknown composition and (b), A1,(SO4),.~H,SO4.24H2O.May, 1978 I N THERMAL ANALYSIS. REVIEW 429 conditions are not observed, then instead of Al,(SO,) ,.18H,O the acid salt is precipitated. When this salt is dehydrated sulphur(V1) oxide is also evolved, In fact, the course of the TG, DTG and DTA curves would not call our attention to this possibility.Further, even an EGD curve would not indicate the water and sulphur(V1) oxide separately. In contrast, the curve denoted SO, in Fig. 15 clearly shows that with the departure of the last two water molecules at 450 "C two thirds of a molecule of sulphur(V1) oxide is also released. Without performing the simultaneous EGA examination, we were not able even to notice the evolution of sulphur(V1) oxide. At the same time, having obtained the curve for the release of sulphur(V1) oxide we not only can determine the amount of sulphur(V1) oxide but may also be prevented from drawing incorrect conclusions regarding the composition of the sample and the kinetics of the process. The sulphur(V1) oxide curve was obtained by applying the technique of thermal gas titrimetry (TGT) (Table IV).With the help of this technique, based on the principle of titrimetric analysis, the amounts of the gaseous decomposition products of inorganic com- pounds can be determined in the presence of one another. TABLE IV EGA METHODS Method combined Method with Group I- TGT .. .. .. .. , . DTGT, DTA, Gas infrared spectroscopy . . TG Absorption liquid conductimetry TG, DTA Absorption liquid thermometry . . DTA DTA TG, DTG Group II- GC . . .. .. .. , . TG DTA TG DTA DTA DTA MS .. .. .. .. .. EGD EGD, DTA EGD, DTA EGD, TG TG EGD, TG, DTA EGD, DTA, TG DTA EGD, DTA, TG, DTG TG DTA TG, GC DTA, TG GC Thermoparticulate analysis . . DTA, TG Thin-layer chromatography .. DTA Date 1966 1967 1964 1963 1963 1963 1966 1968 1969 1975 1975 1965 1965 1966 1968 1969 1969 1969 1969 1969 1969 1969 1971 1973 1974 1960 1967 Workers Reference Paulik et al.52, 17 Keattch 54 Chamberlain and Green 55 Notz and Jaffe 56 Hegediis and Kiss 53 Can0 57 Bandi et al. 58 Chiu 59 Bollin 60 Yamada et al. 61 Mercier 62 Wendlandt and Southern 63 Langer et al. 64 Wendlandt et al. 65 Zitomer 66 Wilson and Hamaker 67 Smith and Johnson 68 Wiedemann 69 Gaulin et al. 70 Redfern et al. 71 Brown et al. 72 Langer and Bradly 73 Chang and Mead 74 Gibson 75 Merritt et al. 76 Doyle 77 Rogers 78 A similar possibility for the selective determination of gaseous decomposition products is offered by infrared spectroscopy and by conductimetric or thermometric measurements on a liquid in which the gases have been absorbed (Table IV, Group I). Examination of Decomposition of Organic Compounds It is known that the decomposition of organic compounds is in most instances a very complex process. Solid, liquid and gaseous decomposition products are formed in the course of the countless reactions taking place in parallel or consecutively.For the selective investigation of these products and processes the methods discussed so far are not suitable. However, these processes can be examined if, for example, gas-chromatographic (GC), mass430 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES Analyst, Vd. 103 spectrometric (MS), thin-layer chromatographic or thernioparticulate analyses are combined with DTA and TG measurements. The earliest of these methods are listed in Table IV, Group 11. Particularly good selectivity can be attained when GC and MS are applied together.In this instance the gaseous decomposition products are separated by means of a gas chromatograph into different fractions, which are then further examined by means of a. mass spectrometer. As these examinations can be performed only stepwise, generally the EGD curve is simultaneously recorded in order to determine the gas-evolution process. The equipment required for these analyses is expensive and requires skill in handling; however, it is indispensable when the kinetics and mechanism of the thermal decomposition of organic compounds are to be studied. TD and ETA We think TD has been neglected up to now, yet it provides useful information about changes occurring in the crystal structure of inorganic compounds, which is an area in which other thermoanalytical methods are not very useful. Two examples of the use of TD are shown in Fig.16. It is known that kaolinite loses its water of constitution between 400 and 800 "C. During the course of a solid-state reaction it transforms first into metakaolinite and then into mullite at about 950 "C. The other example, barium chloride dihydrate, first loses its water of crystallisation and in the temperature interval 350-850 "C a significant recrystallisation process takes place in the anhydrous material. In the vicinity of 900 "C, just before the substance melts, a modification froin the a to p crystal form occurs. I IV I C .- I I w .- 0" TD I a 200 600 1000 0 1 2 E- 3 remperatu re/O C I 0 2 8 4 - 6 8 200 600 1000 Fig. 16. Simultaneous DTA, TG and TD examinations of (a), A120,.2Si0,.2H,0 and ( b ) , BaC1,.2H20. As the shape of the TD curves in the figures proves, the course of all the processes, the solid-state reaction and also the thermal decomposition, the recrystallisation and crystal modification, can be followed by this method.However, so far only in the field of coal, ceramic and silicate chemistry has TD found wide application, and even here mostly as a single method. The reason for the neglect of this technique probably lies in the difficult interpretation of the TD curve, as in general it undoubtedly gives a more complicated pictureNay, 1978 IN THERMAL ANALYSIS. REVIEW 431 than the TG or DTA curves. For example, in thermal decomposition the sample decomposes and an amorphous or microcrystalline phase is formed.The recrystallisation that generally follows is in most instances protracted and overlaps the previous process. While the DTA and TG curves indicate only the first process, the TD curve depicts both processes, which causes difficulties in its interpretation. Therefore, in addition to the derivation of this curve, the simultaneous recording of other thermoanalytical curves, e.g., DTA or DTG (Table V), can greatly contribute to the easier and more reliable evaluation of the TD curve. TABLE V TD METHODS Method combined with Date Workers Reference DTA .. .. . . 1956 Lehman and Gatzke 79 DTA .. . . . . 1958 Koehler 80 DTA .. . . . . 1959 Pearce and Mardon 81 DTD, DTA . . 1961 Paulik et al. 11 DTD, DTA, TG, DTG . . 1961 Paulik et al. 82, 11 DTD, DTA, TG, DTG, TGT,DTGT .. .. 1971 Paulik and Paulik 83 Emanation thermal analysis (ETA) is a method cognate with the previous one (Table 1'1). With the help of this method all those chemical and physical processes can be followed which actually cause the transitional migration of the lattice elements in crystals. However, the results of the two kinds of examination differ from and therefore supplement one another. TABLE VI ETA METHODS Reference Method combined with Date Workers TG . . .. . . .. 1961 Bussiere et al. 84 DTA, TD .. .. .. 1965 Zaborenko et al. 85 DTA, TD .. .. . . 1969 Balek 86 DTA, EGD ., .. 1972 Habersberger and Balek 87 DTA, TG .. .. .. 1973 Emmerich and Balek 88 EC Methods An example of the applicability of EC methods coupled with other thermoanalytical methods (Table VII) can be seen in Fig.17. This figure shows the results obtained with a thallium stearate sample. According to the DTA curve the substance melts immediately after its second polymorphous transition, but it forms an isotropic liquid only after the mesomorph - liquid transition. Now, as the course of the EC curve proves, these last two Method combined with DTA .. .. . . DTA .. .. .. DTA . . .. .. DTA .. .. .. X-ray .. .. TG, DTG, DTA . . .. DTA, TD, ETA . . .. DSC .. .. . . DTA .. .. .. TD .. .. .. DTA .. , . .. DTA, EGD .. .. TG .. .. .. TABLE VII EC METHODS Date Workers Reference 1959 1960 1960 1963 1965 1967 1969 1970 1970 1970 1973 1975 1975 Satava 89 Budnikov et al. 91 Pannetier et al. 92 Bessonov and Ustyantsev 93 Chiu 94 Carrol and Mangravite 95 David 96 Judd and Pope 97 Balek 98 Halmos and Wendlandt 99 Berg and Shlyapkina 100 Juranic et al.101 Berg and Burmistrova 90432 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES AfiaZyst, vol. 103 processes can also be followed by measurement of electrical conductivity. This example is a simple one, but the information-multiplying effect of the method has its real significance in more complicated instances. Also, remarkable results were obtained recently by thermoanalysts who coupled the methods of thermal X-ray analysis, photometric thermal analysis, hot-stage microscopy, thermomagnetic analysis, dynamic reflectance spectroscopy and thermoacoustic analysis with the methods of TG, DTA, EC, MS, EGD and TD (Table VIII). h a 1 v) E 5 0.5 --. V I.’ u 3 -0 0 0 0 j Mesomorph transitions \ Isotropic liquid Pol ymorph transitions / I I I I Iso- I tropic I I Mesophase , liquid I Solid 0 50 100 150 Temperature/”C Fig.17. Simultaneous DTA and EC examinations of thallium stearate.102 Manipulation of Experimental Conditions The picture we intend to give would not be complete without consideration of the manipulation of experimental conditions. For example, the beneficial effect exerted upon the resolution or the selectivity if a vacuum or high pressure are applied is well known. Further, semi-micro techniques can be used to improve results, and many other similar attempts have been made to achieve the same purpose by the selection of experimental conditions. An example was shown above (Fig. 10) to demonstrate that in the course of decomposition, the concentration of the gaseous decomposition products within the sample is continuously changing in an uncontrollable way.This condition greatly influences the whole course of the decomposition. By altering the experimental conditions the concentration of the gaseous decomposition products, and therefore the course of the decomposition, changes too (Fig. 12). Some thermoanalysts wished to eliminate this effect by ensuring a “self-generated” atmosphere. These researchers wanted to solve this problem by use of sample holders whose shape, as shown in Table IX, makes it possible for the gaseous decomposition products However, we shall mention only two subjects here.May, 1978 IN THERMAL ANALYSIS. REVIEW TABLE VIII 433 THERMAL X-RAY ANALYSIS Method combined Method with Thermal X-ray analysis DTA DTA DTA, EC DTA EGD, MS Photometric thermal analysis DTA TG DTA TD Hot-stage microscopy DTA DTA DTA Thermomagnetic analysis TG Dynamic reflectance spectroscopy EGD Thermoacoustic analysis DTA Date 1965 1967 1967 1968 1973 1972 1972 1972 1974 1966 1967 1968 1966 1970 1975 Workers Wefers Ravich Bessonov et al.Barret et al. Wiedemann David Loehr and Levy Barrall and Johnson Chrony Miller and Sommer Dichtl and J eglitsch Van Tets and Wiedemann Simmons and Wendlandt Wendlandt and Bradley Chatterj ee Reference 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 to be released easily, while the diffusion of air from the opposite direction is possibly hindered. They postulated that a pure self-generated atmosphere would immediately be created in the reaction space at the beginning of the decomposition.Thus, the stability of the partial pressure of the gaseous decomposition products would make the course of the decomposition unambiguous. Owing to the poor thermal conductivity of the sample and the rapid increase in temperature, a temperature drop occurs within the sample. Accordingly, the slow heat transfer is also responsible for the delayed transformation, which takes place in a broad temperature interval (Fig. 11). However, as it turned out later, this solved the problem only to a certain extent. TABLE IX SAMPLE HOLDERS FOR A SELF-GENERATED ATMOSPHERE n Piston Bal I-valve Cap i I lary Labyrinth Type* Date P 1960 BV 1960 C 1962 P 1962 BV 1962 BV 1963 C 1964 L 1971 Workers Garn and Kessler Forkel Amiel and Paulmier Claude1 Pannetier et al.Dormieux Lagier et al. Paulik and Paulik Reference 118 119 120 121 122 123 124 125, 126 * P, piston; BV, ball-valve; C, capillary; L, labyrinth. Researchers wanted to overcome this problem by working out a “quasi-static” heating programme (Table X). The idea was that if the temperature increase in the sample was controlled in such a way that the decomposition reaction took place a t a very low and constant rate, then the error caused by the slow heat transfer may be totally eliminated. One, by limiting the rate of mass change, dm/dT, produced quasi-isothermal heating conditions for TG and simultaneous EGA investigations. The other, thermoanalysis with a constant rate of decomposition Two kinds of heating control systems were developed.434 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES Analyst, VoZ.103 TABLE X METHODS WITH QUASI-STATIC HEATING u p1 < ~2 < P3 < 1 atm p = 1 atm = const. PI, P2, P3 = const. Heating Conditions Method Pressure control Date Workers Reference Quasi-static heating TG - dm/dT 1962 Erdey et al. 127 TCRD Vacuum dp/dT 1969 Rouquerol 128 TG PI, pz, p , dm/dT 1971 Paulik and Paulilc 129 Self-generated atmosphere + TG 1 atm dnz/dT 1972 Paulik and Paulik 130 quasi-static heating TG, EGA 1 atm dm/dT 1973 Paulik and Paulik 131 (TCRD), produced a heating control system in vacuum, under which the pressure of the gaseous products became constant, and the correlation between the constant gas flow being evolved and the temperature was recorded.However, the final solution to the problem has been brought about by the simultaneous application of a quasi-static heating programme and a sample holder ensuring a self- generated atmosphere. The gradual progress of this technique and the beneficial effect of its application are shown by the TG curves in Fig. 18. The curves demonstrate the decomposition of calcium carbonate under different experimental conditions. Curves 1 and 2 were recorded a t a heating rate of 10 "C min-l. For curve 1 a conventional open crucible 0 10 \O 0- v; 20 - v) 30 5 40 600 700 800 900 1000 Tern perat u re/ O C Fig. 18. Decomposition of CaCO, investigated under (1 and 2) a dynamic and (3 and 4) a quasi-static heating programme. Sample holder: 1 and 3, conventional; and 2 and 4, labyrinth.was used, while for curve 2 a labyrinth sample holder was used in order to ensure a self- generated atmosphere. Curves 3 and 4 were similarly recorded by using the two different kinds of sample holder, but with a quasi-static heating programme, and the calcium carbonate decomposed under quasi-isothermal conditions, i . ~ . , the temperature of the sample did not change during the decomposition process. It is evident that quasi-isothermal conditions. greatly increase the selectivity and resolution of the examination, as well as the accuracy of the qualitative and quantitative determinations. However, there is a difference even between curves 3 and 4. According to curve 4 the decomposition of calcium carbonate took place at 895 "C, i.e., at its "normal" decoinposition temperature, known from physical chemistry.This situation is of significance from the point of view of standardisation.May, 1978 I N THERMAL ANALYSIS. REVIEW 435 Conclusion In the presentation of the tables our primary intention was to summarise the simultaneous methods that have been reported in a manner that can be easily read. It must be appreciated that a complete survey of the literature is not given here, and the material presented has necessarily been selective. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. ‘33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. References Dejean, F., Revue Me’tall., 1905, 2, 70.de Iceyser, W. L., Nature, L o n d . , 1953, 172, 364. Paulik, F., Paulik, J., and Erdey, L., Hung. Pat., 144 54811954. Erdey, L., Paulik, F., and Paulik, J., Acta Chim. Hung., 1956, 10, 61. Paulik, J., and Paulik, F., Hung. Pat., 143 33211955. Paulik, F., Paulik, J., and Erdey, L., Talanta, 1966, 13, 1405. Lambert, H., Bull. SOC. Fr. Ce’Yavlz., 1955, 28, 23. Waters, P. L., Nature, Lond., 1956, 178, 324. Campbell, C., Gordon, S., and Smith, C. L., Analyt. Chew., 1959, 31, 1188. Freeman, E. S., and Edelman, D., Analyt. Chew., 1959, 31, 624. Paulik, F., Paulik, J., and Erdey, L., Hung. Pat., 150 905/1961. Paulik, F., Paulik, J., and Erdey, L., Hzdng. Pat., 150 906/1961. Wilburn, F. W., and Hesford, J. R., J . Scient. Instrum., 1963, 48, 91. Knofel, G.F., Spreclasaal Keram. Glas Email, 1963, 96, 58. Garn, P. D., “Thermoanalytical Methods of Investigation,” Academic Press, New York, 1965, Rupert, G. N., Rev. Scient. Instvum., 1965, 36, 1629. Paulik, J.. Paulik, F., and Erdey, L., Hung. Pat., 154 372/1966. Paulik, J., and Paulik, F., Talanta, 1970, 17, 1224. PannCtier, G., Revue Chim., Buc., 19G6, 17, 592. Byrd, J . S., United States Atomic Energy Commission, 1969, DP 1211. Price, G. H., J . Phys. E , 1972, 5, 747. Forster, I . B., Lee, J . A., and Tye, F. L., Thevunochim. Acta, 1974, 9, 55. Moskalewicz, R., in BuzAs, f., Editov, “Proceedings of the Fourth International Conference on Paulik, F., Paulik, J., and Erdey, L., Hung. Pat., 145 369/1955. Paulik, F., Paulik, J., and Erdey, L., 2. Analyt. Chem., 1958, 160, 241.Po-cvell, D. A, J , Scient. Instrum., 1957, 34, 225. Papailhau, J., Bull. Soc. Fr. ilfinc’v. CristaLlog./., 1959, 82, 367. Blazelr, A, and Cisar, V., Silikdty, 1960, 4, 52. Reismann, A., Analyt. Chem., 1960, 32, 1566. Piece, R., Schweiz. Miner. Petrogr. Mitt., 1961, 41, 303. Torkar, I<., Lasser, K., and Fritzer, H. P., Sprechsaal Keram. Glas Email, 1962, 95, 212. Formanek, J., and Dykast, J., Silikdty, 1962, 6, 113. Kissinger, EI. E., and Newman, S. B., i n Wine, G. M., Editov, “High Polymers, Analytical Chemistry McAdie, H. G., Analyt. Chem., 1963, 35, 1840. Kriigcr, J. E., and Bryden, J . G., J . Scicnt. Instvu.l.tz., 1963, 40, 178. Wiedemann, €1. G., Chemie-Ingr-Tech., 1964, 36, 1105. Khristianov, A. S., and Korovyatnikov, G. F., Zav. Lab., 1964, 30, 495.Saito, H., Otsuka, R., Iwata, S., and Tsuchiomoto, K., Bull. Sci. Engng Res. Lab. Waseda Univ., Charsley, E. L., and Redfern, J . P., paper presented a t the International Symposium on Thermal Patai, S., Helpern, Y., Estcrman, L., and Weinstein, ill., Israel J . Chew., 1968, 6, 445. Lodding, W., and Hammell, L., Analyt. Chem., 1960, 32, 657. Ayres, W. M., and Bens, E. M., Analyt. Chem., 1961, 33, 568. Garn, P. D., and Kessler, J . E., Analyt. Chein., 1961, 33, 952. Wendlandt, W. W., Agcalytica Chim. Acta, 1962, 27, 309. Teitelbaum, B. V., and Berg, L. G., Zh. Analit. Khim., 1953, 8, 152. Gordon, S., and Campbell, C., Analyt. Chem., 1957, 29, 1706. Charles, R. G., J , Inorg. Nucl. Chem., 1964, 26, 2195. Bancroft, G. M., and Gesser, H. B., J . Inorg. Nucl.Chew., 1965, 27, 1537. Guenot, J., Vallicon, J. L., and Pannetier, G., Bull. Soc. Chim. FY., 1967, 3068. TvIaycock, J. N., and Pai Verneker, V. R., Analyt. Chew., 1968, 40, 1935. Bousquet, J., Blanchard, J. M., Bonnetot, B., and Claudy, P., Bull. Soc. Chim. Fr., 1969, 184. Paulik, J., Paulik, F., and Erdey, L., Mikrochim. Acta, 1966, 886. Hegedus, A. J., and Kiss, B. A,, Magy. Ke‘m. Foly., 1967, 73, 41. Keattch, C. J., “Analysis of Calcareous Materials,” Monograph No. 18, Society of Chemical Industry, Chamberlain, M. M., and Green, A. F., J . Inorg. Nucl. Chew., 1963, 25, 1471. Notz, K. J., and Jaffe, H. H., J. Inorg. Nucl. Chern., 1963, 25, 851. p. 499. Thermal Analysis, Budapest,” Volume 3, AkadCmiai Kiad6, Budapest, 1975, p. 873. of Polymers, ’’ Second Edition, Interscience, New York, 1962, p. 159.1964, 27, 26. Analysis, London, 1965. London, 1964, p. 279.436 PAULIK AND PAULIK : SIMULTANEOUS TECHNIQUES Analyst, V d . 103 Cano, G., Bull. SOC. Chim. Fr., 1963, 2540. Bandi, W. R., Straub, W. A., Buyok, E. G., and Melnick, L. M., Analyt. Chem., 1966, 38, 1336. Chiu, J., Analyt. Chem., 1968, 40, 1516. Bollin, E. M., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume I, Academic Press, New York, 1969, p. 255. Yamada, K., Oura, S., and Haruki, T., in BuzAs, f., Editor, “Proceedings of the Fourth International Conference on Thermal Analysis, Budapest ,” Volume 3, Akad6miai Kiad6, Budapest, 1975, p.1029. Mercier, J . G., in BuzAs, f., Editor, “Proceedings of the Fourth International Conference on Thermal Analysis, Budapest,” Volume 3, Akademiai Kiad6, Budapest, 1975, p. 104. Wendlandt, W. W., and Southern, T. M., Anulytica Chim. Acta, 1965, 32, 405. Langer, H. G., Gohlke, R. S., and Smith, D. H., Analyt. Chem., 1965, 37, 433. Wendlandt, W. W., Southern, T. M., and Williams, J. I<., Analytica Chim. Ada, 1966, 35, 251. Zitomer, F., Analyt. Chem., 1968, 40, 1091. Wilson, D. E., and Hamaker, F. AT., in Schwenker, R. F., Jr., and Garn, P. D., “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 517. Smith, J. W., and Johnson, D. R., in Schwenker, R.F., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 2, Academic Press, New York, 1969, p. 1251. Wiedemann, H. G., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 229. Gaulin, C. A., Wachi, F. M., and Johnston, T. H., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 2, Academic Press, New York, 1969, p. 1453. Redfern, J. P., Treherne, B. L., Aspimal, M. L., and Wolstenholme, W.A., paper presented at the 17th Conference on Mass Spectrometry, Dallas, Tex., USA, 1969. Brown, J . G., Dollimore, J.. Freedman, C. M., and Harrison, B. H., Thermochim. Acta, 1970, 1, 499. Langer, H. G., and Bradly, T. P., in Schwenker, R. I?., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 295. Chang. T. L.. and Mead, T. E., Analyt. Chem., 1971, 43, 534. Gibson, E. K., Thermochim. Acta, 1973, 5, 243. Merritt, C., Sacher, R. E., and Petersen, B. A., J . Chromat., 1974, 99, 301. Doyle, C. D., “Evaluation of Experimental Polymers,” WADD Technical Report 60-283, May, Rogers, R. N., Analyt. Chem., 1967, 39, 730.Lehman, H., and Gatzke, H., Tonind.-Ztg Keram. Rdsch., 1956, 80, 7. Koehler, E., Bull. SOC. Fr. Ckram., 1958, 38, 3. Pearce, J. H., and Mardon, P. G., J . Scient. Instrum., 1959, 36, 457. Paulik, F., Paulik, J., and Erdey, L., Mikrochzm. Acta, 1966, 894. Paulik, F., and Paulik, J., Thermochim. Acta, 1971, 3, 13. Bussiere, P., Claudel, B., Renouf, J. P., Trambouze, Y., and Prettre, M., J . Chim. Phys.. 1961, 58, Zaborenko, K. B., Melichov, L. L., and Portjanoi, V. A., Radiokhimiya, 1965, 7, 315. Balek, V., J . Mater. Sci., 1969, 4, 919. Habersberger, K., and Balek, V., Thermochim. Acta, 1972, 4, 457. Emmerich, W. D., and Balek, V., High Temp. High Pressures, 1973, 5, 67. Satava, V., Colln Czech. Chem. Commun. Engl. Edn, 1959, 24, 3297. Berg, L. G., and Burmistrova, N.P., Zh. Neorg. Khim., 1960, 5, 676. Budnikov, P. P., Gorshkov, V. S., and Titoskaya, V. T., Stroit. Mater., 1960, 6 (12), 30. Pannetier, G., Michel, A., Bergeault, J. M., and Djega-Maridassou, G., Bull. Soc. Chim. Fr., 1963, Bessonov, A. F., and Ustyantsev, V. M., Zav. Lab., 1965, 31, 620. Chiu, J., Analyt. Chem., 1967, 39, 861. Carrol, R. W., and Mangravite, R. V., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Proceedings of the Second International Conference on Thermal Analysis, Worcester, Mass. , USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 189. David, D. J., Thermochim. Acta, 1970, 1, 277. Judd, M. D., and Pope, M. I., J . App2. Chem., Lond., 1970, 20, 380. Balek, V., Analyt. Chew., 1970, 42, 16A. Halmos, Z., and Wendlandt, W.W., Thermochim. Acta, 1973, 7, 95. Berg, L. G., and Shlyapkina, E. N., J . Thermal Analysis, 1975, 8, 417. Juranic, N., Karaulik, D., and Vucelic, D., J . Thermal Analysis, 1975, 7, 119. Halmos, Z., Seybold, K., and Meisel, T., in BuzAs, f., Editor, “Proceedings of the Fourth Inter- national Conference on Thermal Analysis, Budapest,” Volume 2, Akaddmiai Kiad6, Budapest, 1975, p. 429. Wefers. K.. Ber. Dt. Keram. Ges.. 1965, 42, 35. Ravich, G. B., Kododyazhnyi, V. Z., Brodov, V. G., Kuchumov, A. M., and Zhemarkin, A. I., Zh. Neorg. Khim., 1967, 12, 2256. Bessonov, A. F., Ustyantsev, V. M., and Taksis, G. A., Poroshkovaja Metallurgya, 1967, 7, 92. 1960. 668. 1204. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105.May, 1978 I N THERMAL ANALYSIS. REVIEW 437 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. Barret, P., Gerard, N., and Watelle-Marion, G., Bull. SOC. Chim. Fr., 1968, 3172. Wiedemann, H. G., Thermochim Acta, 1973, 7, 131. David, D. J., Thermochim. Acta, 1972, 3, 277. Loehr, A. A., and Levy, P. F., Am. Lab., 1972, 4, 11. Barrall, E. M., and Johnson, J . F., Thermochim. Acta, 1972, 5, 41. Chromy, S., Silikuty, 1974, 18, 105. Miller, R. P., and Sommer, G., J . Scient. Instrum., 1966, 43, 293. Dichtl, H. J., and Jeglitsch, F., Radex Rdsch., 1967, 716. Van Tets, A., and Wiedemann, H. G., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Pro- ceedings of The Second International Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 121. Simmons, E. L., and Wendlandt, W. W., Analytica Chim. Acta, 1966, 35, 461. Wendlandt, W. W., and Bradley, W. S., Thermochim. Acta, 1970, 1, 143. Chatterjee, P. K., i n BuzAs, f., Editor, “Proceedings of the Fourth International Conference on Garn, P. D., and Kessler, J. E., Analyt. Chew,., 1960, 32, 1563. Forkel, W., Naturwissenschaften, 1960, 47, 10. Amiel, J., and Paulmier, C., C.R. Hebd. Se‘anc. Acad. Sci., Paris, 1962, 225, 2443. Claudel, B., Thesis, University of Lyon, 1962. Pannetier, G., Bergeault, J. M., and Guenot, J., Bull. Soc. Chim. FY., 1962, 2158. Dormieux, J. L., C.R. Hebd. Se‘anc. Acad. Sci., Paris, 1963, 259, 579. Lagier, J. P., Quahes, R., and Paulmier, C., Bull. SOC. Chim. Fr., 1964, 5, 1082. Paulik, F., and Paulik, J., Hung. Pat., 163 305/1971. Paulik, F., and Paulik, J., Analytica Chim. Acta, 1972, 60, 127. Erdey, L., Paulik. F., and Paulik, J., Hung. Pat., 152 197/1962. Rouquerol, J., in Schwenker, R. F., Jr., and Garn, P. D., Editors, “Proceedings of the Second Inter- national Conference on Thermal Analysis, Worcester, Mass., USA, August 18-23, 1968,” Volume 1, Academic Press, New York, 1969, p. 281. Thermal Analysis, Budapest,” Volume 3, Akaddmiai Kiadb, Budapest, 1975, p. 835. Paulik, J., and Paulik, F., Analytica Chim. Acta, 1971, 56, 328. Paulik, F., and Paulik, J., Analytica Chim. Acta, 1972, 60, 127. Paulik, F., and Paulik, J., Analytica Chim. Acta, 1973, 67, 437. Received October 24th, 1977 Accepted October 31st, 1977

ISSN:0003-2654    

DOI:10.1039/AN9780300417

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

438 Analyst, May, 1978, Vol. 103, pp. 438-446 Solid-state Mercury( I) Chloride Electrode for Determining 0.1-1.0 pg ml-I Levels of Chloride in Boiler Water and Other High-purity Waters G. B. Marshall and D. Midgley Central Electricity Research Laboratories, Kelvin Avenue, Leatherhead, Surrey, KT22 7 S E A solid-state electrode based on mercury(1) chloride and mercury(I1) sulphide has been developed for determining chloride concentrations of 0.01-1 .O pg 1-1 in boiler water. The greater sensitivity of the electrode compared with that of silver - silver chloride electrodes enables concentrations as low- as 0.01 pg nil-1 to be determined by a simple manual technique. The total standard deviations a t chloride concentrations of 0.5, 0.1 and 0.01 pg ml-l were 0.025, 0.005 and 0.015 pgml-l, respectively.The electrode can be prepared easily in the laboratory from commercially available materials and a RdiiCka Selectrode. The only significant interference i:; from iron(II1) ions and this interference can be eliminated by adding fluoride ions to the sample. Keywovds : Chloride determina,tion ; water anaLysis ; potentiometvy ; chlovide- selectize electrode ; mercury ( I ) chloride electrode In order to avoid acid attack in boilers, the concentration of chloride in boiler waters must be kept at very low levels [in Central Electricity Generating Board (CEGB) practice a maximum of 0.2-2 p g m F , depending on the type of boiler]. The established CEGB manual potentiometric method1 and absorpt iometric methods based on mercury( 11) thio- cyanate are insufficiently sensitive for chloride concentrations much below 0.1 pg ml-l, such as occur in many boilers.The range of the potentiometric method has been extended by using a flow cell whose temperature is precisely controlled2 and that of the absorptiometric method by first concentrating the chloride by a co-precipitation te~hnique.~ In both methods the analytical procedure is fairly complicated. The existing potentiometric method uses electrodes based on silver chloride, but electrodes made from mercury( I) chloride should be more suitable for determining low concentrations of chloride because of the lower solubility of the mercury(1) salt. The difficulty of handling mercury - mercury(1) chloride electrodes has largely restricted them to use in reference electrodes, but an experimental solid-state ion-selective electrode with a membrane composed of a compressed mixture of mercury(I1) sulphide and mercury( I) chloride4 showed promise of being easy to use.We have combined the same membrane materials with a RGiiEka Selectrode5 to produce a solid-state chloride-selective electrode that can be made easily from commercially available components and is suitable for determining chloride at concentrations as low as 0.01 pg ml-l by a simple manual technique. Theoretical As with calomel electrodes, the presence of chloride in a solution disturbs the solubility product equilibrium of mercury(1) chloride [equation (l)], but there is also a simultaneous solubility equilibrium for mercury (11) su1phid.e [equation (2)]. The two solubility equilibria are linked by the disproportionation equilibriiim of mercury(1) ions [equation (3)] :MARSHALL AND MIDGLEY 439 The e.m.f.of an electrode containing mercury(1) chloride and mercury(I1) sulphide both in equilibrium with an aqueous solution of chloride ions should be given by a form of the Nernst equation, which is expressed in terms of either a chloride or sulphide response: E = EoRgaCl4 - k lOg(Cl-} . . .. .. - . (4) .. .. . . . . (4a) k 2 = E o H g S - - log{S2} where k = RTlnlO/F is the slope factor and (S2-1 is the activity of the sulphide ion dissolved from the electrode. From equations (1)-(3) (s2-} = Kggs {C1-}2/KdK,g,a, .. .. * ' (5) and hence from equations (4) and (4a) Experimental Apparatus Potentials were measured with an Orion 801 digital pH meter reading to 0.1 mV and displayed on a Servoscribe 2s chart recorder.When two electrodes were being used simul- taneously they were switched in turn through the pH meter by an Orion 855 automatic electrode switch. The reference electrodes were of the mercury - mercury(1) sulphate type with a ground- glass sleeve liquid junction and 1 niol 1-1 sodium sulphnte filling solution (Electronic Instru- ments Ltd., Type 1380-230). Pvepavntioqz of clzlovide clectvodes A sensitising mixture was prepared by mixing red mercury(I1) sulphide (BDH, Optran grade) and mercury(1) chloride in equimolar amounts and grinding them in an agate mortar until the mixture was uniformly pink. The mercury(I1) sulphide had been washed for 15 min with each of two portions of carbon disulphide, washed with acetone and air dried before use.Because of their toxicity these materials should be handled with care. Radiometer F3012 Universal Selectrodes were impregnated with the above mixture by rubbing the mixture into the exposed graphite surface with a glass rod? Reagents Water. Town mains water was distilled in a stainless-steel still (Manesty Machines Ltd., Liverpool), and the distillate passed through a twin-column mixed-bed de-ionisation unit (Elga Products Ltd., Model BlO6jZ). The conductivity of this water was less than 0.1 pS cm-l a t 20 "C as i t left the unit and was found by a modification of the method of Rodabaugh and Upperman3 to have a chloride content of about 0.7 pg 1-l. This water was used in the preparation of all standard or reagent solutions.A stock solution (1 000 pg ml-l of chloride) was prepared by dissolving 1.649 g of sodium chloride in water and making up to 1 1 in a calibrated flask. Further standard solutions were prepared by successive dilution of this solution. A solution was prepared by dilution of 6.25 ml of concentrated nitric acid (BDH, Aristar grade) to 11. Commercially available mercury( 11) sulphide (BDH, Optran grade) was used for most experiments, but a batch of material was also prepared in the laboratory as follows. Stadard chloride solutioi~s. Nitric acid, 0.1 mol 1-l. Mercury(I1) sul@hide.440 MARSHALL AND MIDGLEY : SOLID-STATE MERCURY(I) CHLORIDE Analyst, VOZ. 103 Sodium sulphide nonahydrate (AnalaR grade) (12 g) was dissolved in about 70 ml of water and the solution filtered through a 0.45-pm Millipore filter.The filtrate was added to a solution of 13.7 g of mercury(I1) nitrate (AnalaR grade) in about 100 ml of water. The precipitate was filtered off, washed with de-ionised water and dried. The material prepared by precipitation was black 13-cinnabar but the commercial product was the red a-cinnabar form. Analytical Procedure E.m.f. measurements were made with the electrodes immersed in stirred 50-ml portions of standard or sample solution to which 5 ml of 0.1 mol 1-1 nitric acid had been added. Between each measurement the electrodes were immersed in a stirred rinsing solution of 0.01 moll-1 nitric acid for about 2 min until the e.m.f. was less negative than that recorded with any of the solutions being analysed (about -30 m'V for solutions below 0.1 pg ml-1 of chloride).The analytical measurement therefore was always obtained from an electrode responding to an increase in concentration, which procedure gave the best over-all response time, and, in addition, contamination of one sample or standard solution by a more concentrated predecessor was prevented. Concentrations of chloride were obtained from a calibration graph prepared by making measurements, as above, with standard solutions. Before the start of each batch of analyses the ground-glass sleeve of the reference electrode was flushed by allowing a few drops of the internal filling solution to flow out. If this precaution was neglected, the e.m.f. tended to drift, as also occurred after a time when reference electrodes with ceramic frit junctions were used. Results Electrode Characteristics Nature of the membrane material Electrodes were sensitised with equimolar mixtures of mercury( I) chloride and either the red or black form of mercury(I1) sulphide.The electrodes with the red form had Nernstian responses (57-59 mV per decade) at concentrations above 1 pg ml-l, while those made with the black form had calibration slopes 1-2 mV per decade lower and were also much slower to respond. When the red mercury(I1) sulphide was washed with carbon disulphide, the sensitivity and response time of the electrodes were further improved. Electrodes prepared with two batches of red mercury(I1) sulphide had almost identical characteristics. Because of its poorer performance, no further tests were made using the black mercury(I1) sulphide. Electrodes prepared with mercury(1) chloride only were still sensitive to chloride, but the e.m.f.difference was only 45 mV between 0.1 and 1 pg ml-l solutions (cf., Table 11) and about 30 min were required for a steady potential to be established; for these reasons no detailed work was done with electrodes of this type. Variations between electrodes The amount of sensitising mixture in the membrane and the pressure exerted in forming it on the end of the electrode did not affect the performance of the electrode. The four Radiometer F3012 electrodes used during the tests gave similar results; the potentials of electrodes in the same solution rarely differed by more than 2 mV, regardless of the concentra- tion of the solution.Removal of half of the membrane and exposure of the graphite substrate had no deleterious effect on the electrode and. did not change the standard potential or the calibration slope. Conditioning of electrodes pared with the same electrode after use for 1 d. chloride solution was effective in conditioning the electrodes. Freshly impregnated electrodes showed a sluggish response and a lower sensitivity com- Immersion overnight in a 10 pg ml-1M a y , 1978 ELECTRODE FOR DETERMINING CHLORIDE IN BOILER WATER 441 Optimum PH f o r chlovide determinations At pH values of 3 or above, the formation of hydroxo complexes of mercury reduces the sensitivity of the electrode at the lower end of the concentration range ((10 pg ml-l). The addition of nitric acid to the sample reduced this interference (Table I) and it was found that there was little practical difference in the sensitivity over the range 10-3-10-2 mol 1-1 of nitric acid.Once enough acid has been added to overcome the hydroxide interference, the shift in the potential with increasing acid concentration at constant chloride concentra- tion is a more likely source of error than small variations in sensitivity; a concentration of mol 1-1 of nitric acid was, therefore, adopted for the analytical procedure, as the greater buffer capacity was considered advantageous. TABLE I EFFECT OF ACIDITY ON ELECTRODE RESPONSE (ELECTRODE E.M.F. IN MILLIVOLTS) Nitric acid Chloride concentration/g ml-I concentration1 I A I moll-1 0" 0.1 1.0 10 100 - 34.5 -83.5 - - 0 - 67 - - 94 - 141 - 204 10-3 - 13 10-2 - 24 -47.5 - 96 - 155 -213 * De-ionised water, no added chloride.E f e c t o f stirring with 5 min in a stirred solution. equilibrium e.m.f. Without stirring, the electrode took 20-30 min to reach an equilibrium e.m.f,, compared The rate of stirring had little effect (<2 mV) on the E f e c t of light The potential decreased (indicating an apparently higher chloride concentration) when the electrode was exposed to more intense light and increased when it was shaded. The effect was reversible in all instances. Switching off the laboratory fluorescent lights on a dull day caused a 2-mV change in potential within 1 min and similar changes could be produced by variations in the intensity of sunlight. For the main body of the tests the electrodes were fixed in an opaque holder which fitted on top of black-painted 100-ml beakers, thus excluding virtually all light from the membrane of the electrode.Resistance The resistance of the cell formed by the chloride electrode and the reference electrode immersed in 50 ml of 0.1 pg ml-l chloride solution to which 5 ml of 0.1 mol 1-1 nitric acid had been added was 30 kQ, as measured with an Avometer. This resistance is small enough to allow the cell potentials to be measured with many types of digital voltmeter; potentials measured with a Solartron LM 1420.2 digital voltmeter were within 0.1 mV of those measured on a pH meter with a high input impedance. Observed and Predicted e.m. f.s The e.m.f. observed when the electrode is immersed in a 1 pg ml-1 chloride solution at 25 "C and pH 2.5 (obtained by addition of 5 ml of nitric acid per 50 ml of solution) can be compared with the values predicted by equations (4) and (4a).The standard potentials6 are E& = - 750 mV and E&ZClo = 268 mV and the chloride ion activity is 2.32 x mol 1-1 (allowing for dilution and using activity coefficients calculated from the Davies equation') ; hence E = E&,a, - 59.16 log{Cl-} = 542 mV From equation (5) and the equilibrium constants6 for equations (1)-(3), and hence {SZ-) = 1.2 x 10-43 mol 1-1 E = E& - 29.58 log(S2-} = 520 mV442 MARSHALL AND MIDGLEY : SOLID-STATE MERCURY(1) CHLORIDE Analyst, VOZ. 103 The observed values, after correction for the :potential of the reference electrode, were in the range 530-537 mV.The agreement between the observed and predicted e.m.f.s is acceptable, particularly in view of the uncertainty in the value of K,,,. Performance Tests The electrodes used for the following tests were made with red mercury(I1) sulphide that had been washed with carbon disulphide and were immersed in stirred chloride solutions containing 0.1 moll-1 nitric acid (5 ml per 60 ml) in darkened glass beakers. Electrodes were conditioned overnight in 10 pg ml-l chloride solution before being used for the first time. Unless otherwise stated, the solutions were brought to a temperature of 25 "C in a water-bath before being analysed. Concentration range The response of the electrode is Nernstian from at least 1 000 down to about 0.35 pg ml-1. Below that level the calibration graph becomes increasingly curved.Fig. 1 shows the curved portion of a typical calibration graph. The slope of the linear portion (1-10 p g ml-l) was -57.8 mV per decade increase in concentration, with a standard deviation of 0.25 mV per decade increase in concentration (estimated from the least-squares fit of the points to a straight line). It is possible to use the non-linear part of the calibration graph to measure chloride concentrations down to 0.01 pg ml-l and the sensitivity can be increased by working at reduced temperatures. C h I or i d e con ce n t r a t i o n / p a m I -- ' 1.O 0.1 @.Of 0.001 - 1 I 10-5 10--6 10--7 C h I or i d e con ce n t r a t i on / m o 1 I -- Fig. 1. Calibration of chloride ion- selective electrode.Precision Over a period of 3 d, five batches of five standard solutions each were analysed in duplicate. The e.m.f. values were normalised with respect to the mean e.m.f. reading for the 1 pg ml-2 standard in each batch and the within-batch, between-batch and total standard deviations were calculated. The results for one electrode are shown in Table 11. A second electrode immersed in the same solutions at the same time gave almost identical results, in which neither the mean values of the normalised e.rx1.f.s nor the standard deviations were signifi- cantly different, at the 5% level, from those in Table I1 (t- and F-tests, respectively). Table I11 shows the recorded e.m.f. values for the two electrodes in batches of 1.0 pg ml-1 chloride solution. The correlation coefficient between the two sets of e.m.f.s was 0.97, showing the high degree of co-variance between electrodes immersed in the same solution at the same time.It is inferred that the changes in e.m.f. are caused less by the variability of the chloride electrodes than by factors that could affect both simultaneously, e.g., small changes in temperature, pH or liquid junction potential.May, 1978 ELECTRODE FOR DETERMINING CHLORIDE IN BOILER WATER TABLE I1 443 PRECISION OF MEASUREMENTS OF CHLORIDE CONCENTRATIONS Standard deviationt Chloride concentration/ A e.m.f./ I A \ pg ml-1 mV* Within-batch Between-batch Total 1.0 0.0 0.63 - - (0.03) 0.5 17.6 0.93 0 0.93 (0.02 5) (0) (0.025) 0.1 51.1 0.83 NSS 0.91 (0.004) (0.0 04) (0.00 5) (0.005) 0.05 59.3 0.74 NS 0.75 0.01 64.6 2.04 0 2.04 (0.0 15) (0) (0.0 15) * E.m.f.normalised with respect to 1 p g ml-l solution, e.g., AOel = EOal - El.o. t Figures in parentheses are standard deviations in concentration units ( p g ml-l). NS = non-significant a t the 5% level. TABLE I11 VARIABILITY OF E.M.F.S OF TWO ELECTRODES IN 1.0 pg ml-l CHLORIDE SOLUTION E.m.f./mV 1 2 3 4 6 r L \ Electrode ,-A-3 r - A - l ,-A-~ c p A - \ number A* B* A B A B A B A B 23 - 106.0 - 106.3 - 103.3 - 103.8 - 103.2 - 105.2 - 102.3 - 101.4 - 100.6 -100.9 24 -102.2 -103.4 -101.4 -101.8 -101.0 -102.3 -99.9 -99.3 -98.3 -98.9 * A and B are the first and second 1 p g ml-1 solutions, respectively, in each batch. Accuracy The accuracy of the analyses using the electrode was tested both by analysing samples of boiler water from five power stations and comparing the results with those obtained by the mercury(I1) thiocyanate absorptiometric method,s and by spiking the samples with 0.1 pg ml-l of chloride and measuring the recovery.The results are shown in Table IV. TABLE IV ANALYSIS OF BOILER WATERS Chloride contentlpg ml-l I Potentiometry Station Location A Boiler 1(A) Boiler 1 (B) B Unit 4(A) Unit 4(B) C Unit 1 Unit 2 D Unit 1 Unit 3 E Boiler 1 Boiler 2 Absorptiometry : sample concentration 0.17 0.18 0.08 0.09 0.11 0.07 0 01 0.01 0.01 0.03 r Sample concentration 0.17 0.17, 0.10 0.10 0.08 0.06 0.01 0.01 0.01 0.01 1 Recovery of 0.1 p g ml-l spike 0.11 0.11 0.11 0.10 0.10 0.11 0.10 0.10 0.10 0.10444 MARSHALL AND MIDGLEY: SOLID-STATE MERCURY(1) CHLORIDE Analyst, VOZ. 103 Response time The response for a change from de-ionised water to 0.1 pg ml-l chloride solution or from 0.1 to 1.0 pg ml-l chloride solution was complete in 5 min, but changes in the reverse direction took longer (15-20 min).A better over-all time (approximately 5 min) for the analytical procedure was obtained by immersing the electrode in stirred 0.01 moll-1 nitric acid solution for about 2 min before the next solution was analysed. Interferences Substances that could occur in power station waters were tested for their interference effects by observing the change in e.m.f. of the electrode when 100-pl portions of concentrated standard solutions of the interferents were injected into a mixture of 50 ml of 0.1 pg ml-1 chloride solution and 5 ml of 0.1 mol 1-1 nitric acid.Dilution of the chloride solution on addition of the interferent solution was calculated to cause a change of less than 0.05mV, which would not have been detectable on the pH meter. The concentrations of interferents tested were generally much higher than those expected in power station waters. The following (separately) caused an interference no greater than 0.001 pg ml-l in the determina- tion of 0.1 pg ml-l of chloride: 2 pg ml-l of 200 pg ml-l of carbon dioxide, 0.43 pgml-1 of silicon dioxide, 120 pg ml-l of CHJOO-, 20 pgml-l of PO:-, 12.6 pgml-l of Cu2+, 20 pg ml-l of Ni2+, 5 pg ml-l of Ca2+ + 5 pg ml-l of Mg2+, 2 pg ml-l of Zn2+, 10 pg ml-1 of Cr3+, 0.1 pg ml-l of Fe2+, 0.02 pgml-l of Fe3+, 10 pgml--l of ammonia, 1 pg ml-1 of hydrazine, 2 pg ml-l of cyclohexylamine or 1100 pg ml-l of morpholine.caused a bias of +0.03 pg ml-l in the determination of 0.1 pg ml-l of chloride and 0.1 pg ml-l of Fe3+ caused a bias of -0,034 pg ml-l, which was eliminated by the addition of fluoride. Anions that form insoluble mercury( I) salts will interfere if present in such concentrations that they can displace chloride from the mei-cury(1) chloride in the membrane. The most strongly interfering ions of this type are sulphide, cyanide, iodide and bromide, but these are not normally present in power station waters. The interference effects of some ions can be eliminated, as in the proposed analyticad procedure, by acidifying the solution, e.g., hydroxide, carbonate and hydrogen phosphate. Sulphate interfered at much lower con- centrations than predicted by this mechanism, probably because the solubility of mercury( I) chloride was enhanced by the formation of a soluble Hg,SO,O complex.Substances that form strong complexes with mercury(I1) ion will promote the dispro- portionation of mercury(1) ion, causing chloride to be released from the membrane and the electrode to indicate a higher chloride concentration than was present in the sample solution. Under the conditions of the analytical procedure, 16 pg ml-l of sulphite caused an inter- ference equivalent to 3 pg ml-l of chloride because of the reaction The addition of 20 pg ml-l of Hg,Cl, + 2S0,2- -+ Hg(S0,),2- + Hgo + 2C1- Cations that form strong chloro complexes will reduce the concentration of free chloride, but this interference can be overcome by adding a substance that will compete with chloride for the interferent but not for mercury(1) ions.The only important interferent of this kind likely to be present in boiler water is iron(IIl[), the effect of which is eliminated by adding fluoride ions, e.g., 0.1 p g ml-l of Fe3+ reduced the reading obtained with 0.1 pg ml-1 of chloride by 0.034 pg ml-l, but with 20 pg ml-L of fluoride present the interference effect was less than 0.01 pg ml-l of chloride. Iron(I1) and other divalent metal ions had no detectable interference effects at the concentrations tested, which are much higher than those likely to be found in boiler water. E f e c t of temperature The Nernstian sensitivity of the electrode increases with temperature, according to the factor RTlnlOIF, but at reduced temperatures,, because the solubility of mercury( I) chloride is suppressed, the linear calibration range is extended to lower concentrations.In addition, there is a shift in the standard potential of the cell formed by the chloride and reference electrode pair to more negative e.m.f. values ,as the temperature decreases. The combined effect of these factors was tested by measurements made at three temperatures (Table V).May, 1978 ELECTRODE FOR DETERMINING CHLORIDE I N BOILER WATER 445 For the analysis of solutions containing less than 1.0 pg ml-l of chloride, the shift in the standard potential is a greater source of error than the change in sensitivity. For measure- ments below about 0.05 vg. ml-l of chloride, it may be worth working at reduced temperatures to obtain greater sensitivity, in spite of the longer response times, e.g., about four times longer at 7-10 "C than at 25 "C.TABLE V EFFECT OF TEMPERATURE ON E.M.F. E.m.f. /mV Concentration of I A > chloride/pg ml-l 25 "C 15 "C 7 "C 1.0 - 87 - 108.5 - 115 0.1 - 39 -59.5 -67.5 0.01 - 25 - 32 - 35 0" -22.5 - 24 - 29 * De-ionised water, no added chloride. Lifetime of the electrode Each application of the sensitising mixture gave a membrane with a lifetime of at least 2 months. When the response of the electrode became sluggish the membrane was shaved off the end of the electrode with the tool provided and the exposed tip re-impregnated with more of the sensitising mixture. The RfiiiEka Selectrode should last for years, its lifetime being limited only by the loss of material each time an old membrane is shaved off.The sensitising mixture can be stored for at least 6 months without ill-effect. Table VI shows the potentials recorded during the life of a typical electrode. As the measurements were made at ambient temperature, the potentials could not be expected to be constant over the period of trial, but no trend in the calibration slope (exemplified by the e.m.f. difference in Table VI) or the standard potential could be discerned. TABLE VI ELECTRODE RESPONSE AS A FUNCTION OF TIME The electrode was prepared on May 25th, 1976. Date of test 26.5.76 10.6.76* 14.6.76 21.6.76 20.7.76 23.8.76 1.9.76 E.m. f .ImV Chloride content Chloride content 1 pgml-l 0.1 pg ml-l f J. \ - 147.2 - 96.5 - 104.2 -53.9 - 103.4 -53.2 - 101.9 -50.8 - 107.7 -57.5 - 108.3 -55.7 - 105.1 - 54.4 E.m.f.difference/ mV 50.7 50.3 60.2 51.1 50.2 52.6 50.7 * Reference electrode changed. Discussion Lechner and Sekerka4 made electrodes with pelleted membranes of the same materials as we have used and obtained generally similar results, although their response times were shorter. The precision of analyses with their electrodeg was poorer than that reported here, e.g., at 0.1 pg ml-l they obtained a relative standard deviation of 14% compared with 4% in this work. The performance of the pelleted membranes depended on the temperature and duration of compression and at best 3 h at 150 "C were required for the preparation of a successful electrode, whereas variations in preparing the impregnated graphite electrode had no appreciable effect on the performance.The solid-state chloride-selective electrode described in this paper is suitable for the analysis of power station waters containing 0.01-1.0 pg ml-l of chloride, using a simple446 MARSHALL AND MIDGLEY manual technique. Higher concentrations can also be determined, if required. At 0.1 pg ml-1 of chloride, better precision can be obtained with this electrode than with the silver - silver chloride type1 (0.004 pg ml-l compared with 0.04 pg ml-l for the total standard deviation), although at 1 pgml-1 the two electrodes are equally precise. At concentrations below 1 pg ml-l the mercury(1) chloride electrode is much more sensitive than the silver chloride types, e.g., the difference between the e.m.f.s observed in 1.0 and 0.1 pg ml-1 chloride solutions is 51 mV for the mercury(1) chloride electrode and 19 mV for silver chloride electrodes and between 0.1 and 0.01 pg ml-l the corresponding differences are 13.5 and 2 mV, respectively.The mercury(1) chloride electrode is slower to reach equilibrium than the silver chloride type, taking 5 min to reach a steady e.m.f. where the latter requires less than 1 min. The mercury(1) chloride electrode is similar in precision at 1 pg ml-1 of chloride to the mercury(I1) thiocyanate absorptiometric method,1° but its precision is better at lower concentrations. Power station waters are unlikely to contain sufficient concentra- tions of interfering substances to bias the results of analyses and generally good agreement was obtained when samples of boiler water from several power stations were analysed using the electrode and the mercury( 11) thiocyanat e absorptiometric method. Good recoveries of spikes were obtained from the same samples and there was no evidence of an effect from iron(II1) ions, the most likely source of interference. The sensitivity of the electrode could make it suitable for monitoring very low concentra- tions of chloride ((0.01 pg ml-l), especially if used in a flow-cell at a reduced and carefully controlled temperature, as has been done with silver chloride electrodes.2 Further work is in hand to assess the performance of the electirode in continuously flowing solutions at low temperatures. This work was carried out at the Central Electricity Research Laboratories and is published by permission of the Central Electricity Generating Board. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. References Torrance, K., Analyst, 1974, 99, 203. Tomlinson, K., and Torrance, K., Analyst, 1977, 102, 1. Rodabaugh, R. D., and Upperman, G. T., Analytica Chim. Acta, 1972, 60, 434. Lechner, J. F., and Sekerka, I., J . Electroanalyt Chem. Interfacial Electrochem., 1974, 57, 317. Hansen, E. H., Lamm, C. G., and RbiiCka, J., .AnaZytica Chim. Acta, 1972, 59, 403. Latimer, W. M., “Oxidation Potentials,” Second Edition, Prentice-Hall, Englewood Cliffs, N. J . , Davies, C. W., “Ion Association,” Butterworths, London, 1962. Florence, T. M., and Farrar, Y . J., Analytica Chim. Acta, 1971, 54, 373. Sekerka, I., Lechner, J. F., and Wales, R., Wat. Res.. 1975, 9, 663. Webber, H. M., Wheeler, E. A., and Wilson, A. L., unpublished work. 1952. Received November 16th, 1977 Accepted December 28th, 1977

ISSN:0003-2654    

DOI:10.1039/AN9780300438

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

Analyst, May, 1978, Vol. 103, @p. 447-451 447 Influence of Ascorbic Acid on the Matrix Interferences Observed During the Carbon Furnace Atomic-absorption Spectrophotometric Determination of Lead in Some Drinking Waters J. G. T. Regan and J. Warren Department of Industry, Laboratory of the Government Chemist, Cornwall House, Stamford Street, Lon don, SE19NQ Nine samples of drinking water taken from a range of locations in England and Scotland have been analysed for lead by using carbon furnace atomic- absorption spectrophotometry. Spiking experiments have been carried out in order to determine the severity of the matrix interference. The suppression of the lead signals ranged from 22 t o 84%. No relationship was found to exist between the hardness of a water sample and its suppression effect.Further spiking experiments carried out in the presence of 1% m / V of ascorbic acid showed that the suppression effect of eight of the water samples was reduced to a level of less than 5%. The remaining water sample gave a suppression of 18%. This water was not the hardest examined, nor did it give the highest suppression in the previous experiment, The natural lead contents of the nine waters were determined both by carbon furnace atomic-absorption spectrophotometry in the presence of ascorbic acid and by a method that involves solvent extraction - flame atomic absorption. Statistical analysis, using a t-test, indicated that there was no significant difference (at the 95% confidence level) in the results obtained by using the two techniques.Keywords : Lead determination ; drinking water ; carbon furnace atomic- absorption spectrophotometry ; matrix interferences ; ascorbic acid The analysis of drinking waters for lead by carbon furnace atomic-absorption spectrophoto- metry is made difficult by severe suppression of the lead signal caused by sample matrix constituents. The use of the method of standard additions appears to be an obvious way of surmounting the difficulty but this procedure is time consuming and, with signal suppressions as high as 84%, may still not yield a result. The addition of sulphuric acid or phosphoric acid partially overcame the effects due to magnesium but increased those due to calcium. Campbell and Ottawayl suggested that analysis in the presence of 10% V/V nitric acid overcomes the suppression of lead by calcium.It was found, however, that while this suppression was reduced to the 10-15% level, a severe loss of sensitivity, attack on atomiser components and spreading on drying, with resultant loss of precision, also occurred. Thompson et al.2 have proposed the use of lanthanum-treated tubes for the analysis of water for lead and cadmium and reported a reduction of suppression effects to the & 7% level. Our earlier paper3 on the use of ascorbic acid showed that its addition at the 1% m/V level overcomes numerous single-element interferences on the lead signal. Some further work has been carried out on the use of ascorbic acid in the determination of lead in some drinking waters in order to study its effectiveness in overcoming multiple interferences arising from a natural, rather than a synthetic, matrix.Various attempts to overcome these matrix effects were made, but without success. Experimental Apparatus A Perkin-Elmer HGA-72 heated graphite tube atomiser installed in a Perkin-Elmer 403 atomic-absorption spectrophotometer fitted with the optical modification was used together Crown Copyright.448 REGAN AND WARREN: INFLUENCE OF' ASCORBIC ACID ON MATRIX Analyst, VoZ. 103 with a Telsec chart recorder of 10 mV f.s.d. and a response time of 0.3 s for full-scale deflec- tion. Reagents Lead nitrate. Nitric acid (sp. gr. 1.42). Ascorbic acid. Specpure (Johnson Matthey Chemicals Ltd.). Aristar (BDH). Standard Laboratory Reagent (Fisons) . Operating Conditions All lead peak-height determinations were made on 50-pl volumes and the mean (coefficient of variation better than 5%) of at least four replicate injections was taken for each solution.Measurements were made using the lead absorbing line at 283.3 nm and a spectral band width of 0.7 nm (slit 4). The thermal programme used was: drying at 100 "C for 45 s; thermal destruction at 450 "C for 30 s; atomisation at 2 080 "C for 7.5 s; and maximum temperature burn-out for 5 s; all temperatures as shown on the instrument indicator. Auto- matic background correction was used throughout. The gas-stop facility was not used. All samples were made 0.015 M with respect to nitric acid on collection. Effect of Sample Matrix on the Lead Signal In order to determine the effect of the sample matrix on the lead signal a spiking experi- ment was carried out.Three millilitres of distilled, de-ionised water were added to an acid-washed, dry, 50-ml calibrated flask and sufficient sample was added to make up to the mark. The spiked sample was prepared by adding 2 ml of distilled, de-ionised water and 1 ml of a 2.5 pg ml-1 solution of lead (as nitrate) to the 50-ml flask prior to the addition of sample. Pasteur pipettes were used to make accurate adjustments of the final volume. The above procedure produced a solution containing 0.050 pg ml-l of added lead. The solutions obtained by this procedure were measured for apparent lead concentration by comparison with lead standard solutions in 0.015 M nitric acid. Some of the matrix constituents of the samples were determined to ascertain whether or not a simple relationship between these and suppression of the lead signal existed.Effect of Addition of Ascorbic Acid To ascertain the effectiveness, if any, of ascorbic acid in overcoming matrix interferences, the experiment was repeated in the presence of 1% m/V of ascorbic acid in samples, spiked solutions and standard solutions. The ascorbic acid concentration of 1% m/V was achieved by replacing 2 ml of the water added to the calibrated flasks with 2 ml of 25% m/V ascorbic acid solution. The resulting solutions were measured for apparent lead concentration by comparison with lead standard solutions in 1% m/V ascorbic acid and 0.015 M nitric acid. No reagent blanks for lead were found during this study. However, other batches of ascorbic acid, even from the same manufacturer, have been found to contain lead.Determination of Lead by Solvent Extraction - Flame Atomic-absorption Spectro- p ho t ome t ry The lead content of the samples was also determined by a method, based on that proposed by the Department of the Environment and the :National Water Council Standing Committee of Analysts, currently used in this Laboratory. The method consists in extraction of the lead as its ammonium tetramethylenedithiocarbamate complex into 4-methylpentan-Z-one, followed by measurement by means of flame atomic-absorption spectrophotometry. Results and Discussion Effect of Sample Matrix on the Lead Signal Table I, exhibit a range of signal suppressions varying from 22 to 84%. The results of the spiking experiment without the addition of ascorbic acid, given in Table I1 givesMay, 1978 INTERFERENCES I N THE AAS DETERMINATION OF LEAD I N WATER 449 some data concerning the matrix constituents of the waters examined, together with the conductivity of each sample.It can be seen that no simple correlation exists between the matrix constituents deter- mined and the degree of observed signal suppression. For example, sample D, with calcium, magnesium, chloride and sulphate concentrations of 98,29, 180 and 170 pg ml-l, respectively, caused a suppression of 44%, whereas sample E, with calcium, magnesium, chloride and sulphate concentrations of only 54, 14, 53 and 76 pgml-l, respectively, gave the highest suppression (84%) found in this study. TABLE I SUPPRESSION OF LEAD PEAK HEIGHT BY DRINKING-WATER SAMPLES Apparent concentration of lead r Sample A B C D E F G H I Sample/pg ml-1 Not.detected 0.044 0.067 Not detected Not detected Not detected Not detected 0.002 8 0.006 4 Sample containing 0.050 pg ml-l of added leadlpg ml-1 0.027 0.081 0.106 0.028 0.008 0.038 0.014 0.0205 0.028 0 Apparent amount of lead added/ pg ml-1 Suppression, % 0.027 46 0.037 26 0.039 22 0.028 44 0.008 84 0.038 24 0.014 72 0.017 7 65 0.021 6 57 Further, although samples B and F gave similar suppressions of 26 and 24%, respectively, they possessed widely different matrices, which could indicate that the effect of a single interferent is modified by other species present in the matrix. TABLE I1 SOME MATRIX CONSTITUENTS (pg ml-l) AND THE CONDUCTIVITY VALUES OF THE DRINKING-WATER SAMPLES EXAMINED Albumi- Sample N nitrogen NH, as noid A 0.03 0.05 B 0.01 0.03 c 0.01 0.01 D 0.05 0.27 E 0.03 0.07 F 0.01 0.01 G 0.01 0.03 H 0.01 0.01 I 0.02 0.03 Total Free residual residual Sample chlorine chlorine Oxygen absorbed Nitrite Nitrate from Chloride Alkalinity Hardness as N as N KRZnO,* as C1 as CaCO, as CaCO, Iron Zinc Copper 0.001 1.0 1.5 7 25 40 0.04 0.1 0.3 0,001 1.0 1.4 4 15 20 0.04 0.1 0.17 0.001 5.8 0.2 11 230 260 0.04 3.5 0.12 0.001 7.0 1.5 180 130 360 0.04 0.05 0.01 0.001 3.4 0.5 53 120 180 0.04 0.1 0.005 0.002 1.0 0.2 22 130 140 0.3 0.6 0.005 0.002 8.8 0.1 25 250 320 0.005 6.5 0.01 0.001 4.0 0.1 47 150 240 0.04 0.05 0.039 0.002 4.6 0.1 27 60 140 0.10 3.2 0.005 Conductivity/ Cadmium Fluoride Calcium Magnesium Sulphate Sodium Potassium pS cm-l A 0.05 0.05 0.001 0.1 10 1.6 1 5 4.4 0.5 120 B 0.05 0.05 0.001 0.1 3.4 1.0 9 12.4 0.7 80 C 0.05 0.05 0.001 0.1 98 2.1 60 18.0 1.8 480 D 3.2 2.0 0.002 0.1 98 29 170 14.9 4.4 790 E 0.05 0.05 0.001 0.60 54 14 76 39.5 3.4 540 F 0.05 0.05 0.001 0.1 52 2.6 34 11.9 2.7 320 G 0.05 0.05 0.001 0.80 112 2.8 24 13.8 2.1 600 H 0.05 0.05 0.001 0.135 48 26 80 47.8 3.6 530 I 0.05 0.05 0.001 0.4 45 2.5 6 12.8 2.2 310 * Oxygen absorbed in 4 h at 26.7 "C from ~ / 8 0 KMnO, solution (an empirical measurement of organic materials).Effect of Addition of Ascorbic Acid The results of the spiking experiment in the presence of 1% m/V of ascorbic acid, given in Table 111, indicate the effectiveness of ascorbic acid in overcoming the observed suppression effects of drinking-water matrices on the lead signal.It can be seen that with the exception of sample H the suppression effects have been reduced to a level of 5% or less.450 REGAN AND WARREN: INFLUENCE 01; ASCORBIC ACID ON MATRIX Analyst, VoZ. 103 The complexity of the matrix interference system is again indicated by the fact that the interference associated with sample D, which exhibited the highest suppression of 84%, was completely eliminated, whereas that associated with sample H, which had a smaller suppression of 65%, could only be reduced to the 18% level by the addition of ascorbic acid, TABLE I11 SUPPRESSION OF LEAD PEAK HEIGHT B Y DRINKING-WATER SAMPLES AFTER ADDITION O F ASCORBIC ACID I Sample A B C D E F G H I Apparent concentration of lead A 7 Sample containing 0.050 pg ml-l of Sample/pg ml-l added lead/pg ml-I Not detected 0.049 0.061* 0.111 0.081* 0.129 Not detected 0.048 Not detected 0.051 Not detected 0.050 0.0202* 0.0702 0.008 7" 0.049 5 0.012 6" 0.060 0 Apparent amount of lead added/ pg ml-1 0.049 0.050 0.048 0.048 0.051 0.050 0.050 0.040 8 0.047 4 Suppression, yo 2 0 4 4 -2 0 0 18 5 * These results should be multiplied by the dilution factor of 50/47 to give natural lead levels based on the original water sample.Accuracy A measure of the accuracy of the results obtained for lead by means of the carbon furnace atomic-absorption spectrophotometric analysis of samples treated with ascorbic acid can be gained from the results of the spiking experiment shown in Table 111. A further measure of the accuracy of these results was gained by comparison with results obtained by using the method involving solvent extraction followed by measurement by means of flame atomic- absorption spectrophotometry .It can be seen from Table IV that the difference in the values obtained by means of the two techniques was less than 10% relative. Statistical analysis of the data, using a t-test, indicated that there was no significant difference (at the 95% confidence level) in the results obtained by using the two different methods. TABLE IV COMPARISON OF RESULTS FOR LEAD IN DRINKING WATER OBTAINED BY THE ASCORBIC ACID METHOD AND THE SOLVENT EXTRACTION - FLAME ATOMIC-ABSORPTION SPECTROPHOTOMETRIC METHOD Concentration of lead + 2a*/pg ml-I A I -l Sample Ascorbic acid method. Solvent extraction - AAS method A Not detected B 0.063 & 0.003 C 0.086 f 0.003 D Not detected E Not detected F Not detected G 0.0215 f 0.0002 H 0.0096 f 0.0007 I 0.0133 4 0.0006 Not detected 0.058 & 0.003 0.092 f 0.002 Not detected Not detected Not detected 0.021 8 f 0.000 6 0.0098 f 0.0006 0.0147 f 0.0038 * 2a values are from a limited number (3-5) of independent determinations carried out at different times and using fresh sample aliquots and standards.Limit of Determination The limit of determination for lead in drinking water using the addition of ascorbic acid is 0.002 pg ml-l with a 5O-pl sample volume a:nd measurement at the 283.3-nm line. This limit is of the same order as that obtained with the solvent extraction-flame atomic-May, 1978 451 absorption spectrophotometric procedure.Further improvement could be expected on using a larger sample volume and the lead absorbing line at 217 nm. The use of the gas-stop facility may also prove beneficial. INTERFERENCES I N THE AAS DETERMINATION OF LEAD IN WATER Conclusions The suppressions of the lead signal exhibited by the nine waters examined, which were gathered from a wide range of geographical locations, bore no simple relationship to the hardness of the water or to the concentration of any of the individual matrix constituents studied. The complexity of this situation demonstrates that suppression effects on lead measured in artificial, single-interferent systems cannot be extrapolated to real, multi- interferent matrices. Also, the effectiveness of ascorbic acid in combating the suppression of the lead signal was not found to be related to the composition of the matrix or the actual degree of suppression caused by the matrix. Although the use of ascorbic acid was not completely successful in overcoming interference in the determination of lead, the suppression effect was greatly diminished, to less than 5% for eight of the water samples and to 18% for the remaining sample. Where suppression by the matrix reduces the lead signal to below the detection limit, the addition of ascorbic acid may be beneficial owing to (1) the reduction of the suppressive interference and (2) the enhancement of the lead signal, as previously reported.3 The authors thank the Government Chemist for permission to publish this paper and also the Water Section of the Laboratory of the Government Chemist for providing most of the matrix constituent data shown in Table 11. References 1 . 2. 3. Campbell, W. C., and Ottaway, J. M., Talanta, 1975, 22, 729. Thompson, K. C., Wagstaff, K., and Wheatstone, K. C., Analyst, 1977, 102, 310. Regan. J . G. T., and Warren, J., Analyst, 19’76, 101, 220. Received September 30th, 1977 Accepted December 6th, 1977

ISSN:0003-2654    

DOI:10.1039/AN9780300447

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

452 Autalyst, May, 1978, Vol. 103, pp. 452-468 Qualification of Estimates for Total Trace Elements in Foodstuffs Using Measurement by Atomic- absorption Spectrophotometry W. H. Evans Department of Industry, Laboratory of the Government Chemist, Cornwall House, Stamford Street, London, SEl9NQ The qualification of results for small concentrations of elements in foodstuffs implies a knowledge of the accuracy of a method when applied to foodstuffs and an assessment of the variation in results that exists in the application of that method. An attempt is made to describe the problems inherent in obtaining such qualifications, and to suggest a standard procedure for accomplishing these aims. From data obtained for a particular method, a statistical appreciation will give confidence limits and detection limits that can be applied to subse- quent results obtained, depending upon the nature of the exercise involved.Keywords : Accuracy and variation of results ; limit of detection ; statistical appreciation ; foodstufls analysis ; atomic-absorjbtion spectrophotometry Little work has been published on the statistical appreciation of analytical methods. General accounts do exist1y2--5 and between-laboratory testing has been described in detail." Exceptionally, certain types of instrumental measurement, such as radiochemical counting techniques' and emission spectrography,8 have been considered at length and some aspects developed, notably by Kaiser and Men~ies,~ for analytical procedures in general. An approach used for instrumental measurlement, however, is frequently not practical when applied to a complete analytical method. The outcome is often a method that is not well qualified, or that is arbitrarily qualified, e.g., in the determination of elements using atomic-absorption spectrophotometric measurement it is usual to quote a limit of detection of twice the instrument background variation.This may be satisfactory as an instrument qualification if the analyst is measuring levels 100-fold greater than this limit, but, if the determination is in the region of the true limit of detection of the method, the quotation of an instrument background variation is of little consequence. Without qualification it is not possible to ascertain the usefulness of a method, compare the results from two different methods or give a practical appreciation of an individual method when results are used for a more profound but tenuous purpose than monitoring.This is particularly so in the determination of trace-element contaminants in foodstuffs and other biological matrices, where they are usually present at levels just within or beyond the capacity of existing methodology . This paper illustrates some of the difficulties in obtaining valid laboratory qualifications, the means of obtaining estimates for them and the reliance that can be placed upon a result for a total element level in foodstuffs. The Problem An analytical method for the measurement of any determinand should be qualified to an extent that enables any subsquent result obtained by an analyst on defined matrices to have some meaning.It is generally accepted that the accuracy of results, i.e., the agreement between a mean of estimates and the correct value, must be tested by defining any systematic bias from the true value throughout the range of determinand concentrations measured. The variation of results obtained must be known, i.e., the standard deviation at the extremity of the concentration determined must be calculated. The level below which the variation creates intolerable uncertainty in determinand values must be deduced, giving a limit of detection for results. Crown Copyright.EVANS 453 Accuracy An analytical method for the determination of trace elements in foodstuffs, using measure- ment by flame atomic-absorption spectrophotometry, involves prior destruction of the organic matrix by, e.g., wet acid digestion or dry ashing, followed possibly by a separation and/or a concentration stage.Such a method may suffer from losses of an element, particu- larly during the chemically violent destruction stage, and hence a check on the accuracy of the method is essential. It is also necessary to ascertain that the laboratory conditions are such that contamination does not cause too high results for small amounts of these elements. Foodstuffs vary widely in organic composition and the digests obtained after destruction of the organic matrix may similarly vary in inorganic composition. The latter variation may introduce errors into the measurement stage because of interference effects. This could apply to the digests obtained from many foodstuffs and for those obtained from foodstuffs that contain an unusually high inorganic composition could cause major inter- ference; in such instances the method would be invalidated.Another aspect that must not be overlooked is the possibility that an element may be present in more than one chemical form. Although this may not apply to all elements, it has been suggested that in samples of marine origin, arsenic may exist as As3+, As5+, as specific methylated arsenic compounds or as complex high relative molecular mass organic compounds of unknown composition.lOJl Similarly, mercury in samples of marine origin occurs mainly as methylmercury compounds, but these occur infrequently in other foodstuffs. The accuracy of a method should therefore be established not only for the method itself, but also in its application to a range of different foodstuffs and ideally to several examples of a particular type of foodstuff.The measure of this accuracy will be the bias obtained from true values for results on homogeneous food samples, and this bias will be systematic or random. In the ensuing account, systematic bias is considered to invalidate results and random bias to increase the variation inherent in the results. Variation Suppose that there exists a foodstuff A, which is an example of foodstuffs A,, A,, A,, . . . If foodstuff A is sampled, there will be a sample B that is representative of A but similar samples B,, B,, B, . . . could be drawn. If an element C is measured in sample B, that element may exist in the sample in forms C,, C,, C,, .. . If element C in sample B of food- stuff A is now determined by analyst D, that analyst is representative of analysts D,, D,, D,, . . . Finally, if analyst D measures C in sample B of foodstuff A, there will be an uncertainty E associated with the measurement. The final result will be an estimate subject to variation, and this variation will depend in part on measurement, analyst, the elemental form in the foodstuff, sampling and the food- stuff itself. The analytical uncertainty E is inevitable; it is the variation inherent in a method (including pre-treatment) reflected by replicate measurement when all other factors are fixed. The main aim of monitoring is not measurement, however, but results, in particular a result starting from a sample B of a foodstuff A and proceeding through a fixed method, and it is confidence associated with that one result that is of interest.It follows that the analytical uncertainty of a result must include also the variation from sampling, and this total uncertainty must be determined by replicate analysis. It will also include variation during pre-treatment of a series of samples and the variation in instrument performance during measurement of such a series. The variance of this analytical uncertainty is the experi- mental error variance, so2, from which the standard deviation, the repeatability so, of an analyst on a matrix may be obtained. In any laboratory it is not usually possible for one analyst to specialise in one type of analysis and hence a laboratory qualifica- tion is necessary, embracing a cross-section of analysts; any new analyst must work within this laboratory qualification.(It should be noted that in collaborative exercises this variation assumes the variation between laboratories.) While it may be inferred what other laboratories could achieve, an analyst’s main preoccupation should be with the laboratory of which he is a member. We have so far considered onlv foodstuff A, but there are unlimited kinds of foodstuffs. No two analysts perform in an identical manner. Let the variance contribution from analysts be sa2.454 EVANS: QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, Vd. 103 This statement as it stands means little, but iE, because foodstuffs differ, variation occurs in the application of a method, or part of it, it can be said that variation occurs because of the nature of the foodstuff.Hence, if some food matrices are such that they take longer to digest than others, this may be a cause of variakion in results for the total element determined in foodstuffs generally. Also, the chemical state of the element to be determined may differ wholly or partly from one foodstuff to another, or even in examples of the same foodstuff. This may introduce variation into the breakdown of molecular species, or hinder release from bindings with high molecular weight organic molecules to the inorganic species measured. There is therefore a measure of uncertainty attached to the application of a method to foodstuffs, which also includes the nature of the element in the food matrix.(It is assumed, within the context of bias, that a method is generally applicable.) Let the variance caused by matrices be denoted by sm2. Depending upon the experimental design, the variation caused by pre-treatment and measurement between series of blanks, standards and one sample, and also the variation inherent in the blank for each series, will be contained within either the variance sa2 or s,2. The variation inherent in the determination of an element can therefore be said to depend on the experimental error, the analyst and the nature of the matrix, and the standard deviation, s, of an estimate on a single sample by any analyst on any food matrix will be .t/so2 + sa2 + .sF2. This expression is similar to that described by Youden and Steiner6 for the reproducibility of inter-laboratory exercises. But, as with accuracy, a range of food- stuffs and a sufficient number of analysts are difficult to encompass.Limit of Detection Two quantities are quoted for the use of atomic-absorption spectrophotometers : the limit of detection is taken to be the concentration equivalent to twice the background noise variation at the wavelength of measurement, and the sensitivity is the concentration corresponding to 1% absorption. Neither is of importance except as a guide to instru- mental behaviour and/or the condition of the element emission sources. Many workers use the variation in the signal or response of the total reagent blank of the method as a basis for calculation. This is hardly tenable if the method involves treatment of the blank in a different manner to the samples, e.g., reduction of the volume of acid used in digestion by boiling is not comparable to the violent oxidation reaction involved in the destruction of organic matter.Also, the varying ionic composition of sample digests may be the cause of random bias, defined as varialion, in the remainder of the method, which will not be reproduced in the sample blank. A few workers have advocated detection limits based upon the variation of the response (or signal) from a determinand at a level similar to the limit of detection (let the standard deviation be sd). In such instances, the variation in response (or signal) of the reagent blank of the method (standard deviation sb) is taken into account and the detection limit is then based upon a multiple of the standard deviation l / s d 2 + sb2 (refs.7, 12 and 13). How- ever, again it is results that are of primary interest and, for this discussion, results for an element in different foodstuffs with different elemental forms and obtained by different analysts. It is reasonable to assume that the limit of detection must be based in some way on the variation in results, or the variation in recovery results, obtained for very small amounts of element and hence must be contiguous with the assessment of variation in estimating results. A Design Before proceeding to a design for obtaining satisfactory qualification, it is necessary to accept the concept of a complete analytical procedure, which has been described in detail el~ewhere.~*g The most important aspects that need to be defined are the food matrices to which the method is applicable and the particular analytical method used.This may appear to be straightforward, but it is impossible to design a method that can be applied to all food matrices and that will iremain an efficient analytical procedure. In the first instance, the technique for initial destruction of organic matter might not be generally satisfactory, e.g., in samples of marine origin a normal wet oxidation may not break down complex compounds of organically bound arsenic that are present to an inorganic species.May, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS 455 Alternative methods of destruction have to be used, such as wet oxidation with vanadium(V) oxide14 or dry combustion.15 Normal wet oxidation will be satisfactory, however, for the determination of arsenic in other foods, as far as is known at present.Similarly, digests obtained from foodstuffs vary widely in inorganic composition, e.g., it is possible for liver or tomato products to contain levels of copper that interfere in the determination of nickel by its extractive chelation into an organic solvent and measurement by flame atomic-absorption spectrophotometry,16 or excessive iron levels may interfere in lead and cadmium determina- tions by a similar method. These examples do not mean that a method should be discarded on account of the few occasions on which it is not applicable, but rather that those occasions should be defined, by a knowledge of the composition of the samples under examination and by careful consideration of the interference effects that apply to the method.The application of an analytical method to foodstuffs should also include a knowledge of the homogeneity of samples, but this cannot be assessed until results from a selection of foodstuffs are obtained. It is not unreasonable to assume that foodstuff samples are homogeneous until proved otherwise. The details of the analytical method must be clearly defined so as to include, for instance, the type of atomic-absorption spectrophotometer, as the performance of an instrument depends upon its design. It has been mentioned that trace contaminant levels are often outside or just within the limits of existing methodology.A calibration graph should define the lower and upper range of measurement and should be linear. Few analysts would be prepared to project a response graph, even if linear, beyond the upper calibration point, yet it is not uncommon to find results derived from readings obtained below the lowest calibration point. I t follows that once a limit of detection has been established, a calibration graph must be prepared that contains a calibration point equivalent to or near to this limit of detection. If difficulty is experienced in measuring such a standard, then clearly the limit of detection is false a t that time. In addition, it must be known whether the experimental pre-treatment and, in particular, instrument measurement are randomised to include standards and samples in any order, or whether the experimental factors involve the determination of a level in a single sample prepared and measured discretely from other samples and standards.Measurements by atomic-absorption spectrophotonietry in some modes suffer from memory effects, which, even if eradicated , may occasionally re-occur either directly or indirectly via interference effects. A standard deviation obtained from a number of single determinations may not reflect that obtained in routine series use. It has been recommended that replicate total analyses of a sample should be carried out, for instance, on different days. However, this will introduce variation attributed to analysts or matrices (through varying instrument performance or varying pre-treatment) into the experimental error variance and distort the ideal of a standard deviation that reflects the performance of an analyst for a sample.I n any proving exercises, replicate total analyses should therefore be carried out within a series but not in consecutive positions within the series. Accuracy Having designed the method to take account of such tenets, and having defined circum- stances where adaptation or alternative methods are desirable, the most important aspect to consider is the accuracy of the results obtained for trace elements in foodstuffs. Only the methodology can be checked for accuracy by measuring the recovery of an added element, in a discrete form, from foodstuffs, and this should always be a preliminary step in establishing an analytical method.The accuracy of results, however, can be confirmed only by obtaining agreement with certified levels in standard reference materials. At present , there are three such materials: NBS bovine liver, NBS orchard leaves and Bowen's kale, for which much information exists.17 It is necessary for these reference materials to be dehydrated in order to maintain a long shelf-life and the mass of sample taken should reflect the wet mass of similar material if material in the latter state is normally examined. Many trace-element levels within the above reference materials do not reflect those normally present in foodstuffs and therefore it is impossible to encompass the range estimated for456 EVANS : QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, VoZ.103 these trace elements. It is hoped that eventually other reference materials will become available that will enable the range of an element normally present to be qualified adequately for various food matrices. Results should be obtained at least in duplicate, by a selection of analysts, on undried material, as it has been reported that some elements in certain food matrices occur as volatile organo-elemental forms that are lost at temperatures below 100 oC.18-20 Therefore, the moisture content should be measured on a separate sample of the reference material and element levels re-calculated to reflect the application of the total analytical method. Variation The accuracy of the analytical method will have been tested initially by measuring the recovery of an element, in discrete form, added to a selection of foodstuffs.The advantages that accrue from this exercise, however, will be wider than a check on the accuracy of the method. From the results it should be possible to estimate the contribution to the total variation from analytical uncertainty, analysts and matrices, to test for significance for these variance sources, to calculate relevant standard deviations and to establish confidence limits. If ten food matrices, varying widely in composition, are taken as representative of food matrices generally, the recovery should be tested at a minimum of three levels. If two of these are at the lower end of the range for measurement, such a system will give information regarding the limit of detection. Ideally, this should be done in duplicate by a selection of analysts (not less than 10) at each of the levels on each food matrix.Few laboratories will have available this statistically desirable number of analysts and will be fortunate to have three analysts competent to deal with new metliodology; even so, each analyst would produce 20 values at each recovery level. This is too many and the criteria described earlier, with one exception, will be satisfied by each of three analysts obtaining recovery values, each on three or four different food matrices at each level. The exception is when contributions to variation from different chemical states of an element occur in different foodstuffs and the above exercise would understate relevant standard deviations. There could be a further penalty, which, however, is less evident when dealing with metallic contaminants, as it is possible to find examples of most foodstuffs in which the level of these is very small.This will not be so with the group of minor nutrient elements, which will invariably be present at levels above trace amounts. Consequently, the naturally occurring levels will distort the recovery figures and any statistical information calculated, because the total recovery will be higher than the fixed amount added. In this instance relevant standard deviations will be overstated. As a check on the above exercise and the possible distortions, it will be necessary to obtain similar statistical information from results obtained on a number of individual foodstuffs, again using the same selection of analysts and carrying out duplicate determinations.The standard deviations obtained will not contain contributions from matrix variation, but may reflect the possible chemical states of an element if sufficient foodstuffs are investigated. Provided that the exercise covers the range of an element normally determined, any gross anomalies in comparison with the first exercke should become apparent and will include, through the repeatability, an assessment of the homogeneity of individual foodstuffs. For this exercise, standard reference materials, CFS mentioned above, or laboratory internal reference materials should be used. Each laboratory should have a series of such materials in order to monitor analyst performance, the method to include instrument performance at the measuring stage, and also to search for possible aerial contamination. Indeed, it is often recommended that any series of determinations should always include duplicate determinations on one of these internal reference materials for this purpose.These single foodstuffs will include some that contain low levels of an element, enabling the limit of detection to be deduced from the variation found and giving a comparison with values obtained from the recovery experiments. Statistical Considerations Before proceeding t o examples of such a design, it is necessary to consider the basic The three factors in analytical methodology that statistical formulae and tests to be used.May, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS 457 require consideration are the mean of results, the standard deviation of results and individual results obtained after the development stage of a method.The true mean of an infinitely large number of defined results normally distributed, or to be more practical a very large number of results, xl, x2, . . ., xi, . . ., xN of size N is denoted by p, the standard deviation is denoted by o and the variance is related to these by the expression N However, the mean of a small sample of the above results, normally distributed, xl, x2, . . ., xi, . . ., x,, size n, is an estimate of p and is denoted by 2, the standard deviation is denoted by s such that NO2 m (n - l)s2 and the variance is In analytical chemistry, only s is calculated and this is an estimate of O, to which uncertainty attaches.A derived term of use for assessing trends is the coefficient of variation, which is the standard deviation of a series of results divided by their mean multiplied by 100%. It is an assumption that a series of defined laboratory results are normally distributed,* but provided that this assumption is accepted distributions derived from the normal distri- bution can also be applied to analytical results. These are the t-distribution for small series of results and the x2 (chi-squared) distribution for standard deviations. If o were known, for a probability p = cc, there is a lOO(1 - 2cc)y0 confidence that any result xi will lie between p 3 p,~, where pa is a factor obtainable from the normalised distri- bution table, e.g., 95% confidence limitst will be p & 1.960.As in the design outlined and in general in analytical chemistry when only series of small numbers of results are possible, only s is known. The corresponding lOO(1 - 2a)% confi- dence limits for a result xa in this instance will be xi 5 tas, where t, is a factor obtained from t-distribution tables. If n = 10 then 95 C.I. for xi will be k 2 . 2 6 ~ ; if n >30 then the t-distribution approximates to the normal distribution. The 95 C.L. of a mean li of ni results of a parent population of size n, standard deviation s, is given by ;Fa & t,s!&where t, is a factor obtained from distribution tables for n results. Irrespective of whether the nl results form part of a population of ;tz results that are not normally distributed, each mean from results will form a distribution of means that will tend to be normally distributed, as the size of each sub-group increases.Even if the parent population of results is normal or assumed to be normal, the distribution of the variance, and hence the standard deviation, s, from n results, will not be normally distributed and the confidence limits for a calculated s are deduced from an asymmetric distribution, the x2 distribution. Thus, for 95 C.L., the true standard deviation, 0, will lie in the interval between s f ( n l ) / ~ i . ~ ~ ~ ] * and s [ ( n - l ) / ~ : . ~ ~ ~ ] * . If n = 10, this means that the true standard deviation lies between 0.69 and 1.83s. This is of fundamental importance, as will be seen in the calculation of limits of detection.and $2 degrees of freedom, a null hypothesis is set up that oI2 and o~~ are the same (the F-test). The ratio s12/s22 for and +2 degrees of freedom will be the measure of the probability, 01, from F-tables, that it is not so. If the alternative hypothesis is that o12 # 02, this is a double-sided test; if the alternative hypothesis is that o12 >oZ2, this is a single-sided test. * Kaisere1 has brought attention to the danger of acceptance of normal distributions to laboratory results and this must be repeated. To compare the values of two variances, cr12 and o ~ ~ , from estimates s12 and sZ2 for Tschebyscheff’s inequality states that [probability (xi - 2) >ko] < l / k 2 When u is known, this gives for k = 2 a probability (2a) of less than 0.25 that a result will be outside the range *220 for any distribution, but for a normal distribution the probability (2a) that a result is outside the range f 2 u is only 0.046.t 96% confidence limits are abbreviated to 95 C.L. hereafter, and 95% confidence interval t o 95 C.I.458 EVANS : QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, VOl. 103 The F-test is of importance in the analysis of variance, where mean squares (variance) from two sources can be tested, whether the results population is normally distributed or not. The denominator variance is always fixed in this instance and the test is therefore always single-sided. Analysis of Variance The technique of analysis of variance is based on the assumption that all results of a large population N are normally distributed and there should be no significance in the difference of variances from groups of results of size n.If there is, it is caused by a difference in the means of the results from groups of size YZ. An analysis of variance of data will permit the following: 1. 2. 3. 4. significance of variances from different sources to be tested; the standard deviations to be calculated singly, or in total from component parts; estimates of the confidence limits of standard deviations to be deduced; the significance of differences in mean values for each source of variation to be tested. Two common types of this system are the hierarchical (nested) and the cross-classification. An example of the first is the system of a bulk sample from which sub-samples are drawn and complete analyses carried out for a determinand on each sub-sample.The result of each analysis is dependent upon an intermediate factor, vix., the sub-sample of the hulk. An example of the second system is a material analysed by three methods in five different laboratories. Each result is classified independently with regard to all sources of variation. In cross-classification systems, only the interaction between two variables can be calculated; in the hierarchical system, with which the design exercises are concerned, the variation of sub-groups within groups is calculated because of dependence. It is also considered that examples of analysis of variance emanating from these exercises are random in nature, i.e., only some foodstuffs are considered from a large number of foodstuffs and, similarly, analysts are considered as representative of a large psopulation of analysts.For fixed or random systems, estimated mean squares will diff er.22 It is not proposed to consider the calculations for this system and reference should be made to suitable Two examples will be described that are typical for data accumulated in exercises detailed in design. Example 1 This example concerns a foodstuff examined €or an element by four analysts. Each analyst makes duplicate determinations, within a series of different samples, starting for each determination from 10 g of the foodstuff and proceeding through a digestion with final measurement upon 100 ml of acidified digest. The results stated as milligrams per kilogram (Table I) are corrected for blank contributions from reagents and working con- ditions by subtraction of the mean of the blanks for each analyst series.The expression of TABLE I EXPERIMENT-4L RESULTS FOR AN ELEMENT I N A FOODSTUFF Analyst r . 1 2 3 4 Resultslmg kg-l . . 12.5, 13.1 12.9, 13.2 13.0, 13.1 12.6, 12.2 results to three digits is sufficient for the calculation of relevant factors in trace element analysis. The mean of the results is 12.57 nlg kg-l and the range is 12.3-13.2 mg kg-1. These data may be treated as a simple one-factor hierarchical classification, which is shown in Table IT. The information to be gathered from Table I1 can be summarised as follows. 1. The mean square obtained between analysts can be tested for significance against The ratio is 4.45 for 3 and 4 degrees of freedom From 17-tables, the 0.05 probability ratio is 6.59.that obtained from experimental error. for the niimerator and denominator.May, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS TABLE II 459 ANALYSIS OF VARIANCE OF TRAXSFORMED RESULTS Degrees of Source of variance Sum of squares freedom Mean square Estimate of Between analysts . . .. .. 0.584 3 Ma = 0.194 7 uo2 + 2aa2 Experimental error - variance . , 0.175 4 M, = 0.043 8 002 Total .. .. .. 0.759 7 (0.108 4) - More than one chance in twenty therefore exists that variation caused by analysts could come froni experimental error, and this is not significant at the 5% level. (If Ma was significant then this exercise would be discredited.) The standard deviation from analytical uncertainty (repeatability) can be obtained from the mean square, Mo, an estimate of uO2.In this account, for a small number of degrees of freedom, the symbol s will always be used in order to avoid confusion. 2. So = d0.043 8 = 0.209 mg k g l AS Ma > 0 but is not significant, Sa2 can be obtained from = 0.075 5 Ma - Mo 2 Sa2 = The standard deviation defined by reproducibility, s, is obtained from S = = 40.043 8 + 0.075 5 = 0.345 mg k g 1 3. The 95 C.L. for a standard deviation can be quoted. Because of the small number of degrees of freedom, for so these will have a wide range, from 0.60 x 0.209 to 2.87 x 0.209 = 0.125-0.600 mg k g l . It may be noted that even with 10 degrees of freedom the 95 C.L. would be 0.146-0.366 mg k g l . Similar confidence limits for Sa can be calculated with the caution that so should have 10 degrees of freedom and, even so, such limits will be approxi- mate in nature.23 4.To ascertain whether each analyst mean falls significantly outside that expected, 95 C.L. can be calculated from 12.87 & bas,/&, where t, = 2.78 for 4 degrees of freedom. The confidence limits for the means are 12.45 and 13.27 and analyst 4 is on the lower limit. The relevant 95 C.I. for a single result can be calculated for the repeatability and reproducibility from taso for 4 degrees of freedom and t,s for an assumed 7 degrees of freedom to give values of -+O.SS and *0.82 mg k g l . The total of these derived factors can be summarised as in Table 111, the standard deviations being quoted to two digits only.5. TABLE I11 REPLICATE ANALYSIS OF A FOODSTUFF FOR AN ELEMENT Parameter Value Parameter Value Foodstuff .. .. .. X Range/mg kg-1 . . . . 12.2-13.2 No. of results .. .. 8 Significance (analysts) . . Not significant % - * .. 1.6 .. 2.1 .. 2.8 3.7 mg kg-1 . . 0.21 l m g kg-l. . 0.37 Reproducibility 1% . . 95 yo confidence . . hO.58 interval/mg kg-l(:o . . 1 0 . 8 2 No. of analysts . . .. 4 Mass of foodstuff/g . . .. 10 Level of element/mg Irg-1 . . 12.9 Amount of element/pg . . 129460 EVANS: QUALIFICATION OF ESTIMATES FOR TOTAL TRACE AaaZyst, VoZ. 103 Example 2 This example is concerned with the recovery of 50pg of the same element, added in inorganic form, to 10-g samples of eight different foodstuffs. As before, the procedure involves a digestion with final measurement on 100 ml of acidified digest.To simulate the design previously described, three analysts are used and this is an example of an unbalanced two-way hierarchical system, which is random in pattern. It is assumed that the element in question is not present in excessive amounts in the foodstuff, but these levels are deter- mined in duplicate in the same series of deterininations as the duplicated recovery experi- ments for each foodstuff, no two foodstuffs being included in any one experimental series. Each level is corrected by the mean of series blanks. The net recoveries, after subtraction of the mean base levels, expressed as a percentage are displayed in Table IV and the analysis of variance in Table V. The mean of the results in Table IV is 100.3y0 and the range is 92-109 yo.TABLE. IV RECOVERY LEVELS FOR AN ELEMENT ADDED TO FOODSTUFFS Analyst A B C 7 --- 7 -7 (-L-- Foodstuff 1 2 3 4 5 6 7 8 Recovery, yo 102, 95 103, 104 92, 93 98, 100 107, 106 99, 97 109, 100 101, 98 Mean, yo 98.5 103.5 92.5 99 106.5 98 104.5 99.5 TABLE V ANALYSIS OF VARIANCE OF TRANSFORMED RESULTS Sum of Degrees of Mean square Estimate of Source of variance squares freedom 0 0 2 + 2ama + 2fioa2 0 0 2 + 2o,2 Doa Total . . .. 351 15 (23.4) - Mml= 39.4 Ma = 21.66 Mm = 46.53 43*33 }276 :}7 Between analysts Between matrices, Experimental error within analysts 232.66 variance 75.00 8 M 0 = 9.36 The mean square Ma in Table V is an estimate of the term shown, where represents the number of matrices per analyst calculated from the equation23 3 i= 1 M2 - Cmi2 - 64 - 22 --- - 2.63 (K - 1)M - 2 x 8 m = The information obtainable from Table V can now be summarised as in example 1.1. Because this is a random pattern, the significance from mean squares between matrices, Mm, is tested against Mo, but the mean square between analysts, Ma, must be tested against Mm. Before proceeding to the calculation, it is necessary to accept certain statistical tenets. If M , is significantly greater than Mo and if .Mm or Ma is greater, but not significantly so, than M , and 111111, respectively, then Sa2 and Sm2 can be calculated according to normal practice. If Ma is smaller than Mm, the best estimate for Sa2 is zero and similarly for s,2 if Mm <M,. In both instances, significance must be further tested by obtaining better estimates for M , or correspondingly Ma if the ratio Mnl/Mo is less than unity.In this example, the latter ratio is 4.97, which is significant for a 0.05 probability level and the ratio of Ma/Mm is less than unity. The best estimate for Sa2 is therefore zero and hence a better estimate, Mml, of 39.4 is obtained by re-calculation. The ratio Mrnl/Mo remains significant for a 0.05 probability level.May, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS 461 2. The relevant standard deviations can be calculated as follows (the use of significant mean squares will be discussed later) : sa, = 0 S = dl5.02 + 9.36 = 4.94% 3. The 95 C.L. for so will be 0.68 x 3.06 to 1.92 x 3.06 or 2.08-5.88~0. 4. The 95 C.L. for matrix means will be 100.3 & t,s/d% where t, is 2.13 for 15 degrees of freedom.5. The relevant 95 C.I. for the repeatability and reproducibility for a single result can be calculated from t,so, where t, for 8 degrees of freedom will be 2.31, and the C.I. &7.1%, and from tas, where t, for an assumed 15 degrees of freedom will be 2.13, and the C.I. & 10.5%. These limits are 89.8 and 110.8%. The total of these derived factors can be summarised as in Table VI. TABLE VI RECOVERY OF ELEMENT ADDED TO FOODSTUFFS Parameter Value Parameter Value No. of foods examined . . 8 No. of results .. .. 16 Amount of element added/pg 50 Range of recovery, yo . . 92-107 No. of analysts . . .. 3 Mean recovery, % . . .. 100 Matrix, yo . . 5 Analysts, yo Not significant Significance { f % - . .. 3.1 . - Repeatability pg .. .. 1.5 I m g kg-l . . 0.15 mg kg-I . . 0.25 95% confidence I s o .. 30.36 interval/mg kg-lxs . . f0.53 % 4.9 2.5 Discussion It is assumed for the purposes of this discussion that series of laboratory results around a mean are normally distributed. Firstly, the standard deviation will be considered. Secondly, the accuracy of a result and how the standard deviation, indirectly through an analysis of variance, may be an aid in assessing accuracy will be considered. Lastly, the limit of detection, the magnitude of which is decided by the standard deviation of results when low levels are determined, will be considered. Having assessed each of these in the evaluation of a method, the influence of the standard deviation and the limit of detection on the reporting of a result (or a mean of results) obtained subsequently by the method will be considered.Standard Deviation Of the factors that define an analytical method, the standard deviation is that for which most information is nornially gathered. In the examination of foodstuffs for trace elements in general, the standard deviation for the variation between results by an analyst on a foodstuff can be defined as repeatability, signified by so in examples 1 and 2. Clearly, a measure of this factor cannot be obtained for each foodstuff by each analyst over the range normally determined. Hence, values must be obtained for separate points in the range, at best, for analysts concerned with foodstuffs constituting a total exercise or by known analysts462 EVANS : QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, VoZ.103 on representative foodstuffs. It is within this context that the performance of new analysts must be tested to the standards originally set. Values for this factor will include any random bias associated with the design and with the experimental technique, and will be derived from the recovery of an element in discrete form added to a cross-section of food- stuffs and for an element in natural form in a number of foodstuffs, both covering the range of levels of the element normally present. The standard deviation, defined as reproducibility, s, for different points in the range normally determined, will be the projection of the variation in results for competent analysts in general on foodstuffs in general to which the method is applicable. That is, it will include random bias within the context of an exercise involving different foodstuffs and different analysts (and including between-series variation and blank variation), but not systematic bias that invalidates the method in application.If the variance component from food matrices is significant, statisticians could formally discredit the standard deviation, s, but in the very definition of the problem random bias niight be expected to happen, as in example 2. Accordingly, s must be calculated so as to take account of matrix variance whether significant or not within limits that will be discussed later. The same argument does not apply to significance from analysts. There are, therefore, for any method in general application, two sets of standard deviations which encompass the range of element normally determined.* In practice, which set should define the variation of results in subsequent exercises will depend upon the nature of the exercise undertaken.If the design in this paper is followed, the variation will always be based upon s, i.e., for a single result on a foodstuff obtained by a known analyst or for a foodstuff not normally examined. If comparison is sought between foodstuffs or between discrete samples of the same foodstuff, and these are examined in separate series of deter- minations and even though with known analysts, the variation must also be based upon s. In examples 1 and 2, it was made clear that any calculated standard deviation is in itself an estimate and the 95 C.L.for this estimate will be decided by the asymmetric distribution, the x2 distribution. The smaller the number of degrees of freedom, the wider the 95 C.L. becomes. Thus, in the two examples, for 4 and 8 degrees of freedom these limits for the repeatability are 0.60 to 2 . 8 7 ~ ~ and 0.68 to 1.92s,, respectively. In a set of calculated standard deviations in the range normally determined, any difference in the particular standard deviation for a point in that range should not be considered abnormal or observa- tions made without consideration of these limits. This will apply especially when comparing standard deviations obtained at similar levels, by recovery exercises and replicate analyses of individual foodstuffs, with the object of seeking distortion, from matrix effects, differing elemental forms and inhomogeneity.(It should be noted that attempts to achieve homo- geneity in foodstuffs, when it is possible to improve the texture, carry with them considerable risks of contamination at the low levels of element determined in foodstuffs when a wide range of different elements are being monitored.) Often, in qualification of a method, samples of varying masses, particularly reference materials, are used in order to simulate the mass of wet foodstuff or to enable an acceptable amount of element to be determined. Clearly, the assessment of trends by comparison of standard deviations a t different levels requires an expression in terms of the actual amount, the coefficient of variation being also helpful, and this applies also to subsequent application of the method.Comparison of the standard deviation of different techniques for an element must be made, however, in terms of concentration. Levels of lead in foodstuffs seldom exceed 0.40 mg kg-l, which coincidentally is a level seldom exceeded in * It should be noted these standard deviations will be for levels in foodstuffs that approach the limit of detection of existing methods. For concentrations 'cwo orders of magnitude greater than the detection limit, a graph of the coefficient of variation against concentration becomes asymptotic, i.e., the coefficient of variation tends to a limiting value. For concentrations below this magnitude, the graph becomes a sharply rising curve.25 A general linear equation relating any standard deviation s to concentration C , such as s = aC + b,26 where a and b are constants, is not necessarily tenable for low element concentra- tions and s cannot be calculated from estimates s, and s, for points intervening between concentrations C , and C,.I t must also be accepted there could be a limiting value for the absolute standard deviation of a result as the concentration decreases, and hence the range normally encountered will decide the number of points a t which the standard deviations are calculated. In Tables 111 and VI, each standard deviation has been expressed in three ways. The last coniparison is of interest in monitoring foodstuffs for lead.May, 1978 ELEMENTS IN FOODSTUFFS USING MEASUREMENT BY AAS 463 human whole-blood samples ( ~ 4 0 pg per 100 ml of blood).Several methods, all of which involve measurement by atomic-absorption spectrophotometry, have been evolved that include stages varying from having no yre-treatment to prior destruction of organic matter. Measurements can be made on a solvent extract after c h e l a t i ~ n , l ~ s ~ ~ by basically the Delves cup technique,*gJ9 or a carbon rod at~niiser,~O a heated graphite f ~ r n a c e ~ l - ~ ~ or hydride generation can be usedS4; the references cited here are a selection only. With the exception of the last method, which is novel and for which insufficient data are available, together with the direct application to foodstuffs of measurement with the heated graphite furnace, the standard deviation has been well documented for blood samples and a number of foodstuffs.Invariably, the determination of approximately 0.40 mg kg-l of lead in such samples is attained to a standard deviation within the range 0.015-0.037 mg k g l for 5-12 replicates, irrespective of the method used. In the absence of experimental information to the contrary, this is taken to define the reproducibility, and for 12 replicates is the confidence range for a standard deviation of 0.020 mg k g l ( ~ 2 pg per 100 ml of blood). What does this imply for levels obtained in routine use in similar exercises subsequently ? On the basis of quoted figures, it seems reasonable to assume that when a single level of 0.40 mg k g l of lead in foodstuffs or in whole blood samples is determined, a result will lie in the range 0.356- 0.444 mg k g l (3644 pg per 100 ml of blood).How that single result is reported now becomes a matter of interest. The 95 C.I. decrees that the digit in the first decimal place cannot be guaranteed, still less the second decimal place. If it is to exist as a single figure, it must be reported as 0.4 mg k g l by taking a reporting interval of similar magnitude to the confidence interval, suitably rounded. (If the 95 C.I. had been A0.022 mg k g l , then the reporting interval would be to the nearest 0.05 mg k g l . ) If replicate measurements are made, the standard deviation for a result will be only marginally affected.6 If replicate total analyses are carried out, in triplicate, the standard error of the mean obtained in the above example will be O.O2O/l/3 = 0.012 mg k g l , and the 95 C.I.of this mean will be &0.025mg k g l . (The standard deviation and t, factor will remain based upon that obtained originally for 12 replicates.) The 95 C.L. for such a mean will therefore be 0.37, and 0.42, (37.542.5 pg per 100 ml of blood), which is not a great improvement. When results are used for other purposes, such as in a survey of the levels of an element in samples of a particular foodstuff, it could be the mean that is of interest. If only the mean is quoted, this could be obtained from individual results calculated to three digits. If the standard deviation was independent of concentration, a mean of approximately 0.4 mg k g l of lead for 25 samples of a foodstuff, determined independently, could be quoted to an analytical 95 C.I.of &0.044/4z = 50.009 or as 0.40, 0.42 mg k g l , etc. A problem arises when individual levels are required together with the mean, in order to display, for instance, the distribution of results. If the number of results is small, the quoting of individual results within the ideal of confidence limits could cause truncation of the distri- bution, giving a poor estimate of the mean. In such circumstances, provided that it is clearly understood that each single result is an estimate, reporting intervals other than &t,s could be adopted. Thus, in the example of lead in foodstuffs, reporting intervals could be halved to 0.05 mg kg-1, but this problem is of greater importance in the region of the limit of detection. The 95 C.I. for a single result will be 2.2 x 0.020 = h0.044 mg k g l .Accuracy It will have become increasingly evident from the above that total qualification of trace- element levels for a range of foodstuffs is difficult to accomplish or unattainable in practice. The most important of these qualifications, and the hardest to ensure, is the accuracy of a defined method in application to foodstuffs generally. In example 2, the mean recovery of 50 pg of the element added in ionic form is, by chance, lOOy/,. The recovery of the method is assured for a cross-section (a small range) of food- stuffs at this level. Similar assurance would be required for other amounts at the extrenii- ties of the levels of this element likely to be encountered. However, this would not imply that the element is recoverable to this extent from all food matrices or that the element in unknown form in any food matrix is recoverable. The only certain method of ensuring the464 EVANS : QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, VOl.105 accuracy is by agreement with standard reference materials with an authenticated content of the element. If the results detailed in example 1 were fr0.m a reference material with a certified level of 13.00 mg k g l of element, the data could be abridged, as detailed el~ewhere,~ to show the bias i.e. 12.87 -13.0 = -0.13 mg k g l , with 95 C.I. for the mean based upon 8 readings from t,s/l/s, which would be h0.29 mg k g l . In such an instance, the agreement would be satisfactory, but the accuracy of the method in application to foodstuffs generally would remain unproved on the basis of one or of a small number of such agreements.If the certified value had been 13.50mgkg-l, the bias would be -0.63mgkg-1 and the mean obtained would be significantly different from the expected value. Even if agreement was attained with certified values from other authenticated reference materials, until the cause had been ascertained the application of the existing methodology would lead to uncertain results. Indirectly, data obtained for interference effects will be an aid in defining the application when the ionic composition of the foodstuffs is known; the technique of assessing interference in analytical methodology has been described. 5 9 3 5 However, for these uncertain circuni- stances, recourse to any additional information on the accuracy of results is welcomed.For this the analysis of variance is important and, in particular, the interpretation of significance from different sources of variance that could enable one to decide whether there is systematic or random bias within the context of this design. For instance, if experiments on the recovery of an element in discrete form using a defined method consistently display no analyst significance, whereas replicate analyses of individual foods conversely display significance, then the defined method is faulty in application to foodstuffs. As different analysts are not equivalent, an element that exists in different forms could be revealed by analyst significance. Random bias would emanate from many different sources that are experimental in origin or tenable within the design outlined, The objective of trace-element analysis of foodstuffs is to obtain a method that is as generally applicable as possible.It will not be surprising, therefore, if random bias introduces significance to variance when applying a method to different food- stuffs, but it would not be expected to happen for results at low levels in the region of minimum detection as the repeatability, so, expressed as the coefficient of variation, will be high. The experimental error variance, so2, could therefore be high in relation to the variance from matrices. As levels increase to 100-fold of this region of minimum detection, the repeatability expressed as the coefficient of variation would be low and random bias in relation to the lowered repeatability could become distinguishable by significance testing (this cannot apply to analyst significance).In example 2, the matrix significance must not be considered abnormal but, in order to define any unusual source of the over-all matrix significance, the means of each duplicate should be tested as described using the repro- ducibility, s, as the relevant standard deviation (in example 1, so would be the relevant standard deviation). In this example, no foodstuff is significantly outside the expected range, but this exercise could have been repeated on two different levels of element giving 24 duplicate means, and from random statistical considerations for this number at least one and perhaps two would be expected to exceed relevant statistical ranges for means.Some care is therefore required in the interpretation of significance testing. Another aspect for judgement, which would be an aid in considering accuracy, is the assessment of trends in the standard deviations with concentration in recovery exercises and in replicate analyses of a series of different foodstuffs for an element. Accepting that each standard deviation is an estimate subject to an asymmetric range of values, anv serious discrepancy in a decreasing gradation, expressed as the coefficient of variation, with increasing amount determined in these factors (taking account of possible heterogeneous behaviour for so) should merit further consideration. It will therefore be seen that the initial ideal of proving accuracy by comparison with standard reference materials, in the present circumstances, has to be elaborated so as to use whatever information is available.The mathematical probabilities will give rise to deductions that will involve logic or opinion, and will not be based upon readily recognised factors. At present, this is not possible because of their small number. This would be an example of systematic bias invalidating the method.May, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS 465 Limit of Detection A definition of the limit of detection is “the smallest amount of a component to be deter- mined, which is still large enough to be detected,”l and is generally expressed as amount or concentration. In the ensuing account it is accepted that instrumental signals are readily converted into concentration, or preferably amounts, by a linear calibration function.Kaiser and Menziesg while basing this limit upon the variation of blank measures, extended the limit of detection to a more general case of the variation of the measure of small but discrete amounts in the region of the limit of detection, X, and, taking into account the mean value of the method blank also, gave a general equation for this limit of detection: .. .. .. ’ (1) .. If one assumes that by measures, responses or signals are implied, then using a high probability (0.0014 to give a one-sided confidence limit of 99.86%) a factor of 3 will eliminate errors that originate from random variation of noise being mistaken for an analytical signal.Roos13 pointed out that for a concentration above a certain critical level, the criterion of detection, in addition to a risk of concluding that a determinand is present when it is not, there will be a risk of concluding that a determinand is absent when in fact it is not. As such, the true detection limit is double that shown in equation (1) and, excluding the mean blank level, is similar to the limit of guarantee of Kaiser and Menzies, which is x b l + GO.* Two terms, the critical level L, and the detection liniit LD, have also been defined,’ which are essentially equivalent to Kaiser and Menzies’ two definitions, but are based only upon the variation of a signal from small amounts of determinand and do not take in account the magnitude of the signal from the blank.This is reasonable as the signal from the method blank is automatically subtracted from a gross signal to give the sample signal, but implies acceptance of the variation being measured in the region of the limit of detection that is of interest. This acceptance is important as the standard deviation may vary in absolute terms with the amount of element measured. There is interest only in a limit of detection being exceeded and hence any confidence limit is one-sided. For a probability of 0.05 corresponding to a 95 C.L. for an infinite population, this limit is defined by 1.640, not 1.960. In the examples quoted,’ for such a probability for a signal response, S + I?, con- taining some of the substance being determined, measured with a paired blank response B , or a sample containing none of the substance being estimated, L, and L, are described as follows: Lo = 1 .6 4 d 0 ~ ~ + + cr2B and L, = 3.29d02, + + o~~ . . - (2) If both S + B and B are in the region of the limit of detection, these expressions become 1 . 6 4 4 % ~ ~ and 3 . 2 9 4 % ~ ~ . The application of equation (2), however, assumes that 0 is known and in general only s can be estimated in analytical chemistry. The systems described are based generally upon the variation of measurement, related to the determinand by a calibration factor, but the limit of detection of a method essentially defined by the critical level L, must be based upon the variation inherent in the complete analytical method in application and not on the measurement stage of that method alone.If it depends upon a standard deviation, it has been shown in this text that more than one standard deviation defines a method, i.e., repeatability or reproducibility, and so the limit of detection is no longer invariate. As there may be different confidence limits for a single result, depending upon whether it is obtained on representative foodstuffs with known analysts or on unknown foodstuffs with any analyst, there may be different limits of detection for the two sets of circumstances, For the design described in this paper, the reproducibility should be used for calculation, unless both so and s are the same or similar, in which event statistical practice would decree which standard deviation should be used. This factor may be further influenced, depending on whether replicate measurements or replicate total analysis are involved, in a manner similar to their influence upon the standard deviation.Good reason has already been given why this factor should not be based upon the variation of a total reagent blank. This does not mean that a limit, so calculated, might not be a valid estimate, but it makes the assumption that random bias from sources other than466 EVANS: QUALIFICATION OF ESTIMATES FOR TOTAL TRACE Analyst, vol. 103 measurement will not contribute to variation and this is an untenable assumption to make until proved, when it would be too late, if true, to accept. (It should be noted that any calculation based upon the variation of the reagent blank must take account of variation of this blank measured at random in a series of blanks, standards and samples.) The liniit of detection niust therefore be based upon the variation of results obtained by a method in application to foodstuffs containing low levels (of the element.In dealing with results, the variation of these results will include automatically also the variation inherent in blanks paired with each result. As it must be clear that ranges of foodstuffs, levels and analysts are required for qualification of trace-element levels in foodstuffs, the number of results will necessarily be lower than desired. The t-factor must therefore be applied at some level of probability to an equation for the detection limit involving the standard deviation, i.e., t,s. There is one proviso, namely that replication of samples with a mean below the detection limit will give a skewed distribution of results a3 zero is the least figure that can be reported and such samples will understate the standard deviation calculated from that sample.To exemplify the following argument for an acceptable probability level, suppose that so has been estimated for a level of element at thl? limit of detection for 8 degrees of freedom (the calculation of confidence limits for s is more complicated and reference should be made to advanced statistical texts). Suppose a probability level of 0.025, corresponding to a one-sided confidence limit of 97.572, was selected. For 8 degrees of freedom the t-factor is 2.31 and hence the detection limit would be t0.0:'5 so =L 2.31S0. However, so is an estimate, in itself subject to variation, as is the total standard deviation, s.Kaiser and Menziesg brought attention to this point, which is of fundamental importance. Thus, for this example the 95 C.L. for so itself are 0.68 and 1 .92so. If, as could happen once in 20 times, the true value u0 was greater than 1.92s0, then the limit of detection calculated on 2 . 3 1 ~ ~ would be equivalent to (2.31/1.92)00 =I l.21oo, and instead of the detection limit being spuriously exceeded once in 40 results through random chance when the element was not present, it would be exceeded once in every 9 results if the series of results was large. If one accepts an equal probability of an element at a level the same as the detection limit being declared absent when it is in fact present, seven out of nine results only could be guaranteed to be greater than zero.When detection limits are calculated from small numbers of results, it is preferable to give a higher probability, 0.005, corresponding to a one-sided confidence limit of 99.5%, which is readily referred to in t-tables. The detection limit for this example would then become 3 . 3 6 ~ ~ and would be equivalent to (3.3611.92)~~ = 1.750~ and so, at worst, the detection limit would not be exceeded for a large number of results by chance more than once in 25 times and 11 out of every 12 results would be guaranteed as being greater than zero. In the determination of lead in both foodstuffs and blood samples, for levels of approxi- mately 0.4 mg k g l , a standard deviation of 0.02 nig kg-1 was indicated, in general, representing the reproducibility based upon 12 replicates.If for similar circumstances the same standard deviation had been obtained at 0.1 mg kg-l, a calculated limit of detection would be 3.11 x 0.02 = 0.062 mg k g l . There would be a high degree of confidence that such a result implies it would be greater than zero and that is all. The nature of the exercise and the relevant confidence limits will decide how such a level will be reported. The 95 C.L. for this level would be 0.018 and 0.104 mg kg-l and a single result of 0.062 mg kg-1 would therefore be reported as 0.1 mg k : g l and a result of (0.062 mg k g l would be reported as <0.1 mg kg-1. If a survey of values for lead included many results in this region (0-0.1 mg k g l ) , the mean could be quoted as for higher values from individual results calculated to three digits but, if individual results are required together with a mean, great care would be required in order to avoid truncation of data.If, as detailed under Standard Deviation, reporting intervals could be halved to 0.05 mg kg-l, the intervals 0-0,025 -+ 0, 0.026-0.074 -+ 0.05 and 0.075-0.124 -+ 0.10 woul'd have to be adopted, which discredits the ideal of a limit of detection for a single result. Clearly, a limit of detection must be calcu- lated in the evaluation of a method and its use must be in defining whether levels subsequently obtained have a meaning within the context of the exercise undertaken. The factors obtained for too05 for 8 and 16 degrees of freedom are 3.36 and 2.92, respectively.This range is very similar to the empirical factor used by Kaiser and Menzies, but it refers to variation in results and not to variation of a signal response, whether it be caused by discrete element or reagent blank. It will be possible to assess the effectiveness of suchMay, 1978 ELEMENTS I N FOODSTUFFS USING MEASUREMENT BY AAS 467 calculations for estimates of the limit of detection, together with other qualifications, in succeeding publications that describe the determination of cadmium, lead and nickel by a method involving chelation from acidic solution and extraction into organic solvent with measurement by ff ame atomic-absorption spectrophotometry16 ; and for arsenic, antimony and tin by measurement with an electrothermal atomisation technique after prior evolution from digests as the respective hydrides.The observations contained in this account are based upon results obtained with these methods and their application subsequently in obtaining many thousands of results. The order in which this account and its applications are presented should not be taken as being deficient of much practice. It is equally certain that circumstances will exist that invalidate the system, so it should not be considered definitive. Conclusions With the design suggested, provided that a strictly defined analytical procedure is used, the accuracy of a method for the determination of trace elements in foodstuffs can be assessed. This should be achieved at the extremities of the range normally determined and should include as many types of foodstuff as possible and a selection of competent analysts.If the former, the method is invalidated for part of its application. If the latter, and signifi- cance testing of variance components through means indicates only random bias, this becomes part of the variation of a method. Such a system will involve the calculation of two standard deviations for several points in the range normally estimated. The first, emanating from an analyst (or known analysts) upon a matrix (or representative matrices), will be the repeatability; the second standard deviation will be the variation likely from any analyst upon any matrix to which the method is applicable and will be the reproducibility. From these, confidence limits for a single result can be calculated that will define how that single result is reported, and will depend upon the exercise involved.From these two standard deviations obtained for low levels of an element, the limits of detection for the element in foodstuffs generally can be deduced, completing the qualifications by the analytical procedure for any result and establishing the validity of subsequent results. On the basis of information obtained, bias can be treated as systematic or random. This paper is published with the permission of the Government Chemist. 1 . 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. References Eckschlager, K., “Errors, Measurement and Results in Chemical Analysis,” Van Nostrand Reinhold, Wilson, A. L., Talanta, 1970, 17, 21. Wilson, A. L., Talanta, 1070, 17, 31. Wilson, A. L., Talanta, 1973, 20, 725. Wilson, A. L., Talanta, 1974, 21, 1109. Youden, W. J., and Steiner, E. H., “Statistical Manual of the Association of Official Analytical Currie, L. A., Analyt. Chem., 1968, 40, 586. Kaiser, H., Spectrochim. Acta, 1947, 3, 40. Kaiser, H., and Menzies, A. C., “The Limit of Detection of a Complete Analytical Procedure,” Lunde, G., J . Sci. Fd Agric., 1973, 24, 1021. Braman, R. S., and Foreback, C. G., Science, N.Y., 1973, 182, 1247. Wilson, A. L., Analyst, 1961, 86, 72. Roos, J. B., Analyst, 1962, 87, 832. Uthe, J. F., Freeman, H. C., Johnston, J. R., and Michalik, P., J . Ass. Off. Analyt. Chem., 1974, Leblanc, P. J., and Jackson, A. L., J . Ass. Off. Analyt. Chevn., 1973, 56, 383. Evans, W. H., Read, J. I., and Lucas, B. E., Analyst, 1978, 103, in the press. Bowen, H. J., J . Radioanalyt. Chew., 1974, 19, 215. Strohal, P., LuliC, S., and JelisavCid, O., Analyst, 1969, 94, 678. Fourie, H. O., and Peisach, M., Analyst, 1977, 102, 193. Behne, D., Bratter, P., Gebner, H., Hube, G., Mertz, W., and Rosick, U., 2. Analyt. Chem.. 1976, Kaiser, H., Analyt. Chem., 1970, 42(4), 26A. Bauer, E. L., “A Statistical Manual for Chemists,” Academic Press, London, 1961. London, 1969. Chemists,” Association of Official Analytical Chemists, Washington, D.C., 1975. Adam Hilger, London, 1968. 57, 1363. 278, 269.EVANS Davies, 0. L., and Goldsmith, P. L., “Statistical Methods in Research and Production,” Oliver Brownlee, K. A., “Industrial Experimentation,” Chemical Publishing Co., New York, 1953. Thompson, M., and Howarth, R. J., Analyst, 1976, 101, 690. Nalimov, V. V., “The Application of Mathematical Statistics to Chemical Analysis,” Pergarnon Yaeger, D. W., Cholak, J., and Henderson, E. W., Envir. Sci. Technol., 1971, 5, 1020. Delves, H. T., Analyst, 1970, 95, 431. Cernik, A. A., BY. J . Ind. Med., 1974, 31, 239. Kapur, J. K., and West, T. S., Analytica Chim. Acta, 1974, 73, 180. Ealy, J. A., Bolton, N. E., McElneny, R. J., and Morrow, R. W., Am. Ind. Hyg. Ass. J., 1974, 35, Gross, S. B., and Parkinson, E. S., Atom. Absorption Newsl., 1974, 13, 107. Pagenkopf, G. K., Newmann, D. R., and Woodriff, R., Analyt. Chem., 1972, 44, 2248. Vijan, P. N., and Wood, G. R., Analyst, 1976, 101, 966. Maurice, M. J., and Buijs, K., 2. Analyt. Chem., 1969, 244, 18. Boyd, Edinburgh, 1972. Press, London, 1963. 566. Received August 26th, 1977 Accepted December 20th, 1977 468 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 36.

ISSN:0003-2654    

DOI:10.1039/AN9780300452

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

Analyst, May, 1978, Vol. 103, p p . 469-474 469 Analysis of Metals Using a Glow-discharge Source with a Fluorescent Atomic Vapour as Spectra I =I i ne Isolator H. G. C. Human, N. P. Ferreira, R. A. Kruger and L. R. P. Butler National Physical Research Laboratory, CSIR, P.O. Box 395, Pretoria 0001, Republic of South Africa A compact and rugged spectral-line isolator based on atomic fluorescence from atoms generated by a low-pressure gas discharge has been constructed. The device is bolted to a standard glow-discharge source. A pulsed electrical current generates the atoms and the fluorescence is measured by means of a gated integrator. Several types of metals have been analysed, e.g., steel, cast iron, aluminium and gold. Good precisions and accuracies have been obtained. Keywords : Metal analysis ; atornic-fluorescence spectrometry The value of a low-pressure gas-discharge sputtering cell as a means of isolating resonance emission lines has been adequately demonstrated by Walsh and co-worker~,~-~ who used it for the measurement of the primary source intensity in atomic-absorption analyses.In their measurements, the primary source radiation was focused on to an atomic cloud of the analytical element generated by the discharge. The energy absorbed and re-emitted as fluorescent radiation is highly selective if the cathode of the glow discharge is pure. It is also directly proportional to the intensity of the resonance analytical line(s) emitted by the primary source. Gough et aL4 used such a sputtering cell as both atom reservoir and resonance-line isolating device in conjunction with high-intensity lamps as sources of irradiation in atomic-fluorescence analyses.Butler et al.5 showed that a sputtering cell can be used for the isolation and detection of relatively weak emission from a Grimm-type glow-discharge emission source. Less than 100 p.p.m. of copper could be determined in aluminium using a lock-in amplifier to discrimi- nate between the d.c. emission from the sputtering cell and the analytical signal induced by the modulated radiation from the glow-discharge lamp. In a later refinement of this work,6 an important advance was made in that the current of the sputtering cell was pulsed while being irradiated continuously by the glow-discharge source. A gated integrator was used to measure the fluorescence signal several milliseconds after the sputtering cell current had been switched off.At this stage, the emission from the cell had subsided to zero while there were still sufficient numbers of free neutral atoms in the irradiated region above the cathode to cause absorption of the analytical resonance line and subsequent fluorescence. In this way, the measuring system could not “see” the high emission intensity from the cell and so the shot noise caused by this high signal level was eliminated from the analytical signal and detection limits could be lowered significantly. In the last two experiments mentioned above, the sputtering cell was a separate loose Pyrex sealed-off unit, which presented problems concerning optical alignment and de-gassing of the spectral-line isolating lamp.In fact, although the principle of operation and potential of the system were well illustrated, it was not robust or sturdy enough to be used in a factory or a t a smelter works site. This paper describes the further development of the spectral-line isolator (SLI) as an instrument suitable for industrial use. The apparatus has been applied to several problems of industrial interest, e.g., the determination of copper and magnesium in nodular cast iron (S.G. iron), silver in gold and copper in aluminium. Experimental The glow-discharge emission source (RSV Prazisionsmessegerate, Hechendorf, West Germany) is of the type described by Grimm.’ A schematic diagram of the SLI is shown in Fig. 1. It consists of a brass cylinder (110 x 50 mm) containing a shallow hollow cathode of the analytical element, a lens (f = 50 mm, d = 40 mm) for focusing the radiation from the470 HUMAN et al.: ANALYSIS OF METALS USING A GLOW-DISCHARGE Analyst, VoZ. 103 glow-discharge lamp directly above the hollow cathode, a gas inlet and a pumping port. A side-arm attached at right-angles to the brass cylinder in line with the hollow cathode allows observation of the fluorescent radiation. In order to improve the light-collection efficiency, another lens of focal length 50 mm (d = 40 mm) focuses the active part of the atom cloud on to a diaphragm in front of the photomultiplier tube. The SLI is fixed to the glow-discharge source by three screws so that no alignment p:roblems arise. Side view Position of side arm f Gas ClU? Standard glow discharge lanp Ga Cathode I Lens Atomic cloud Front view Fig.1. Schematic diagram of apparatus. The SLI is continuously pumped, sharing one of the two pumps of the standard glow- discharge emission source. In order to ensure optimum speed and stability under factory conditions, a separate pump is advisable. A needle valve controls the argon pressure in the SLI and a separate supply of argon is used to ensure that the pressure remains constant. The pressure is adjusted so that the ratio of the fluorescence to emission signal is a maximum (about 200 Pa). The flow-rate of gas through the lamp is restricted by an orifice of diameter 1 mm to approximately 10 cm3 min-l. As a result of electrical time constants in the measuring circuits, a small emission signal from the atomic cloud persists for some time after the termination of the current pulse app1iLed to the hollow cathode.The shallow hollow cathode of the SLI has an external diameter of 10 mm, a recess 7 mm in diameter and a depth of 2 mm. The hollclw cathode gives results similar to a cathode with a flat surface but has the advantage that its burning voltage is lower for the same current. Both the anode and the cathode are electrically insulated from the body of the lamp. The cathode is enclosed with electrically insulating material to prevent the discharge from burning to its sides. Gas flows in between cathode and sheath upwards through the region where the light from the emission source is focused (15mm above the top of the cathode), to the pumping port.The heart of the electrical measuring system is a dual-gated integrator (Molectron Corp., Sunnyvale, Calif., USA), which is operated as follows. At the beginning of the measuring cycle, the discharge to the detector cathode is switched on for 4 ms. The first gate of the integrator measures between 6 and 8 ms while the second gate is set to measure for the same length of time (2 ms) just before the end of the cycle (the cycle period is either 40 or 60 ms). The difference in signal between the first and second gates represents the fluorescent intensity. At the time when the second gate is measuring, all of the atoms generated by the discharge pulse have diffused away and there is no more fluorescence. This measurement is used to correct the signal for stray light of the (d.c.) emission source reaching the photomultiplier.The current applied to the SLI cathode is adjusted so that approximately 40% absorption takes place during the first measuring period. The currents necessary for cathodes of copper and silver are 15 and 10 mA, respectively. These relatively low currents result in the SLI cathode having a very long life. If the abscrption exceeds 50%, the fluorescence signal decreases as a result of excessive re-absorption of fluorescent light. The absorption andMay, 1978 SOURCE WITH A FLUORESCENT VAPOUR AS SPECTRAL-LINE ISOLATOR 47 1 therefore the concentration of ground-state atoms in the light path gradually decays to zero in 3040 ms. The rate of decay determines the period of operation. For copper and silver a period of 60 ms was used, but for magnesium a shorter period of 40 ms could be used.It was found that operation of the gated integrator synchronous with the mains frequency (in multiples of the mains period of 20 ms) resulted in a large reduction in electrical noise. The high emission signal emitted by the SLI while the current pulse is applied is short-circuited to earth by an FET switch in order to avoid saturation of the pre-amplifier. Results Magnesium in Steel I n the determination of magnesium in steel, a problem encountered was that with a pure magnesium cathode in the SLI, no steady discharge could be obtained. The discharge tended to burn to small spots on the cathode surface, which became very hot, and was probably caused by an oxide layer on the surface of the pure magnesium.This problem was overcome by using an alloy of approximately 20% magnesium - SOY0 zinc for the cathode, which gave a steady discharge. The current (4-ms pulse) necessary for 35% absorption of the primary beam was 40 mA, but even this relatively higher current caused negligible electrode erosion. A solar-blind photomultiplier (Hamamatsu R166) was used as the light detector. Radiation emitted by the glow-discharge lamp and reflected from the end window and walls of the SLI caused a high background signal and consequent noise on the analytical signal. This could be avoided to a great extent by using an interference filter (peak at 285.2nm €or magnesium) in front of the photomultiplier. As the stray light is mainly gas and matrix- element (iron) emission lines for this application, this radiation is then prevented from reaching the photomultiplier.However, the low transmission of the filter ( 10-20~0) rendered the signal so low that a poor signal to noise ratio was again obtained. The only alternative was to omit the filter and to reduce the stray light as much as possible by using baffles in the incident beam and lining the walls and end window of the SLI with black absorbing material. A set of analysed samples containing between 0.022 and 0.10% of magnesium was available for testing the performance of the SLI in the application where magnesium in nodular (S.G.) cast iron had to be determined. The points are average intensities of three measured values. Except for the point marked No.39, the individually measured values deviated from the calibration graph by not more than 3%. The signs of these deviations differed, showing them to be random measurement errors, and the value of 3% can be considered as a measure of the accuracy of determination for a single measurement. Sample No. 39 showed a systematic deviation and the true magnesium concentration was later found to be 6% higher than that given. A pre-amplifier with a gain of 30 was used before the gated integrator. Fig. 2 shows the analytical graph obtained. " 200 400 600 800 1000 Magnesium concentration/pg g-' Fig. 2. Calibration graph for mag- nesium in S.G. iron. I I in I - +--Time Fig. 3. Signals recorded for different samples, showing different burn-off characteristics of samples (magnesium in S.G.iron). Numbers on peaks denote magnesium con- centrations (pg s-').472 HUMAN et al. : ANALYSIS OF METALS USING A GLOW-DISCHARGE Analyst, VoZ. 103 The accuracy of determination could probably be improved to better than 2% if the burn-off characteristics of the glow-discharge lamp (emission source) could be improved. Fig. 3 shows the direct recordings of the signals from the detector. It is clear that different samples show different burn-off characteristics, which makes assessment of the signal level uncertain. This characteristic of the glow-discharge source is well known and can be minimised by using a more energetic burn-in discharge condition. Reproducibility of measurement was determined by measuring 22 values of the fluorescence signal given by a sample containing 0.063% of magnesium.A new burn spot was selected for each measurement, and as only four spots could be accommodated on the 35-mm diameter sample, the sample had to be re-surfaced after each set of four burns. A relative standard deviation of 1.4% was obtained. The sample burning time was 2 min. The average value of the signal during the last minute was measured. Fig. 4 shows five consecutive values of the signal as registered on the chart recorder for this sample. No long-term drift was noticeable over the l i - h measuring period. For all of these measurements the glow-discharge lamp current was 100 mA. The source was current stabilised and the burning voltage varied between 1100 and 1200 V for individual measurements. The risk therefore existed that if zinc were present :in the sample, it could cause fluorescence at 213.9 nm in the vapour of the magnesium - zinc cathode.To avoid this, a nickel sulphate solution filters that transmitted about 90% at 285.2 nm but absorbed totally at 213.9 nm was used in front of the photomultiplier. Commercially available glass filters, such as the Schott UG 11 filter (0.5 mm thick), would achieve the same purpose. Other resonance zinc lines did not interfere. This was tested by measuring absorbance with a zinc hollow-cathode lamp as primary source instead of the glow-discharge source. No absorption of the zinc 307.6-nm line could be measured. It has been mentioned that a solar-blind photomultiplier was used as the detector. I 2 min U I +-Time Fig. 4.Consecutive recordings of signal illustrating reproducibility of measure- ment. Copper concentration, % Fig. 5. Calibration graph for copper in steel. Copper in Steel and Aluminium The graph is linear up to 0.56% of copper, the highest standard available. A pure copper cathode was used in the SLI. A current of 15mA (4-ms pulse) proved sufficient to cause 4040% absorption in the vapour cloud so generated. A 1P28 photomultiplier with a 324.7-nm narrow-band interference filter was used in order to eliminate stray light. Again, the graph is linear up to the highest standard available, viz., 4.4% of copper. A matrix effect well known in glow-discharge emission analysis is noticeable, viz., the deviation of the reading from the graph for samples that have high magnesium contents.Fig. 5 shows the analytical graph obtained for copper in steel. Fig. 6 shows the calibration graph obtained lor copper in aluminium.May, 1978 SOURCE WITH A FLUORESCENT VAPOUR AS SPECTRAL-LINE ISOLATOR 473 Copper and Silver in Gold For the fluorescence detection of silver in gold, a pure silver cathode was used. The optimum current in this instance was only 10 mA. A 1P28 photomultiplier with a 338.0-nm interference filter was used. Fig. 7 shows the calibration graph obtained for silver concentra- tions of up to 0.2%. Copper in gold exhibited a similar linear graph for concentrations of up to 0.2%. The measuring period for both copper and silver was extended to 60ms, as these atoms seem to take longer to diffuse out of the region where the light from the emission source is focused.8 15 .- I= 3 : 10 + .- f t I 0 0.1 0.2 Copper concentration, % Silver concentration, % Fig. 6. Calibration graph for copper in aluminium. Fig. 7. Calibration graph for silver in gold. Samples containing up to 5% of copper and silver showed that the calibration graphs in these regions are curved, the curvature being evident at the 1% concentration level. This effect is due to self-absorption and reversal in the glow-discharge emission source, which would be expected in view of the high sputtering rate of the gold metal and the fact that the resonance line intensities are measured. It can, of course, be reduced by decreasing the energy of the glow-discharge source. Detection Limits The detection limits obtained were estimated from the noise and signal levels of low- concentration samples and blanks when they were available.Defining the detection limit as that concentration which gives a signal equal to twice the peak-to-peak noise level as registered on a chart recorder, the values obtained are as given in Table I. TABLE I DETECTION LIMITS OBTAINED Element Matrix Detection limit, p.p.m. Magnesium . . . . Steel 20 Copper . . . . Steel 10 Copper * . . . Gold 2 Copper . . . . Aluminium 25 Silver . . .. . . Gold 3 Conclusion The fluorescent spectral line isolator proved to be a sensitive yet robust, non-dispersive spectrometer for the isolation of resonance lines emitted by the glow-discharge emission source. Its resolution is very high, being determined by the absorption line width in the cell, which is mainly determined by Doppler broadening.It is mechanically stable and is insensitive to temperature and humidity changes and can therefore be used in almost any environment. Preliminary experi- ments, however, showed that very little signal is lost by pumping the lamp from ports not situated directly opposite the cathode. It is therefore possible to position a second cathode The SLI is at present used for single-element determinations only.474 HUMAN, FERREIRA, KRUGER AND BUTLER directly opposite the first. Pulsing the two cathodes alternately and using a two-channel electronic measuring system should enable dual-element measurements to be made. If a single photomultiplier tube is used for measurement of the fluorescent radiation intensity, the use of separate filters for each element will not be possible and a solar-blind photomultiplier will have to be used.Such an instrument is under construction and will be evaluated soon. A fact that should be borne in mind is that this type of detector is limited to the measure- ment of resonance-line intensities, and analyses a t high concentrations of analyte will not be possible, e.g., by selecting a non-resonant einission line as is done in ordinary emission spectrometry. As the absorption line profile in the SLI is very narrow (gas temperature probably lower than 500 K and negligible pressure broadening), the peak of the emission line is essentially measured and the slightest amount of self-absorption in the emission source will be detected. This problem can be avoided to a great extent, however, by using low- power conditions in the glow-discharge emission source. The very short optical path and the fact that both the sample and SLI are in low-pressure argon atmospheres suggest the use of this instrument for the determination of carbon and other elements that have their resonance lines in the vacuum ultraviolet region. This aspect is being investigated and will be reported in the near future. References 1. 2. 3. 4. 5. 6. 7. 8. Sullivan, J. V., Spectrochim. Acta, 1965, 21, 727. Sullivan, J . V., and Walsh, A., Spectrochim. Acta, 1966, 22, 1843. Walsh, A., P w e Appl. Chem., 1970, 23, 1. Gough, D. S., Hannaford, P., and Walsh, A., Spectrochim. Acta, 1973, 28B, 197. Butler, L. €2. P., Laqua, K., Hagenah, W., and Kroger, K., Preprints of 17th Colloq. Spectrosc. Int., Butler, L. R. P., Kroger, K., and West, C. D., Sj7ectrochifiz. A d a , 1975, 30B, 489. Grimm, W., Spectrochim. Acta, 1968, 23B, 443. Kasha, M., J. Opt. SOC. Am., 1948, 38, 929. Florence, 1973, Vol. 2, p. 552. Received September 6th, 1977 Accepted Xovember lst, 1977

ISSN:0003-2654    

DOI:10.1039/AN9780300469

出版商:RSC    

年代:1978

数据来源: RSC

摘要:

Analyst, May, 1978, Vol. 103, pp. 475-481 475 Radiochemical Neutron-activation Analysis of Sulphide Ores Using Zinc Diethyldithiocarbamate as Extraction Reagent* E. Pernicka,t P. A. Schubiger* and 0. Muller§ Max-Planck-Institut fur Kernphysik, Postfach 103980, 0-6900 Heidelberg, West Germany A procedure for the analysis of lead sulphide and mixed sulphide ores for silver, arsenic, gold, cadmium, copper, manganese, antimony and zinc was developed with emphasis on the determination of the low gold to silver and arsenic to antimony ratios. Radiochemical neutron-activation analysis was necessary and a solvent-extraction technique has been developed. In the first separation step arsenic(II1) chloride was extracted from the ore solution with benzene. The results are compared with the values obtained after separation of arsenic by distillation.Gold (111) , silver( I), copper (11) , cadmium (11) and several other trace elements were extracted with zinc diethyldithiocarbamate in chloroform, whereas antimony(V) remained in the aqueous phase. The activities of the samples were counted on a germanium(1ithium) well-type detector and compared with those of known volumes of standard solutions. Chemical yields were determined by re-activation. The combination of conventional arsenic separation and this newly developed diethyldithiocarbamate extraction technique proved to be a very efficient and reliable method for the analysis of sulphide ores. Keywords ; Sulphide ore analysis ; neutron-activation analysis ; yadiochemical separation ; zinc diethyldithiocarbamate ; gamma-ray spectrometry I n the course of studies concerned with the chemical composition and origin of ancient Greek silver,ls2 we have analysed argentiferous lead sulphide ores and mixed sulphide ores from Laurion near Athens and some islands of the Aegean Sea with the aim of determining the provenance of the ancient silver used for coins.This paper deals with the radiochemical procedure for sulphide ore analyses. The ore samples and deposits, mainly from the site of Laurion, Attica, Greece, were characterised by their mineralogical and elemental compositions. The main aim of this study was to establish whether argentiferous ores and metallic silver can be linked by distinct trace-element concentrations or elemental ratios, which are considered typical for a sulphide ore deposit.For this reason and considering the trace elements detected in the silver coinsJ3 the emphasis of this work was on the gold to silver ratios of ores compared with those of coins, because the ratio of these elements is probably least affected by the smelting and refining processes. We have also studied whether any other trace-element contents, such as arsenic, antimony or copper, or the ratios of those elements might be used to determine the provenance of ancient Greek silver. To test the validity of this assumption, we started with the well documented example of archaic Athenian silver, which originated from the famous mines of Laurion. Most of the analysed Athenian silver coins that belong to the Asyut hoard (time of burial about 475 BC) contain only a few hundred parts per million of g ~ l d .~ , ~ The highest silver content of the lead sulphide ores (galena) from Laurion studied in this work is about 0.5%. As the occurrence of gold is not geochemically associated with galena, we expected rather low concentrations of gold in those ores. Neutron-activation analysis, with its high sensitivity for the elements of interest and especially for gold, is a suitable method for solving this problem. However, after irradiation * Presented a t the Fourth SAC Conference, Birmingham, July 17th to 22nd, 1977. t Present address : Institut fur Analytische Chemie der Universitat, Wahringerstrasse 38, A-1090 $ Present address : Eidgenossisches Institut fur Reaktorforschung, CH-5303 Wiirenlingen, Switzerland.5 To whom correspondence should be addressed. Vienna, Austria.476 PERNICKA et al. : NEUTRON-ACTIVATION ANALYSIS OF SULPHIDE Analyst, Vd. 103 of the ore samples with thermal neutrons, the main activity is due to copper-64 and antimony-122. The latter isotope interferes, especially with the 559-keV photopeak of arsenic-76, and increases the Compton background substantially so that gold-198 (y-radiation, 412 keV) cannot be determined instrumentally with a germanium(1ithium) detector. There- fore, a radiochemical procedure has to be applied for the separation of the interfering elements. Principle of the Method The diethyldithiocarbamate anion (DDC) is widely used in analytical chemistry as a chelating agent for extractions both in supra- and in sub-stoicheiometric amounts.436 Successive sub-stoicheiometric extractions from a multi-element system result in a certain extraction order, which has been established b y several worker^^-^ at various pH values, for example by Wyttenbach and Bajo1° for 0.1 N sulphuric acid: gold(III), mercury(II), thallium(III), silver(I), copper(II), bismuth(IlJ), antimony(III), tellurium(IV), molyb- denum(VI), selenium(IV), indium(III), arsenic(III), lead(II), cadmium(I1) and zinc(I1).Arsenic(V) and antimony(V) are not extracted by DDC. Wyttenbach and Bajo1° showed that this extraction order also represents the inverse order of replacement. Therefore, the extraction with supra-stoicheiometric amounts of MDDC as reagent gives a fraction with a high yield of the elements to the left of M in the above order in the organic phase (where M = metal cation).By applying this method with different MDDC compounds, even samples of complex composition can be analysed rapidly with good separation efficiency and with a minimum of chemical operations. A detailed outline of the theory of this method was given by Wyttenbach and Bajo.lo Applications to water,ll biological material,12 silicate rocks13 and silver coins1* have been reported. We have used Zn(DDC), to separate gold, silver., copper and cadmium from antimony. Copper-64 has a short half-life of 12.70 h; therefore. it does not interfere seriously with the gamma-ray spectrometric measurements after a few days' cooling time. Arsenic, however, under these conditions remains together with antimony in the aqueous phase and has to be separated in a specific procedure.Firstly, the classical separation of arsenic by distillation from an acidic solution of high hydrochloric acid content at 110 "C; with the exception of germanium, no other e1e:ments are distilled under these conditions. Secondly, arsenic in the trivalent state can be extracted quantitatively into benzene from an aqueous phase that is more than 9~ in hydrochloric acid.16 Only germanium(1V) is also extracted into the organic phase, while antimony(II1) and tin(1V) remain in the acidic phase.lG The solvent-extraction method has proved to be less time consuming, and the separation of arsenic from antimony was as good as with the distillation technique. Two methods were considered for this separation.Experimental Preparation of Zinc Diethyldithiocarbamate Reagent A solution of 14 g of sodium diethyldithiocarbamate (NaDDC.3H20) in 100 ml of ethanol was slowly added and the mixture stirred for 30 min. The precipitate was filtered off, washed with 500 ml of water and dried at 70 "C overnight. The crystalline product was dissolved in 400-500ml of chloroform and filtered. The filtrate was mixed with 150ml of ethanol and covered with a watch-glass. After evaporation of about half of the solvent, the crystals were filtered off, washed twice with absolute ethanol and dried in air. A 0.03 M solution of Zn(DDC), in chloroform was prepared. Ten grams of zinc sulphate (ZnS04.7H20) were dissolved in 150 ml of water. Sample Preparation and Neutron Irradiation Approximately 30-g amounts of lead sulphide ore and mixed sulphide ore were selected from samples and each was ground in an agate blall mill, sieved through a 60-pm sieve and mixed thoroughly.Amounts of about 100 mg o:E each prepared sample and two standards were packed in highly pure polyethylene containers and irradiated in the carrousel position of the Heidelberg TRIGA reactor at a neutron flux of about 2 x lo1, neutrons cm-, s-1 for 2 h. The standards were prepared by pipetting chemical standard solutions on to filter-May, 1978 ORES USING ZINC DIETHYLDITHIOCARBAMATE AS EXTRACTION REAGENT 477 paper. Gamma-ray self-shielding is not important under these conditions (less than 2% at 511 keV and about 1% at 1116 keV). After a cooling period of 1 d, the samples were measured instrumentally on top of a germanium(1ithium) well counter of resolution ~ 2 .7 keV in order to determine copper, manganese, antimony, zinc and in some instances silver. The photopeaks were accumulated and stored and the data were processed on-line with a Digital Equipment Corp. (DEC) system consisting of a PDP 11/40 computer, a cartridge disk, a magnetic tape unit and a fast ~ r i n t e r . ~ Immediately after this measurement, the radio- chemical separation was started (Fig. 1). One run with two ore samples and two standards lasted about 4 h. For re-activation, 50-pl amounts of the Zn(DDC), extracts were pipetted on to filter-paper situated in polyethylene containers for the determination of chemical yields. Together with pipetted carrier solutions (1OOyo yield), they were irradiated for 4 h at a flux of about 2 x 10l2 neutrons cm-2 s-l in the Heidelberg TRIGA reactor.Packing of 100- rng samples in polyethylene containers; irradiation for 2 h a t flux of 2.10’~ neutrons cmA2 s-’ 1 1 1 Standards (a) Sb, Cu, Zn, Mn, Ag (b) As (c) Au, Ag, Cd, (Cu) 4 Sample preparation: hand.picking, grinding, sieving Dissolving in HN03 - HC104 (1 + 1); evaporation; reduction with KI + ascorbic acid in 2 N HCI; extraction with benzene from a solution 9 M in HCI 7 4 Carrier solutions Count 2 in calibrated flasks As Evaporation; heating in 0.5 rnl of aqua regia; extraction with 0.03 M Zn(DDC)? solution in CHC13 from a solution containing 0.5 M HC104 Au, Ag, Cd, (Cu) Post-irradiation of 50 pl of organic phase in polyethylene containers for 4 h a t 2.1012 neutrons cm-2 s-1 sarn p I es Fig.1. activation. Scheme of analysis for lead sulphide ores and mixed sulphide ores by neutron478 PERNICKA et al. : NEUTRON-ACTIVATION ANALYSIS OF SULPHIDE Analyst, YUZ. 103 Distillation of Arsenic The amounts of each carrier element added varied from 30pg to 2 mg; they were adjusted so as to yield reasonably strong activities of all of the elements after re-activation. Carrier solutions, 1 ml of concentrated perchloric acid and 1 ml of concentrated nitric acid were pipetted into a distillation flask. The ore sample was added and the flask was immediately connected with a distillation apparatus consisting of a dropping funnel, a thermometer, a Liebig condenser and a gas inlet. A conical flask cooled with ice and containing 5 ml of water served as a receiver.On gentle heating, the reaction started and the heating was continued until white fumes appeared. Concentrated hydrochloric acid (5 ml) was added to the distillation flask, the (contents of which were evaporated to half of the volume, and 5 ml of concentrated hydrobromic acid were added. Distillation was continued at 110 "C until the temperature began to increase. A continuous flow of nitrogen was passed through the apparatus during the entire procedure. The solution in the receiver was transferred into a 25-ml calibrated flask, diluted to 25 ml with water and the flask was put on the germanium(1ithium) detector in fixed geometry and counted. Extraction of Arsenic Amounts of carrier solutions as described above, 0.5 ml of concentrated perchloric acid and 0.5 ml of concentrated nitric acid were pipetted into a 50-ml extraction vessel of length about 15 cm, equipped with a hollow plug, and the ore sample was added.The size of the plug was about 60 x 15 mm, and the ground-glass portion was connected to the extraction vessel after sample decomposition.l'l On gentle heating of the glass vessel, the sulphides were easily dissolved and a white precipitate of lead sulphate was formed. The heating was continued until all of the nitric acid was fumed off. In order to reduce arsenic(V) to arsenic(III), 100 mg of potassium iodide arid 25 mg of ascorbic acid were dissolved in a few millilitres of 2 M hydrochloric acid and added to the sample solution, which was kept on a boiling water bath for 15 min. After cooling, 25 ml of benzene were pipetted into the extraction vessel and enough ice-cold concentrated hydrochloric acid was added to make the acid solution at least 9~ in hydrochloric acid.The mixture was shaken vigorously for about 10 min to extract arsenic(II1) into the benzene and subsequently centrifuged. At this stage, the organic phase showed the purple colour of dissolved iodine. An aliquot of 20ml of the organic phase was pipetted into a. 25-ml calibrated flask, diluted to the mark with benzene and counted, The arsenic standard was dissolved in 1 M nitric acid in a calibrated flask of the same size and counted in the same geometry on the germanium(1ithium) detector (Fig. 1). Extraction with Zinc Diethyldithiocarbamate After addition of 1 ml of concentrated nitric acid, the aqueous phase from the extraction vessel was evaporated almost to dryness by use of an oil-bath and cooled.To ensure that all of the gold had dissolved, the residue was treated with 0.5 ml of aqua regia, 0.25 ml of concentrated perchloric acid was added and the nitric acid carefully fumed off. The solution was diluted with 10 ml of 0.5 M perchj!oric acid and gold(III), copper(II), silver(1) and cadmium(I1) were extracted with 5 ml of 0.03 M Zn(DDC), in chloroform, when most of the antimony(V) remained in the aqueous phase. The mixture was shaken for 10 min and subsequently centrifuged. A 4-ml volume of the organic phase was pipetted into a poly- ethylene counting tube and measured in the well of the germanium(1ithium) detector.The errors caused by antimony-122 and antimony-:124 might be reduced by further treatment of the organic portion with an antimony hold-back carrier solution (see Fig. 3). Results and IMscussion The efficiency of the separation of arsenic fro:m antimony with benzene was tested over a range of arsenic to antimony ratios varying by a factor of lo3. Almost lOOyo of arsenic but less than 0.4% of antimony are extracted into the benzene phase under the given conditions (Fig. 2). If there was any need, the amount of antimony could be reduced by another factor of 100 by washing the benzene slolution with concentrated hydrochloric acid. In Fig. 3 the gamma-ray spectra A and B of sulphide ore C3 (W 110) from Laurion con- taining 95% of lead sulphide and 0.48% of silver, in addition to other constituents, are shownMay, 1978 ORES USING ZINC DIETHYLDITHIOCARBAMATE AS EXTRACTION REAGENT lo4 g - Fig.2. Trial separation of As from Sb with benzene for five samples of various Sb/As ratios. As and Sb standard solutions were pipetted, dried and irradiated, and the activities of 76As and lz2Sb counted on a Ge(Li) detector after benzene extraction. Line B shows the amounts calculated. Line A shows the Sb/As abundance ratio before extraction. 1 2 3 4 5 Sample number 479 before and after radiochemical processing. The spectra show that a good separation with Zn(DDC), has been achieved, and the photopeaks of silver-llOm, gold-198 and cadmium-115 show up in spectrum B of the extracted organic phase. The chemical yields for the extracted elements were gold >95, silver >SO, copper >80 and cadmium >70%.The pH of the aqueous phase that has been chosen is too low for a quantitative extraction of cadmium, as cadmium has been shown17 to be extracted quantitatively from perchloric acid only up to 0.2 M. Gold(III), silver(1) and copper(I1) are known to be extracted with satisfactory efficiency from solutions with the acidity used in this work.1°J4 It may also be possible lo6 - 105 - 1 o4 tn 4 d C 3 0 1 o3 1 o2 10’ 0 500 1000 1500 E nergylkev 2 000 Fig. 3. y-Ray spectrum of neutron-activated ore sample C3 (W 110) from the site of Laurion. Sample mass 100 mg. A, Before radiochemical separation; B, after extraction with Zn(DDC),. Sum peaks are not marked. The strong peak of 1022 keV in spectrum B, which has been measured in the well of the Ge(Li) detector, is due to the sum of annihilation radiation.480 PERNICKA et d.NEUTRON-ACTIVATION ANALYSIS OF SULPHIDE Analyst, VO, 103 that the MDDC complexes are slowly decomposed in strongly acidic solutions, which would lower the chemical yields. It has been shown recently that an extraction time of 2min would be sufficient. We also determined the arsenic contents in three Laurion sulphide ores by benzene extraction and by distillation, and found good agreement between the two methods (Table I). TABL:E I DETERMINATION OF ARSENIC BY EXTRACTION WITH BENZENE AND BY DISTILLATION Arsenic content, p.p.m. Arsenic extracted Sulphide ore sample with benzene as Arsenic distilled as from Laurion arsenic(II1) chloride arsenic(II1) chloride C3 (E7) 4.9 & 0.4 5.1 f 1.7 A4 0.9 & 0.1 0.6 & 0.1 A5 16.8 f 1.5 16.8 1.3 In Table I1 the results of replicate analyses far two lead sulphide ore samples from the site of Laurion are shown.On average, the precision was equal to of- less than 5% for copper, antimony, manganese and zinc, equal to or less than 10% for silver, arsenic and cadmium, and 25% for gold. The precision for gold and silver could possibly be improved by replacing Zn(DDC), with Ni(DDC), in which instance only mercury(II), silver(1) and gold(II1) would be extracted. However, it was more important for this study to characterise the ores with as many minor and trace elements as possible i n one run. We have been able to show that the gold to silver ratio can be a valuable indicator for the determination of the provenance of ancient silver, while no conclusions can be drawn from the arsenic to antimony ratio at present.TABLE: I1 REPLICATE ANALYSES OF LEAD SULPHIDE ORE SAMPLES FROM THE LAURION SITE Copper, Arsenic, Antimony, Manganese, Cadmium, Sample Silver, p.p.m. Gold, p.p.m. p.p.m. p.p.m. % p.p.m. Zinc, yo p.p.m. C3 (W110) 4 800 f 360 0.071 f 0.004 870 f 60 45.3 f 1.2 1.07 * 0.01 870 f 30 0.37 & 0.01 60 f 4 C3 (E7) 4 400 f 400 0.047 -f 0.009 1460 f 70 5.0 f 0.1 0.92 f 0.05 186 f 6 t0.02 26 f 9 The great interest of Professor W. Gentner in this work is appreciated. We are grateful to Ms. Karin Trauner for many chemical experiments and to Mr. Klaus Oberfrank for assistance in gamma-ray spectrometry and data handling.We thank Priv.-Doz. Dr. A. Wyttenbach and Dip1.-Chem. S. Bajo, Eidgenossisches Institut fur Reaktorforschung, Wiirenlingen, Switzerland, for valuable discussion and advice, and the staff members Mr. Jiinger and Mr. Anasensl of the TRIGA reactor, Deutsches Krebsforschungszentrum, Heidelberg, for the neutron irradiations. Financial support by the Stiftung Volkswagenwerk is gratefully acknowledged. References 1. 2. 3. Muller, O., Schubiger, P. A., and Gentner, W., “Symposium on Archaeometry and Archaeological Prospection, Oxford,’’ 1975. Schubiger, P. A., Muller, O., and Gentner, W., “Proceedings of the 1976 International Symposium on Archaeometry and Archaeological Prospect ion, Edinburgh,” in the press. Schubiger, P. A., Muller, O., and Gentner, W., “Proceedings of the 1976 International Conference on Modern Trends in Activation Analysis, Miinchen, 1976,” Volume 11, p. 1176, and J . Radio- analyt. Chem., 1977, 39, 99. 4. Thorn, G. D., and Ludwig, R. A., “The Dithiocarbamates and Related Compounds,” Elsevier, New York, 1962. 6. RbiiEka, J ., and Starf, J. , “Substoichiometry in Radiochemical Analysis,” Pergamon Press, Elmsford, N.Y., 1968. 6. Bode, H., and Tusche, K.-J., 2. Analyt. Chem., 1957, 157, 414.May, 1978 ORES USING ZINC DIETHYLDITHIOCARBAMATE AS EXTRACTION REAGENT 7 . 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. Wickbold, R., 2. Analyt. Chern., 1956, 152, 259. Eckert, G., 2. Analyt. Chem., 1957, 155, 22. Starjr, J., and Kratzer, K., Analytica Chirn. Acta, 1975, 47, 1813. Wyttenbach, A., and Bajo, S., Analyt. Chern., 1976, 47, 2. Rauter, R., EIR-Bey., 1977, No. 313. Wyttenbach, -4., Bajo, S., and Hakinnen, A., Be&. Tabakforsch., 1976, 8, 247. Bajo, S., and Wyttenbach, A., Analyt. Chern., 1975, 47, 1813. Schubiger, P. A., and Muller, 0.. Radiochem. Radioanalyt. Lett., 1976, 24, 353. Beard, H . C., and Lyerly, L. A., Analyt. Chern., 1961, 33, 1781. Fischer, W., and Harre, W., Angew. Chem., 1954, 66, 165. Bajo, S., and Wyttenbach, A., Analyt. Chem., 1977, 49, 158. 481 Received Sefiternber 19th, 1977 Accepted October 31st, 1977

ISSN:0003-2654    

DOI:10.1039/AN9780300475

出版商:RSC    

年代:1978

数据来源: RSC