ANALYTICAL CHEMISTRY.IN this year’s Report on Analytical Chemistry it has been decidedto include separate sections on spectrography and on silicate analysis,as it seemed that both subjects, although perhaps somewhat special-ised, were sufficiently interesting and important to justify a specialaccount of the recent advances made in them. With silicate analysissuch an account has, indeed, been long overdue. Other subjects,such as the use of the polarograph and the development of micro-chemical technique and apparatus, especially in Organic Chemistry,are also due for special consideration, but must be deferred to afuture Report. Except for a brief treatment of the subject ofstandardisation in volumetric analysis the rest of the present surveydeals only with qualitative analysis on the micro-scale.The past year has seen the publication of many important papersconnected with gravimetric and volumetric analysis, but their survey,and also a report on organic analysis prepared by Dr.R. W. West,must be omitted owingt80 the limitations of space.L. S. T.1. SPECTRUM ANALYSIS BY TIIE EMISSION SPECTRUEd (1933-1 937).IN this summary it is assumed that the reader is aware of the stateof the art of spectrum analysis as applied to chemistry five yearsago. Spectrum analysis was then already in regular use for theexamination of almost every kind of substance-metals and alloys,liquids, powders, ore8, soils, glasses, slags and other vitreous sub-stances, refractories, precipitates, and residues, and animal andvegetable tissues.At that time the determination of the metallic(and some non-metallic) constituents of a mixture or compoundcould in general be made by the spectrograph to within 10-15%of the constituent, whether the quantity present was large or small.Progress since then may be epitomised as the achievement of approxi-mately four times this accuracy. As a consequence there are manymore analyses to-day than there were five years ago of which it canbe said that the spectrographic is better than the chemical methodof analysis. This result has been reached by replacing visualobservation of the photographed spectrum by the photometry ofthe lines and by studying the best way t o produce the radiation(the arc, spark, or oxy-hydrogen flame).The methods tend to resolve themselves into a limited numbeTWYMAN : SPECTRUM ANALYSIS BY EMISSION SPECTRUM.466each of which is widely applicable ; and in this Report most attentionhas been given to methods which have become of establishedutility in analysis. Those interested in the variety of applicationsshould refer to the Bibliography (p. 463). Although, doubtless, theRpectroscope continues to be used, almost all the new developmentsare concerned with the spectrograph.The quantitative accuracy here spoken of, is only attainable whenthe substance to be detected is present in a small amount. Whatthat amount should be naturally varies greatly with differentsubstances, but as a very rough rule one may place it a t approxi-mately 4%.Principles of Procedure which have become generally accepted-The results of many investigations have established that : (a) Anymixture or combination of elements can be made to emit aline spectrum characteristic of the elements; ( b ) in such spectrathe intensities of the lines of the various elements have a relationto each other depending only on the percentages present; (c)small variations in the percentage of any of the elements presentcause no variation in the intensities of the lines of the other elements.Quantitative spectrum analysis resolves itself, therefore, into ameasurement of the intensities of spectral lines.Since statements (a), ( b ) , and (c) are the foundations of quantita-tive spectrum analysis, it is desirable to consider briefly the limitswithin which they are true.Statement ( a ) requires no qualificationwhatever, except on the score of practical considerations, for as G. R.Harrison has recently pointed out, all the elements have sensitivelines, i e . , lines which are in evidence when only minute quantitiesare present. The metals and a few of the non-metals have thesesensitive lines in the range of spectrum lying between 2,000 and10,000 A., which is accessible to the ordinary quartz spectrographworking in air. Sensitive lines of most of the non-metals, on theother hand, are of wave-lengths shorter than 1850 A., at which airbecomes almost completely opaque. This is why practical spectrumanalysis is confined to the metals, including antimony and bismuth,the metalloids arsenic, selenium, and tellurium, and the non-metalsboron, phosphorus, and silicon ; since, although very many vacuumspectrographs are in use for physical investigations, the cost of theirconstruction and the difficulties of vacuum technique preclude theiruse in analysis,The ratio of the intensitiesof two lines of a substance (whether emitted by the same element orby two different ones) may vary greatly according to the means ofexciting the radiation (arc or spark).One should, therefore, always1 Ilfetals and Alloys, Nov., 1936,Statement (b) requires qualification456 ANALYTICAL CHEMISTRY.keep these conditions constant (e.g., maintaining the same sparkingcircuit, current, and distance apart of the electrodes).Very little has been published on the effect of the factor (c), butthe effect is an important ~ n e .~ g ~ For this reason the best resultsare obtained by comparing the samples to be analysed with specimensof known and not very different composition.The methods which yield the best results all adopt a principlewhich largely avoids errors arising from variations in the arc orspark discharge, vix., that the measurement should be based on acomparison of selected spectral lines of a minor constituent withthose of the main constituent-a principle called by W. Gerlachthat of the ‘‘ internal standard.” Analysis based directly on thisprocedure has been widely used by him and other^,^ who havecompiled lists of pairs of lines of the main and minor constituentswhich are of equal intensities at stated percentages of the latter.Theory and observation alike support the recommendation ofW. Gerlach and E, Schweitzer to use for the measurements whereverpossible only lines whose intensities relative to each other do notvary with even widely varying conditions of the arc or spark; inthis method those authors confine themselves to the use of lines whoseintensities remain equal to each other under widely different con-ditions of excitation, for as Gerlach points out,6 such lines remainof equal intensity even when they are over- or under-exposed. Heuses these “ homologous line pairs,” as he calls them, as the basisof his system of analysis.* To do so is a severe restriction on thechoice of lines, and, working with the microphotometer, one can usealso line pairs of unequal intensity, though these should still beselected so that the ratio of their intensities does not alter withchanges in the conditions of excitation.Various workers recommend lines suitable for particular analyses,and reference to these can be found in Smith’s bibliography and otherbooks mentioned on p.463.Methods of Pducing the Radiation.-In the period under reviewmany variations of the three main ways (flame, arc, and spark)have been put forward to meet the special requirements of those whohave devised them. Only the most important of them will beF. Twyman and C. S. Hitehen, Proc. Roy. SOC., 1931, A, 133, 72.0. S. Duffendack,lnd. Eng. Chem. (Anal.), 1935, 7, 410.“ Foundations and Methods of Chemical Analysis by the EmissionSpectrum ” (Adam Hilger, Ltd.).D.35. Smith, J . Inst. Metals, 1934, 55, No. 2, 283; 1935, 56, No. 1, 257;E. H. S. van Someren, ibid., 1934, 55, No. 3, 265; M. Milbourn, ibid., 1934,55, No. 2, 275.6 op. cit., p. 67.* Valuable lists of homologous line pairs are given in Gerlach and Ried’sbook (see p. 463)TWYMAN : SPECTRUM ANALYSIS BY EMISSION SPECTRUM. 457mentioned here, some preference being given to those which have beenused by workers other than those introducing them.H. Lundegiirdh continues to use the method describedby him for the analysis of solution^.^ He sprays the liquid into anacetylene-air flame from an atomiser,8 and claims that this pro-cedure is superior in constancy to either arc or spark; he has usedit with success in combination with a microphotometer in soiland plant analysis, and in physiological and pathological in-vestigations.Although, in general, the spark and arc yield a moresensitive detection of many of the metals, Lundegiirdh’s methodusually suffices in this respect for those which are of the greatestinterest in such fields of work, and the accuracy attained (using amicrophotometer) is very satisfactory. LundegSirdh says that theerrors of determination never exceed 5-7 yo, and with the alkalis,calcium, and manganese are not more than P--2yo.* The method hasbeen adopted by others with satisfactory results, and R. L. Mitchellclaims an accuracy in determining the alkalis and alkaline earthsof & 47,’,’, which can be improved by replication of the spectro-grams.Determinations such as those of strontium in the presenceof large excess of calcium, which are scarcely feasible otherwise on aroutine scale, present no difficulties.H. Ramage lo continues to use his flame method in the examinationof biological material. He puts the material to be analysed in afolded filter-paper, and places this in an oxy-hydrogen flame. Themethod is simple, although it does not approach the quantitativeaccuracy of Lundegiirdh’s (probably because Ramage uses no photo-metric means of measuring line intensities). It has been used ina survey of the micro-constituent elements of biological material.11With easily oxidisable or low-melting-point metals andalloys it is sometimes exceptionally difficult to obtain an arc whichis steady in either position or character, and to remedy thisW.Gerlach introduced the interrupted arc, in which the electrodesare repeatedly brought into contact and separated ; they do not gethot enough to be oxidised, and one thus avoids the almost continuousbackground, due to the complex molecular spectra, which causesdifficulty in using the arc spectra of, e.g., aluminium.Flame.Arc.“Die Quantitative Spectralanalyse der Elemente,” Parts I and I1See Ann. Reports, 1934, 31, 288. J. 8oc. Chem. Ind., 1936, 55, 267.lo Nature, 1936, 137, 67; H. Ramage, J. H. Sheldon, and W. Sheldon,l1 D. A. Webb and W. R. Fearon, Sci. Proc. Roy. Dublin SOC., 1937, 21,* m7here accuracy is stated the Reporter means in every case the percentage(Gustav Fischer, Jena; 1929 and 1934).Proc. Roy.SOC., 1933, B, 113, 308.487, 505.of the total amoiint of constituent present458 ANALYTICAL c3HEMIsTRY.S. Judd Lewis has extended the use of his pellet methodla (inwhich an ash or other powder is compressed into a pellet and placedon the lower pole of a copper or silver arc) by synthesising a secondpellet (whose composition is thus known) so that its spectrumexactly matches that of the pellet to be analysed. This in itselfmay not be new, though Judd Lewis seems to have been the firstfully to realise how many difficulties and instrumental complexitieswere thus avoided,* but he has added to this a very useful principle.By chemical determination of the main ingredient in the specimento be analysed (e.g., a portion of a plant) and spectrographic deter-mination of the ratio of the minor ingredient to that of the main onein the ash, he is able at once to state the percentage of the minoringredient in the original substance.He calls this the “Ratioquantitative ” method. The principle can, of course, be used withother means of producing the radiation, and it was used withRamage’s flame arrangement by Webb and Bearon.ll No quantita-tive method of spectrum analysis of like accuracy requires less in theway of apparatus than this, and although synthesis is not feasiblein the routine analysis of metals and alloys, the principle can beused with advantage in laboratories dealing with a run of similaralloys.M. Milbourn l3 pointed out that enhanced sensitivity in the detec-tion of minute amounts of impurities in copper is attained byusing a globule of the sample (0-2-0-5 9.) as electrode in place of asolid rod.Suitable lines and their intensities are given for thedetection and determination of Bi, As, Pb, Fe, Ni, Ag, Sb, and Snby this method. S. Pina de Rubies and J. Doetsch,14 desiring todetect extremely minute percentages, effected a concentration of aspecimen placed in the arc by allowing it to melt there and evaporateoff the more volatile constituents. Finally, K. Cruse and H.Schubert,f6 in determining lead in blood, effect a concentration of thelead by electrolysis, a method which was used many years ago byDupr6 in detecting the presence of mercury in gun-cotton ; electro-lytic concentration has also been used by A.Schleicher and(Frl.) N. Kaiser.16Spark. It is with the spark that most progress has been made inquantitative analysis in the last five years, It does not appreciablyconsume the material, or overheat it, and most important of all, itis more constant than the arc both in position and in character.A few new modes of producing the spark may be mentioned.l2 “ Spectroscopy ” (Blackie & (20.).l4 2. anorg. Chem., 1934, 220, 199.l5 2. anal. Chem., 1936, 105, 241.* For example, quantitative analysis can be achieved without the use of18 LOC. cit., ref. ( 5 ) .16 Ibid., p. 393.any photometric measurement of the intensities of the linesTWYMAN : SPECTRW ANALYSIS BY EMISSION SPECTRUM. 459W.Gerlach,l7 analysing very small areas of non-metallic specimenssuch as thin sections of animal organs (liver, etc.), mounts the speci-mens on slides and produces a high-frequency (Tesla) dischargebetween a point and the selected portion of the specimen.K. Ruthardt 18 has analysed metallurgical specimens in this way.0. E’eussner l9 modified the usual spark oircuit, to make it more con-stant, by introducing & rotary interrupter and resonant circuit.W. Seith and E. Hofer 20 blew an air blast from twin jets on the gapof the ordinary spark to remove the outer layer of comparativelycold metallic vapour which absorbs the radiation of certain of thespectrum lines. 0. S. Duffendack, for the analysis of solutions,added to the Twyman and Hitchen sparking tube for liquids2an exhaust for drawing o f f the fine spray caused by the spark.Later, with F.K. Wiley and J. S. OwensY3 he used a Pyrex or silicatube.It does not appear that any of these special sparking devices areneeded for the ordinary run of work in chemical or metallurgicalanalyses, and it remains the general practioe to use the simpleoscillating spark, with condenser and self -induction coil.21Devices for Collecting the Radiation.-Most frequently one desiresthat the spectrum should truly represent the whole of the radiationfrom the arc or spark, and that the spectrum lines shall be of uniformintensity throughout their length. To enswe this, a convex lensof quartz i s used, preferably near the slit, to form a real image ofthe light source on or near the prism of the spectrograph.Thereappears no need in spectrum analysis to continue the use of the oncepopular sphero-cylindrical condenser.The radiation is not uniform from pole to pole of the arc or spark,some spectrum lines being short, and only originating quite cloaeto the pole, while others are continuous from pole to pole. Thespectrographer may sometimes derive useful information fromobserving these differences, and he then changes the position of hiscondenser lens so as to produce an image of the light source on theslit; e.g., V. M. Goldschmidt,22 in the analysis of rocks, minerals,etc., observes the spectrum in the layer immediately near the cathode,thereby attaining greater sensitiveness, especially when small quan-l7 “ Clinical and Pathological Applications of Spectrum Analysis ” (Adamla Mstallw., 1934, 13, 869.Is 2.tech. Phgsik, 1932,18, 573 ; Arch. Eisenhuttenw., 1933, 0, 551.21 “ The Practice of Spectrum Analysis,” F. Twyman and others, 6th edtn.,28 See L. W. Strock, “ Spectrum Analysis with the Carbon Arc CathodeHilger Ltd.).8. Elektrochern., 1934, 40, 313.p. 20 (Adam Hilger Ltd., 1935).Layer ” (Adam H,ilger Ltd., 1936)460 ANALYTICAL CHEMISTRY.tities of material are utilised for the analyses. This method, devisedby R. Mankopff and C1. Peters, was improved by L. W. Strock, 23who forms on a screen an enlarged image of the arc; that portionof the image whose spectrum is desired (i.e., the cathode layer)passing through a hole in the screen.A further condensing lensplaced near the slit serves to produce that uniform illumination whichis essential for photometric measurement of line intensities.R. Breckpot 24 (using a stepped sector and microphotometer)points out how the readings of the microphotometer can be falsifiedowing to the definition of the line falling off towards the ends whenthe long slit is employed, which is necessary in working with thesector, and for this kind of work it seems desirable to take care thatwhen such a sector is used only the central part of the spectrographlens system is utilised.Means of determining the Intensities of the Lines.-A very simplemeans of quantitative analysis, and one which has been verywidely u ~ e d , ~ * ~ ~ is provided by making lists of homologous pairs oflines, lines, namely, of minor and major constituents of a substancewhich at known percentages are of like intensity.When theminor constituent is present in only slight traces, most of the linesof the major constituent will be considerably over-exposed, butusually weaker lines will be found which can be utilised. There arecases, however, e.g., when magnesium is the major constituent, inwhich no suitable weak lines are to be found, particularly as it isdesirable, in order that comparison may be made accurately, thatthe lines to be compared should be very close together. Variousdevices have been employed to deal with such cases ; e.g., there maybe an impurity easy of determination by chemical means which pro-vides lines of a suitable low intensity, or a known quantity of stsubstance may be added to serve as an intermediate comparison.26When a synthetic mixture to match the sample can be prepared,a quantitative determination can be made without any photometricappliance, and the necessary instrumental equipment is reducedto a minimum. I n the routine analyses of metals and alloys sucha procedure is not feasible, and even where methods of synthesisare applicable, time can be saved in the chemical operations andincreased accuracy can be obtained if methods of photometry areavailable.A number of devices have been applied to this purpose,but only two, the logarithmic sector and the microphotometer, havebecome a t all widely used.The logarithmic secfor.This device, due to G. Sclieibe and A.23 op. cit.25 W. Gerlach and $1. Riedl, op. c i f .s6 qT. Seith and A , Jieil, Z. E'lektrochem., 1936, 42, 299.24 Ann. Soc. sci. Bruxelles, 1937, 57, 129TWYMAN SPECTRUM SNALYSIS BY EMISSION SPECTRUM. 461Ne~haiisser,~~ is a disc whose periphery is cut to a logarithmic curve,and which is rotated in front of the slit, so that the exposure ismade to vary along the slit. This results in the lines on the spectro-gram being of different lengths, which are thus a measure of theirrelative intensities.The logarithmic sector has been used by a number of workers inthe last five years 28 with both the arc and the spark. To give oneexample only, M. Slavin 29 used it in determining cobalt, iron,copper, cadmium, thallium, germanium, and lead in concentratedzinc sulphate solutions, and iron, copper, cadmium, and lead inmetallic zinc.His procedure differs from that of P. Twyman andA. A. Pitch30 only in that he used salts packed into holes drilledin pure carbon electrodes instead of the solid metals ; he achieves anaccuracy within 10% for the majority of tests, with occasionalerrors amounting to 20%. The speed permits the method to be usedin the works control of such operations as involve the removal ofmetal below arbitrary concentration. Slavin uses the dried saltssince it is then easy to make mixtures of known composition,It appears that an accuracy of within 10% in determining a minorconstituent is usually obtained by the use of the logarithmic sector.The method is simple, but it is very difficult to define precisely theend-points of the lines.J. S. Poster and C. A. Horton 31 find thisdifficulty reduced by the use of a special optical comparator wherebythe two lines are brought adjacent to one another and moved untilthey taper off together. These authors also replace the sector bya graduated wedge film of aluminium.The microphotometer. This gives an accuracy distinctly greaterthan that of the logarithmic sector, and by using the microphoto-meters of the non-recording type which have been designed for thiskind of work, routine analyses can be carried out very quickly.Among those who have published results obtained with this methodare H.Lundegiirdh (solutions and acetylene flame),' 0. Findeisen(lead alloys),32 0. S. Duffendack et al. (solutions used with the ~ p a r k ) , ~R. Breckpot (general analyses using the stepped sector) 24 and H. K.Whalley (aluminium alloys with the spark) .33I n confirmation of the superiority of the microphotometer to themethods of wedge or stepped spectra (logarithmic sector, etc.) onemay instance the work of J. Cholak 34 on the quantitative spectro-2 7 2. angew. Chem., 1928, 41, 1218; see also ref. (21).28 For references, see Smith, Zoc. cit., ref. (5).29 Eng. Min. J . , 1933, 134, 509.31 Proc. Roy. SOC., 1937, B, 123, 422.33 2. Metallk., 1933, 25, 12.33 J . SOC. Chem. Ind., 1937, 56, 1 8 0 ~ .34 J . Amer. Chem. Soc., 1935, 5'9, 104; Ind.Eng. Chern. (Anal.), 1935, '7, 287.30 J . Iron Steel Inst., 1930, 122, 289462 ANALYTICAL CKEMISTRY.graphic determination of lead in biological material. In his earlierwork, he used a rotating logarithmic sector, later he preferredmicrophotometer measurements.The microphotometer is a thermoelectric or photoelectric instru-ment with which the density of blackening of the lines on a spectro-gram can be measured. It was first used in astrophysical investi-gations, and the problem of deriving from such measurementsthe relative intensities of the radiations themselves was solved byphysicists by sound but laborious methods.An instrument with which measurements of blackening of theplate can be made 10 times as accurately as by eye could not longescape the notice of spectrographers, and five years ago it was alreadybeing applied by them.Some of the claims to very high accuracythen made were scarcely justified, for the authors had neglected tomake use af the experience of the physicists; but in 1933, 0. S.DuKendack was (in exceptionally favourable cases) achieving a1 yo accuracy in the analysis of solutions. The work is now develop-ing into a widely applicable routine which, while sound, is also simpleand quick when applied to routine analyses.The astrophysicists who used the microphotometer for measuringthe intensities of spectral lines found it necessary to calibrate eachspectrum plate by imposing upon it a set of exposures of thesame light source but of differing and known relative intensities,thus producing stepped spectra.This principle was adopted byDuffenda~k,~ for example, with excellent re~ults.~5 It may benecessary wherever it is important to push the method to its highestachievable accuracy (although the Reporter doubts it), but it isvery laborious and may be avoided by the procedure describedbelow.The following procedure adopted in the Reporter’s laboratory,in which he thinks an accuracy of 2-5y0 is attained in the routineestimation of minor constituents of metals and alloys by theirspark spectra, is the result of endeavours extending over someyears to reduce the methods of other workers to their simplestterms consistent with reliability. A number of standard specimensare prepared or collected, the composition of which is like that of thesamples to be tested, but in which the proportions of the variousconstituents vary over the permitted range.Spectrograms of thewhole range of specimens are taken on the one pZate. The plate isput on the microphotometer, and galvanometer deflections areobtained with selected pairs of lines, one of the main and one of36 See also D. H. Follett, J. Sci. Instr., 1936,13, 221 ; J. S. Foster and D. R.Foster and Horton, however, McRae, Proc. Roy. SOC., 1936, A , 163, 141.reverted to a wedge method, see ref. (31)TWYMBN : SPECTRUM ANALYSIS BY EMISSION SPECTRUM. 463each minor constituent. A graph is then plotted for each minorconstituent with logarithms of the ratios of the galvanometerdeflections as abscissae and logarithms of the ratios of the per-centages of the minor to those of the main constituent as ordinates.In analysis, spectrograms of the samples to be analysed are takenin the same way.Any desired number can be taken on the plate,but there must also be included spectrograms of at least two of thestandard specimens. Microphotometric measurement of the spectro-grams of the standard specimens and comparison of the results withthe graph serve to check that the graph is applicable to the conditions*obtaining a t the time, and if so, the percentages of the variousconstituents can be immediately read off from the graphs. If not,provided the discrepancy be not great, it is only necessary tomultiply the percentages obtained from the graphs by a factoreasily derived from the measurements on the spectra of the standards.Development along these lines seems likely to become the acceptedprocedure for quantitative spectrum analysis.Direct Photometry of the Lines.-A few workers have used directphotoelectric comparison of the intensities of the lines without theintermediary step of taking a photograph, and doubtless where onekind of determination has to be made very frequently such anarrangement may be quick and accurate.There are few labora-tories, however, where so specialised a method would be worthwhile.36Bibliography and TubZes.-A number of books and tables have beenpublished in the last five years, of which the following are perhapsthe most useful :W. and We. Gerlach, “ Clinical and Pathological Applications ofSpectrum Analysis ” ; translated by Joyce Hilger Twyman (AdamHilger, Ltd., 1934) : Describes methods applicable to the examin-ation of biological samples and their applications to the detection ofspecial elements in organs, secretions, etc., with notes on applicationsin mineralogy and other special chemical problems. Contains verymany valuable notes : e.g., reference to coincidence lines, vix.,lines of different elements which may be confused one with another.W. Gerlach and E. Riedl, “ Die Chemische Emissionspektral-analyse,” Part I11 (Leopold Voss, 1936) : Very valuable tables ofthe most useful lines of a number of metals, each table includinglists of coincidence lines and of homologous line-pairs.Adam Hilger, Ltd., “The Practice of Spectrum Analysis,” 6th36 W.H. Jansen, J. Heyes, and C. Richter, 2. physilcal. Chem., 1935, 174,* Such as kind of plate, temperature and kind of development, condition291.of electrodes, exposure, etc464 ANALYTICAL CHEMISTRY.edtn. (1935) : A condensed but comprehensive review of themethods of spectrum analysis.fifitroductory Books.-S. Judd Lewis, “ Spectroscopy in Scienceand Industry ” (Black & Son, 1933) : An adequate introduction tothe methods, including those of absorption spectrography.P. Swings, ‘( La Spectroscopie Appliquke ” (George ‘Shone, 1935).Finally may be mentioned the indispensable classified “ Biblio-graphy of Literature on Spectrum Analysis,” by D. M. Smith(British Non-Ferrous Metals Research Association, 1935).This isconcerned mostly with metallurgical applications, but includesalso references to the most valuable general reviews of the subject.F. T.2 . THE ANALYSIS OF ROCKS ANY MINERALS.Since the publication of the last editions of W. F. Hillebrand’sand H. S. Washington’s classic treatises on rock analysis, a con-siderable amount of work has been done in this field, but most of ithas consisted either of a critical examination of existing standardmethods, or the adaptation of previously-known ones to the specialproblems involved in rock analysis. During recent years, in-creasing importance has been attached to the distribution of theless common elements in rocks and minerals on account of theirgeochemical significance.E. Troger has discussed this questionwith reference to the igneous rocks, and stressed the desirabilityof the determination of the minor constitueflts in rock analyseswhenever possible. Owing to the very small proportions in whichthese elements occur in the majority of rocks, the ordinary chemicalmethods are frequently inapplicable, and have been largely replacedby spectroscopic processes. H. Moritz 4 has enumerated the mostsuitable lines for use in the spectroscopic detection and determinationof the various elements in ores and rock-forming minerals, but theprincipal worker in this field has been V. M. Goldschmidt. Heand his co-workers have developed a special technique in theapplication of quantitative spectroscopy to the determination oft’he proportions in which the rarer elements, such as gallium,“ The Analysis of Silicate and Carbonate Rocks,” Washington, 1919.Most of the subject-matter of this monograph has been incorporated inHillebrand and Lundell’s book, “ Applied Inorganic Analysis,” New York,1929.“The Chemical Analysis of Rocks,” New York, 1930.Other worksdealing with this subject are : “ Anleitung zur chemischen Gesteinsanalyse,”by J. Jakob, Berlin, 1928 ; “ Gesteinsanalytisches Praktikum,” by E. Dit,tler,Berlin, 1933 ; “ Silicate Analysis,” by A. W. Groves, London, 1937.Chiem. Erde, 1934, 9, 286. Ibid., 1933, 8, 321BARWOOD : ANALYSIS OF ROCKS AND MINERALS. 465germanium, the platinum metals, etc., occur in the rocks andminerals constituting the crust of the earth.5The distribution of a large number of elements has been investig-ated by them, with surprising results in many cases, such as theunsuspected presence of germanium to the extent of 1% in the ashof certain coals.The spectroscopic methods in general affordquantitative figures for percentages of the element under investig-ation ranging from 0.01 to 0*0001 yo, and thus begin where ordinarychemical methods leave off.The tendency of modern petrological research to attach greaterimportance to the chemical composition of the individual mineralspresent in the rocks rather than to the figures obtained from thebulk analysis of the rock as a whole, has led to the demand for achange in analytical technique. Although the amount of materialavailable for the analysis of a rock is usually ample, the case isdifferent where the constituent minerals are concerned.Thesemay in certain cases form only a small fraction of the whole rock,and their isolation in a pure state is often a difficult and laborioustask, so that the analyst is ultimately faced with the problem ofcarrying out a complete analysis, involving the determination ofpossibly 13 or 14 constituents, on 1 g. of material or even less.In order to cope with demands of this kind, attempts have beenmade during the past few years to adapt the existing methods ofrock analysis so as to permit of the work being carried out on asemi-micro scale.A detailed account of the modified procedure employed in suchcases is given by W.C. Guthrie and (Miss) C. C. Miller,6 who haveused it in the analysis of several igneous rocks. Samples weighing0.1 g. were taken for the various determinations, a total of about0.7 g. being required for the whole analysis. Two analyses, eachcomprising 13 constituents, could be completed in four days, andthe figures obtained were in excellent agreement with those affordedby check analyses carried out by the usual methods. Work onsimilar semi-micro lines has also been carried out by W. Janczak,'who employed it for the analysis of dolomite and chemical glass.A start has also been made with the application of purely micro-chemical methods to the analysis of minerals and rocks. Oneof the earliest analyses of this kind was made by A. Benedetti-Pichler and H.Thurnwald,s who analysed kolbeckite, a mineral6 See numerous papers in the Nach. Qes. Wiss. GGttingen, and SET. NorskeVidenskaps-Akad., Matemat.-Naturv. Klasse, €or 1930-1935, and summaryin J., 1937, 655, where a bibliography will be found.6 Min. Mag., 1933, 23, 405. Rocz. Chem., 1935,15, 304; 1936,16, 377.8 Mikrochem., 1932,11,200; see also A. Benedetti-Pichler and F. Schneider,ibid., 1930, Emich Festschrift, 1466 ANALYTICAL CHEMISTRY.containing beryllium silicate and phosphate, in addition to iron,aluminium, calcium, and magnesium. Analyses of a number ofgarnets, isolated from various rocks, have been made on a micro-chemical basis by H. Hueber,g 30 mg. of the mineral being requiredin each case.*The chief worker in this field is, however, F.Hecht, who hascarried out much excellent and painstaking pioneer work, includingmicrochemical analyses of several radioactive minerals, such aspitchblende, thorianite, and monazite, using 20-50 mg. of material.1°The same worker subsequently extended similar methods to silicaterocks, and has analysed a glassy material separated from transfusedquartz.ll In a later paper,12 Hecht instances the advantages ofemploying microchemical methods of precipitation, filtration, etc.,in the determination of the minor constituents, such as manganeseand nickel, in rocks, and describes the procedures with full experi-mental details. Two analyses of basalts are given in illustration,in which the minor constituents have been determined in this way.More recently still, a complete scheme for micro-silicate analysishas been published by the same author.13 The methods givenpermit of the determination of the 16 principal constituents ofsilicate rocks, a total of eight portions of rock powder, varyingin weight from 10 to 20 mg.each, being required for the wholeanalysis. The procedure followed differs from that employed onthe macro-scale in several respects, notably in that silica is deter-mined in one separate portion of rock powder, whilst in a secondportion alumina, titania, total iron, lime, magnesia, and manganeseoxide are determined; in the latter case, the old unsatisfactorymethod of precipitating the manganese as sulphide and igniting tooxide is retained. Five analyses carried out by the above methodare cited in the paper.Tech.Min. Petr. Mitt., 1932, 43, 84.10 Mikrochem., 1931, 10, 46; 1932, 12, 193; F. Hecht and W. Reich-Rohrwig, ibid., p. 281; F. Hecht, Amer. J . Sci., 1934, 27, 321; F. Hecht andE. Kroupa, 2. anal. Chem., 1935,102,81; 1936,108,82; in this last paper thePb/U ratio is computed from the analyticaI data, and applied to the calculationof the geological age of the minerals.11 See A. Holmes and H. F. Harwood, " Petrology of the Volcanic Area ofBufumbiro, Uganda Geological Survey Memoir, 1937, 3, Part 2, 254, wherethese analyses are given.la 2. anal. Chem., 1937,110, 385.l3 F. Hecht, Mikrochim. Acta, 1937, 2, 188. * See also H. Alber and C . Benedicks, Arkiv Kemi, Min. Qeol., 1933, 11A,nr. 6 ; R. Treje and H.Alber, Jernk. Ann., 1933, 117, 457 ; M. Shioiri and T.Nagahara, J . Imp. Agric. Exp. Sta., 1933, 2, 161 ; M. Shioiri and S. Kane-matu, J . Japanese Ass. Min. Petr. Econ. Beol., 1934,12, 180; J. Iwasaki, ibid.,1935,13, 93HARWOOD ANALYSIS OF ROCKS BND MINERALS. 467The results obtained by Hecht in his analyses of minerals and rockson a purely microchemical basis are tolerably satisfactory, but inthose cases where a comparison analysis on the macro-scale wascarried out at the same time divergencies are often noticeable, thevalues obtained microchemically for lime and magnesia in par-ticular being frequently erratic. Experiments made with an“ artificial pitchblende,” synthesised from standardised solutions,14showed similar discrepancies, and it is quite evident that the micro-chemical analysis of complex minerals md rocks in which a largenumber of separations are involved cannot yet yield results corn-parable in accuracy with those obtained by ordinary macro-methods.The present tendency to extol microanalysis in all cases at theexpense of the mecro-methods, is, in the Reporter’s opinion, entirelyunjustifiable. A number of micro-methods, especially those in-volving organic compounds, such as the micro-Kjeldahl determin-ation, are superior to the corresponding macro-methods, in that fheyeffect a great saving of time and material without any sacrifice ofaccuracy.In mineral analysis, however, the case is entirelydifferent. Here, a number of determinations have to be madesuccessively on a single weighed portion of material, involving a,series of complex separations, whilst the constituents to be de-termined are usually present in widely differing amounts.Muchstress is frequently laid upon the saving of time effected by micro-methods in comparison with macro-ones. In a recent paperdealing with the appliances and methods used for the analysis ofminerals on a rni~ro-scale,~~ Hecht rightly gives a warning againstthe very real danger that any attempt unduly to shorten the timerequired for a microchemical analysis must inevitably result in aloss of accuracy. The saving of time is, moreover, often largelyillusory when all the relevant factors are taken into account.Weighings on a micro-balance consume much more time than thosemade on an ordinary chemical balance (especially on a modernaperiodic type), and micro-analytical work requires close attentionthroughout, so that it is not easy to keep a number of analyses inprogress simultaneously, as can readily be done on the macro-scale.The oarrying out of microchemical analyses also demands ahighly specialised technique and excellent working conditions, thelatter not always readily to be attained. At the present time it isimpossible t o carry out successfully all types of mineral and rockanalyses on a micro-scale, since, as Hecht states, the necessaryconditions for the separations have in many cases not yet beenworked out. Furthermore, when coarse-grained porphyritic rocks14 F. Hecht and H. KrafYt-Ebing, Mikrochem., 1934, 15; 39.15 Osterr.Chem. Z$., 1937, nr. 10, 1468 ANALYTICAL CHEMISTRY.are being investigated, composed of minerals differing widely infriability and hardness, and especially when micaceous mineralsare present, it seems doubtful if a sample weighing only 30 mg.or so would be truly representative.16In view of the above facts, it is the Reporter’s opinion that rnicro-analytical methods should be employed for rocks and mineralsonly in those cases where the small amount of material availablerenders their use imperative, for otherwise any saving of timeand materials effected is more than offset by the diminished accura,cyof the results.l7On the other hand, serni-micro methods do not suffer to such amarked extent from the disadvantages inseparable from puremicro-methods. They can be made to embody the best points ofboth the micro- and the macro-procedure, and offer a fruitful fieldfor development in their application to this branch of analysis.Recent work in connection with the macro-analysis of silicaterocks and minerals will now be discussed under the headings of theindividual determinations.8ampEing.-The variations shown in the figures obtained insuccessive analyses made from the same rock mass, and due todifferences in sampling, are discussed by F.F. Grout,18 who con-cludes from numerous analyses that, although for fine-grained rocksa single hand specimen is sufficient to provide a representative samplefor analysis, yet a different procedure is necessary for coarse-grainedrocks.It is recommended that the latter should be sampled bytaking a large number of chips at random from a fresh outcrop, upto a total of 60 lbs. or more in weight. The material thus obtainedis then crushed and quartered in order to obtain the final samplefor analysis. SuScient consideration is often not given to thispoint, and much subsequent careful analytical work may therebybe invalidated.Silica.-In connection with the determination of silica in silicates,W. F. Hillebrand and G. E. F. Lundell l9 point out that the thirdl6 See, however, J. Mika (2. anal. Chevb., 1928, 73, 257) and B. Baule andA. Benedetti-Pichler (ibid., 74, 442), who deduce mathematicalIy from thetheory of probability that a 5 mg. sample should be fully representative,This requires a particle size of 0-001 mm., and the effects of such fine grindingon the oxidation of the ferrous compounds and sulphides present in the rockcannot be ignored.l7 F.Hecht, in a private communication to the Reporter, has endorsedthe above, stating that in his opinion it is not possible t o carry out a complexrock analysis by purely micro-methods with the same degree of accuracy as isattainable on the macro-scale.la Amner. J . Sci., 1932, [v], 24, 394.l o “ Applied Inorganic Analysis,” 1929, pp. 540, 722HARWOOD : ANALYSIS OF ROCKS AND MINERALS. 469evaporation formerly recommended is unnecessary, as equilibriumis reached after two treatments, and a third evaporation yieldsno more silica, the small amount still remaining in solution havingto be recovered subsequently from the R20, precipitate.Nonumerical data are furnished, however, in support of this contention,which is not in accordance with the Reporter's own experience.The inconvenience resulting in some cases from the presence oflarge amounts of sodium salts in the solution obtained after thefusion of the rock with an excess of sodium carbonate as usuallyprescribed, is overcome by a method put forward by A. N. Finn andJ. F. KlekotkaY2O in which 0-5 g. of the silicate is intimately mixedwith only 0.6 g. of sodium carbonate, and the whole heated in aplatinum crucible in an electric muffle furnace for two hours at875". The mass is then digested with a little water, acid is added,and the analysis finished as usual.Two new papers dealing with the determination of silica in thepresence of fluorine have appeared, by J.I. Hoffman and G. E. F.Lundell 21 and W. T. Schrenk and W. H. Ode 22 respectively. Thefirst deals with an improvement of the old Berzelius method, zincnitrate being substituted for ammonium carbonate in the removalof the main bulk of the silica; and in the second it is claimed thatfluorspar in the presence of silica can be completely decomposedwithout loss of silica if heated with perchloric and boric acids, all thefluorine being evolved as boron trifluoride. The residual silica maythen be determined as usual.A difference method, basedon the decomposition of the sample withhydrofluoric and sulphuric acids, followed by conversion of the sul-phates into orthophosphates, by ignition with sodium meta-phosphate, is described by G.T. Galfajan and W. M. T a r a j a ~ ~ . ~ ~The results are said t o be accurate to O*lyo, even in the presence of29% of alumina. The above method has been adapted for micro-chemical procedure by K. Sch~klitsch.~~" Free Silica."-The determination of the amount of " free silica "(quartz) present in a rock is of importance owing to its presumedbearing on the question of silicosis. The old Lunge and Millbergprocess gives unsatisfactory results, but for coal-measure rocks anaccuracy of 1% is claimed by A. S h a ~ , ~ ~ using a modification ofW. A. Selvig's process,26 and this is confirmed by L. R. D ~ n n . ~ 'A. Knopf 28 used hydrofluosilicic acid to attack the silicates of the2 1 Ibid., 1929, 3, 581.23 2.anal. Chem., 1933, 92, 417.25 Analyst, 1934, 59, 446.20 Bur. Stand. J . Res., 1930, 4, 809.22 I n d . Eng. Chern. (AnaZ.), 1929, 1, 201.24 Nikrochem., 1935, 18, 144.2 6 Carnegie Inst. Tech. Min. Met. Invest. Bull., 1928, No. 21.27 Analyst, 1935, 60, 36.28 U.S. Pziblic Health Reports, 1933, 48, 183470 AXALYTICAL CHEMISTRY.rock, but W. R. Line and P. W. Aradine 29 recommend the replace-ment of this by hydrofluoboric acid as having less action on thequartz present. They claim that an accuracy of approximately1% is attainable in the determination of the quartz. The errorcaused by the solubility of the quartz in hydrofluosilicic acid in-creasing with its fineness of division has also been pointed out byC.B. M ~ k e . ~ ~It has been shown 31 that micrometric estimations of the amountof quartz present in thin sections made from igneous rocks agreewithin 2% with the results obtained by chemical methods. Withsedimentary rocks the figures obtained are too low, and the methodcannot therefore be applied to shale or slate.A modified petrographic immersion method for free silica inrock powders or dusts has been described by H. L. Ross and F. W.Seh1,32 and a calorimetric method for the determination of ths“ active silica ” in puzzuolana has also been evolved.%A2umina.-The customary procedure hitherto baa been to takethe alumina by difference after determination of all the other oxidescomposing the “ ammonia precipitate.’’ A direct determinationof this constituent has now been rendered possible by the applic-ation of 8-hydroxyquinoline.Iron, titanium, and zirconium areremoved by an excess of sodium hydroxide, and the filtrate, contain-ing aluminium (and beryllium, if present) is acidified with hydro-chloric acid. The aluminium may then Be precipitated by theaddition of an acetic acid solution of 8-hydroxyquinoline, followedby the addition of 2N-ammonium acetate, the beryllium remainingin solution. By suitably modifying the conditions, a separationfrom phosphoric acid can also be attained.3* The above method hasbeen applied by Knowles 35 and by Knowles and J. C. Redmond 36to the determination of alumina in felspar. It has also beenutilised by C.0. Harvey3’ in the analysis of an apatite rock, andby P. P. Budnikov and S. S. Shukovskaja38 in that of bauxites,clays, and earthenware. The errors which result in this method iftitanium is not previously removed have also been pointedThe 8-hydroxyquinoline reagent has also been used by K. Schok-litsch40 for the precipitation of iron, aluminium, and magnesium29 Id. Eng. Chem. (Anal.), 1937, 9, 60..w A. Shaw, Bull. Inst. Min. Met., 1936, No. 385.Ind. Eng. Chem. (And.), 1935, 7 , 30.s3 P. P. Budnikov and L. G. Gulinova, Kolloid-Z., 1935, 70, 100.84 H. B. Knowles, J. Res. Nat. BUT. Stand., 1935, 15, 87.86 L O C . cit.37 Analyst, 1936, 61, 817.a9 P. Koch, Ber. deut. Iceram. Ges., 1935, 16, 118.40 MikTochem., 1936, 20, 247.30 J .I d . Hyg., 1936, 18, 299.a6 J . Arner. Ceramic Soc., 1936, 18, 106.88 J . Appl. Chem. Ru8sk, 1936, 9, 2079HARWOOD : ANALYSIS OF ROCKS AND MINERALS. 471in the micro-analysis of silicates. Contrary to the generallyaccepted view, 0. V. Krasnovski41 states that alumina can beaccurately determined by the usual methods (precipitation withammonia in presence of methyl-red as indicator) in aluminium-bearing borosilicates without previous removal of the boric oxide,even when as much as 30% of the latter is present. With more than10% of alumina, however, a double precipitation is necessary.The replacement of ammonia by pyridine * has been recommendedfor the precipitation of the R,O, oxides in the analysis of cobaltand manganese ores, but does not appear to have been tested yetfor rocks, although from a preliminary experiment carried out in theReporter's laboratory the method seems promising.Ferrous and Ferric OxidM.-In the determination of ferrous oxidein rocks, V.Smirnov and N. Aidinjan43 decompose the samplewith hydrofluoric and sulphuric acids under a layer of toluene.The resulting solution is poured into water, and the determinationcompleted as usual.In order to avoid any transference of the ferrous iron solution,as is inevitable in the usual methods of procedure, B. A. Soule44effects decomposition of the sample in a Pyrex flask with hydro-fluoric and sulphuric acids ; the titration is made electrornetricallyin the same vessel with ceric sulphate solution. In these circum-stances reducing substances (arsenious oxide) dissolved from theglass are without influence, which is not the case if permanganateis used.A number of rock-forming minerals such as tourmaline,ilmenite, and some varieties of garnet are very imperfectly decom-posed by hydrofluoric acid; in such cases Pratt's method giveslow results for the ferrous oxide content, but a method worked outby H. P. Rowledge45 can be employed. The rock or mineral isfused at 950" in a Pyrex tube with sodium borofluoride, the melt dis-solved in sulphuric acid in absence of air, and the solution titratedwith permanganate. A. T. Tscherni 46 determines ferrous oxidein rocks and minerals containing manganese peroxide by heatingthe sample in carbon dioxide for one hour a t 100" with standardisedferrous sulphate solution in the presence of sulphuric and hydro-fluoric acids; the excess of ferrous salt is then titrated with per-manganate.In another sample the active oxygen present is de-termined by the usual method, and the ferrous oxide content canthen be calculated.2. anal. Chern., 1929, 79, 175.dz See Ann. Reports, 1936, 33, 442, ref. 88.43 Compt. Tend. Acad. Sci. U.R.S.S., 1937, 14, 353.44 J . Amer. Ghem. Soc., 1928, 50, 1691; 51, 2117.45 J . Roy. SOC. Western Australia, 1933-1934, 20, 165.46 Ukrain. Chern. J., 1936, 11, 15472 ANALYTICAL CHEMISTRY.In general, the amount of ferric oxide in a rock is obtained bydifference, after determinations have been made of the ferrous ironand total iron content.It may, however, be determined directlyby a method due to 0. Kackl.47 The solution obtained by dis-solving the rock in hydrofluoric and sulphuric acids is titrated withtitanous sulphate solution, after addition of boric acid to combinewith the excess of hydrofluoric acid. The results are always slightlytoo high, due to oxidation taking place during the decompositionof the sample, but the process is useful whenever organic matteris present (e.g., in bituminous shales) which would interfere withthe customary permanganate titration.Magnesium.-If manganese has not previously been removed, itwill be precipitated with the magnesium, and L. A. Dean and E.Truog 48 have shown that this precipitation is quantitative. Theseworkers precipitate the two metals together as phosphates, weighor titrate the precipitate, and then determine the manganese presentby the bismuthate method, the magnesium being found bydifference.*8-Hydroxyquinoline has been applied by several workers to thedetermination of magnesium in silicates.49 A point which does notappear to have been fully investigated in this connection is thebehaviour of any manganese, not previously removed, in the solu-tion a t the time the precipitation is made, and the necessity forintroducing a correction for its possible presence in the precipitate.According to K. Schoklit~ch,5~ who worked on a micro-scale, thewhole of the small amount of manganese present was precipitatedwith the magnesium complex, and could be determined colori-metrically after destruction of the oxine with nitric acid and hydro-gen peroxide.Only two results are quoted, however, and in eachcase the amount of manganese found was only half that given in acorresponding macro-analysis.Potassium and Sodium.-As the blank on the reagents used inLawrence Smith’s process comes almost entirely from the calcium47 2. anal. Ch,em., 1925, 66, 401.4 8 Ind. Eng. Chem. (Anal.), 1935, 7, 383.49 J. I. Hoffman and G. E. F. Lundell, Bur. Stand. J . Res., 1930, 5, 299;6o Milcrochem., 1936, 20, 247.* In the Reporter’s laboratory, it has been the practice for years to deter-mine the amount of manganese present in the magnesia precipitate after thishas been weighed, and correct the weight accordingly. It has been foundthat in general about two-thirds of the total manganese present is precipitatedwith the magnesia, and the balance with the R,03 oxides, onIy a negligibleamount accompanying the lime.This method has proved more satisfact’orythan separating the manganese as sulphide in the course of the analysis,J. Robitschek, J . Arner. Ceramic SOL, 1928, 11, 587, et alHARWOOD : ANALYSIS OF ROCKS AND MINERALS. 473carbonate, E. R. Caley 51 prefers to dissolve 2 g. of this in hydro-chloric acid, and to determine the sodium directly in the solutionwith magnesium uranyl acetate. Whether it is permissible toignore the amount of alkali introduced from the glass of the washbottles, etc., during the operations is a moot point, although thework of Miller and Traves (see below) seems to support the con-tention.No provision is made, however, for the possible presenceof potash, a small amount of which is generally obtained in a blankrun, according to the Reporter’s own experience.A modification of the Lawrence Smith method which permitsof the determination of the alkalis without previous removal of thecalcium has been worked out by (Miss) C. C. Miller and (Miss) F.Traves.52 The solution obtained by the usual procedure is made upto a known volume, and sodium is determined in an aliquot part bydouble precipitation with zinc uranyl acetate. In a second portion,potassium is determined as perchlorate after a preliminary separa-tion as cobaltinitrite. Lithium, if present, is determined by extrac-tion of the lithium chloride (and calcium chloride) with isoamylalcohol or acetone, and subsequent determination as lithium zincuranyl acetate under carefully controlled conditions.It wasfound by these workers that a direct determination of the sodiumcontent of the calcium carbonate used, made by a double precipit-ation with magnesium uranyl acetate, checked well with the amountactually obtained in a blank run.The well-known Berzelius method has been modified by E. W.Koenig,53 who removes aluminium, iron, and magnesium fromthe solution of the rock by the addition of lime, the remaining silicabeing simultaneously eliminated as silicofluoride . The excess ofcalcium is then removed with ammonium carbonate, and theanalysis finished as usual.Cmsium and Rubidium.-These two elements are present inappreciable quantities in many naturally occurring silicates, suchas beryl and zinnwaldite (see V.M. Goldschmidt 5), but have rarelybeen determined owing to the difficulties attending their separation.A procedure for the separation and determination of all the alkalimetals has now been worked out 54 which enables small quantitiesof rubidium and c*sium to be determined in minerals in additionto lithium, sodium, and potassium ; the principal reagents requiredare platinic chloride, alcohol, ether, and ammonium sulphate. Iflarge amounts of rubidium or cesium are present (as in pollucite)some modification of the original method is ne~essary.~5I n d . Eng. Chern. (Anal.), 1929, 1, 191.53 Ind. Eng.Chern. (Anal.), 1935, 7, 314.54 R. C. Wells and R. E. Stevens, ibid., 1934, 6, 439.66 J. C. Hillyer, ibid., 1937,9,236; R. C. Wells and R. E. Stevens, ibid, (reply).62 tJ., 1936, 1390474 ANALYTIUAL CHEMISTRY.Wader.-The usual methods for the determination of water insilicates give untrustworthy results in the case of many micas, andan improved procedure, which is a combination of Penfield's andthe sodium tungstate method, has been devised by K. Wiskont andI. Alimarin 56 for use in the analysis of micas. The fact that epidoteis one of the few minerals which do not give up the whole of theirwater when heated to 1000" in a current of dry air has been pointedout by A. F. S m e t h ~ r s t , ~ ~ who obtained only 0.35% of water inthis way from a specimen of the mineral, whereas fusion with sodiumtungstate gave 1.32%.It has been frequently confirmed in theReporter's laboratory also that; with many rocks fusion with sodiumtungstate yields a higher percentage of water.F. Hecht 58 determines water in minerals microchemicdly byignition in a current of dry air, using a modified Pregl combustionapparatus; sulphur dioxide and hydrochloric acid if present are re-moved by a layer of lead monoxide and dioxide. The determin-ation of water in rocks and minerals by micro-methods has also beencarried out by E. Dittler and H. H ~ e b e r . ~ ~Carbon Dioxide and Carbon.-The use of perchloric acid forliberating carbon dioxide in the analysis of carbonates has beenrecommended 6O and an apparatus for the purpose described.M.H. Hey has devised an apparatus in which the liberatedcarbon dioxide is absorbed in barium hydroxide solution, and theresulting barium carbonate filtered off and washed in the absenceof atmospheric air before being weighed. Although doubtlesssatisfactory, the method seems unnecessarily elaborate by corn-parison with the excellent and much simpler one in which G. T.Morgan advocated the use of phosphoric acid.62 A micro-methodfor the determination of carbon dioxide in minerals such as Icelandspar is given by E. Dittler and H. Hueber 63 : 3-5 mg. of the mineralare decomposed by hydrochloric acid, the oarbon dioxide evolvedis collected in barium hydroxide solution (containing a little bariumchloride to diminish the solubility of the barium carbonate formed),and the excess baryta is titrated with 0.01N-hydrochloric acid.An improved wet-combustion method for the determination oforganic carbon in rocks and minords has been worked Thisis based on Morgan's process, but an additional flask is used, con-taining chromic acid, phosphorio acid, and a, little mercuric oxide,to ensure completeness of the oxidation.Carbon dioxide and56 2. anal. Ckem., 1929, 79, 271.6 8 Mikrochim. Aca, 1937, 1, 194.59 2. anorg. Chem., 1931, 195, 41; 199, 17.6o C. A. Jackson and J. W. Haught, Ind. Eng. Chem. (Anal.), 1930, 2, 334.61 Min. Mag., 1935, 24, 76. 62 See Ann. Reports, 1936, 38, 447.c3 2. anorg. Chem., 1931, 199, 26. O4 B. E. Dixon, Analyst, 1934, 59, 739.6 7 Min. Mag., 1935, 24, 173HXRWOOD : ANALYSIS OF ROCKS AND MINERALS.475carbon are determined successively in the mme aample of material,and the method is especially well suited for the determination oforganic carbon in rocks which contain considerable amounts ofcarbon dioxide.Titanium Dioxide.-The colorimetric determination of titaniumby means of hydrogen peroxide has been studied in detail by H.Gin~berg,~~ who finds that as little as 0.07 mg. of titanium dioxidein 100 ml. of solution can be determined with an accuracy of 10%if a Pulfrich photometer is used. The well-known bleaching of theyellow pertitanate colour due to phosphate ion may be prevented,according to I?. G. GermuthYG6 by the addition of 1 ml. of 0.1%uranium acetate solution for each 0.1 mg.of titanium dioxidepresent.An extremely useful method for the gravimetric determinationof titanium 67 affords a direct separation from iron, aluminium,manganese, and phosphoric acid, and can consequently be applieddirectly to the R,03 precipitate. The latter is dissolved in sulphuricmid, and the titanium precipitated from the acid solution by theaddition of tannin and phenazone, an orange-red precipitate beingformed, which on ignition yields titanium dioxide. The Reporterhas obtained promising results with this method in the determinationof titanium in rocks when the amount present was too large for thecolorimetric method to be used.Phosphoric Oxide.-The well-known effect of titanium in hinder-ing the quantitative precipitation of phosphoric acid as ammoniumphosphomolybdate may be offset G8 by the addition of considerableammonium nitrate or chloride together with free acid, providedthat less than 35 mg.of titanium are present ; with more than thatamount, quantitative precipitation of the phosphorus cannot beattained. The complete analysis of a phosphate rock requiresmodification of the usual methods, and a scheme for use in such acase as exemplified by apatite rock is given with full practical detailsby C. 0. Har~ey.6~ The determination of phosphoric oxide by 8direct precipitation with magnesia mixture in the presence ofammonium citrate to prevent interference by iron and other metalshas been used successfully by J. I. Hoffman and G. E. F. Lundell 70in the analysis of phosphate rock, silicate cements, eto., and itsemployment in referee analyses of these materials is advocated.2irconia.-The selenite method for the determination ofa@ 2.anorg. Chsm., 1931,198, 162; 1932, 209, 106; 1933, 211, 401; 1935,66 J. AmeT. Chern. Soc., 1928, 50, 1910.6 7 L. Moser, K. Neumayer, and K. Winter, Monatsh., 1930, 55, 86.68 G. HergBrd, Z. anal. Chem., 1933, 9S, 329.69 Analyst, 1936, 61, 817.226, 57.7* J. Res. Nat. BUT. Stand., 1937, 19, 59476 ANALYTICAL CHEMISTRY.zirconium 7 1 has now been combined with the older phosphateprocess, and applied to 0res.72 A further modification consists inthe precipitation of the zirconium as arsenate, and conversion intozirconia by ignition, the arsenic in the precipitate being removedby ignition with carbon (filter-paper).73Hillebrand’s original method for the determination of zirconiumin rocks, which is somewhat tedious, has been modified by H. F.Harwood. 74Fluorine.-0. Hack1 75 uses a duplication method based onSteiger’s colorimetric process with titanium sulphate and hydrogenperoxide. R. E. Stevens 76 uses the Berzelius method of attack,and treats the solution containing sodium chloride and fluoridewith gelatin, alcohol, and calcium chloride. Calcium fluoride isformed as a protected colloid, and may be determined nephelo-metrically ; an accuracy of 1 yo on the fluorine is claimed. The fadingin colour produced by fluorine in solutions of iron acetylacetone isutilised by W. D. Armstrong 77 for the determination of this element ;differences due to variations in acidity are eliminated by a duplica-tion method.F. Specht 78 has applied the lead chlorofluoridemethod to the analysis of cryolite, and the determination of fluorinein glasses and enamels is the subject of a paper by J. I. Hoffmanand G. E. F. L~ndell,’~ who substitute zinc nitrate for ammoniumcarbonate in the removal of the silica,* and determine the fluorineas lead chlorofluoride. Fluorine may be determined in cryoliteby a volumetric method due to F. J. Frere,80 in which the titrationis effected with yttrium nitrate or cerous nitrate solution; the leadchlorofluoride method was found to give low results. A numberof volumetric and gravimetric methods for the determinationof fluorine in zinc ores are discussed by I,.P. Taylor.81The bleaching action of fluorine on the reddish-violet colourgiven by alizarinsulphonic acid with soluble zirconium salts hasbeen utilised by several workers. H. Leitmeier and F. Feigl 82and also I. P. Alimaring3 have applied the reaction to thedetection of small amounts of fluorine in minerals and rocks.7 1 S. G. Simpson and W. C. Schumb, J . Amer. Chem. A~OC., 1931, 53, 921.72 Idem, Ind. Eng. Chem. (Anal.), 1935, 7, 36.75 W. C. Schumb and E. J. Nolan, ibid., 1937, 9, 371.54 Tidsskr. Kjemi Berg., 1932, 12, 23. 75 2. anal. Chem., 1934, 97, 254.713 Ind. Eng. Chem. (Anal.), 1936, 8, 248. 7 7 Ibid., 1933, 5, 300.78 2. anorg. Chem., 1937, 231, 181.Ind. Eng. Chem. (Anal.), 1933, 5, 17.82 Tsch. Min. Petr.Mitt., 1929, 40, 6. * This method for the removal of silica has been adopted in the Reporter’slaboratory for use in the Berzelius method of determining fluorine, and foiindto be a decided improvement, efferting an appreciable saving of time.Bur. Stand. J . Res., 1929, 3, 581.81 I n d . Chem., 1937, 13, 221.83 2. anal. Chem., 1930, 81, 8HARWOOD : ANALYSIS OF ROCKS AND MINERALS. 477The last worker mixes the powdered rock with boric oxideand heats it in a hard-glass tube, as in Penfield’s method forwater. If fluorine is present, hydrofluoboric acid collects withthe water in the bulb, and the fluorine may then be detectedby the zirconium-alizarin reaction. It is claimed that O . O O S ~ Oof fluorine can thus be detected in 0.3 g. of rock.According toF. Feigl and E. Rajmann,84 the sensitivity of the above tests maybe greatly increased by the replacement of zirconium-alizarin byzirconium p-dimethylaminoazophenylarsonate : 0.25 pg. of fluorinecan then be detected. Attention has been directed 85 to the fact,that the volatilisation method for fluorine (as silicon tetrafluoride)yields untrustworthy results for slags and certain mineral phosphateswhich contain silicates decomposable by acid. The error is ascribedto the formation of a non-volatile compound of fluorine, probablySiOF, .Vanadium, Chromium, Molybdenwm-The methods employedfor the determination of vanadium in rocks are discussed by K.Jost,86 and the distribution of chromium in rocks has been studiedby both ordinary and X-ray spectroscopic methods by several~orkers.~7 An entirely new process for the determination of theabove three elements in rocks, based on colorimetric methods, hasbeen worked out by E.B. Sandell; 88 0.001% of chromium orvanadium, and O - O O O l ~ o of molybdenum can be determined in thesame 1-g. sample, which is decomposed by fusion with sodiumcarbonate. Vanadium is determined colorimetrically with phospho-tungstic acid in an aliquot part of the aqueous extract of the melt,after it has been previously separated from chromium by the ex-traction of its 8-hydroxyquinoline compound with chloroform.Chromium is determined with diphenylcarbazide after removal ofvanadium, and molybdenum by the stannous chloride-thiocyanate-ether process. The Reporter finds that the method gives quitesatisfactory resultls, and effects a considerable saving of time andmaterial over the older process of Hillebrand, which required 5 g.of rock.A new study of the colorimetric phosphotungstate methodfor vanadium, as used in the above process, has been made byE. R. Wright and M. G. Mel10n.~~Nickel.-A direct method for the determination of nickel oxide inrocks has been worked out by H. F. Harwood and L. S. Theobald.9084 Mikrochem., 1932, 12, 133.85 D. S. Reynolds and K. D. Jacob, Irtd. Eng. Chern. (Anal.), 1931, 3, 371.86 Chern. Erde, 1932, 7, 177.V. M. Goldschmidt and C1. Peters, Nach. Ges. Wiss. GBttingen, 1933, 278;G. voii Hevesy, A. Mesha, and K. Wiirstlin, 2. anorg. Chem., 1934, 219, 192,88 Ind.Eng. Chem. (AnaE.), 1936, 8, 336.89 Ibid., 1937, 9, 251. O0 Analyst, 1933, 58, 673478 ANALYTICAL OHEMISTRY.The rock is dissolved in hydrofiuoric and sulphuric itcid8, andthe nickel determined directly in the resulting solution by precipit-ation with dimethylglyoxime or a-furildioxime, citric acid beingadded to prevent the precipitation of the R,03 oxides; 0-0025~0of nickel oxide can thus be detected when working on 2 g. of rock.Tho same authors have investigated the behaviour of nickel in rockanalysis. In the ordinary course of analysis some nickel is pre-cipitated with the R,O, oxides, and when the total nickel exceedsO-OEi% a correction must be made for this. The whole of the re-mainder passes into the filtrate from the magnesia, and may bedetermined in this after the removal of ammonium salts.Ncmganese.-The addition of various oxidising agents to thesolution during the precipitation of the R,03 oxides by ammoniahag been studied by E.V. Holt and H. F. MarwoodYg1 who find thatthe addition of bromine water simultaneously with the ammoniaensures complete co-precipitation of the manganese, provided thatnot more than 60 mg. of manganous oxide be present. Potassiumpersulphate, or hydrogen peroxide, is not satisfactory, as it is thenimpossible to obtain the whole of the calcium in the atrate, evenwhen two precipitations are made. 0. HackIy92 however, appearsto have overlooked the latter point, and precipitates manganesetogether with the R,O, oxides by the addition of ammonia andhydrogen peroxide.The precipitate is dissolved in cold 50%nitric acid, and reprecipitated. The manganese in the ignitedprecipitate is determined calorimetrically after fusion with potasaiumand sodium pyrosdphates ; titanium and iron may be subsequentlydetermined in the same solution. Unfortunately no numericaldata are adduced in support of the completeness of the separationfrom the alkaline earths.Burium and 8trontium.-The inaccuracy of the usual method forthe determination of strontiuin in rocks has been pointed out by€I. IF. Harwood93 and by W. NolI.94 The principal source of errorlies in the incompleteness of the precipitation of strontium as oxalatealong with the caIcium, especially when two precipitations are made ;according to Noll, an error of 5% on the strontium may result fromthis cause. Most of this strontium can be recovered by the additionof a strontium-free calcium salt to the filtrate from the second pre-cipitate of calcium oxalate, and determination of the strontiumin the precipitate so obtained.The nitric acid method of W.No11 95 is preferable to the ether-alcohol process in the separationO1 Min. Mag., 1927, 21, 318.sa 2. anal. Chem., 2936, 105, 81, 182; 1937,110, 401.Og aeol. Mag., 1033, 78, 142.95 2. anorg. Chem., 1931,199, 193.B4 Chem. Erde, 1933-1934, 8, 507HARWOOD : ANALYSIS OF ROCKS AND MINERALS. 479of small amounts of strontium from Euch calcium. For thegreatest accuracy, however, a spectroscopic method is preferred to apurely chemical one where strontium is concerned, C.J, vanNieuwenberg and R. H. Dewald,96 on the other hand, have testedfive different rocks of various types for their strontium content, bothby chemical and by spectroscopic methods. The two sets of valuesagreed well; hence they conclude that the apparent anomaly in thegeochemical frequency of barium and strontium (which occur inrocks in approximately equal amounts, whereas in most othergroups of elements the geochemical frequency falls with increasingatomic weight) is not due to analytical under-estimation of thestrontium present.The errors in the usual chemical methods for the determinationof barium in rocks are discussed by W. von EngelhardtY97 and thesomewhat disturbing conclusion is reached that the results areuntrustworthy for small amounts of barium (less than O.lyo).Aspectroscopic method is advocated.Borofi.-Boron is rarely determined in ordinary analyses of rocks,but V. M. Goldschmidt and C1. Petersgs find that, although thiselement is usually present in igneous rocks only to a very minuteextent (O~oO05-0~OOl yo), yet a large number of determinationscarried out on sedimentary rocks (clay-slates) showed that thesehad an average content of 0.1% of boric oxide. Whenever rocksof this class are being analysed, a determination of boric oxide isconsequently clearly desirable.Rare Earths.-A new method for the determination of thoria inmonazite sand, based on the separation of this oxide from the ceriumearths by means of hexamethylenetetramine, is given by A.M.Ismail and H. F. H a r ~ o o d . ~ ~ The existing method of W. F. Hille-brand for the determination of rare earths in silicate rocks is longand tedious; a shorter process is much to be desired, but has sofar not appeared.Copper, Lead, Zinc.-The colorimetric method for copper withsodium diethyldithiocarbamate has been applied by A. W. Grovesto the determination of copper in rocks ; 0.001 yo of copper can bedetermined in a 2-g. sample of rock. A procedure €or the de-termination of copper, lead, and zinc in silicate rocks, based onFischer’s use of dithizone for the detection of traces of the heavymetals, has been worked out by E. B. Sandell : O.OOOZ~o of leador copper and o.oo3~0 of zinc can thus be detected. The limit96 Rec.Trav. chim., 1935, 54, 633.g 8 Nach. Qes. Wiss. QGttingen, 1932, 402.O9 Analyst, 1937, 62, 185.Ind. Eng. Chem. (Anal.), 1937, 9, 464.97 Chem. Erde, 1935-1936, 10, 187.Min. Mag., 1935, $4, 35480 ANALYTICAL CHEMISTRY.of accuracy of the determination is about 20% when working on0.25 g. of rock powder.Beryllium.-The detection and determination of beryllium inrocks has been discussed by G. Reinii~ker,~ who states that anapproximate idea of the beryllium content of a rock may be obtainedby means of Fischer's quinalizarin reaction, the solution beingdiluted until the limit of sensitivity is reached; 3-5 pg. give apositive, and 2 pg. a negative, reaction. A direct colour comparisonis not possible with this process. One of the main difficulties in thedetermination of beryllium in rocks has hitherto lain in the separationof the small amount of beryllium from the titanium present.Thishas now been overcome by B. E. D i ~ o n , ~ who precipitates thetitanium with p-chloroaniline, the beryllium being subsequentlyprecipitated by ammonia.H. F. H.3. INORGANIC ANALYSIS.S'tandards for Volumetric Analysis.This subject has been dealt with in the last two Reports1 anddeserves to be mentioned once again. G. F. Smith and G. F. Croadhave been unable t o confirm the work of Waldbauer et a.L3 concerningthe stability of anhydrous sodium carbonate at temperatures up to450"; they have shown that at temperatures above 300" there isappreciable decomposition, and the resulting product is unsuitableas an acidimetric standard.This, of course, is in harmony with thegenerally accepted procedure, but the lack of agreement overpoints such as this seems to be typical of the whole subject ofstandardisation.The use of borax as an acidimetric standard was highly recom-mended in 1926: but chemists have been reluctant to change fromthe sodium carbonate method in spite of many objections that canbe raised against it. Now that the findings of H. Menzel5 thatborax can be kept indefinitely over a solution saturated with respectto sucrose and sodium chloride have been confirmed,6 it seems thatthe last objection to the use of borax has been removed. This,and the superiority over the carbonate in other respects, makethe change-over eminently desirable.3 2.anat. Chem., 1932, 88, 29. Analyst, 1929, 54, 268.Ann. Reports, 1935, 32, 452; 1936, 33, 433.I n d . Eng. Chem. (Anal.), 1937, 9, 141.See Ann. Reports, 1935,32, 453, ref. (15).I. M. Kolthoff, J . Amer. Chem. Soc., 1926, 48, 1453.2. anorg. Chem., 1935, 224, 10.F. H. Hurley, jun., Ind. Eng. Chem. (Anal.), 1937, 9, 237; cf. G. Kilde,Dansk Tidsskr. Farm., 1936, 10, 273THEOBALD : INORGANIC ANALYSIS. 481The advantages of using copper vessels for the storage of standardsolutions of acids and alkalis have been discussed as well as theold problem of the preservation of solutions of sodium thiosulphate. 8In the latter case it is now found that 0.1N-solutions containingno preservative are stable if prepared with " aged " distilled waternot more than six months old.The stability of 0-1N-potassiumthiocyanate, recently pointed out by I. M. Kolthoff and J. J.Lingane, lo has received confirmation ;ll the addition of amyl alcoholas a preservative is not recommended.In the standardisation of ceric sulphate with potassium iodideby the acetone method it has been found l2 that the titre varies withthe concentration of acid, but within the range ~~~-2-7"-sulphuricacid, the results are accurate to 0.1%.Ammonium oxalate precipitated from aqueous solution by ethylalcohol and dried at 85-90' has been recommended for the standard-isation of potassium permanganate,13 but I. M. Kolthoff, H. A.Laitinen, and J. J. Lingane l4 now prefer potassium iodide andarsenious oxide as primary standards, rather than sodium oxalate.In agreement with R.M. Fowler and H. A. Bright,15 they findthat McBride's procedure gives high normalities and the best resultswith sodium oxalate are afforded by Fowler and Bright's method,but even here there is a positive error referred to potassium iodide.Finally, H. A. Bright l6 has compared the values obtained in thisstandardisation against arsenious oxide, using potassium iodideor iodate as catalyst, with those given by the new oxalate meth0d.l'For 0.1N-solutions the normalities agreed to within 1 part in 3000,showing that arsenious oxide is suitable as a direct primary standard.The Nicrochemical Detection of Cations and Anions.General.-It is probably true to say that colour reactions for theinorganic ions have now established their worth in analytical chem-istry, and that they should occupy a permanent place in the tech-nical equipment of every analytical chemist.The present positionJ. Lindner, Mikrochem., Molisch Festschr., 1936, 301.See Ann. Reportu, 1935, 32, 454.* P. Horkheimer, Pharm. Ztg., 1936, 81, 1184; cf. E. Tschirch, ibid., 1937,lo See An%. Reports, 1936, 33, 434, ref. (12).l1 F. H. Campbell and G. R. Hook, J . Xoc. Chem. Ind. Victorh, 1936,88,1106.12 D. Lewis, J . Arner. Chem. Xoc., 1937, 59, 1401; cf. Ann. Reports, 1936,13 M. M. Kirilov, J . Appl. Chem. Russia, 1936, 9, 2065.11 J . Amer. Chem. SOC., 1937, 59, 429.1 5 J . Re$. Nut. Bur. Stand., 1935, 15, 493; see Ann. Reports, 1935, 32, 453.16 I n d . Eng. Chem.(Anal.), 1937, 9, 577.17 Fowler and Bright, Zoc. cit.82, 450.33, 447, ref. (36); and ref. (14) below.REP.-VOL. XXXIV482 ANALYTXCAL CHEMISTRY.requires consideration, however, and it is well that some attentionis being directed to the development of existing methods as distinctfrom the breaking of new ground. For inBtance, H. Fritz haspublished further work on methods involving electro-drop analysis,lbut perhaps the best example of development is afforded by thework of 33. L. Clarke and H. W. HermanceY2 who have obtained amuch increased sensitivity in many well-known reactions by amodification of technique. Anyone using these reactions on papersoon discovers that the sensitivity of a test is much affected by theway in which it is carried out, and these authors have emphasisedthis fact.They have aimed a t reducing to a minimum the areacovered by the products of reaction, so as to increase the sensitivity,and they achieve this by means of a capillary burette to control therate of spreading, and thin, close-textured papers impregnated withreagents of low solubility. A factor of considerable importanceis the selection of the precipitating reagent; e.g., a paper impreg-nated with zinc ferrocyanide instead of the more soluble potassiumsalt provides a much improved test for iron. Again, cadmiumxanthate is preferable to the potassium salt as an impregnatingreagent in teat papers for the paper then remains sensitive for months ;moreover, it is more selective and gives sensitive reactiom only withcopper and molybdenum.By such devices as these, a 5-100-foldincrease in sensitivity has been obtained for many familiar reactions,H. Yagoda3 uses another method in order to confine the area ofreaction. He embeds a ring of paraffin wax in the fibres of the paper,leaving a clear space of known size in which to carry out the test.An approximation to the amount of an ion present can thus be moreeasily made, and it is claimed that the metal content of pure copperand nickel salts, for example, can be determined with an accuracyBy removing the silver from glossy or semi-matt bromide paperwith sodium thiosulphate and then soaking the paper with theappropriate reagent, I. M. Korenman also obtains a much increasedsensitivity with various tests.The same medium of silver-freephotographic paper or gelatin has been used 5 in the microchemicalanalysis of very small particles of material. The grains are spreadon a glass plate, attacked by exposure to the vapour of nitric orhydrochloric acid, and then brought into contact with the impreg-nated paper or gelatin, whereupon each grain leaves a coloured spot,of 1-370.1 Mikrochem., 1935,19, 6; 1936, 21,47; 1937, aR,34,168; 23,61; Molischa Ind. Eng. Chem. (Anal.), 1937, 9, 292.6 M. Rey and M. Zeicher, Bull. SOC. chirn. BeEg., 1937, 48, 173.Festschr., 1936, 125 ; 2. anal. Chem., 1929,78, 418.Ibid., p. 79. 4Mikr~hRm., 1936, 21, 17THEOBPLLD : INORGANIC ANALYSIS. 483which, in favourable cases, gives an indication of the amountpresent.The method is reminiscent of that used for the detectionof phosphate in rocks,6 and has been employed by these authorsin the investigation of mineral particles which cannot be identifiedby examination under the microscope.An increased sensitivity is also reached in reactions carried out incapillary tubes by a method described by T. A. Thomson.7 Nosuperiority is claimed in general for the method over drop analysis,but it may have useful practical applications in microbiochemistry.As examples of some recent applications showing the steadyextensions of these reactions in many and varied fields, we maymention the identification of drugs, reactions for some 270 of whichhave been given,s the detection of traces of nickel carbonyl in oilsand gases by virtue of its direct reaetions with dimethylglyoximeand dithizoneYg the distribution of injected heavy metals in celltissues and in the cell contents of plants,lO or that of the chlorideion on the surface of wood or contaminated fabric by means of silverchromate in a gelatin sol a8 a reagent,ll and the detection of thesolubility corrosion of metals, especially zinc, copper, tin, brass, andgun-metal, by sea-wster.12Separation into Groups.-Several schemes that combine a separa-tion of cations into groups with the use of drop reactions have beenput forward from time to time,13 but they appear to have beenneglected. The tendency has been to make short cuts and to avoida separation, a course of action that has many dangers.Indeed,the more one’s experience of these tests in applied work widens themore one realises how real these dangers are and how unsoundthis course must be.It, is a healthy sign that attention is againbeing direeted to the necessity for a preliminary separation in manycases. Thus, a procedure recently given for the qualitative analysisof the commoner cations on a semi-micro scale separates them intotheir usual groups and then applies drop reactions directly for the6 See F. Feigl, “ Qualitative Analyse mit Hilfe von Tupfelreaktionen,”1935, p. 456.7 Mikrochem., 1937, 21, 209.8 C. A. Rojahn et al., Pharna. Zentr., 1937, 78, 81, 127, 146.9 B. Steiger, Mikrochem., 1937, 22, 216. See idem, Petroleum, 1937, 33,No. 27, Motoreenbetr., 10, 3, for the detection of tetraetlqd-lead and nickelcarbonyl.10 S.Pritt, Mikmchem., Molisch Festschr., 1936, 342; see also B. Broda,Wiado9la. fm., 1936, 63, 0, 15.11 F. G. Lennox, J . Proc. Austral. Chem. Inst., 1936, 3, 313.12 W. R. G. Atkins, Trans. P a r h y SOC., 1937, 33, 431.13 See F. Feigl, op. cit., p. 345; C . J. Engelder, T. H. Dunkelberger, andW. J. Schiller, “ Semi-micro Qualitative h l y s i s , ” J. Wiley and Sons, NewYork, 1936484 ANALYTICAL CHEMISTRY.detection of the ions concerned.la Much time can thus be saved,especially in Group 11. The most important work, however, inthis connexion is the continuation of a series of papers by A. A.Benedetti-Pichler and W. F. Spikes l5 mentioned in last year’sReport.16 Details of a complete scheme for the analysis of theammonium sulphide group, including gallium, indium, and therare earths, etc., on a milligram scale using less than 1 ml.ofsolution and following, in the main, the macro-separations ofNoyes and Bray, have been worked out. Unusual difficulties wereencountered in the analysis of the chromium group and in the detec-tion of tungsten, and the authors are of opinion that, at present, nochemical method for the detection of small amounts of this elementin the presence of relatively large amounts of chromium, uranium,vanadium, and phosphate ion can be recommended. Many differentschemes for the separation of these elements were investigated, andare summarised in a second paper; l7 both these papers are ofconsiderable interest. *This work has been done on a milligram scale, but in a later paperA.A. Benedetti-Pichler l8 has described a general working techniqueapplicable to the qualitative analysis of 1 pg. of solid material.The chemical separations are carried out in cones of 0.5 c.mm.capacity, and the transfer of solutions is effected by means ofmicrurgical pipettes operated by a hypodermic syringe. Most ofthe manipulations have to be done under observation with a low-powered microscope. Working procedures are described in theanalysis of 0-01 c.mm. of a solution containing 0.1 p.g. of antimonyand 0.01 pg. of bismuth. The author states that the separationsobtained have the same sharpness as in the analysis of larger amounts,and although the operations at present are rather time- consuming,they have the merit of demonstrating that even on this small scalethe familiar tests are still valid.For the micro-detection of siZver, theblue colour formed with o-tolidine is said to develop in the presenceof not leas than 3 x 10-8 g.of this metal.l9 To the numeroussensitive tests already available for copper must be added the darkred colour given by cupric salts with potassium bromide and sul-14 J. H. Winkley, L. K. Yanowski, and W. A. Hynes, Mikrochern., 1936,21, 102.Drop Reactions.-Cations.1 5 Mikrochem., Molisch Festschr., 1936, 3. l6 P. 453.1 7 A. A. Benedetti-Pichler and TY. F. Spikes, Mikrochem., Molisch Festschr.,18 I n d . Eng. Chem. (Awal.), 1937, 9, 483.1s L. M. Kulberg and S. B. Serebriani, J .Gen. Chem. Russia, 1936, 6, 1335.* See also A. A. Benedetti-Pichler and J. R. Rachele, I n d . Eng. Chem.1936, 36.(Anal.), 1937, 9, 589, for the selenium group of Noyes and BrayTHEOBALD : INORGANIC ANALYSIS. 485phuric acid,20 the yellowish -brown precipitate obtained with anilinethiocyanateY2l the pink to purple colour obtained with -urobilin,22and the catalytic acceleration of the atmospheric oxidation andrecoloration of phen~lphthalin.~~ The last test is carried out in thepresence of potassium cyanide and is reasonably selective : it isrecommended by the authors for the detection of copper in naturalwaters.Satisfactory drop reactions for cadmium are scarce, but the needfor a good reagent seems to have been met by F. P. Dwyer,24 whoutilises the red lake formed between cadmium hydroxide andp-nitrodiazoaminoazobenzene (" Cadion ").Used as a drop reactionon paper, the test is selective and highly sensitive ; only silver andmercury ions have previously to be removed, and other interferingelements are provided against by the addition of Rochelle salt.*The solution of the reagent showed no sign of deterioration after2 months' keeping.Osmium, as the tetroxide, gives sensitive colourations on filterpaper impregnated with an acetic acid solution of potassium ferro-cyanide or benzidine, and by boiling the solution and directing thevapour containing the tetroxide on to the reagent, traces of osmiumare said to be detectable in the presence of all other cations.25Less sensitive colours, blue or orange, respectively, are obtained whena solution of sodium osmate is treated with pyrogallol or withephedrine hydrochloride .26Rhenium is detected by utilising the fact that reduction oftellurate by stannous chloride is catalysed by ReO,'.A blackprecipitate is obtained with 2.5 x lo-* g. of rhenium in 0.05 ml.Vanadium, tungsten, arsenic, selenium, osmium, and molybdenuminterfere above certain limiting concentration^.^^The red colouration observed in blue light when hydroxynaphth-acenequinonesulphonic acid in sulphuric acid is added serves as asensitive reaction for germanium.28The reduction of the selenite ion by ammonium thiocyanate in the20 S. Augusti, Mikrochem., 1937, 22, 139.21 F. P. Dwyer and R. K. Murphy, J . Proc.Austral. Chern. Inst., 1937,4, 334.22 G. Bertrand and L. de Saint-Rat, Mikrochim. Actu, 1937, 1, 5.23 I. M. Kolthoff and J. J. Lingane, Mikrochem., Molisch Festschr., 1936,274.24 J . Proc. Austral. Chem. Inst., 1937, 4, 26; J . SOC. Chem?. Ind., 1937, 56,25 N. A. Tananaev and A. N. Romanjuk, 2. anal. Chem., 1937, 108, 30.26 S. 0. Thompson, F. E. Beamish, and M. Scott, Ind. Eng. Chem. (Anal.),37 N. S . Poluektov, J . Appl. Chem. Russia, 1936, 9, 2312.2 8 Idem, ibid., p. 2302.* The Reporter now prefers this test t o that with dinitrodiphenylcarbazide7 0 ~ .1937, 9, 421.and it has proved very satisfactory486 ANALYTICAL CHEMISTRY.presence of hydrochloric acid has been employed as a qualitativetest for selenium. Reduction to the red form is rapid and completein 6~-hydrochloric acid at the boilkg point.Interference is causedby ferrous iron, tervalent antimony, and stannous tin, but not byferric iron, lead, copper, or mercuric mercury, e t ~ . ~ ~ Por the sameelement F. Feigl and V. Demant 30 utilise the red to reddish-violetcolour resulting from the oxidation of as-diphenylhydrazine in aceticacid by the selenite ion. Other oxidising agents such as iodate,permanganate, and peroxides must first be destroyed by treatmentwith hydrochloric acid, and the effect of those such as ferric iron,copper, molybdate, and tungstate, must be prevented by conversioninto a complex oxalate. Tellurates and tellurites do not react.The test is applicable to elementary selenium, to selenides or toselenates after conversion into selenite.*2 : 3 : 7-Trihydroxy-9-methy1-6-fluorone has been suggested as areagent for antimony. At a p H of approx,4, ter- and quinque.valent antimony ions give bright red precipitates in a hydrochloricacid solution or in 10% nitric acid in the presence of tartaric acid.It is claimed that antimony can be detected with certainty in mineralscontaining arsenic, bismuth, et c.31A neutral or slightly acid solution containing as little as 4 x lO-7g.of moZybdenumt in 0.04 C.C. gives a reddish-violet colouration with2 : 2‘-dipyridyl in ethyl alcohol and stannous chloride.32 Rhenium,vanadium, and quinquevalent arsenic do not interfere, but in thepresence of tungsten, tartaric acid should be added. Iron must,of course, be absent.In an investigation of the serviceability ofother methods suggested for the detection of molybdenum, A. C.Rice and L. A. Yerkes 33 find that the thiocyanate-stannous chloridemethod is the most sensitive, with the potassium ethylxanthatemethod ranking next to it. In both ca,ses sensitivity is increasedby extraction with ether.29 H. A. Ljung, Ind. Eng. Chem. (Anal.), 1937, 9, 328.30 Mikrochim. Acta, 1937, 1, 322.31 R. Duckert, HeZv. Chim. Acta, 1937,20,362 ; P. Wenger, R. Duckert, and52 A. S. Komarovski and N. S . Poluektov, J. Appl. Chern. Russia, 1937,33 U. S , Bur. Mines, Rept. Invest. 3328, 1937, p. 37. * The danger of loss in selenium analyses owing to volatilisation on evapor-ation with hydrochloric acid solutions of an acid concentration greater than6~ appears to the Reporter to have been overlooked in the procedure recom-mended by the authors for the detection of selenium in minerals (for data,see Hillebrand and Lundell, “ Applied Inorganic Analysis,” 1929, p.259,and K. Briickner, 2. anal. Chem., 1933, 94, 305). t For a fluorescence reaction with cochineal see L. Szebelledy and J. J b d ~ ,Mikrochim. Acta, 1937, 1, 46.C1. P. Blmcpain, ibid., p. 1437.10, 565; Mikrochim. Acta, 1937, 1, 264THEOBALD INOBGANIC AHALYSIS. 487Traces of gold can be detected by means of the blue colour that isgiven with leuco-nitrobrilliant-gree11.3~ In the presence of colouredions, the blue colour is rendered visible by extraction with chloro-form in which it is soluble. The reaction has been applied tocolorimetric determinations.phsanilic acid forms a reddish-brown to pink colour with cericions in dilute acid solution, and may meet the need for a good testfor this element.Cerous ions, the rare earths, and many otherions (e.g., those of titanium, ferric iron, manganese, tungstate,molybdate) do not interfere. Chromium and cobalt may beinimical on account of their colour, and fluoride must be absent.Zirconium is also precipitated, and excess of reagent tends to pre-cipitate thorium .s5 The chocolate-coloured precipitate obtainedwith both ter- and quadri-valent cerium and morphine hydro-chloride in the presence of an excess of ammonia can also be usedas a drop reaction.362 : 4-Dihydroxyacetophenone is an addition to the qualitativereagents for iron.37 The sensitivity (0.002 mg.per ml. of solution)compares well with that of ammonium thiocyanate or potassiumferrocyanide, and, unlike 2 : 2'-dipyridyl, the reagent is neithercostly nor difficult; to prepare. An alcoholic solution produces a,red colour with ferric iron in a weakly acid solution and the inter-ferences from other ions are not marked. Phosphates (and, presum-ably, fluorides) must be absent.A test for aluminium is provided by the orange-red fluorescenceobtained with Pontachrome-blue-black-R in ultra-violet light.Highly-coloured ions should be removed by means of sodiumhydroxide, but beryllium does not interfere.3* The sensitivity ofthe reaction of aluminium with eriochromcyanin-R in the presenceof other ions has also been rec0rded.3~ In this case beryllium doesgive a similar reaction, but by a suitable procedure g.ofaluminium can be identified in the presence of 500 times as muchberyllium.The sensitivity of the potassium periodate test for manganese ismuch increased by the introduction of p-tetramethyldiamino-diphenylmethane,m which is catalytically oxidised to a blue productby traces of permanagnate formed from the manganese ion and theperiodate. The same effect has been attained by the substitution34 L. M . Kulberg, Zuvod. Lab., 1936, 5, 170.35 J. F. Miller, I n d . Eng. Chem. (Anal.), 1937, 9, 181.36 F. M. Schemjakin, Compt. rend. Acad, Sci. U.R.S.S., 1937,14, 115.37 S. R. Cooper, Ind. Eng. Chem. (Anal.), 1937, 9, 334.58 C.E. White and C. S. Lowe, ibid., p. 430.3Q E. Eegriwe, 8. anal. Chem., 1937, 108, 268; 1929, 76, 440.40 F. Feigl, op. cit., p. 228488 ANAL1 TICAL CHEMISTRY.of p-phenetidine for the tetramethyl base. A violet-red colourresults, and chloride, iron in the presence of sodium fluoride, cobalt,and chromium are said not to interfere. With chromate present,the sensitivity is reduced and more periodate is consumed.41The same base in the form of its hydrochloride furnishes, togetherwith potassium ferricyanide, a sensitive drop reaction forA large excess of the alkaline earths, magnesium, sodium, potassium,or ammonium has no effect on the blue colouration produced.New work on magnesium concerns the blue colour obtained withbenzoazurin-G-a test that is subject to interference by manyions, h0wever.4~Anions.-In the fluorescein test 44 for bromide in the presence oflarge amounts of chloride, aqueous chromic acid is preferred45to potassium permanganate and sulphuric acid 46 since it causesless general bleaching of the fluorescein and is easier to handle.Successive extraction of a solution with chloroform and hydrogenperoxide, which liberates only iodine, and with aqueous nitrousacid and chloroform permits the detection of 1 x g.of iodideand bromide, respectively.The xylenol method for nitrate 47 has been modified so as to detect0-001 mg. of NO,’ by the yellow to red colouration formed when5-nitro-m-xylenol is distilled into dilute sodium hydroxide. Per-oxides and nitrites must first be destroyed, and alkali sulphidesremoved by means of copper sulphate.I n a modification of the Griess-Ilosvay test for nitrites, substitu-tion of a-naphthylamine by dimethylaniline leads to formation ofmethyl-orange, which, of course, is turned red by the acid present.48The sensitivity and selectivity of the alizarin test for boric acid 49are both said to be improved by examination in ultra-violet light.50According to E.SchroerY5l sulphur in all its compounds is reducedto hydrogen sulphide by nascent hydrogen. Organic compoundsmay be converted into mercaptans, detectable in minute amountsby their odour. The hydrogen sulphide is best detected by theblue luminosity that it confers on the flame of burning hydrogen.4 1 L. Szebellhdy and M . Bhrtfay, Z .anal. Chem., 1936, 106, 408.42 L. SzebellBdy and S. Tanay, ibid., p. 342.43 E. Eegriwe, ibid., 1937, 108, 34.44 Cf. F. Feigl, op. cit., p. 277.4 5 R. G. Aickin, J . Proa. Austral. Chem. Inst., 1937, 4, 267.4 6136, 90.4 7 F. Werr, 2. anal. Chem., 1937, 109, 81.4 8 J. C. Giblin and G. Chapman, Analyst, 1936, 61, 686.49 F. Feigl and P. Krumholz, Mikrochem., Pregl Festschr., 1929, 79.60 L. Szebellhdy and S. Tanay, 2. anal. Chem., 1936,107, 26.51 Mikrochem., 1937, 22, 338.Cf. R. Lorenz, E. Grau, and E. Bergheimer, 2. anorg. Chem., 1924THCEOBALD : INORGANIC ANALYSIS. 489Selenium and tellurium appear to be the only elements likely tobe seriously detrimental to the test.The Use of the Term “ Speci$c.”-In the above-mentioned dropreactions, the limiting amount of an element or ion that can bedetected is generally less than g.per drop. The actualsensitivities found are those for a solution containing only the ionin question, but it must always be remembered that this is themost favourable case, and that in practice the presence of otherions usually necessitates a modification of procedure which, moreoften than not, involves a loss in sensitivity. Almost withoutexception each test is subject to interference from the presence ofother ions, and it cannot be too strongly emphasised that the possi-bility of these interferences occurring must be taken into considerationbefore a test is applied. The ideal of a reaction that is “ specific ”in the true sense of the word is far from being realised in many ofthese reactions; indeed, it is hardly to be expected that it should.This point has been emphasised in the two preceding Reports,52and as a result of the last note the following announcement has beenmade, by the Committee appointed by the International Unionof Chemistry for the study of new analyticaJ reagents, concerningthe use of the terms “ specific ” and “ selective ” : “ The Committee.. . has decided to differentiate between specific and selectivereactions (and reagents) and recommends this convention for generaluse. Reactions (and reagents), which under the experimentalconditions employed are indicative of one substance (or ion) only,are designated as speci$c, whilst those reactions (and reagents)which are characteristic of a comparatively small number of sub-stances are classified as selective. From this it follows that it ispermissible to describe reactions (or reagents) as having varyingdegrees of selectivity; on the other hand a reaction (or reagent)can be only specific or not specific.” 53It is to be hoped that this recommendation will be followed, foronly by some such means can a correct statement of the facts,which is at present lacking, be attained.It would be deplorableif, through laxity of expression or a misunderstanding of terms,some of these reactions should undeservedly fall into disrepute,and as a result their utility in chemical analysis should not beexploited to the full.The Behrens Tests.-As distinct from colour reactions, thesetests, involving the recognition of crystalline precipitates under themicroscope, still find favour in certain quarters, especially amongmineralogists in the United States.These, as well as certain62 Ann. Reports, 1936, 82, 471; 1936, 33, 453.63 See Analyst, 1937,62, 568; Chem. and Ind., 1937,56, 636490 ANALYTICAL CHEMISTRY.chemists whose work is connected with mineralogy, will welcometwo recent papers concerning these tests. In the first,64 the inter-ferences in size, shape, or colour of the crystals produced in certainstandard tests of the U.S.. Geological Survey’s scheme 55 for thedetection of elements in ore minerals are hlly described, and arediscussed with the object of avoiding the pitfalls attendant onthe use of such methods.These pitfalls are numerous and mayoften lead to error for, as the authors point out, in skilled andexperienced hands changes in colour or form, or possibly a negativetest, occur and make interpretation difficult. This strengthens anopinion expressed in an earlier Report,56 and for the general run ofminerals, at least, it is preferable in most cases to use colour reactionsinstead of the Behrens tests.*With the precious metals the case is different. The colour re-actions t for these need further development and the Behrenstests are still of great vdue, all the more so as these metals,with the exception of palladium, form an easily-segregated groupby virtue of their insolubility in dilute hydrochloric or nitricacid.The second of the papers 57 mentioned above deals first with theindividual reactions of platinum, palladium, rhodium, ruthenium,iridium, osmium, and gold with ammonium dichromate, benzidine,thiourea, pyridine hydrobromide, etc., and secondly with theinterferences that may be produced by one of these metals in thepresence of another.The most suitable tests for the identificationof these metals in their minerals and alloys are indicated.The characteristic habits of the silver compound with hexa-methylenetetramine, of copper sulphate and potassium bromide withantipyrine, and of barium hydroxide with caffeic acid are amongthe tests described 58 for numerous substances both organic andinorganic.I. M. Korenman 59 employs the Behrens tests for the detection ofcopper, tin, lead, antimony, bismuth, silver, and nickel in alloys,64 H. J. Praser and R. M. Drayer, Amer. Min., 1937, 22, 949.6 6 Bulletin No. 825.66 Ann. Reports. 1935, 32, 471.67 H. J. Fraser, Amer. Min., 1937, 22, 1016; cf. aleo Ann. Reports, 1935,68 L. Rosenthaler, Mikrochem., 1937, 21, 215.68 Zavod. Lab., 1937, 6, 308. * The Reporter’s experience is that these tests are not so suitable, even forpost-graduate students, as the colour reactions, for which previous experienceis not essential.t The utility and efficiency of some of these tests are discussed by S. 0.Thompson, F. E. Beamish, and M. Scott, Ind. Eng. Chern. (Anal.), 1937,9, 421.32, 474, ref. (66)THEOBALD INORGANIC ANALYSIS. 491and with S. S. Messonshnik 6o records data for the limiting amountsof lead detectable with the well-known triple nitrite test and withother reagents.Other investigations deal with tetraethylammonium iodide as ilreagent for quinquevalent antimony (purple, hexagonal plates) andbismuth (dark amber, triangular plates) ,61 the identification oftellurium by the formation of yellow crystals of the bromide orblack crystals of the tetraiodide, 62 and the crystalline Precipitatesof characteristic habit formed with sparteine, ammonium thio-cyanate, and cobalt or ferric ~ a l t s . ~ 3W. Geilmann and W. Wrigge 64 describe the crystalline doublesalts, suitable for identification microscopically, given by rhenium(as ReCl, or H,ReCl,) with reagents such as potassium, rubidium,or casium chloride, and the hydrochlorides of certain organic bases.After a discussion of many confirmatory tests for beryllium which,with the exception of the quinalizarin test, they consider to beunsatisfactory, A. A. Benedetti-Pichler and W. F. Spikes 65 preferto convert beryllium into its acetate which, on sublimation, yieldseasily-recognisable, well-formed crystals of the basic acetate. Asa good confirmatory test for gallium they recommend the formationof an alum with caesium chloride containing a trace of potashalum.In order to differentiate between chromate and dichromate ions,advantage is taken of the fact that with [Co(NH,),]Cl, these ionsyield [Go( NH,) ,]Cr 04c1 and [ co( NH,),] ( cr,o,)s, respectively, ofvery different crystalline form.66 The same reagent providesrapid tests for dithionate, ferrocyanide, ferricyanide, iodate, sulpho-salicylate, etc., and gives characteristic precipitates also withphosphomolybdate, phosphotungstate, tellurite, cobaltinitrite,nitroprusside , and other ions more common thanBook-A new edition, the ninth in English, of the qualitativesection, Vol. I, of Treadwell and Hall's text-book has recentlybeen published. In bringing this well-known work up-to-date,many changes and improvements have been made.W. R. Schoeller has published his monograph on tantalum andniobium under the title " The Analytical Chemistry of Tantalum and60 Zavod. Lab., 1936, 5, 168.61 F. T. Jones and C. W. Mason, I n d . Eng. Chem. (Anal.), 1936, 8, 428.62 G. DenigBs, Compt. rend., 1937, 204, 1256.68 A. Martini, Mikrochim. Acta, 1937, 1, 164.64 2. anorg. Chem., 1937, 231, 66.6 5 Mikrochem., 1937, 21, 268; c f . idem, ibid., Molisch Festschr., 1936, 23.6 6 M. G. Malko, L. K. Yanowski, and W. A. Hynes, Mikrochem., 1936,67 W. A. Hynes and L. K. Yanowski, Mikrochem., 1937, 23, 1.21, 57492 ANALYTICAL CHEMISTRY,Niobium. The Analysis of their Minerals and the Application ofTannin in Gravimetric Analysis ” (Chapman and Hall Ltd., 1937).It is a work that will be needed whenever an anatlysis of theseminerals has to be undertaken.L. S. T.H. F. HARWOOD.L. S. THEOBALD.F. TWYMAN