MITCHELL APPLICATION OF SPECTROGRAPHIC ANALYSIS TO SOIL INVESTIGATIONS 361 Applications of Spectrographic Analysis to Soil Investigations BY R. L. MITCHELL IN the work of the Macaulay Institute we are interested in the changes occurring during the processes of soil formation and in the relationship of the'soil to the yield composition, and health of the plants and animals associated with it as well as in the analysis of the soil itself. The recent realisation of the importance of certain trace constituents to the health both of plants and animals has led to a search for methods for their determination and arc methods of spectrographic analysis have proved suitable. The quantities involved often of the order of parts per hundredmillion are generally smaller than those considered important in metallurgical analysis although the demand for accuracy is possibly not quite so great, and at present we are satisfied with errors of the order of f 1 0 per cent.Flame spectro-graphic methods applicable to the determination of the alkalis and alkaline earths in solution 362 MITCHELL APPLICATIOK OF SPECTROGRAPHIC ANALYSIS TO SOIL IXVESTIGATIOX s are being employed particularly for the routine determination of potassium in extracts of soils and plants. A soil consists essentially of an inert framework of silicate minerals together with an active portion of weathered colloidally active clay minerals and some organic matter. The colloidally active fraction carries those ions which can be termed “available” to the plant, whilst the reserves of most constituents are bound up in the crystal lattices of the silicate minerals and are only slowly released by weathering.A knowledge of the ultimate chemical composition is of little practical use as far as the major constituents are concerned and determinations of these by spectrographic methods have not been attempted at the Macaulay Institute. TABLE I TRACE ELEMENT CONTENTS OF THREE TYPICAL SOILS FROM K.E. SCOTLAND DETERMIXED BY THE SEMI-QUANTITATIVE CATHODE LAYER AKC SPECTROGRAPHIC I< b Li Ba Sr Cr sc A h co Ni Zr Zn La Y c u v 310 Be Ga Sn Pb T1 Ge Ag METHOD AS PARTS PER MILLION 5011 deriyed from (--L--- 7 A Sensitivity Granite Xorite Sandstone IYavelength Old Red 600 30 “0 I 20 .50 i800.2 20 6707.8 1 4934.1 5 2000 ]I 000 700 4607-3 10 300 500 300 4254.3 1 5 50 1’00 4246-8 10 - 2 0 I0 4030.8 10 ion 2000 .300 3453.5 2 - 30 18 3414.8 2 10 1 5 60 3392.0 10 3CNO ? 1 3 1 :joo 3345.0 300 30 3337.5 30 30 30 3327.9 30 30 3280.7 1 -3274.0 < 10 10 40 3185-4 5 20 300 I00 3170.3 1 -3131-1 10 2943.6 1 25 “0 20 2 840.0 6 5 2833.1 ~ 10 2 0 2767-9 50 2651.2 10 -- _- __ .__ --- -- _- -_ - -- __ - - _-- __ -A dash indicates that the content is less than the sensitivity quoted. A determination of the total content in a soil probably has more significance for the trace constituents than it has for the major constituents since the amounts of the trace constituents may vary upwards of one thousand fold from soil to soil whilst the major constituents seldom vary more than five to ten fold.Because of these large variations and because the relation-ship between total content and plant availability is not exact; a semi-quantitative deter-mination giving an accuracy of &30% under favourable conditions is employed for such determinations. This is the original cathode layer arc technique as described by Mannkopff and Peters,l P r e u s ~ ~ ~ Strock,* Tongeren5 and Mitchell.6 A mixture of the finely ground material with carbon powder is filled into a deep narrow boring (8 mm. x 0.8 mm.) in a thin (2.8 mm. diam.) carbon electrode and burnt as the cathode in a 9 amp. direct current arc. The arc is imaged on the slit in order to take full advantage of the cathode layer effect there being marked increase in sensitivity in the immediate neighbourhood of the cathode for elements with low ionisation potentials.The spectrograms are compared in a spectrum comparator with those for standard mixtures similarly treated and the approximate content estimated. It is possible by this means to get an indication of any trace constituents present in deficient or excessive amount. This method has been used for the determination of the trace constituents in different layers of soil profiles,-it has shown for instance that cobalt and nickel do not follow iron during podzolization,-and it is also being used in the investigation of rocks and their individual minerals. This follows the lines of Goldschmidt’s work for which the method was developed, and supports his ideas on the distribution of.trace elements in rocks and their constituen MITCHELL APPLICATIONS OF SPECTROGRAPHIC ANALYSIS TO SOIL INVESTIGATIONS 363 minerals. The important factors arc the ionic radius and valency of the trace constituent and the possibility of its taking the place in the crystal lattice of one of the constituent ions. Thus Co" (0-82A) and Ni" (0.78~) replace Mg" (0.78~) or Fe" (0.83~) in ferromagnesian minerals but are absent from the felspars in the same rock as there is no suitable lattice substitution. As an illustration of the results obtainable by this semi-quantitative cathode layer arc method in Table I are given the wavelengths of suitable lines the sensitivities and the results for three typical soils from North-east Scotland. The higher values for cobalt nickel, chromium and vanadium in the soil derived from the more basic rock and for rubidium, zirconium lead and barium in that from the acidic rock are explained by arguments similar to the above.Analyses of soils along the same general lines have been reported by Allison and Gaddum Ballard * and Guelbenzu Ruiz and Azcona.9 For the analysis of plant ash a similar technique even if it were accurate enough would be possible only for certain elements owing to the dilution effect of the high alkali alkaline earth and phosphate contents and generally a chemical pre-treatment to remove these and to effect a concentration is necessary. It has been found possible to adopt the same concen-tration procedure for trace constituents extracted from plant ash after sodium carbonate fusion as for those removed from soils by the less energetic extractants (such as acetic acid or ammonium acetate) whose action simulates to some extent that of the plant.A concentra-tion method has the advantage that it allows quite different types of material to be submitted to the same spectrographic treatment as the composition of the major constituents of the resultant concentrate can be standardised. ScottlO has pointed out the effect that the com-position of a material may have on the intensity of the lines of a trace constituent or even on the ratio of the trace constituent to internal standard intensities. Thus chromium vanadium and molybdenum in calcium carbonate give lines of the same strength as those given by 5 to 10 times as much in a silica base and this intensification in presence of calcium may have marked effect on the direct determination of these elements in agricultural materials unless a standard of precisely the same composition is used for comparison.Plant ashes in particular are very variable in composition depending on the plant species and the part analysed. When a chemical concentration method is adopted it is possible to bring the elements recovered into a base of standard composition. Alumina has proved to be the most suitable base for the analysis of trace constituents concentrated from plant materials and from soil extracts. It is easily manipulated for filling into electrodes and its behaviour in the arc is satisfactory. We have found concentration to be carried out most conveniently by pre-cipitation with organic reagents 8-hydroxyquinoline precipitates cobalt nickel molybdenum titanium zinc and copper together with iron and aluminium in ammonium acetate solution, at pH 5-1.The amounts of aluminium and iron in the solution can be adjusted with previous experience of similar samples to give a precipitate weighing between 30 and 50mg. after ignition at 450" C. and containing a suitable amount of iron for use as spectrographic internal standard. Generally iron has to be added to soil extracts and aluminium to plant materials. This method described by Scott and Mitchel1,ll has recently been developed further12 by the use of mixed precipitants to bring down other elements and a simultaneous precipitation by 8-hydroxyquinoline tannic acid and thionalide recovers in addition to the elements men-tioned chromium vanadium tin lead beryllium and germanium and probably also silver, gallium thallium and other elements although these have not yet been studied fully.Cadmium is recovered and is added to the solution in known amount to serve as a second spectrographic internal standard. For the quantitative analysis of these precipitates a modified cathode layer arc technique as described by Davidson and Mitchell13 is employed. The previously described electrodes are used and into each about 4mg. of precipitate mixed with an equal weight of carbon powder is filled. The use of such electrodes with a carbon powder mixture results in a very steady arc which strikes to the sample and the inner rim of the electrode.The gain in stability of the arc more than outweighs the slightly increased work in filling the narrow boring. Admixture with carbon helps to prevent spluttering and for certain types of material the ratio may even be increased with advantage. It should be noted that the electrodes are carbon and not graphite as the latter owing to its greater thermal conductivity burns away rather slowly. The carbon electrode is burnt to the full depth of the boring in about 3 minutes at 9 amperes and the exposure is continued for the full period. A lens at the slit of the Hilger Large Quartz Spectrograph produces an image of the arc at the collimator where 364 MITCHELL APPLICATIONS OF SPECTROGRAPHIC ANALYSIS TO SOIL INVESTIGATIO& S mask isolates light from the cathode tip and the adjoining one third of the arc column.By this means the slit would be evenly illuminated from end to end were not a rotating step sector situated in front of the slit to divide the spectrogram into strips which receive ex-posures increasing in steps of 2. This enables all measurements of intensity to be made at the same photographic density. In the simplest means of photometry what in fact is measured is the exposure time relative to that for the internal standard required to give a certain density. In order to obtain working curves from which contents can be determined series of standard mixtures containing several trace constituents in an alumina base with a fixed iron content are prepared by grinding the necessary oxides in an agate mortar contents from 1 per cent.down to 1 part per million being obtained by dilution with the base in steps of 410. Separate dilution series are prepared with different contents of iron the internal standard a range of Fe,O from 2 to 40 per cent. being covered. The method of photometry involves for each line the measurement of 3 steps in the spectrogram adjacent to the density chosen for measurement. This is generally a density of 0.4 which with a clear plate setting of 4 = 60 as is usual on a Hilger Microphotometer, corresponds to a galvanometer deflexion of i = 20. Thus three steps giving deflexions between 5 and 30 are generally measured. From each value of i is obtained the optical density log (&/i). A Table giving this for values of I between 0 and 50 has been prepared and reprints are available.The same procedure is followed for the internal standard line, and where background can be disregarded the values of log (@) for both lines are plotted against the logarithm of the exposure given by the corresponding step of the step sector as in Fig. 1 ~ . Figs. 1 and 2 illustrate generally the types of curvcs obtained. Fig. 1 ~ . Blackening curves for Fig. 1 ~ . Working curves not Fig. lc. Correction curve for lines of trace element (T) and corrected for background for displacement of working curve internal standard (S) ignoring different Fe,O contents. due to change in l+.e,O, background. content. The separation of the two curves at any density is a measure of the ratio of the exposures which the lines have to be given when producing that density.This separation for a series of standard mixtures plotted against the logarithm of the trace element content gives the working curve (Fig. 113). The working curve is displaced vertically along the separation axis by changes in the content of the internal standard and this is the basis of the variable internal standard method of Davidson and Mitchell.l3 From the graph of the displacement against change in iron content (Fig. lc) a correction can be obtained for any iron content, enabling the standard working curve at a convenient iron content to be utilised for all deter-minations. The working curves are approximately straight lines except at low trace element contents where the effect of background introduces a pronounced toe and for accurate deter-minations in this region a correction for background is necessary.The accuracy is rather better over the whole range when this correction is made. The photometric procedure is then to measure in addition to three steps of each line, two or three steps of the background again covering a density of 0.4 (Fig. 2A). The logarithms of the relative intensities of the background analysis line + background and internal standard line + background are given by the points at which the blackening curves cut the selected density. From these values the relative intensities of the two lines freed from the effect of background can be obtained. This is a somewhat tedious calculation if ordinary logarithms are used but can be shortened by using gaussian or subtraction logarithms and i MITCHELL APPLICATIOXS OF SPECTROGRAPHIC ANALYSIS TO SOIL INVESTIGATIONS 365 greatly simplified14 by means of a Table derived from subtraction logarithrhs which we have prepared and of which reprints are a~ai1able.l~ This method makes the assumption that in practice the Eberhard effect can be disregarded and the results which we have obtained would appear to justify this.I . . . . . / Lcj I or !q Relative Exposur-c Fig. %A. Blackening curves for background correction method, for lines of trace element (T + B) and internal standard (S -!- R) andthebackground(B). Log Concentration Fig. 213. Working curves cor-rected for background for different Fe,O contents. 1;ig. 2c. Correction curve for displacement of working curve clue to change in Fe,O,content, after background correction, on logarithmic scale.When the logarithm of the ratio of the relative intensities after background correction, is plotted against the logarithm of the concentration (Fig. 2 ~ ) the working curve for most trace constituents is a straight line at the theoretical angle of 45" and the toe disappears. At very high contents there is some falling off in slope owing to self reversal effects as the lines which have to be employed in order to obtain adequate sensitivity are just those which, ending on the ground state are liable to self reversal. The correction curve for changes in iron content (Fig. 2c) also approaches its theoretical slope when the displacement is plotted against log iron content. This allows a reasonably accurate determination of any element for which working curves are not available and where their complete experimental deter-mination would not be justified for the purpose in view.Empirical working and correction curves can be drawn with slopes of unity through values obtained for one standard mixture close in composition to the unknown and from these results of sufficient accuracy are often obtainable as Scottls has shown. He obtained angles of slope of 45" rt 0.5" for chromium, cobalt nickel vanadium molybdenum beryllium gallium tin and lead in alumina and sodium chloride bases and similar values for the corresponding iron internal standard lines. The curves before correction for background had angles of 35-42". Marked changes in the composition of the base generally affect the relative intensities of the analysis and internal standard lines.Thus whilst the Co Fe ratio is practically independent of the base materia1,lO Cr Fe or Mo Fe ratio is very sensitive to changes as already mentioned but such effects can be eliminated by a concentration method. Another method for the control of this effect is the addition of a spectroscopic buffer in large amounts to each sample examined but generally the dilution effect of this method would reduce the sensitivity too far in our type of sample. It is being employed for copper in plant material where contamination effects make the concentration method difficult and adequate sensitivity is available. The recovery of some elements from solution by the concentration method involving precipitation with 8-hydroxyquinoline tannic acid and thionalide followed by spectrographic analysis of the precipitate is illustrated in Table 11.The errors of the whole process Will be seen seldom to exceed &lo% over the range of contents shown. A content of 50 p.p.m. in the precipitate analysed corresponds to 2 micrograms in the solution or with our normal aliquots one part in ten million in the plant material or soil. By the concentration process a 600- to 1000-fold concentration can be effected. The amount in the electrode is about one-tenth of the precipitate or 0-2 pg. upwards. For cobalt amounts as low as 0.04 pg. can be determined with error"s of the 10% order. The essential requirement of any such concentration method is the simultaneous quan-titative recovery of as many of the trace constituents as possible with the removal of the major diluents chiefly the alkalis alkaline earths and phosphate.There would appear t o be no other practical method of achieving this at present available. Extraction method 366 MITCHELL APPLICATIOXS OF SPECTROGRAPHIC ANALYSIS TO SOIL INVESTIGATIONS such as those with dithizone tend to be rather selective,l7J8 whilst adsorption methods such as those with synthetic resins as used by Eastmond,l9 separate cations and anions but the separation of different cations does not appear practicable. TABLE I1 DETERMINATION OF VARIOUS ELEMENTS BY CONCENTKATIOX FROM SOLUTION WITH 8-HYDROXYQUINOLINE -/- TANNIC ACID + THIOXALIIIE FOLLOWED BY SPECTROGRAPHIC ANALYSIS Present parts per million . . 50 115 250 SO0 12.70 Cobalt .. . . . . 45 126 24!1 471 12!)0 Nickel . . . . . . 52 135 273 520 1375 MoIybclenuni . . . . 57 124 341 503 1120 Chromiuiii . . . . 50 126 233 432 1220 Vanadium . . . . 53 1.37 241 497 1175 Found 9 p p 7 1 9 : (2 in Present parts per million . . I00 350 500 1000 2300 5UOi) 10006) (1 O 1 Found 9 9 9 ? 9 9 : Berylliuiii . . . . 117 263 490 1023 2660 Gernianiuni . . . . IS2 269 525 1014 2850 Tin . . . . . . 295 537 1088 2340 Lead . . . . * . 344 956 2611 Titanium . . . . 110 342 1100 5433 10700 (1.07‘%,) Present Zinc per cent. . . 0.1 0.3 1.0 5.0 10.0 Found 9 ) 9 ) 9 ) . . 0.56 1.02 5.07 10.16 Some typical results showing the amounts of trace constituents extracted by dilute -5%) acetic acid from the soils for which total contents are given in Table I are to be found Table 111.It will be seen that whilst certain of the elements for instance cobalt show TABLE I11 TRACE CONSTITUENTS EXTRACTED BY DILUTE ACETIC ACID FROM SOILS OF DIFFERENT GEOLOGICAL ORIGINS AS PARTS PER MILLION co Ni Mo V Cr Ti Sn Pb Zn OIdRedSandstone 0.41 1-04 0.02 0.13 0.17 0.21 0.4 0.4 4 Xorite . . . . 1.49 0.56 0.02 0.19 0.11 0.38 0.3 0.3 16 Granite . . . . 0.12 0.42 0.07 0.12 0.14 0.47 6.8 1.6 22 quite large variation from soil to soil others are relatively constant despite appreciable variation in total content, TABLE IV TRACE CONSTITUENTS IN A PASTURE HERBAGE AS PARTS PER MILLION OF DRY MATTEN Sample Co Ni Mo V Cr Ti Sn Pb Zn Cu 1;c Mixed . . 0.19 1.4 0.44 0.09 0.13 1.7 0.4 1.1 43 18.2 83 Red clover .. 0.21 1-6 0.35 0.10 04XI 8.1 0.2 1.6 41 18-6 85 Rye grass . . 0.09 0.1 0.72 0.10 0.21 2.7 1-1 34 6.1 41 Cocksfoot . . 0.05 0.1 0.78 0.18 0.12 2.0 - 1-5 22 11.7 26 Blank . . <0.01 0-05 0.01 0.01 0.04 0.1 0.1 0.3 <1 The results for plant material in Table IV indicate the contents of a mixed pasture herbage and of its chief constituent species. One practical point which these results bring out is the necessity for careful sampling of mixed pastures. We have been particularly interested in the cobalt content of soil extracts and plant materials because of a cobalt deficiency disease in sheep in various parts of Scotland. This occurs when the diet is low in cobalt the limiting value in the herbage being of the order of 5 to 10 parts per hundred million of the oven-dry material corresponding to a content in the soil of about 26 parts of cobalt per hundred million extractable by 2.5% acetic acid.This disease when it occurs on normal arable soils can be prevented by qddition of 2 lbs. of cobalt chloride per acre (about 0.25 p.p.rn. of Co) and in Table V are given some data for herbage contents showing the effects of adding cobalt-rich fertilizer and lime to the soil. The in-fluence of lime affecting principally the soil acidity on the uptake of trace constituents by the plant is well shown. The uptake of cobalt and nickel is decreased whilst that of molyb-denum is increased. Where cobalt deficiency is serious deaths among lambs are common MlTCHELL API'LIC'ATIONS OF SPECTROGRAPHIC ANALYSIS TO SO1L INVESTIGATIONS 367 whilst cobalt manuring may produce live-weight increases compared with those of lambs which survive without cobalt manuring of some 20 to 30 lbs.at the age of six months.20 Other investigations are covering animal diseases involving copper and molybdenum, And the effects of other trace constituents on plants and animals are being studied. TABLE V EFFECT OF COBALT ADDITIOK (2lbs. of CoC1,.6H20 per acre) AXD LIMING (6 tons of ground limestone per acre) ON UPTAKE BY RED CLOVER (as parts per million of dry matter) So cobalt Cobalt f \ -v co Xi ill0 Fe C O XI 310 Fe Unlimed 0.22 2.0 0.3 Ci 3 0.81) 1.6 0-3 .i 7 Limed . . 0.18 1.4 1 . 3 75 0.53 1.0 1.0 68 A LundegGrdh $ame method-The use of this method21922 represents an entirely different application of spectrographic analysis to soil investigations.It is employed as a routine method for the determination of the alkali and alkaline earth metals magnesium and man-ganese in solution. The source is an air-acetylene flame the gas pressures of which are accurately controlled. The solution to be analysed is introduced into the flame as a fine spray in the air supply. In view of the ease of preparation of standard solutions it is usual to carry these together with a series of solutions to be analysed on each plate. Spectro-grams of 16 unknowns and six standards in duplicate can be prepared in a little over one hour. The low energy of the flame source gives a simple spectrogram iron in moderately high concentration showing only a few lines so that large dispersion is not necessary but the ultimate lines of the alkali and alkaline earth metals and a few other elements are very sensitive.A simplified method of photometry taking merely the ratio of the galvanometer de-flexions for the line and the flame background on which it is superimposed is generally employed. This ratio in fact gives the transmission of the line itself (free from background), and is plotted against the concentration in the solution. In extracts of soils and plant materials determinations of potassium sodium calcium magnesium manganese and strontium can generally be made with spectrographic errors not exceeding &5% when duplicate spectrograms are taken whilst greater accuracy is obtainable by increased re-plication or by the use of an internal standard. The use of the method for soil work has been principally the determination of potassium in acetic acid extracts in connection with the soil advisory service for farmers and for this purpose it would appear that direct photometric determinations such as are now being widely employed in the United States with electron photomultiplier tubes could well be used in conjunction with the Lundeg5rdh flame.Suitable equipment does not however appear to be available in this country as yet. Determinations of exchangeable cations extracted from soils by ammonium acetate, and of the alkalis in plant ash (after hydrochloric acid extraction) and in rocks and minerals (after the initial stage of the Lawrence Smith extraction) are other applications of the method. As an indication of the sensitivity Table VI gives suitable ranges of contents in milligrams TABLE VI SUITABLE RANGES OF CONTENTS (as mg.per 50 ml.) FOR DETERMINATION BY THE LUNDEGARDH FLAME METHOD Element K Na Li Ca iug Sr hln Fe A A 4044 3302 6708 4227 2852 4607 403 1 3860 Range 0.5 -10.0 0.5 -10.0 0.005- 0.1 0.025- 0.5 0.3 - 6.0 0.025- 0.5 0.03 - 0.6 1.0 -20.0 per 50 ml. This volume is a convenient unit to work with; determinationican be made QII much smaller volumes but on a routine scale use of smaller volumes leads to filtration and washing difficulties. Interference effects are found remarkably seldom; in fact the only serious effect reported is that of aluminium on calcium and strontium,23 where marked depression occurs. Thi 368 SMITH SPECTROGRAPHIC ANALYSIS OF RARE AND HIGH PURITY MATERIALS makes precautions necessary when dealing with soil extracts made with strong acids.The cause of the depression is not clear but it is possible to overcome it for one of the elements affected by addition of excess of the other. Some elements notably potassium are slightly depressed in the presence of hydrochloric acid stronger than tenth normal but this effect is not observed if the acidity conditions in the solutions are standardised. Generally inter-ference troubles can be overcome by preparing standards containing the same extraneous substances. This short description can only indicate the lines along which we are working. Generally, however it can be said that our applications of spectrographic methods are for the accurate quantitative determination of those constituents in which we are interested.Our use of the method for purely qualitative purposes has been very limited and it is in its quantitative applications that bve have found its value. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. If). 20. 21. 22. 23. REFERENCES Blannkopff R. and Peters C. 2. Physik 1931 70 444. Preuss E. Chern. Erde 1935 9 365. - 2. angew. Min. 1938 1 167. Strock L. W. “Spectrum Analyszs with the Carbon Arc Cathode Layer,” 1936; London Hilger Tongeren W. van. “Contributions to the Knowledge of the Chemical Composition of the Earth’s Crust I . The Spectrographic Determination of the Eleinents according an the East Indian A rclzipelago. to Arc Methods in the Range 3600-5000~,” 1938; Amsterdam D. B. Centen’s. Mitchell R. L. J . Soc. Chew,. Ind. 1940 59 210. Alltson R. V. and Gaddum L. W. Proc. Soil Sci. SOC. Florida 1940 2 68. Ballard S. S. J . appl. Phys. 1940 11 760. Guelbenzu M. 9 Ruiz A. S. and Azcona J. I!!. L. An I m t . Edafol. Ecol. Fisiol. Veg. 1944,3 3!)1, Scott R. O. J . SOC. Chewz. Iizd. 1945 64 189. Scott R. O. and Mitchell R. L. Id. 1943 62 4. Mitchef’l K. L. and Scott R. O. in preparation. Davidson -4. M. &I. and Mitchell R. L. J . Soc. Chenz. Ind. 1940 59 213. Mitchell R. L. Scott R. O. and Farmer V. C. Nature 1946 157 193. - __ _- Macaulay Inst. Ann. Rep. 1943-44. Scott R. O. J . S O ~ . Chem. Ind. in press. Rohner F. Helv. Chinz. Acta 1938 21 23. Cholak J. and Hubbard D. M. Ind. Eng. C h e w Anal. Ed. 1941 16 333. Eastmond E. J. J . Opt. Sac. Amev. 1946 36 57. Stewart J. Mitchell R. L. and Stewart A. B. EmP. J . Exp. Agric. 1942 10 57. LundegBrdh H. “Die quantitative Spektralanalyse der Elements,” I I1 ; 1929 1934; Jena Gustav Mitchell R. L. J . SOC. Chew. Ind. 1936 55 267~. Mitchell R. L. and Robertson I. M. Id. 1936 55 2691 Fischer. ~TACAULAY INSTITUTE FOR SOIL RESEARCH CRAIGIEBUCKLER ABERDEE