410 OSBORN AND JOHNS: THE RAPID DETERMINATION OF SODIUM AND [VOl. 76 The Rapid Determination of !Sodium and Potassium in Rocks and Minerals by Flame Photometry BY G. H. OSBORN AND H. JOHNS (Presented at the meeting of the Physical Methods Group on, Friday, October 6th, 1950) A method is outlined'for the rapid and accurate determination of sodium and potassium in rocks and minerals by means of the flame photometer. It is shown that the method gives results that compare favourably with those obtained elsewhere by classical procedures, and that the saving in time required for an analysis is very great. It :is also shown that the sodium and potassium may be determined, if necessary, when only very small amounts of material are available. THE usual method for the chemical analysis of sodium and potassium in rocks and minerals is that of Lawrence Smith,l which consists essentially in fusing an intimate mixture of one part of ground rock with one part of ammonium chloride and eight parts of calcium carbonate, extracting in water and filtering to remove the silicates and aluminates of calcium, and the carbonates of iron, calcium and magnesium.The alkalis pass into the filtrate as chlorides. The rest of the procedure consists in the precipitation of the excess of lime by means of ammonium carbonate, expulsion of ammonium salts by heating the evaporated filtrate, removal of the last traces of lime, conversion of the traces of alkali sulphates to chlorides, weighing the mixed alkali chlorides and, finally, the separation and weighing of the potassium either as the perchlorate or chloroplatinate, with the estimation of sodium by difference.The drawbacks to this method are : (a) the time required by even an experienced operator is excessive-Haslam and Beeley,2 in a critical review of the method, state that they found it difficult to complete the sodium and potassium determinations in less than three days; (b) great manipulative skill is required; (c) a blan:k determination must be made for all the reagents; ( d ) there is great risk of loss of the alkalis by occlusion when filtering from the insoluble silicates, aluminates and carbonates, so that several reprecipitations are required for highly accurate work; (e) there is risk of loss of alkalis at the final volatilisation of ammonium salts. Despite its limitations this method is still very widely used in mineral analysis, as other chemical methods have not been shown to possess any outstanding advantages.Haslam and Beeley2 also proposed a modification of the method for which they claimed increased accuracy, but stated that whilst the sodium determination could normally be completed in one day, the following day was required to complete the potassium determination. In this modification sodium is determined as the zinc u.ranyl acetate complex and the potassium, if present in small amounts, as the cobaltinitrite or, if in large amounts, as the perchlorate.July, 19511 41 I Various authors, including Mitchell3 and M~ller,~ in an attempt to simplify and speed up the determination, have proposed spectrographic methods.Lundegardh and Mitchell in early experiments introduced a suitable solution of the mineral into an air - acetylene burner by means of an atomiser, passed the emitted light through a spectrometer and photographed the spectrum; the density of the spectral lines was measured by means of a photometer and compared with a composite standard solution analysed in the same manner. The results were promising, but the technique was intricate and time-consuming. Mitchell and Robertson5 noted that the intensity of the light emitted by an ion was not a simple function of its concentration in solution, but varied in a complicated manner depending on the presence or absence of certain other ions in solution. With the Lundegardh apparatus these authors observed that calcium and strontium flames were strongly depressed by aluminium, but that in the presence of an excess of calcium the depression of strontium by aluminium was considerably diminished. In controlled conditions this depression could be made use of for the indirect determination of aluminium.Many other similar interferences have been observed, both with anions and with cations, e.g., the emission from alkali metals is reduced slightly by sulphates and tartrates and very much by phosphates. More recent developments of spectrophotometric analysis have been largely concerned with studies of these and other interferences and with attempts to reduce them. The direct reading photo-electric photometer made it possible to carry out analysis much more rapidly, with consequent reduction in the errors caused by fluctuations in gas and air pressure.In some of these instruments a desired waveband is scanned through selective filters, but the difficulty of obtaining spectrally pure filters makes the more versatile instruments that are provided with monochromators much to be preferred. Two different methods have been used for relating intensity of light to concentration of the ion in solution. In the absolute method a calibration curve is drawn for solutions of known concentrations, gas, air and oxygen pressures being kept constant within fine limits. The concentrations of unknown solutions can then be determined by interpolation on the graph. The objection to this method is that interference from foreign ions may affect the sample, but not the calibrating solution, and that these interferences may be entirely unsuspected.Attempts have been made to obviate these effects by preparing calibration solutions that imitate as closely as possible the composition of the sample t o be examined. For the determination of alkali metals in cement, a method has been described6 in which a blank solution of calcium oxide in hydrochlbric acid is used as a base’ for the calibration solution. Excellent results are claimed, and it is possible that this method is satisfactory for simple routine analysis, but for more complex materials, such as rocks and minerals, containing an unknown and probably large number of constituents in unknown concentration, such an artifice would be impracticable. A recent paper7 described the addition of “radiation buffers” to calibration solutions and to samples under test.The addition consists of a large excess of an interfering ion, so that further effects due to trace constituents in the sample are negligible. A serious objection, however, to all calibration methods is that light intensity depends on many variable factors that cannot always be reproduced after an interval of time. Possibly the most difficult factor to control is rate of atomisation, which depends on air pressure, the viscosity and temperature of the solution and the width of the orifice. Slight clogging of the orifice changes the flame response profoundly. The internal-standard method, inherited from arc spectrography, has been claimed to be less affected by interferences.* A known small amount of a foreign ion is deliberately added to the material to be examined and the light intensity due to this ion is then compared with the light intensity due to the ion sought.A dual optical system is used so that fluctua- tions in pressure, viscosity and temperature are avoided. The internal standard is normally chosen so that its properties resemble those of the element sought, e.g., lithium is used in determination of sodium and potassium. Interfering effects are, however, not necessarily the same for two different ions, even if they are closely related, and in examining unknown solutions a further uncertainty is introduced by the internal-standard method, as the material may already contain some of the reference element.6 For these reasons it is difficult to apply the internal-standard method with precision to the analysis of rocks, although it is POTASSIUM IN ROCKS AND MINERALS BY FLAME PHOTOMETRY412 OSBORN AND JOHNS: THE RAPID DETERMINATION OF SODIUM AND [Vol.76 certainly satisfactory for such determinations as the sodium and potassium content of blood serum. With the shortcomings of the two methods in general use in mind, the possibility was considered of determining alkali metals by a direct addition method, i.e., by measuring the response of the sample and then measuring the increase in light intensity on adding a small amount of standard alkali to a suitable aliquot, I t was thought that, by taking readings for several aliquots containing different quantities of standard, a graph could be constructed and extrapolated to determine the amount of alkali present in the sample.The usual interference effects should in this manner be entirely obviated. Extrapolation is satisfactory only if the graph is perfectly regular .and preferably linear. Experimental work demonstrated that potassium gave an almost linear graph a t low con- centrations (0 to 10 p.p.m.) but that sodium gave a slightly curved graph even at this very small concentration. Calcium gave a linear response up to 500 p.p.m. Mr. B. S. Cooper has suggested that the parabolic shape of these graphs is due to absorp- tion of light from the rear part of the flame by the vapour of the substance in the front of the flame. It is possible that a more linear response would result from the use of a relatively flat flame of the “fish-tail” type, and experiments are being conducted to investigate this point.Alternatively, saturation emission may be prevented by working at very low con- centrations or by the addition of foreign ions known to depress emission. It has already been mentioned that certain anions depress the amount of light emitted by alkali metals, and that the phosphate ion in particular causes a large depression. The effect of phosphate ions on the shape of the graph of sodium concentration versxs light intensity was therefore investigated. The results showed that for conce:ntrations up to 20p.p.m. the graph was perfectly linear, and very much less steep than in the absence of phosphate. Similar results were obtained for potassium. The reduction of light emission in the presence of foreign substances may be due to restricted ionisation, leading to a molecular, rather than an ionic spectrum, but other factors such as absorption and alteration of flame temperature may play a part.Nitrogenous organic substances may, for instance, be expected to lower the flame temperature if they are present in a large excess, and a similar effect is possible in the presence of halogenated paraffins. However, in a series of experiments, no reduction in light intensity was observed on addition of pyridine or chloroform to a sodium solution, although a small decrease was brought about by addition of an excess of urea. Other organic substances such as acids and phenols were also found to have little or no effect, and it was concluded that at the temperature of the flame organic molecules are, €or the most part, completely destroyed.It has been reportedg that light emission is actually increased by the addition of alcohol and acetic acid, and these results have been verified by experiment. It is probable that this is a physical effect resulting from an increased rate of atomisation owing to the lowered surface tension of the test solution. Inorganic ions that were found to suppress emission of light from sodium include phosphate, borate and molybdate, which are all about equally effective, and may in some circumstances reduce the light intensity by more than half. A smaller effect is produced by addition of an excess of sulphate, but nitrate, chromate and halides give the same response as the free base.The effect of some of these anions is shown in Figs. 1 and 2, which were prepared by plotting readings on the Beckman photometer against concentration of sodium or potassium in parts per million. The curves did not pass through the origin as the galvanometer is slightly deflected even when no sodium is intentionally vaporised in the flame. This effect is due to flame background and to traces of sodium in the air and the water used for preparing solutions, and is measured by vaporising pure distilled water into the flame. In order to investigate the effect of cations, series of readings were taken to compare pure dilute solutions of sodium and potassium salts with solutions of various salts containing the same quantity of added alkali.In the first series the equivalent of 20 parts per million of sodium, as chloride, was added to various 1 per cent. solutions of metal chlorides. It was observed that the intensity of light at 5893 A was reduced by about 20 per cent. by ammonium, cupric, zinc, cadmium, magnesium, ferric , cobalt and nickel chlorides, with no appreciable difference in effect between any of these ions. Simaller effects were observed with bariumJuly, 19511 413 and calcium chlorides, the depression being about 15 per cent. In the presence of lithium and potassium chlorides no decrease in intensity of sodium emission was detected. POTASSIUM I N ROCKS AND MINERALS BY FLAME PHOTOMETRY Sodium, p.p.m. Potassium,, p. p. m. Fig. 1. Effect of anions on sodium Fig. 2. Effect of anions on potassium Curve A, free base; curve B, solution in Curve A, free base; curve B, solution in 1 per cent. boric acid; curve C, solution in 1 per cent.ammonium molybdate; curve D, solution in 1 per cent. ammonium phosphate Approxi- mately uniform depression of about 20 per cent. was noted when the sodium was determined in 1 per cent. solutions of silver, aluminium, lead, cerium, lanthanum and uranium as nitrates. Similar results were obtained on repeating these investigations with potassium instead of sodium. It appears from these preliminary results that any solids, other than alkalis and alkaline earths, have the effect of depressing the intensity. These findings were applied to the analysis of rocks and minerals by finely grinding a number of samples and analysing them on the Beckman flame spectrophotometer, model DU.PROCEDURE Use wavelengths of 5893 A for sodium and 7670 A for potassium; these are the principal lines for these elements. With a slit width of 0.1 mm, pass propane into the burner until a pressure of 2 cm is registered on the manometer. Then start the oxygen supply and increase it until a non- luminous flame is produced. At the wave- lengths specified the flame background will be found to be very slight. Weigh 0.1 g of finely ground material in a platinum dish and moisten it with 1 drop of sulphuric acid. Evaporate to dryness with 2 ml of hydrofluoric acid. If decomposition is not complete, repeat the evaporation with hydrofluoric acid. Dissolve the residue by boiling it with 10 ml of 5 N hydrochloric acid.Dilute the solutions to 500 ml with a 1 per cent. solution of ammonium phosphate. Then take four 100-ml aliquots and add to them 0, 0.1, 0.2 or 0.3 ml respectively of 0.1 N alkali according to the elements being determined. The ammonium phosphate must previously have been examined for sodium and potassium on the flame photometer. Normally, the amount present should not exceed 0-1 p.p.m. in a 1 per cent. solution. MEASUREMENT- units as follows- 1 per cent. ammonium phosphate In a second series, sodium as nitrate was determined in other nitrate solutions. Operate the atomiser at a pressure of 251b. When different solutions are vaporised, the flame intensity is measured in arbitrary (a) Prepare a reagent blank by evaporating hydrofluoric and sulphuric acids in the quantity used for the decomposition of the rock specimen and treat as described above.(b) Use a prepared solution of the sample under examination. (c, d and e) Use solution (b) containing 0.1, 0.2 and 0.3 ml respectively of 0.10 N sodium hydroxide in 100 ml.414 OSBORN AND JOHNS: THE RAPID DETERMINATION OF SODIUM AND [VOl. 76 If the increase in flame intensity between (c), (d) and (e) does not indicate a rectilinear relationship between concentration and light emitted, all the solutions are diluted with water and the determinations repeated. When the dilution has been so adjusted that a constant difference is found between (c), ( d ) and (e), the percentage of Na,O equals- (:+;) x 31/106 percentage of solution. Owing to the limifed linearity of the curve, greater accuracy is obtained if the value of (b) is kept fairly small in the above equation.If a very large deflection is obtained it is advisable to dilute the sample solution 10- or 100-fold and to repeat the determination. Some results by the above technique are shown in Table I and compared with results for the same specimens found by Dr. Max Hey of the British Museum of Natural History by the classical Lawrence Smith method of weighing the mixed chlorides of sodium and potassium, determining potassium as the perchlorate and calculating Na,O by difference. TABLE I COMPARISON OF RESULTS BY SPECTROPHOTOMETRIC AND CHEMICAL METHODS FOR SODIUM Na,O by chemical Na,O by flame Sample analysis, spectrophotometer, % Yo 1516 Rhyolite, Lupata Gorge, Nyasaland . .1.97 2.16 2358 Black manganese ore (46% of MnO), Benallt, N. Wales . . .. .. 0.27 0.3 1 2253 Limestone, Nyasaland . . .. .. 0-03 0.15 2194 Quartz - alkali - syenite, Lion Rock Gully, Nyasaland . . . I .. 3.88 4.09 2361 Chloritic mudstone, country rock of 2193 Felspar - pyroxene rock, Nkalonge Hill, Grey phosphatic manganese ore (17% of Biotite-bearing algerine - augite foyaite, manganese ores, Benallt, N. Wales . . 0.14 0.20 N yasaland .. 0.27 0.36 2018 Phenolite dike, Maize Hill, Nyasaland 10.73 11-00 2359 MnO, 10% of P,06), Benallt, N. Wales nil 0.07 2067 Basalt, Teliki Volcano, Kenya . . .. 4.14 4-13 1907 Nyasaland , . .. .. .. 8-25 8.36 Difference, % +0-19 + 0.04 +0*12 +0.21 + 0.06 + 0.09 + 0.27 + 0.07 - 0.01 +Owl1 The results with the flame spectrophotometer are usually slightly higher than those by chemical methods, but this is satisfactory if the probable loss by occlusion or volatilisation in the gravimetric method is accepted.When no sodium was detected gravimetrically we are convinced it was present in small amounts. Moreover, the concordance is much better than that of the figures quoted by Hillebrand and LundelP for sodium determinations on the same rock made gravimetrically by different analysts. The corresponding figures for potassium are shown in Table 11. TABLE :I1 COMPARISON OF RESULTS BY SPECTROPHOTOMETRIC AND CHEMICAL METHODS FOR POTASSIUM Sample 1516 2358 2194 2361 2193 2018 2359 2067 1907 K,O by chemical K:,O by flame analysis, spec.trophotometer, % % 4.86 4.86 nil 0.1 1 6-28 6-23 0.48 0.59 14.68 14.70 4.72 4.53 nil 0.04 1-91 1.90 7-53 7-64 Difference, % nil +0*11 - 0.05 + O * l l + 0.02 -0.19 -!- 0.04 - 0.01 1-0-11July, 19511 POTASSIUM IN ROCKS AND MINERALS BY FLAME PHOTOMETRY 415 It will be observed that the agreement is even better than that for sodium, and that although the deviation is random in direction, it is, as before, mostly in the positive direction, so agreeing with the findings for sodium.As before, when no potassium was reported by the gravimetric method, some was recorded by the flame photometer. Again the concordance is better than that between figures for potassium in the same rock determined gravimetrically by different analysts. CONCLUSIONS Consideration of the above results indicates that by working with suitably dilute solutions in the presence of an excess of phosphate, the graph of flame intensity versus sodium or potassium intensity is sufficiently nearly linear to permit analysis by a direct addition method, the results comparing favourably with those of the classical chemical procedure.The time taken for the analysis of an average sample of acid-soluble rock or mineral, including all preparatory work, is about one hour. If a number of samples are being examined simul- taneously, the average time would be less. All the rocks examined were found to be acid soluble after treatment as described on p. 413, but if the rocks or minerals to be examined are very refractory, e g . , tourmaline, beryl, biotite or topaz, and will not dissolve directly in acid, then, after the Lawrence Smith fusion, the mixture can be extracted with acid and, after the removal of the silica, the sodium and potassium can be determined directly on an aliquot of a measured volume. The presence of large amounts of calcium in the acid solution of the product obtained from the opening-up process will not influence the method in any way when the ammonium phosphate is added in the course of the flame photometry procedure. If only very small amounts of material are available, it would be possible to use as little as 0.005 g of material if the amount of sodium is above 0.5 per cent., or 0.02 g if the amount of sodium is 0.5 per cent.or less. Hence the method can be used as a micro-method without loss of accuracy. The authors thank the Directors of the British Drug Houses Limited for permission Thanks are also due to Dr. Max Hey of the British Museurn (Natural to publish these results. History) for providing the chemical analyses of the rocks and for helpful discussion. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Lawrence Smith, J., Amer. J. Sci., 1870, 50, 269; Ann. Cheni. Pharnz., 1871, 159, 82. Haslam, J . , and Beeley, J., Analyst, 1941, 66, 185. Mitchell, R. L., J , SOC. Chem. Ind., 1936, 55, 267. Muller, R. H., Anal. Chem., 1947. 19, part 8 (August), 21n. Mitchell, R. L., and Robertson, I. M., J. Soc. Chem. Id., 1936, 55, 269. Chemical Age, 1950, 62, 857. West, P. W., Folse, P., and Montgomery, D., Anal. Chem., 1950, 22, 667. Berry, J . W., Chappell, D. G., and Barnes, R. B., Ind. Eng. Chem., Anal. Ed., 1946, 18, 19. Parkes, T. D., Johnson, H. O., and Lykken, L., Anal. Chem., 1948, 20, 827. Hillebrand, W. F., and Lundell, G. E. F., “Applied Inorganic Analysis,” John Wiley and Sons, New York, 1929, pp. 5, 850, 861 and 874 et seq. ANALYTICAL DEPARTMENT THE BRITISH DRUG HOUSES LIMITED LABORATORY CHEMICALS GROUP POOLE, DORSET