Analyst August 1983 Vol. 108 pp. 925-932 926 Determination of Caesium and Rubidium by Flame and Furnace Atomic- absorption Spectrometry Z. Grobenski D. Weber B. WeIz and J. Wolff Bodenseewerk Perkin-Elmer & Co. GmbH Postfach 1120 0-7770 Oberlingen Federal Republic of Germany Caesium and rubidium have been determined by flame and furnace atomic-absorption spectrometry. It was found that both techniques are reasonably free from interferences and accurate enough for the routine analysis of various types of samples. Keywords ; Caesium determination ; rubidium determination ; atomic-absorption spectrometry Several general problems arise in the determination of caesium and rubidium by atomic-absorption spectrometry (AAS). The primary resonance lines at 852.1 nm (Cs) and 780.0 nm (Rb) are at the beginning of the infrared region of the spectrum and therefore at the limit of normal atomic-absorption instruments that utilise only one grating and a wide-range photo-multiplier.Nevertheless almost every commercially available atomic-absorption spectro-photometer has a specified wavelength range up to 870nm. However there is a great difference between spectrophotometers for determinations in this region of the spectrum. It can generally be said that double-grating monochromator instruments having one grating optimised for the UV range and a second for the visible range of the spectrum are much better. Ruled and illuminated areas blaze wavelength and other characteristics of the visible-range grating are of importance. The performance of an instrument can be considerably improved for the determination of rubidium and caesium by selecting a photomultiplier that is more sensitive in the visible range near-infrared part of the atomic-absorption spectrum.The next very important contribution comes from the radiation sources used. For most elements metal-vapour lamps have lost significance in atomic-absorption work except for rubidium and caesium. Their radiant intensity is considerably higher than that of the corresponding hollow-cathode 1amps.l Electrodeless discharge lamps for caesium and rubidium have an even higher radiant intensity thus improving the signal to noise ratio and detection limits. Even more important metal-vapour lamps emit extensively broadened lines because of self-absorption and self-reversal while electrodeless discharge lamps emit narrow lines with very little self-absorption.Sensitivities and linearity of the analytical curves obtained with the latter are therefore considerably better. A cut-off filter (opaque to wavelengths below 650 nm) is usually inserted in front of the monochromator thus improving further the signal to noise ratio. Ionisation potentials for caesium (3.893 eV) and rubidium (4.176 eV) are very low2 and very pronounced ionisation interferences therefore occur in flame atomic absorption. To suppress ionisation it is essential to spike standard and analyte solutions with a large excess of other easily ionised elements (e.g. 2-5mgml-l of potassium). Ottaway and Shaw3 found an extensive ionisation of alkali metals in electrothermal atomisation using atomic emission.This strongly influences the atomic line emission intensity and ionisation buffers should be added to prevent it.394 In contrast no ionisation interferences were found in atomic ab~orption.~ This can be explained by the fact that ionisation in the graphite furnace occurs only after the atomic-absorption signal has already been measured and thus has no practical influence on the analyte signal. Only a few papers dealing with the determination of caesium and rubidium by atomic-absorption methods have been published. To determine these elements in silicate rock samples by flame AAS a separation from the bulk of the matrix by coprecipitation with ammonium 12-molybdophosphate was recommended.6 In the determination of caesium in river and sea waters by graphite furnace AAS interferences from cobalt and iron were observe 926 GROBENSKI et al.DETERMINATION OF CAESIUM Analyst VoZ. 108 and chromatographic separation on a strong cation-exchange resin was recommended.’ For silicate rocks atomisation of the solid samples was investigated8 and compared with the analysis of dissolved samples using graphite furnace and flame techniques. Both papers7s8 discussed briefly the requirement for background correction in the visible range without really applying it. Some modern instruments provide such background correction by means of a tungsten lamp. Experimental Instrumentation Perkin-Elmer Model 5000 and 4000 atomic-absorption spectrophotometers were used. These are double-beam instruments having Czerny - Turner monochromator design with a focal length of 408 mm.The spectral range is a t least 180-900 nm covered by two gratings, each used in the first order. The visible grating has 1440 lines mm-1 with a large ruled area (84 x 84 mm) and blazed at 580 nm. A cut-off filter is automatically inserted into the optical path for measurements above 650 nm. The photomultiplier detector has a multi-alkali metal cathode with a UV-transmitting window covering a wavelength range from 185 to 930 nm. Electrodeless discharge lamps for caesium and rubidium were used and the slit employed had a spectral band width of 1.4 nm (with a graphite furnace the same slit with a reduced height was used). For electro-thermal atomisation a Model HGA-500 graphite furnace and an AS-40 autosampler were used.The background signal (using the tungsten lamp as a continuous light source for background correction) was recorded simultaneously with the analyte signal on a Perkin-Elmer Model 56, two-pen recorder. Integrated or peak absorbance was printed using a Model PRS-10 printer sequencer. Primary resonance lines at 852.1 nm (Cs) and 780.0 nm (Rb) were used. Reagents All reagents used were of the highest purity available (e.g. Merck Suprapur acids). Work-ing solutions of the required concentrations were prepared from stock standard solutions (Merck Titrisol) as required. Results and Discussion “State of the art” limits of detection are presented in Table I. Different pre-analysed samples were measured by flame and furnace AAS depending on the element concentration in the particular sample.TABLE I DETECTION LIMITS FOR DIFFERENT TECHNIQUES IN ATOMIC SPECTROSCOPY Technique Rb/mg 1-1 Cs/mg 1-1 Flame AAS with vapour discharge lamp . . . . 0.002 0.05 Graphite furnace AAS with electrodeless discharge lamp (100 pl) . . 0.00005 0.000 05 Flame AAS with electrodeless discharge lamp . . . . 0.001 0.01 (contamination problems) FlameAES* . . 0.0003 0.008 Graphite furnace AESS . . . . 0.000 1-0.0002 0.02-0.07 * AES atomic-emission spectroscopy. Determinations by Flame Atomic-absorption Spectrometry An oxidising lean blue air - acetylene flame was used. Determinations by AAS are simple and straightforward if sample decomposition is complete and a sufficiently high concentration of an ionisation buffer is added to samples and standards.Rubidium measured in different pre-analysed rock samples after conventional decomposition with hydrofluoric-sulphuric acid serves as an example. To 0.2-g samples 10-15 ml of hydrofluoric acid (40%) (depending on the silicon concentration) and 0.5 ml of sulphuric acid were added in a platinum vessel and the mixture was slowly heated to dryness. Hydrochlori August 1983 AND RUBIDIUM BY FLAME AND FURNACE AAS 927 acid (4 ml) was added to the residue and heated until a volume of about 0.5 ml remained. The solution was transferred into a 50-ml flask and diluted to volume with 2000 mg 1-1 potassium solution. For the blank 2000mg1-1 potassium solution was measured and there was no difference from the signal obtained on aspirating de-ionised water.Direct calibration against aqueous standards was applied and generally good agreement with certified values was found (Table 11). Only with the basalt BCR-1 was too high a value found; there is no evident explanation for the difference from the reported values. Only in the grandiorite sample MA-N of the International Working Group (IWG) was the caesium level very high and therefore measured by flame AAS. The values found of 547 and 536 pg g-1 were slightly too low (proposed value13 640 pg g-l). Higher values have been reported for flame emission a1~0.l~ Caesium was further measured in a soil extract with nitric acid using the analyte addition technique (method of additions 10 x scale expansion) and 0.15 mg 1-1 of caesium was found.This result was later confirmed using the graphite furnace technique. TABLE I1 DETERMINATION OF RUBIDIUM BY FLAME AAS IN STANDARD ROCK SAMPLES Each reported value represents the mean value of multiple determinations for separate decomposition. Sample G-2 granite USGS . . USGS GSP-1 granodiorite IWG MA-N granodiorite USGS AGV-1 andesite . . USGS BCR-1 basalt . . IWG BE-N basalt . . Found/pg g-l 171 172 112 . 260 259 266 . . 3352 3242 65 67 I1 61 69 48 50 Reported values P Value/pg g-l Reference 168 9 166 6 161 10 170 11 114 12 254 13 260 11 265 9 3 600 14 67 9 69 6 67 11 61 13 47 11 46.9 9 41.6 6 46.6 13 41.6 14 (3 120-4600) (24-61) Determinations by Graphite Furnace Atomic-absorption Spectrometry Conditions for both elements were established using electrodeless discharge lamps uncoated and pyrolytically coated graphite tubes conventional and maximum power (fast) heating for atomisation and different argon flow-rates (Table 111).A sensitivity check is the concentra-tion of the element in milligrams per litre that would generate a signal of 0.2 A when a 20-4 aliquot is dispensed. Values were established with an argon flow-rate of 50 ml min-l during atomisation. The quoted concentrations serve as reference values that should be obtained to within approximately 25%. A further sensitivity increase could be obtained for the normal heating mode by increasing the atomisation temperature. This would however decrease the tube life (especially for pyrolytically coated tubes).The use of fast heating provides higher sensitivities (up to a factor of 2) compared with normal heating. A pyrolytically coated tube provides higher sensitivities (4-10 fold) and this effect is even more pronounced when the gas stream is interrupted during atomisation. Atomisation off the L’vov platform was introduced into graphite furnace AAS to achieve freedom from vapour-phase interferences.16-19 A L’vov platform was inserted into a pyro-lytically coated graphite tube and similar conditions were used to those for atomisation off the wall of a pyrolytically coated tube with maximum power heating (Table 111) 928 GROBENSKI et d. DETERMINATION OF CAESIUM Analyst Yol. 108 TABLE I11 GRAPHITE FURNACE CONDITIONS FOR CAESIUM AND RUBIDIUM WITH 0.2% OF NITRIC ACID AS DLLUENT Conditions A pyrolytically coated tube with platform maximum power heating; B, pyrolytically coated tube off the wall maximum power heating; C pyrolytically coated tube off the wall normal heating; D uncoated tube off the wall maximum power heating; and E uncoated tube off the wall normal heating.Platform conditions can be recommended as the best. Element cs * . TemperaturelOC --- Conditions Pre-treatment Atomise A 900 1900 B 900 1900 C 900 2 700 D 900 2 200 E 900 2 700 Characteristic amountlpg m A r peak height peak height 4.0 -5.0 (PA) 3.6 7 4.5 12 30 41 46 77 No. of mgl-l (20 p1) for 0.2 absorbance 0.005 0.01 0.03 0.09 0.18 Rb A 800 1900 1.4 - 0.003 B 800 1900 1.3 2.7 0.008 C 800 2 700 1.6 4.5 0.01 D 800 2 100 6 10 0.03 E 800 2700 10 17 0.04 2.3 (PA) A 105-fold excess of rubidium was added to a 5 ng 1-1 caesium standard (and vice versa) to check for the presence of ionisation interferences.There was no difference in the signal with and without the ionisation buffer. A 105-fold excess of caesium increased slightly the signal for rubidium but this was attributed to the contamination obvious from the blank measurement. Rabidiam Rubidium was determined in different pre-analysed samples (Fig. 1) and it was found that in instances of signal depression application of the L'vov platform is beneficial. The analyte 0.3 0.2 8 e n rn a 0.1 0 B A C - I r I I -E E -OI E 0 0 -m m I - ' I ' - E Fig 1. Recorder tracings for the determination of rubidium in (A) NBS SRM bovine liver (10 p1 standard + 10 pl sample); (B) orchard leaves (10 pl standard + 10 p1 sample); and (C) acid standard (20 pl).A L'vov platform and maximum power atomisation at 1900 "C under gas interrupt (0 ml min-1 argon) were used. An autosampler was used for the automatic addition of sample into the tube and calculation of results was done off-line with a desk mini-computer August 1983 AND RUBIDIUM BY FLAME AND FURNACE AAS 929 addition technique was applied to check for the absence of depression and integrated absorbances were evaluated using linear regression. As there was no depression direct calibration was valid. It was found that a thermal pre-treated temperature of 900 "C was too high and minor losses were observed when longer thermal pre-treatment times were applied.This was observed only in the presence of a matrix. Because of this a temperature of only 800 "C can be recommended for thermal pre-treatment. A fast heating rate of up to 2000 "C s-l was used routinely for atomisation off the L'vov platform in a pyrolytically coated graphite tube. Biological and botanical samples [0.2 g of NBS Standard Reference Materials (SRMs) bovine liver spinach and orchard leaves] were decomposed in an autoclave (Perkin-Elmer Autoclave 3) with 6 ml of nitric acid (1 + l) and made up to 50 or 100 ml. Results for rubidium are presented in Table IV. Good agreement with certified values was obtained. Without the L'vov platform the values found were systematically lower.In addition, rubidium was determined in two US Geological Survey standard rock samples (AVG-1 and BCR) previously measured by flame AAS (see Table 11) and a value of 69 pg g-l was found for andesite (AGV-1) and 66 pg g1 for basalt (BCR-1). TABLE IV RESULTS FOR THE DETERMINATION OF RUBIDIUM USING GRAPHITE FURNACE AAS Sample Certificate valuelpg g-l Found/pg g-l NBS SRM 1577 bovine liver 18.3 f 1.0 18.9 f 0.8 NBS SRM 1571 orchard leaves . . 12 f 1 11.6 f 0.9 NBS SRM 1570 spinach . . . . 12.1 f 0.2 12.2 f 0.7 11.8 f 0.8* * Via solid sampling. The sensitivity for rubidium with pyrolytically coated tubes is so high that environmental contamination may cause a problem. This is also the main reason why the detection limit for rubidium is not better than that for caesium (see Table I).Rubidium was determined in water samples taken from Lake Constance (stabilised with 0.1 % of nitric acid) using 50-pl sample aliquots pyrolytically coated tubes atomisation off the wall and direct calibration against acid standards. The concentration found was 1.24 pg 1-1 (Fig. 2). Solid sampling was applied to the determination of rubidium in NBS spinach.lB The half as sensitive line at 794.7 nm maximum power to 2 100 O C atomisation off the wall and an argon flow-rate of 200 ml min-1 during atomisation were applied. In routine analysis carbon build-up was avoided using an ashing step at 500 "C with oxygen as alternate gas before going to the maximum pre-treatment temperature of 800 "C. Direct calibration against ordinary standards gave excellent agreement with the certified value when integrated absorbance values 0.2 E 4? 8 0.1 a m 0 +lo0 pg Rb Fig.2. Recorder tracings for the determination of rubidium in lake water. Sample aliquots 60.p1, pyrolytically coated tubes maximum power atomsa-tion at 1900 O C and a flow-rate of 30 ml min-l of argon were used 930 GROBENSKI et al. DETERMINATION OF CAESIUM Analyst Vol. 108 were used (found 11.8 & 0.8 pg gl; certified 12.1 If more than one element has to be determined in spinach and a number of samples are available, determination after decomposition is faster than solid sampling and more suitable for automa-tion. 0.2 pg gl) (see Table IV). Caesiwn Caesium was determined inUSGS rock sample (Fig. 3) and the results are presented in Table V.The decomposition method was the same as previously described for flame determinations. Pyrolytically coated graphite tubes and maximum power atomisation were applied. A signal depression caused by the matrix was observed for atomisation off the wall and integrated absorbance could only partially compensate for it. Only after 10 pl of sulphuric acid (96%, Suprapur) had been added to 1 ml of decomposed rock sample solution was depression no longer observed and direct calibration could be applied. The same amount of sulphuric acid was added to the calibration solutions but there was essentially no influence on their peak height and area values. For atomisation off the L’vov platform the addition of sulphuric acid was still required. The same water samples from Lake Constance were also analysed for caesium.With a 10-fold scale expansion pyrolytically coated tubes a sample aliquot of 50 p1 and maximum power atomisation at 1900 OC a concentration of lower than 0.1 mg 1-1 was found (Fig. 4). Caesium was measured by graphite furnace AAS in the same nitric acid soil extract as with the flame technique. Because of much lower sensitivity (10-fold) uncoated graphite tubes TABLE V DETERMINATION OF CAESIUM IN STANDARD ROCK SAMPLES BY GRAPHITE FURNACE AAS Each value represents a mean value of multiple determinations for a single decomposition. Reported value USGS GSP-1 granodiorite . . 0.9 1.1 1.0 1.0 USGS AGV-1 andesite 1.23 1.24 1.23 1.6 1.4 1 .o 1.3 IWG BE-N basalt . . 0.7 0.7 Sample USGS G-2 granite .. Found/pg g-l Valuelpg g-l 1 .o 2.1 0.9 1.3 1.0 1.4 1.4 1.0 1.23 1.24 1.33 1.22 1.7 1 .o 1 1.0 0.8 0.88 1.00 1.03 0.98 1.4 1.2 1 .o 1.27 1.26 1.34 1.23 0.96 0.96 1.2 1.22 0.96 1.03 0.96 0.8 Reference 6 8 13 11 20 21 22 23 24 6 13 11 20 21 22 23 24 16 13 20 21 22 23 24 16 13 11 20 22 23 24 16 1 August 1983 AND RUBIDIUM BY FLAME AND FURNACE AAS 93 1 Fig. 3. Recorder tracings for the determination of caesium in standard rock samples of USGS granite G-2 granodiorite GSP- 1 and andesite AGV- 1. Pyrolytically coated tubes and maximum power atomisation a t 1900 O C under an argon flow-rate of 50 ml min-’ were applied. were used and sulphuric acid was added for the reasons mentioned above.The result obtained of 0.15 mg 1-1 was in excellent agreement with that obtained by the flame technique. All determinations were performed using the tungsten lamp background corrector. Owing to the absence of background however it was not required for the samples investigated. Only when the possibility of analyte modification for caesium with phosphoric acid was investigated were non-specific signals observed and had to be corrected. With the addition of a 10000-fold excess of cobalt and iron no spectral interferences for caesium were observed. 0.04 I 0.03 8 +50 pg Cs 4 L Fig. 4. Recorder tracings for the determination of caesium in lake water. Sample aliquots 50 pl pyrolytically coated graphite tubes maximum power atomisation a t 1900 “C under gas interrupt (0 ml min-l argon) were used.Conclusion The satisfactory agreement with pre-analysed samples indicates that the determination of caesium and rubidium by AAS is reasonably free from interferences and is accurate enough for the routine analysis of the various types of samples. An essential prerequisite is good instru-mentation optimised for analysis in the near-infrared region of the spectrum as well as intense and stable spectral line sources For flame AAS control of ionisation interferences by using effective buffers is required in addition. Graphite furnace atomisation off the L’vov platform using maximum power heating and peak-area integration helps to eliminate vapour-phase interferences.References 1. 2. 3. Welz B. “Atomic Absorption Spectroscopy,” Verlag Chemie y n h e i m and New York 1976. Weast R. C. Editor “Handbook of Chemistry and Physics, Ottaway J. M. and Shaw F. AppZ. Spectrosc. 1977 31 12. Fifty-third Edition CRC Press, Cleveland OH 1973 932 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. GROBENSKI WEBER WELZ AND WOLFF Ottaway J. M. Hutton R. C. Littlejohn D. and Shaw F. Wiss. 2. Karl-Marx-Univ. Leipzig, Sturgeon R. E. and Berman S. S. Anal. Chem. 1981 53 632. Das A. K. Banerjee S. and Chowdhury A. N. 2. Anal. Chem. 1978 290 319. Frigieri P. Trucco R. Ciaccolini I. and Pampurini G. Analyst 1980 105 651. Langmyhr F. J. and Thomassen Y. 2. Anal. Chem. 1973 264 122. De Laeter J. R. and Rosman K. J. R. Geostand. Newsl. 1977 1 35. Govindaraju K. Mevelle G. and Chouard C. Chem. Geol. 1971 8 131. Abbey S. Geostand. Newsl. 1978 2 141. Flanagan F. J. Geochim. Cosmochim. Acta 1969 33 81. Flanagan F. J. Cosmochim. Acta 1973 37 1189. Govindaraju K. Geostand. Newsl. 1980 4 (l) 49. Goguel R. Geostand. Newsl. 1981 5 95. Slavin W. and Manning D. C. Anal. Chem. 1979 51 261. Vollkopf U. Lehmann R. Grobenski Z. and Welz B. Appl. At. Spectrosc. 1980 26. Grobenski Z. Lehmann R. Tamm R. and Welz B. Microchim. Acta 1982 1 115. Kaiser M. L. Koirtyohann S. R. and Hinderberger E. J. Spectrochim. Acta Part B 1981 36 773. Govindaraju K. Analusis 1975 3 164. Randle K. Chem. Geol. 1974 13 237. Katz A. and Grossmann L. U.S. Geol. Surv. Prof. Pap. 1976 No. 840 49. Steinnes E. J . Radioanal. Chem. 1972 10 65. Terashima S. and Mita N. Geostand. Newsl. 1981 5 71. Math.-Naturwiss. Reihe 1979 28 367. Received June llth 1982 Accepted February 25th 198