Communication Evaluation of internal standardisation in electrothermal atomic absorption spectrometry Bernard Radziuk,a Natalya P. Romanovab and Yngvar Thomassenc a Bodenseewerk Perkin-Elmer GmbH, Postfach 101761, D-88647 � Uberlingen, Germany b Department of Analytical Chemistry, St. Petersburg State Technical University, St. Petersburg 195251, Russia c National Institute of Occupational Health, P.O. Box 8149 DEP, N-0033 Oslo, Norway Received 20th November 1998, Accepted 10th December 1998 Significantly improved performance in electrothermal atomic absorption spectrometry is possible using an internal standardisation technique.A Perkin-Elmer SIMAA 6000 simultaneous multielement spectrometer was used to study the correlation between two integrated absorbance signals. The behaviour of Pb (analyte) in different urine, blood and placenta samples was compared to that of Bi or Tl used as the internal standards. All samples were spiked with known amounts of Pb and Bi or Tl.A satisfactory signal correlation (r = 0.94) between the integrated absorbances for spikes of the analyte and internal standard was observed with Bi as the internal standard. After signal correction, the relative standard deviation of the integrated absorbance for Pb spikes reduced from 29 to 7% for urine, from 19 to 2% for blood and from 22 to 4% for placenta. The mean difference between Pb concentration found in analysed samples by the method of additions and using an internal standard was 10%.Introduction Although classical internal standardisation is a well established technique in multielement emission spectroscopy, very few studies have been carried out for electrothermal atomic absorption spectrometry (ETAAS). Internal standardisation in atomic absorption spectrometry was introduced in 1965 by Massmann1 when he reported on the use of an internal standard to reduce the variability of sample introduction. In those experiments, he used a laboratory-made multi-channel spectrophotometer.The scatter among results arising from inaccurate introduction of samples was slightly reduced for certain combinations of elements and increased for other combinations. Later, the concept of internal standardisation was used by Katskov and L’vov2,3 for determining trace elements in powdered samples, in particular, in zirconium dioxide and powdered graphite. When zirconium dioxide was being analysed, zinc was used as the internal standard, and when graphite was analysed, silver was used.These elements were introduced into the samples in the form of solutions. The use of internal standards made it possible to allow reliably for the weight of the powder samples and to obtain a standard deviation of 6.2% in the determination of cadmium. The recent development of commercially available simultaneous multielement atomic absorption spectrometers has made it practicable to apply internal standardisation techniques in order to improve analytical performance in ETAAS. In this work, we have tested Tl and Bi as internal standards when measuring Pb in blood, placenta and urine samples.Many authors use aqueous lead standards for calibration while analysing lead in biological materials. However, these matrices sometimes cause a significant decrease in the absorption signal of Pb. In this work we investigated the effectiveness of internal standardisation under these circumstances.We have studied the causes of the variation in the lead signal and have made an attempt to improve both precision and accuracy by internal standardisation. The criteria used for the selection of the internal standard were similarity of chemical/ physical properties to those of the analyte and that the internal standard concentration was negligible in the sample population. Table 1 summarises some chemical/physical parameters of Bi, Pb and Tl. Since these three elements have very similar characteristics we could expect both Tl and Bi to be appropriate internal standards for Pb.Experimental Instrumentation A Perkin-Elmer SIMAA 6000 ( � Uberlingen, Germany) simultaneous multielement atomic absorption spectrometer equipped with a Perkin-Elmer AS-71 autosampler and transversely heated graphite atomiser (THGA) was used for all measurements. A hollow cathode lamp was used for Pb and electrodeless discharge lamps for Bi and Tl. The measurement wavelengths were 283.3, 223.1 and 276.8 nm, respectively.End-capped graphite tubes with integrated L’vov platform supplied by Bodenseewerk Perkin-Elmer ( � Uberlingen, Germany) were used (Part No. B-300-0655). For the measurement of major constituents in the urine samples (Ca, K, Mg, Zn, P, S) a Perkin-Elmer Optima Model 3000 inductively coupled plasma atomic emission spectrometer was used (Perkin-Elmer, Norwalk, CT, USA). Reagents All standard and modifier solutions were prepared by dilution of 1 mg ml21 Pb, Bi, Tl and Pd Stock Standards (Spectrascan CertifiedTM, Teknolab AS, Drøbak, Norway).For dilution and Table 1 Some physical and chemical parameters for elements under study and their oxides4,5 Parameter Tl Pb Bi Atomic number 81 82 83 Molar mass/kg m23 0.204 0.207 0.209 Melting point/K 577 601 545 Boiling point/K 1748 2018 1825 Heat of vaporization/kJ mol21 180 196 199 Tl2O3(s) PbO(s) Bi2O3(s) Activation energy for oxide atomisation/ kJ mol21 225 268 257 Dissociation energy of MO(g)/kJ mol21 230 372 339 Dissociation energy of MCl(g)/kJ mol21 368 297 301 Anal.Commun., 1999, 36, 13–16 13digestion of urine, heparinised whole blood and placenta samples, water purified by reverse osmosis and deionisation and ultrapure 65% nitric acid (Scan Pure, Chemscan, Elverum, Norway) were used. Samples Whole blood and urine specimens were obtained from male workers exposed to lead; placenta samples were obtained from healthy mothers using protocols which conformed to the ethical guidelines of the Declaration of Helsinki.We tested the sampling equipment by leaching with 0.5% nitric acid; no detectable lead contamination occurred ( < 1 mg l21). All samples were also checked for possible contamination of bismuth and thallium using ETAAS and the content of these two candidate internal standards was below the detection limits ( < 1 mg l21). This is in good agreement with the results reported by e.g.Schramel et al. who documented very low physiological levels of bismuth ( < 0.01 mg l21) and thallium (0.25 mg l21) in urine.6 Analytical procedure The placenta tissue samples were after homogenization, freezedried in a standard laboratory system. Two and a half ml HNO3 were added to 0.3 g placenta weighed in a 13.3 ml polypropylene tube. After the 1.5–2 h necessary for the completion of active reaction, tubes containing the samples were heated at 95 °C for 90 min in a laboratory oven.After cooling, samples were diluted with H2O to final volume (13.30 ml ± 0.04). Two ml of a heparinised blood sample were transferred using a positive displacement micropipette to a 13.3 ml polypropylene tube and 2.5 ml HNO3 were added. The rest of the procedure (heating and dilution) for the preparation of blood samples was the same as for placenta tissue. Urine samples were diluted 1 : 1 with 0.5% HNO3. In all experiments, the sample and modifier aliquots were 10 ml.Modifier solution (0.1% Pd for urine and 0.05% Pd for other samples) was taken first, followed by sample. To check the accuracy of internal standardisation the method of standard additions was used routinely for determination of Pb in all the materials mentioned above. The final concentrations of spiked elements in the sample solutions to be analysed were 25 mg l21 for Bi and Pb, and 10 mg l21 for Tl. The experimental conditions for the graphite atomiser are given in Table 2.In order to better demonstrate the effectiveness of the internal standardisation, we intentionally selected the higher than recommended (850 °C) pyrolysis temperature for Pb. It led to increased scatter of results (see below). Results and discussion Possible causes of interference on Pb and selection of internal standard The integrated absorbance signal for Pb is suppressed in all three matrices under study here. In order to shed light on the ori this interference, the signals were correlated with the concentrations of major components in a series of urine samples with both low, medium and high matrix salt content.The best correlation (r = 0.81) is obtained with the concentration of phosphorus, representing the phosphate content of the urine sample (Fig. 1a). Thus, an appropriate internal standard should be affected in the same way as lead is, especially by phosphorus. The depression of the signal in the presence of other major constituents such as creatinine and sulfur may not be component specific, but rather the result of physical expulsion due to the rapid formation of molecules during the atomisation phase.This indicates a second important characteristic of the internal standard, i.e., the appearance temperature should be the same as that of the analyte. Based on this consideration, and on the general properties listed in Table 1, Bi and Tl were selected as the most likely internal standard candidates.Initially, qualitative studies on the effectiveness of these two elements were carried out. The correlation between Pb and Tl signals in different urine samples is shown in Fig. 2. It illustrates the variation of the signals for Pb and Tl spikes in different samples during the simultaneous detection of Pb and Tl. It is evident from this figure that these elements differ significantly in their behaviour and for this reason Tl cannot be used as internal standard for Pb.Bismuth gives much better correlation with Pb for urine samples (Fig. 3) and placenta and blood samples (Fig. 4). Calculation scheme When internal standardisation is used, all calculations are based on the supposition that sensitivity of analyte (SPb) and internal standard (SBi) depend identically on sample matrix and Fig. 1 Correlation between Pb-spike signal and concentration of P in urine samples: a, uncorrected Pb-spike signal; b, Pb-spike signals after correction by Bi.Concentrations of spike in samples are 25 mg l21 of both Pb and Bi. Table 2 THGA heating program used for the SIMAA 6000 spectrometer Step Temperature/°C Ramp/s Hold/s Dry 110 1 30 Dry 130 15 30 Pyrolysis 1100Bi/600Tl 10 20 Cool-down 20 1 5 Atomisation 1600Bi/1800Tl 0 3Bi(7aqua)/5Tl Clean 2450 1 3 14 Anal. Commun., 1999, 36, 13–16uncontrolled variations of conditions during pyrolysis and atomisation steps. Therefore, the ratio of sensitivity of aqueous solutions (Sa) must be related to that of real samples (Ss) as S S S S Pb a Bi a Pb s Bi s = (1) Assuming that the integrated absorbance, Q = º A dt, is proportional to mass, m, of element in the tube, we may express the sensitivity as S = Q/m (2) In this case, we can use for the calculation of the mass of Pb in the sample the expression m Q Q Q Q m Q k Q m Pb s Bi a Bi s Pb s Pb a Pb a Pb s Bi s Pb a = ¥ ¥ = ¥ ¥ (3) which follows from eqn.(1), (2) and mBi a = mBi s . Here k Q Q = Pb a Bi a (4) If the masses of Pb and Bi, spiked in samples and aqueous solutions, are equal, eqn. (1), (2) and (4) may be rewritten as: (QPb s )cor = QPb s + k (QBi a 2 QBi s ) (5) where QPb s and (QPb s )cor are the original and the corrected signals for Pb, respectively.Eqn. (5) will be used for comparison of corrected signals of Pb in different samples. Evaluation of results The cross-correlation function calculated by the SPSS statistical software package (SPSS Inc., Chicago, IL, USA) revealed reasonable correlation of Bi and Pb spike signals for all the materials investigated.Correlation coefficient equals 0.94. This led to significant improvement in the variation of Pb signals corrected with eqn. (5). Fig. 1b shows that the dependence of Pb sensitivity on phosphorus concentration was much reduced when Bi was used as internal standard. However, the slight upward trend in the corrected results indicates that the effect of phosphorus on Bi is somewhat stronger than that on Pb under the pyrolysis conditions used in this study.The improvement in the results for the determination of Pb in real samples is demonstrated in Table 3. In general, the use of the internal standard technique provides a 4- to 9-fold improvement in precision for different samples. Comparison with the integrated absorbance for the spikes in aqueous solution shows that recovery, i.e. accuracy of the determination, is also improved dramatically. The signals for Pb spikes are comparable after correction, irrespective of sample matrix (urine, placenta and blood).In addition to the above calculations of integrated absorbance signals for Pb in spikes, the concentrations in the samples were determined and compared with those obtained using the method of standard additions (Table 4). The sample numbers in Table 4 correspond to those in Figs. 3 and 4. Concentrations of Pb in several samples of urine (4, 8, 9, 12, 15, 17, 18, 24, 25, 29 and 30) were close to or below the detection limit and hence were Fig. 2 Correlation between Pb (—) and Tl (- - -) spike signals in urine samples. Concentrations of spike in samples are 25 mg l21 of Pb and 10 mg l21 of Tl. Fig. 3 Correlation between Pb (—) and Bi (- - -) spike signals in urine samples. Concentrations of spike in samples are 25 mg l21 of both Pb and Bi. Fig. 4 Correlation between Pb (—) and Bi (- - -) spike signals in blood and placenta samples. Concentrations of spike in samples are 25 mg l21 of both Pb and Bi.Table 3 Statistical evaluation of integrated absorbance signals (s) for 25 mg l21 of Pb corrected by Bi as internal standard Urine (n = 36) Placenta (n = 5) Blood (n = 5) All samples (n = 46) Parameter Original Corrected Original Corrected Original Corrected Original Corrected Mean signal 0.0171 0.0308 0.0158 0.0301 0.0146 0.0291 0.0167 0.0305 Min signal 0.0047 0.0272 0.0107 0.0292 0.0116 0.0282 0.0047 0.0272 Max signal 0.0230 0.0348 0.0205 0.0322 0.0180 0.0299 0.0230 0.0348 s 0.0049 0.0021 0.0035 0.0012 0.0028 0.00064 0.0046 0.0020 RSD (%) 29 7 22 4 19 2.2 28 6 Recovery (%) 52 94 48 92 45 89 51 93 Anal.Commun., 1999, 36, 13–16 15omitted in Table 4. It can be seen from the comparison of these data that the difference between concentration values includes both random and systematic errors. The last error is probably dominated by the uncertainty in the determination of the k value. In our experiments, the k value (1.4 ± 0.1, n = 3) was estimated as the ratio between the integrated absorbances of Pb (QPb a ) and Bi (QBi a ) in aqueous solution. At k = 1.3, chosen as an empirical example, the systematic discrepancy practically disappears and the mean difference value reduces from 10 to 5%.Conclusions Thallium and bismuth have been tested as internal standards when measuring lead in blood, placenta and urine samples. Bismuth has been found to be the best match in thermal volatility and atomisation behaviour to Pb in different matrixes.It was used to compensate for changes in Pb signals resulting from the matrix effects. This made it possible to obtain good recoveries for Pb in various matrices without the use of the method of standard additions. Nevertheless, it will be necessary to continue this study and to investigate the effectiveness of the internal standard technique under optimum sample pretreatment conditions and to improve the reliability of the determination of the sensitivity ratio (QPb a /QBi a ) during the analysis process.It should be stressed, however, that the selection of the internal standard in ETAAS is not a straightforward task. There must be a close coincidence in chemical and physical properties between analyte and internal standard. An inappropriate selection of internal standard can even be detrimental to the quality of an analysis. Acknowledgement We gratefully acknowledge Professor Boris L’vov, St.Petersburg State Technical University, for his active participation in the discussion of the results, which was essential to the success of this work. References 1 H. Massmann, in Second International Symposium Reinststoffe in Wissenschaft und Technik, ed. G. Ehrlich, Academie-Verlag, Berlin, 1966, pp. 297–308. 2 D. A. Katskov and B. V. L’vov, Zh. Prikl. Spektrosk. 1969, 10, 382 (in Russian). 3 B. V. L’vov, Atomic Absorption Spectrochemical Analaysis, Adam Hilger, London, 1970, p. 247. 4 Tables of Physical Quantities. Handbook, ed. I. K. Kikoin, Atomizdat, Moscow, 1976 (in Russian). 5 Disruption Energies of Chemical Bonds. Ionization Potentials and Electron Affinity Handbook, ed. V. N. Kondratiev, Nauka, Moscow, 1974 (in Russian). 6 P. Schramel, I. Wendler and J. Angerer, Int. Arch. Occup. Environ. Health, 1997, 69, 219. Paper 8/09096C Table 4 Lead concentrations in urine measured by method of addition and internal standardisation Concentration/mg l21 Sample Method of addition Internal standard Difference (%) Urine 1 5.0 4.4 11 2 4.2 3.8 11 3 4.7 4.2 12 5 5.2 5.4 23 6 3.3 3.2 4 7 5.3 5.1 4 10 5.5 5.1 7 11 7.4 6.8 9 13 10.4 10.1 3 14 6.4 5.9 7 16 8.8 7.2 18 19 5.3 4.6 14 20 6.4 5.7 12 21 3.2 2.8 11 22 5.1 4.5 11 23 4.8 4.2 12 26 3.6 3.0 16 27 5.8 6.1 25 28 4.6 4.3 7 31 4.5 4.4 4 32 5.9 5.4 8 33 4.4 3.9 9 34 3.7 3.4 8 35 4.5 3.9 13 36 4.5 4.6 22 Concentration/ng g21 Placenta 1 0.21 0.21 23 2 0.38 0.33 13 3 0.50 0.43 14 4 0.26 0.24 8 5 0.27 0.23 13 Concentration/mg l21 Blood 6 63.1 52.8 16 7 51.2 43.8 14 8 54.1 47.5 12 9 48.3 40.9 15 10 53.6 45.0 16 mean 10 16 Anal. Commun., 1999, 36, 13–16