INTER-LABORATORY NOTE Determination of Cadmium in Very Low Concentration Urine Samples by Electrothermal Atomic Absorption Spectrometry JAMES P. SNELLa , SUSANNE SANDBERGb AND WOLFGANG FRECH*a aDepartment of Analytical Chemistry, Umea° University, S-901 87 Umea° , Sweden bDepartment of Occupational and Environmental Medicine, Umea° University, S-901 87 Umea° , Sweden Urine samples were collected that contained concentrations of A number of previously published papers outline method Cd between 0.05 and 0.5 mg l-1.When using ETAAS for the development for this analyte–matrix combination. For analysis of these samples, spectral interference is present that example, Ediger and Coleman,5 Yin et al.6 and Smeyerscan obscure low masses of Cd and deteriorate the precision Verbeke et al.7 have demonstrated the benefits of using and accuracy of the analysis. Instrumental advances such as NH4NO3 modifier and for the last two, in combination with electrodeless discharge lamps, the use of echelle Pd.Fraile et al.8 used a controlled atomization temperature to monochromators, solid-state detectors and end-capped separate temporally the vaporization of Cd and sodium chlorspatially isothermal atomizers have been combined to reduce ide. Our study included specimens from children and diabetics. detection limits. Spatially isothermal atomizers provided with As the physiological dilution of urine may be greater for these end-caps and modified contact cones were found to double the two groups, a significant number of samples contained much sensitivity for Cd, compared with a standard atomizer lower concentrations of Cd than those previously reported2 or configuration.A combined chemical modifier of Pd and used for method optimization.6,8–10 These papers give detection NH4NO3 can reduce interference from molecular alkali limits from 0.05 mg l-1,10 with typical concentrations between halides, on atomization of Cd. Samples were decomposed with 1 and 10 mg l-1 Cd.In our case, without careful optimization boiling HNO3 to destroy carbon compounds and ensure of procedures the very low Cd signals of the order of 0.01 homogeneity. Calibration was by standard additions to ensure absorbance could be disturbed by background absorption, accuracy in the determination; however, standard calibration which is not sufficiently reduced by chemical modification or may be used with a deterioration in accuracy of about 6%.For the use of a spatially isothermal transversely heated graphite a sample containing 0.1 mg l-1 Cd, the method including atomizer (THGA). sample pre-treatment gives a precision of about 15%. The Instrumental noise levels for the technique have been con- accuracy of the method was established by standard additions tinuously lowered by advances in instrumental AAS design. to samples and the analysis of a urine reference material; the Electrodeless discharge lamps are known to increase light instrumental detection limit is 0.008 mg l-1 Cd in urine.intensity. Echelle spectrometers and solid-state detectors were Keywords: T ransversely heated graphite atomizer; end-capped recently introduced into commercial instruments11–13 (for a tube; modified contact cone; matrix interference; cadmium; review see ref. 14). For the spectral resolution required, the urine echelle spectrometer has a more intense light flux than a conventional Czerny–Turner monochromator.At the Cd wave- Cadmium is a toxic element at relatively low concentrations length of 228.8 nm, the solid-state detector used has improved for which urine provides a reliable guide to the body’s load.1–4 quantum efficiency over photomultiplier tubes.13 In addition, Urine is also a convenient sample to collect. The accurate the solid state detector gives less noise than photomultiplier determination of Cd in urine is now possible, using modern tubes at the measured light intensities typical of absorption instrumentation, at such low levels that most samples are spectrometry.13 measurable. This allows the monitoring of the Cd load on For these state-of-the-art instruments, the modification of people who excrete low levels of the metal. The subjects of our furnace geometry can further improve sensitivities.To increase study are residents of Sweden, where natural exposure to Cd analyte peak area for a given mass, THGAs with end-capped is low, and some are children who excrete more diluted urine furnaces are used to increase analyte residence time in the than adults and have had less time to accumulate Cd.In furnace14–18 and in-house modified contact cones19 can further addition, the subjects are diabetics who may also excrete more increase residence time at lower atomization temperatures. As diluted urine than people with normal pancreatic function. It well as limiting the rate of diffusion, a reduced atomization is known that diabetes may lead to nephropathy and that this temperature is suitable for Cd determination as it achieves a can reduce the kidneys’ tolerance to the toxic effects of Cd.It partial temporal separation of Cd atoms from sodium chloride is therefore particularly important to determine Cd at low diatomic vapour. concentrations for correlation to the health of diabetics. Our study was performed using these improved conditions For the determination of Cd in urine, ETAAS is the chosen to lower detection limits and adopted methods of validation. method of many analysts because of its relative simplicity, the Standard additions were made to samples to give the recovery possibility of automation and the low instrumental detection of the preparation procedure.Sample digestion was repeated limits given. With this technique, however, matrix interferences, to check in-batch precision. Reference materials were used to particularly from various concentrations of sodium chloride, remain that deteriorate accuracy and precision.check method accuracy. Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 (491–494) 491EXPERIMENTAL injection port diameter of 1.8 mm. Two sets of graphite contact cones were used: the standard Perkin-Elmer models and modi- Reagents and Sample Preparation fied cones with a groove cut around the injection port17 (Fig. 1). Recovery tests were performed by adding 500 pg of Cd to Sample aliquots of 10 ml were used; these were introduced with 10 ml of chemical modifier and 10 ml of either diluent or approximately 2.5 g of urine, before acid decomposition.Calibration graphs were constructed using a digested urine standard containing 0.5% m/m HNO3. The chemical modifier used was 6 mg of Pd, and 0.5 mg of NH4NO3 dissolved in sample with a Cd concentration below the instrumental detection limit. The autosampler of the spectrometer was pro- 0.9% m/m HNO3.Cadmium standards were prepared daily from a 10 mg l-1 stock solution and contained 0.5% m/m grammed to pipette various volumes of Cd standard and diluent with 10 ml of the urine sample. Sample measurements HNO3 . All sample containers were new and acid-cleaned prior to were made consecutively. The instrumental detection limits for Cd in urine are given use. Samples were collected in 100 ml poly(ethene) bottles. Care was taken to avoid contamination at all stages3,4,20 and in Table 2 and are defined as 3s of ten replicates of a low concentration sample, the Cd content of which is then deter- volunteers were requested not to smoke before collection or touch the inner surfaces of the container.Midstream, morning mined by standard additions. urine was voided directly into the container, the amount recorded and the sample frozen. On receipt the samples were RESULTS AND DISCUSSION thawed, transferred into 20 ml poly(propene) scintillation tubes (Beckman Instruments, Fullerton, CA, USA) and acidified to Sample Preparation below pH 2 with HNO3 to prevent analyte loss.9 The samples Samples were decomposed by heating with HNO3 to dissolve were then frozen at -20 °C until determination to minimize solid particles and oxidize organic compounds, which was not deterioration.achieved by adding acid to a cold sample. Once a sample had Solutions were prepared from water purified with a Milli-Q been evaporated to near dryness, it was reconstituted with system (Millipore, Bedford, MA, USA) and HNO3 purified water to about 0.8 times its original mass.Further concen- was with a sub-boiling still (Acidest, Heraeus Quartzschmelze, tration is not desirable as a 10 ml injection into the graphite Hanau, Germany). Palladium modifier (pro analysi quality) furnace gives an optimum signal-to-background absorbance. and NH4NO3 (ACS reagent quality) were supplied by Merck While larger injection volumes will increase the analyte signal, (Darmstadt, Germany).Seronorm reference material used was with the matrix present the gain in signal may be less than the low values urine, batch No. 101021 (Nycomed Pharma, that predicted for the greater analyte mass. The accompanying Oslo, Norway), with a Cd concentration of 0.35 mg l-1. This was reconstituted according to the manufacturer’s instructions and prepared as a normal sample. Samples were prepared in laminar flow clean benches (Ultramare, Stockholm, Sweden), which provide a Class 100 working environment. Aliquots of 2–5 g of thawed urine samples were weighed into quartz conical flasks and about three times the sample mass of HNO3 was added.The mixture was boiled until the flasks were almost dry; they were left to cool and about 0.8 times the sample mass of water was added and weighed. The mass ratio of the original and digested samples was taken to calculate the sample concentration. This procedure avoids the use of volumetric equipment, which could increase the risk of sample contamination.The densities of the digests were measured independently for some samples and were typically 0–2% higher than those of the original samples. Therefore, the error introduced by measuring sample mass rather than volume is negligible. Prepared samples were trans- Fig. 1 Diagram of the modified contact cones after Hadgu and ferred into 20 ml poly(propene) scintillation tubes and stored Frech.19 in the dark at 4 °C until use.Table 2 Performance comparison of furnace adaptations Instrumentation Atomizer Characteristic Detection limit in A Perkin-Elmer SIMAA 6000 spectrometer (Bodenseewerk configuration Light energy* mass/pg† urine/mg l-1‡ Perkin-Elmer, U� berlingen, Germany) was run in single element Standard cone 455 2.01 0.015 mode with a Perkin-Elmer Series 2 electrodeless discharge Open furnace lamp for Cd. The furnace temperature programme is given in Standard cone 431 1.20 0.013 End-capped Table 1.Baseline offset correction was applied for 2 s and furnace taken 3 s before atomization. Graphite tubes were open or Modified cone 452 1.28 0.018 fitted with end-caps (3 mm aperture, 2 mm length) and had an Open furnace Modified cone 425 0.89 0.008 End-capped Table 1 Furnace temperature programme furnace Temperature/°C Ramp time/s Hold time/s * Transmitted light energy in arbitrary units, recorded by the spectrometer. 90 1 10 † Characteristic mass was calculated by standard additions to a 120 90 30 urine sample, with a natural Cd content below the detection limit. 600 15 30 ‡ Detection limits are 3s of ten replicates of a urine sample containing 1300 0 6 0.059 mg l-1 Cd. 2300 1 2 492 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12rise in background absorbance will increase measurement error light attenuation by background species is reduced as these are now vaporized over a longer period of time and, more and thus deteriorate the detection limit.21 significantly, an increase in sensitivity is observed at lower temperatures as the atomic residence time is increased.The Drying latter effect is significantly enhanced by using an end-capped tube with grooved contact cones (Table 2). For a 30 ml injection volume, a slow temperature ramp as part With the standard design of the THGA, the base of the of a 2 min drying step is necessary to preserve the graphite injection port of the cone is in close proximity (0.5 mm) to the tube, which is prone to corrosion around the injection port tube.When a stream of inert sheath gas is passed through this under rapid drying conditions. space, it creates a volume of reduced pressure around the injection port of the tube. The difference in pressure forces Pyrolysis gases through the injection port of the tube and thus permits efficient removal of vapour during drying. On the other hand, Because Cd is a fairly volatile element, the possibility of with this type of cone, analyte atoms are removed more quickly complete removal of matrix constituents during pyrolysis is due to this convective flow, particularly at low atomization limited.In complex sample matrices such as urine, careful temperatures.19 For the grooved cone, the space around the optimization and temperature control of pre-treatment protoinjection port of the tube eliminates convective flows,19 thus cols, as well as the use of a suitable modifier, are therefore increasing sensitivity.required. Here, a mixture of Pd and NH4NO3 was added to The improvement in sensitivity gained by the removal of the sample as NH4NO3 facilitates removal of chlorides6,7 and convective flow is greatest when using end-capped tubes. While Pd is known to ensure the stability of Cd during pre-treatment end-capped tubes increase sensitivity by reducing the rate of at 600 °C. It was also important to apply a fairly slow ramp analyte diffusion, this increase is further enhanced by virtually during the pyrolysis step to prevent sputtering of the sample removing convection.constituents. After lower pyrolysis temperatures, the shape of the background absorbance during the atomization of Cd typically shows signals in two time domains (Fig. 2). By 1 teff = 1 tD +1 tC decomposing the urine in HNO3, the background obtained during the first 2 s is decreased significantly, indicating that Here teff, tD and tC are the effective, diffusional and convective this absorbance originates from organic compounds.mean atomic residence times, respectively.18 As tC tends to The spectrum for the later background signal correlated in high values, the residence time of atoms in the atomizer part to that of diatomic sodium chloride.6,22 This shows that becomes entirely dependent on diffusion. the reaction of the NH4NO3 modifier with the sodium chloride However, with the SIMAA 6000 instrument, end-capped in urine in the presence of Pd is not complete at the temperature furnaces reduce the light throughput, which can increase noise. used here.This agrees with the results of Yin et al.6 For The size of the end-caps, therefore, has to balance the increase optimum S/N, it was essential to keep the background as low in sensitivity with the increase in noise. Table 2 shows the as possible and to maintain the set pre-treatment temperature difference in light throughput and sensitivity for standard open accurately.There is always some uncertainty and variation and end-capped furnaces. While light throughput is not signifi- between the set and real pre-treatment temperatures. To avoid cantly affected, the detection limits from the two furnace types losses of Cd, the temperature chosen here was about 100 °C demonstrate the benefits of end-caps for this analysis. lower than the maximum permissible. While the detection limit can be reduced by improving the sensitivity of measurements, it has been shown14,21 that longer integrationtimes, a consequence of lower atomization tempera- Atomisation tures, will increase noise and thus deteriorate precision.The atomization temperature of 1300 °C is 100 °C lower than However, at higher atomization temperatures, the matrix the recommended value and gives two benefits. The maximum vapour causes light attenuation that coincides with the analyte absorbance peak.23 The lowered light transmission through the furnace will deteriorate the stability of the analyte baseline, increasing noise.L’vov et al.21 demonstrated that, by using a longer time for baseline offset correction (BOC) as well as a shorter time for absorbance peak integration, the S/N of measurements is improved and detection limitscan be considerably reduced. While increasing the BOC time may be adntageous, this setting cannot be changed on the SIMAA 6000 instrument. In summary, the possible deterioration by noise of the instrumental detection limit for Cd in urine, given by reduced light throughput due to end-caps and a longer atomization time, is more than compensated for by the increase in sensitivity brought about by increasing the atomic residence time.Method Validation To check the accuracy of the sample preparation method, recovery tests were performed by adding Cd to samples before acid digestion. Three samples with different concentrations were chosen for the test and three digestions were made of one sample (Table 3).The recovery values given are reasonably close to 100%, suggesting that Cd is completely recovered. Fig. 2 Zeeman corrected atomic absorbance and background signals The deviation seen is determined by instrumental precision at for Cd in Seronorm urine. Ashing temperature, 300 °C; atomization temperature, 1300 °C. low concentration, which will be discussed later. Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 493Table 3 Recovery tests Sample+addition/ Sample/mg l-1 mg l-1 Recovery (%) n 0.101 0.193 95.7±14.6 3 0.144 0.227 84.0 1 0.161 0.253 96.8 1 * Concentration values were determined by standard additions calibration; the number of repetitions is given in the final column and the uncertainty for the first sample is the standard deviation of the three sample digestions. Recovery tests were made by adding approximately 250 pg of Cd to approximately 2.5 g of each urine sample before digestion in acid.The precision of the digestion method was checked by repeating the digestion of five different samples (Table 4). The standard deviation of the two observations was calculated for each sample and the square root of the sum of the variances is given. The value of 0.003 mg l-1 is below the instrumental Fig. 3 Atomic peak area for masses of Cd introduced with urine. detection limit, which suggests that the digestion process and Percentage relative standard deviation for three replicates is given for each point.instrumental determination is repeatable. Seronorm reference material was used to check the accuracy The authors thank Bodenseewerk Perkin-Elmer for the loan of determination. It was prepared according to the manufacof the SIMAA 6000 spectrometer, Yngvar Thomassen of the turer’s instructions, then digested as a normal sample and Norwegian Institute of Occupational Health for supplying the analysed with a batch of samples (Table 5).On six occasions, Seronorm reference material and for valuable discussions and Seronorm was supplied without the analyst’s knowledge and Karin Olsson of Umea° University for assistance with labora- labelled as a normal sample. Both mean values given are tory work. within 5% of the recommended concentration, which suggests that determination is accurate at this level. Calibration by standard additions was used for the urine REFERENCES samples in this study; however, standard calibration may be 1 Stoeppler, M., Spectrochim.Acta, Part B, 1983, 38, 1559. used without a significant deterioration in accuracy. The 2 Herber, R. F. M., Stoeppler, M., and Tonks, D. B., Fresenius’ deviation of the characteristic mass between 20 samples on J. Anal. Chem., 1990, 338, 279. one day was found to be 6%, which is only slightly higher 3 Nordberg, G. F., and Nordberg, M., in Biological Monitoring of than that expected from the instrumental deviation.One T oxic Metals, ed. Clarkson, T. W., Friberg, L., Nordberg, G. F., and Sager, P. R., Plenum Press, London, 1988, pp. 151–168. addition of 0.2 mg l-1 is normally made to the sample. The 4 Diamond, G. L., in Biological Monitoring of T oxic Metals, ed. absorbance-to-mass of analyte relationship must be linear for Clarkson, T. W., Friberg, L., Nordberg, G. F., and Sager, P. R., this method of calibration. As a test for linearity, different Plenum Press, London, 1988, pp. 515–529. masses of Cd were pipetted into the furnace with a urine 5 Ediger, R. D., and Coleman, R. L., At. Absorpt. Newsl., 1973, 12, 3. sample showing no detectable Cd. Fig. 3 is the calibration 6 Yin, X., Schlemmer, G., and Welz, B., Anal. Chem., 1987, 59, 1462. produced from three replicates of ten standards up to 7 Smeyers-Verbeke, J., Yang, Q., Penninckx, W., and Vandervoort, F., J. Anal. At. Spectrom., 1990, 5, 393. 0.5 mg l-1. Evidently, the absorbance-to-mass relationship is 8 Fraile, R., de Benzo, Z.A., and Velosa, M., Fresenius’ J. Anal. linear, even below 0.010 absorbance, and gives a correlation Chem., 1992, 343, 319. coefficient of 0.999. To demonstrate instrumental precision, the 9 Halls, D. J., Black, M. M., Fell, G. S., and Ottaway, J. M., J. Anal. relative standard deviation is included on the plot. At. Spectrom., 1987, 2, 305. 10 Dube, P., Krause, C., and Windmu�ller L., Analyst, 1989, 114, 1249. 11 Harnly, J. M., and Radziuk, B., J.Anal. At. Spectrom., 1995, 10, 197. Table 4 Repeatability of digestions 12 Radziuk, B., Ro�del, G., Stenz, H., Becker-Ross, H., and Florek, S., J. Anal. At. Spectrom., 1995, 10, 127. Mean/mg l-1 ±/mg l-1 13 Radziuk, B., Ro�del, G., Zeiher, M., Mizuno, S., and Yamamoto, K., 0.039 0.001 J. Anal. At. Spectrom., 1995, 10, 415. 0.046 0.001 14 Harnly, J. M., Fresenius’ J. Anal. Chem., 1996, 355, 501. 0.096 0.006 15 Hadgu, N., Ohlsson, K. E. A., and Frech, W., Spectrochim. Acta, 0.123 0.001 Part B, 1995, 50, 1077. 0.130 0.001 16 Hadgu, N., Ohlsson, K. E. A., and Frech, W., Spectrochim. Acta, Part B, 1996, 51, 1081. Mean±/mg l-1 0.003 17 Hadgu, N., and Frech, W., Spectrochim. Acta, Part B, 1994, 49, 445. 18 L’vov, B. V., and Frech, W., Spectrochim. Acta, Part B, 1993, * Two aliquots of five different samples were digested and their Cd 48, 425. contents were determined. The mean of two observations is shown; 19 Hadgu, N., and Frech, W., Spectrochim. Acta, Part B, in the press. mean uncertainty is the square root of the sum of the variances. 20 Cornelis, R., Heinzow, B., Herber, R. F. M., Molin Christensten, J., Paulsen, O. M., Sabbioni, E., Templeton, D. M., Thomassen, Y., Vahter, M., and Vesterberg, O., Pure Appl. Chem., 1995, 67, 1575. Table 5 Quality control with Seronorm reference material* 21 L’vov, B. V., Polzik, L. K., Borodin, A. V., Dyakov, A. O., and Novichkhin, A. V., J. Anal. At. Spectrom., 1995, 10, 703. Run type Mean/mg l-1 ±/mg l-1 n 22 Culver, B. R., and Surles, T., Anal. Chem., 1975, 47, 920. Known 0.364 0.058 16 23 L’vov, B. V., Polzik, L. K., and Fedorov, P. N., Spectrochim. Acta, Blind 0.343 0.032 6 Part B, 1992, 47, 1411. * Results for Seronorm Batch 101021; recommended Cd concen- Paper 6/06150H tration 0.35 mg l-1. Shown are the mean, standard deviation and Received September 6, 1996 number of determinations. Determinations were made on different days; blind samples were supplied without the operator’s knowledge. Accepted January 20, 1997 494 Journal of Analytical Atomic Spectrometry, April 1997, V