JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 23 Use of Inductively Coupled Plasma Mass Spectrometry for the Determination of Ultra-trace Elements in Human Serum Hans Vanhoe and Richard Dams Laboratory of Analytical Chemistry University of Ghent Institute of Nuclear Sciences Proeftuinstraat 86 6-9000 Ghent Belgium Jacques Versieck Department of Internal Medicine Division of Gastroenterology University Hospital De Pintelaan 185 6-9000 Ghent Belgium A method for the determination of 11 ultra-trace elements (Li B Mo Cd Sn Sb Cs Ba Hg Pb and Bi) in human serum by inductively coupled plasma mass spectrometry is described. Sample preparation was kept to a minimum serum samples were diluted 5-fold with 0.14 mol I-’ HNO and suitable internal standards (Be In and TI) were added to correct for matrix effects and for ion signal instability. Special attention was given to optimization of the electrostatic lens settings and the nebulizer gas flow rate.Detection limits between 0.007 ng ml-‘ (for Bi) and 0.5 ng ml-’ (for B) could be obtained taking into account a 5-fold dilution of the serum sample. Memory effects which can be experienced with the conventional methodology for sample introduction leading to positive errors were reduced to a negligible level by the use of a short (2 min) clean-out procedure. With the exception of B (1-2 ng ml-’) and Pb (0.08-0.15 ng mi-’) blank levels were shown to be below 0.1 ng ml-‘. Results are given for a ‘second-generation’ biological reference material Freeze-Dried Human Serum (University of Ghent) and for Human Serum SRM 909 from the National Institute of Standards and Technology (for Li Cd and Pb).Finally serum samples from healthy individuals were analysed in order to determine typical element concentrations for normal human serum. Keywords Inductively coupled plasma mass spectrometry; ultra-trace element determination; human serum Because of its extreme importance for the human organism and easy accessibility human blood plasma or serum has been selected by many clinical and analytical scientists for the determination of trace and ultra-trace elements. The concen- trations of the trace elements to be determined vary from 10 pg ml-’ to less than 0.1 ng ml-’. Some of these are essential like Co and Mo or toxic like Hg and Pb and are therefore an interesting group for intensive investigation.In order to determine elements at the ultra-trace level (less than 10 ng ml-’) sensitive analytical methods are required. Neutron activation analysis (NAA) electrothermal atomic absorption spectrometry (ETAAS) inductively coupled plasma atomic emission (ICP-AES) and mass spectrometry (ICP-MS) and proton induced X-ray emission spectrometry (PIXE) have often been used for these purposes. So far radiochemical and instrumental neutron activation analysis (RNAA and INAA)’-’ and ETAAS+” are almost the only techniques that have been employed for the determination of ultra-trace elements in human serum.12 Since ETAAS is hampered by the occurrence of matrix effects and has a single-element character and because for NAA long waiting times or elaborate radiochemical separations are often involved the development of an alterna- tive analytical technique was desirable.Therefore the use of ICP-MS which combines the exceptional characteristics of the ICP for the atomization and ionization of the injected sample with the sensitivity and selectivity of MS was evaluated. Since its commercial introduction in 1983 ICP-MS has been exten- sively used in the determination of elements at the trace and ultra-trace level in biological material~.’~-~ Recently a review of the capabilities of ICP-MS for the determination of trace elements in body fluids and tissues was published.24 Besides the direct analysis of biological certified reference materials by laser ablation ICP-MS2’ and the use of flow injection (FI) for the analysis of undiluted urine,” all appli- cations devoted to biological fluids involve more or less extensive sample preparation prior to analysis.Special care has to be taken during sample preparation to minimize con- tamination of the samples and losses of analytes by volatiliz- ation adsorption or precipitation. A first sample preparation method is based on the destruction of the organic material present in the sample. Most often digestion with one or more acids e.g. HNO HClO and/or H2S04 is used. Disadvantages of this technique are the possible introduction of significant blanks from the acids and the loss of volatile elements such as Hg. The latter problem can be avoided by the use of high-pressure bombs or microwave oven digestions.This procedure has been applied for the analysis of urine,” human blood plasma2c28 and human blood by ICP-MS.28-31 A suitable alternative is ashing of the biological material. Lyon et a!.’’ and Wang et ~ 1 . ~ ~ applied this method to urine whereas Serfass et ashed blood plasma at 480°C before the measurement of Zn-isotope ratios. Finally Smith et a!.34 reported the determination of B in blood plasma after a sodium carbonate fusion and separation of B from the matrix compo- nents using a selective ion-exchange resin. A second method that limits the sample preparation is dilution of the biological fluid with a suitable diluent. Although this method necessarily introduces a deterioration of the detection limits accurate determination of several elements is still possible.Mulligan et aZ.35 showed the capability of ICP-MS for the analysis of urine after 10-fold dilution with 0.14 mol I-’ HNO (for the determination of Cd Sb and Hg). Ba~mann,~ and Vanhoe et a1.37 described the rapid and accurate determination of I in milk after dilution with 1% v/v ammonium solution whereas Dean et demonstrated the accurate determination of Pb isotope ratios in milk after an approximately 40-fold dilution with 0.1 YO Triton X-100 solution. Another biological fluid that has already been analysed after simple dilution is blood. A roughly 25-fold dilution with a 5% Triton X-100 solution appeared to be sufficient to determine accurately both the total Pb content and the Pb isotope In order to minimize contamination and losses of analyte in this laboratory an attempt was always made to limit the sample preparation to a 5-fold dilution with 0.14 moll-’ HNO,.From the composition of human serum it is obvious that apart from the water and the protein content there is a considerable salt content with Na K and Ca as easily ionizable elements and S and C1 as major interfering elements. The salt24 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 content is equivalent to 9 mg ml-' of NaCl or 0.9%. This can lead to signal suppression of up to more than 3O-40Yi4' In order to correct for these matrix effects and for signal instabil- ity several internal standards (Be Co In and T1) were added to the sample solution. In addition some proteins may deposit in the central part of the torch and salts may clog the sampling cone or the nebulizer.Lyon et a/.'' analysed synthetic protein solutions after a 10-fold dilution and achieved reasonable results for Al Cr Mn Fe Ni Cu Zn Mo and Ba. Besides for Mo (at a concentration of about 7 ng ml-') the concen- tration of the elements mentioned ranged from 80 ng ml-' (for Ba) to 7 pg ml-' (for Fe). Park et determined Cr Cu and Pb in National Institute for Standards and Technology (NIST) Standard Reference Material (SRM) 909 Human Serum after a 10-fold dilution. It should be realized that the concentration of elements in the serum samples mentioned above are signifi- cantly higher (up to a factor of 1000) than expected in normal human serum. In order to evaluate the potential of ICP-MS for the determination of ultra-trace elements in human serum a serum reference material with trace element levels comparable to those found in normal human serum was employed namely a 'second-generation' biological reference material Freeze- Dried Human Serum identical to the material used for the determination of Fe Co Cu Zn Rb," Sr,43 Br44 and I.37 This reference serum was prepared by Versieck et a1.45 under rigorously controlled conditions in order to avoid the addition of extraneous species.Experimental Instrumentation The ICP mass spectrometer used was a VG PlasmaQuad (VG Elemental Winsford Cheshire UK) equipped with a Gilson-2 peristaltic pump a Meinhard concentric glass nebulizer (type TR-30-A3) a Scott-type double pass spray chamber with surrounding liquid jacket made of borosilicate glass and a Fassel-type torch.Details of the operating conditions are summarized in Table 1. Reagents and Standards High-purity water was obtained with a Millipore Milli-Q water purification system (resistivity of 18 MR cm). Concentrated nitric acid (14 moll-') was purified by sub-boiling distillation in a quartz still using analytical-reagent grade nitric acid (Pleuger) as feedstock. External calibration using single-element standard solutions with a concentration of 0.5 1 5 and 10ngml-' to produce Table 1 ICP-MS operating conditions ~~~~ ~~ Stage Parameter Plasma R.f. power Forward Reflected Gas flows Plasma Intermediate Nebulizer Peristaltic pump Nebulizer Spray chamber Ion sampling Sampling cone Skimmer cone Sampling depth Vacuum Expansion stage Intermediate stage Analyser stage 1.35 kW <low Conditions 13 1 min-' 1 1 min-l Variable Minipuls 2 (Gilson) pumped at 0.9 ml min-' Meinhard Tr-30-A3 concentric glass nebulizer Double-pass Scott type water-cooled (10 "C) Nickel 1.0 mm orifice Nickel 0.75 mm orifice 10 mm (from load coil) 2.4 mbar 1.0 x mbar 4.0 x lop6 mbar single-element calibration curves was employed to calculate the concentration of the analyte elements.Special attention was given to the purity stability and accuracy of the standards used. Table 2 gives for each element the product used for the preparation of the standard solutions. The single-element standard solutions with concentrations between 0.5 and 10 ng ml-' were freshly prepared before each analysis sequence. For the internal standards use was made of commerically available AAS standard solutions Be and In (Janssen Chimica) and T1 (Alfa Products).Sample Preparation Blood samples from healthy subjects working in this labora- tory were collected and processed according to the sampling protocol developed at the institute and described in detail by Versieck and Cornelis.12 Briefly in order to avoid significant contamination at the sampling stage blood was taken with a polypropylene intravenous catheter mounted on a metallic needle [Intranule 110 16 (Vygon)] and collected in high-purity quartz tubes with stoppers made of poly(tetrafluoroethy1ene) (PTFE). No anticoagulant was added. After collection samples were immediately placed in a thoroughly cleaned plastic box and transported into a clean laboratory where all further sample handling was performed.After clotting serum was separated by centrifugation (3500 rev min-' for about 30 min) and decanted into polyethylene screw-cap containers. Afterwards the serum samples were stored at -25 "C prior to analysis. After defrosting and homogenization with a quartz stirring- spoon (the whole process took about 2 h) 5 ml of liquid serum taken with a polyethylene pipette (Kartell) were transferred into a 25 ml polyethylene calibrated flask. After addition of 2.5 ml of a multi-element solution (100 ng ml-' of Be In and T1 used as internal standards) the solution was adjusted to volume with 0.14 moll-' HNO,. A blank solution was pre- pared in the same way as the serum solutions but without the addition of sample. For lyophilized serum the form in which the 'second-generation' biological reference material Freeze- Dried Human Serum is available first the sample was reconsti- tuted with Millipore Milli-Q water in a PTFE beaker ( 6 ml of water for 500 mg of sample) before being quantitatively trans- Table 2 Standards used for the analysis of human serum Element Standard Li B Mo Li,C03 (powder) IRM-016 (CBNM Geel) solvent H,O; diluent 0.14 moll-' HNO H,BO (powder) IRM-011 (CBNM Geel) solvent and diluent 0.14 moll-' HNO AAS standard solution (Johnson Matthey) solvent 5% HCl and traces of HF diluent 0.1 moll-' HCl solvent 14 mol I-' HNO,; diluent 0.14 mol I-' HNO solvent 1:l HNO (14 moll-') HCl(l0 moll-'); diluent 1 moll-' HC1 (first step) 0.1 moll-' HCl diluent 0.1 mol I-' HC1 diluent 0.14 moll-' HNO solvent and diluent 0.14 mol 1-' HNO solvent 0.7 moll-' HNO and 0.025% K,Cr,O,; diluent 0.14 moll - HNO solvent 0.14 moll-' HNO solvent 14 moll-' HNO,; diluent 0.14 moll-' HNO Cd Sn Cd metal (Goodfellow Metals 99.95%) Sn wire (Goodfellow Metals 99.99%) Sb Cs Ba Hg Hg metal AAS standard solution (Alfa Products) AAS standard solution (Alfa Products) Ba(NO,) (powder) (Merck analytical reagent grade) Pb Bi Pb( NO,)z (UCB analytical-reagent grade) Bi metal (Vieille Montagne 99.999%)JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL.9 25 fered into a calibrated flask with the diluent (0.14moll-1 HNO,). All manipulations were performed on a clean-bench. Human Serum SRM 909 was reconstituted with diluent water as described in the NIST certificate. A dilution factor of 5 (for Cd and Pb) or 50 (for Li) was used by diluting respectively 5 ml or 500 p1 of the reconstituted solution to 25 ml with 0.14 moll-' HN03.Optimization of the Instrumental Parameters In order to obtain maximum signal intensity necessary for the determination of elements at ultra-trace levels in human serum several parameters have to be optimized. Both gas flow rates and electrostatic lens settings were studied in more detail. Nebulizer gasflow rate It is generally accepted that the signal response is strongly dependent on the nebulizer gas flow rate. At one particular radiofrequency (r.f.) power each M+ signal shows a maximum intensity at a certain nebulizer gas flow rate. In addition as described by Vanhaecke et the optimum nebulizer gas flow rate is mass dependent with the heavier elements having a lower optimum nebulizer flow rate at a particular r.f. power.Therefore in order to determine elements at the ultra-trace level it is necessary to use different nebulizer gas flow rates depending on the analyte elements as shown in Fig. 1. For the analysis of human serum three nebulizer gas flow rates were employed 0.7801min-1 for the light elements (Li and B) 0.725 1 min-l for the mid-mass elements (Mo Cd Sn Sb Cs and Ba) and 0.700 1 min-' for the heavy elements (Hg Pb and Bi). It is worth mentioning that the intermediate gas flow rate shows a less pronounced effect and was set at 1 1 min-' (Table 1). Lens voltages In addition to the nebulizer gas flow rate the lens voltages can significantly influence sensitivity.In the optimization study a different behaviour was noticed for various nuclides as illustrated in Fig. 2 which shows the relation between the normalized ion signals for 'Be "CO "'In and 238U and the voltage on lens 2 which is positioned behind the photon stop. Firstly all lenses were adjusted to obtain a maximum signal intensity for 1151n. Afterwards all voltages were kept constant except the voltage on lens 2. From Fig. 2 it is clear that the ion signals for the four nuclides do not behave in the same way. These observations are in agreement with the conclusions of Schmit and Chtaib,47 Vaughan and H~rlick,~' and Tanner49 who reported that there is a mass dependency of the trajectories that the ions describe through the electrostatic lens system.Therefore before each analytical sequence the lens settings 700 I 600 = 500 600 700 800 900 Nebulizer gas flow rate/ml min-' Fig. 1 A 9Be; B "'In; and C '05Tl Influence of the nebulizer gas flow rate on the ion signal for 0' I I 1 . 1 1 I I 1 I -160 -140 -120 -100 -80 -60 -40 -20 0 20 Lens 2 voltageN Fig.2 Influence of the lens 2 voltage on the ion signal for A 9Be; B 59C0; C Ir51n; and D 238U were optimized in order to obtain a maximum ion signal intensity for the internal standard used. Choice of Internal Standard In previous work18,43,44 it was reported that the signal intensity is influenced by the serum matrix. In addition it was noticed that the heavy elements are suppressed to a larger extent than the lighter elements for 4 mg ml-' of NaCl the signal suppres- sion increases from 33% for 7Li to 56% for 238U.Therefore in order to obtain accurate results a method must be developed to correct for these matrix effects which are mainly caused by easily ionizable elements such as Na K and Ca. The method applied was based on the use of internal standardization. Because the signal suppression was established to be mass dependent the choice of suitable internal standards was exam- ined. Serum solutions which were diluted 5-fold were spiked with a multi-element standard solution (each element at a concentration level of 100 ng ml-') and compared with a blank solution containing the multi-element solution in 0.14 moll-' HNO,. The results are summarized in Table 3 the experimental recoveries range from 98.7 to 102.6% using a suitable internal standard.It can be concluded that a suitable internal standard with a mass close to that of the analyte element accurately corrects for matrix effects in the case of human serum. For that reason three internal standards (at a concentration level of 10ngml-') were employed beryllium (9Be) for the light elements indium ("'In) for the mid-mass elements and thallium ('''Tl) for the heavy elements. Moreover the internal standard corrects also for instability of the ion signal so that a relative standard deviation (RSD) on the results of 4% or better can be expected in cases with sufficient counting statistics. Method of Analysis and Calculations Several experiments showed that when the measuring time is increased an improvement in the detection limit proportional to the square root of the increase of the measuring time can Table 3 Results of recovery experiments Nuclide 7Li 1°B 9 8 ~ ~ '"Cd lz0Sn lZ1Sb 133cs 138Ba 202Hg zosPb 209~i Internal standard 9Be 9Be lIsIn '"In 1 1 5 1 ~ 1 1 5 1 ~ "51n 1 1 5 1 ~ 2 0 5 ~ 1 2 0 5 ~ 1 20sT1 Recovery( YO) 102.6 -t 2.7 99.6 & 2.5 101.0+ 1.4 99.6 k 2.5 98.7 k 2.6 101.4+ 1.3 99.4 & 0.8 100.8 & 1.9 102.0k2.1 100.2k0.9 99.2 f.2.026 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 Table 4 Scanning conditions for the ultra-trace element analysis of human serum; internal standards used are given in parentheses Elements Mass range/u Dwell time,@ Channels Sweeps Li B (Be) 4-1 1 320 512 400 Cd Sn Sb (In) 109-124 320 512 400 Hg Pb Bi (Tl) 199-21 0 320 512 400 Mo (In) 96-1 17 320 512 400 Cs Ba (In) 113-140 320 1024 200 be expected. For this reason to perform an ultra-trace element analysis of human serum it is necessary to limit the number of analytes determined simultaneously to a few elements with comparable masses.Therefore the mass range was divided into several parts consisting of 20-30 u (Table 4). In order to obtain a correct integration of the peaks each peak must consist of at least 20 channels. Hence 512 or 1024 channels with a dwell time of 3 2 0 p were used so that short-term fluctuations on the ion signal intensity were largely eliminated. Finally the mass range was scanned 200 times (for 1024 channels) or 400 times (for 512 channels) so that one measure- ment lasted about 1 min. Five replicates were made on each solution.Memory effects which are encountered for elements such as Li,50 B,51 Mo," Sn Hg and Bi,52 were reduced to a negligible level by the use of the following analysis sequence first a blank for the samples was measured then several serum samples next a blank for the standard and only at the end of the sequence several single-element standard solutions with con- centrations up to a maximum of 10ngml-'. Moreover the sample introduction system was rinsed for at least 2 min with 0.14 mol 1-' HNO after the measurement of each solution and the blank level was controlled with a rate meter. Finally before each analysis sequence the sampling and skimmer cones were cleaned and all parts of the sample introduction system were leached with concentrated nitric acid. For each solution (blank sample standard) the signal (peak area integrated over 0.8 u around the peak maximum) of each nuclide was normalized to the signal of the internal standard.The mean and standard deviation of the five resulting nor- malized signals of each solution were calculated. The average normalized signal of the blank was subtracted from that of the serum solutions. External calibration (calibration curve) was employed to calculate the corresponding concentrations. Results and Discussion Blank Levels and Detection Limits In order to determine very low concentrations it is necessary that the blank values are as low as possible. The elevation of the blank signal can have several origins. One of these is contamination of the reagents used for sample preparation. For that reason use was made of de-ionized water which was purified by a Millipore Milli-Q system so that a resistivity of 18 MR cm was obtained.In addition nitric acid was purified by sub-boiling distillation to remove impurities present in the feedstock. Table 5 gives the concentration levels for the analyte elements found in a blank solution consisting of 0.14 moll-' HNO,. Except for B and Pb all concentrations are below 0.1 ng ml-'. Experiments showed that most of the boron is present in the Millipore Milli-Q water whereas the sub-boiled nitric acid contains traces of Sn Ba and Pb. The blank signal for Hg originates from impurities present in the Ar gas. For the other elements the blank signals observed are due to the continuous background present over the whole mass range.Since the background signal is stable over several hours (Fig. 3) it is possible to correct for the contribution of the background which is not negligible for ultra-trace element determinations by measuring a blank solution. Table 5 gives detection limits (34 a blank solution was measured ten times) obtained with the scanning conditions Table 5 Blank levels and detection limits Element (nuclide) Li ('Li) B ('OB) Mo ('*Mo) Cd (l14Cd) Sn ('"Sn) Sb (121Sb) c s (l33Cs) Ba ('38Ba) Hg (202Hg) Pb ('08Pb) Bi ("'Bi) Blank value/ ng ml-' 0.02-0.04 1-2 0.01-0.03 0.01-0.02 0.02-0.06 0.01-0.02 0.005-0.015 0.02-0.05 0.04-0.08 0.08-0.15 0.001-0.003 Concentration in Detection limit*/ human serum/ ng i d - ' ng ml-l 0.05 1 0.5 5-50 0.04 0.3-1.2 0.02 0.1-0.2 0.05 0.03 < 0.5 0.03 1 0.2 0.1-4.8 0.1 0.02-4 0.007 <0.1 0.40-0.64 0.04 ?( < 10) * Taking into account a 5-fold dilution of the serum sample.100 UJ v) 3 0 L m C v) U C 3 P Y m 4-l (J 10 .- 2 m r T I IJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 27 already summarized in Table4 and compares these with the element concentration expected in normal human serum. The detection limits vary from 0.007 ngml-' (for '09Bi) to 0.5 ng ml-' (for 'OB) taking into account a 5-fold dilution of the serum sample. For Li B Mo Sn Cs and Ba the detection limit is at least a factor of 10 lower than the concentration in human serum whereas for Cd Sb Hg Pb and Bi the serum concentration is in some cases lower than the detection limit. In order to check the accuracy and precision of the method developed two serum reference materials were analysed namely a 'second-generation' biological reference material Freeze-Dried Human Serum (University of Ghent) and Human Serum SRM 909 (from NIST).Results for the 'second- generation' biological reference material Freeze-Dried Human Serum are given in Table 6 whereas Table 7 gives results for Human Serum SRM 909 (for Li Cd and Pb). As an application serum samples from healthy individuals were analysed. The concentration levels obtained are given in Table 8. The results for each element are discussed below in more detail. Lithium Because of the lower detection limit and the smaller influence of variations of the isotopic composition of Li determinations were made using the most abundant isotope 'Li (92.5%). They resulted in a mean Li concentration of 15.04 ng g-' (corre- sponding to 1.37 ng ml-' for serum obtained after reconsti- tution of the lyophilized form). The RSD on the measurements varies from 2.3 to 4.2% whereas the contribution of the blank signal to the total signal in human serum amounts to a maximum value of 13%.The result obtained is in good agreement with that of Ab~u-Shakra~~ (using ICP-MS). The value using ETAAS54 (no background correction) is somewhat higher. The results for Human Serum SRM 909 (Table 7) are in good agreement with the certified value. It should be noted that the concentration of Li in this reference serum is about a factor of 8000 higher than that of the 'second-generation' biological reference material Freeze-Dried Human Serum.For normal human serum a range between 0.25 and 0.87 ng ml-' with a mean Li concentration of 0.60 ng ml-' was found. Literature values mainly obtained by ETAAS range from 0.2 to 44 ng ml-'. Since the concentrations of Li in human serum are low and higher Li values were only obtained with less sensitive techniques such as flame AAS probably the normal concentration of Li is less than 5 ng ml-' (a mean of 1 ngml-') which is in agreement with the result obtained here. Determinations of Li in human serum and other biological materials have been discussed in more detail in a previous p ~ b l i c a t i o n . ~ ~ Boron In order to avoid the overlap from the intense ''C' peak with the "B' peak,51 'OB (20%) was used for the determination of B.A mean B concentration of 227 ng g-' was obtained (corresponding to 20.6 ng ml-' for liquid serum). This result is in excellent agreement with that of Ab~u-Shakra~~ (using ICP-MS). Although the contribution of the blank signal to the total signal in human serum is relatively high (between 20 Table 6 Results (ng g- ' dry mass*) for the 'second-generation' biological reference material Freeze-dried Human Serum. Values can be recalculated into ng ml-' of original liquid serum by dividing them by 11 i.e. loo/( 1.025 x 8.87) (1.025 being the density of the fresh serum and 8.87 the mean percent residue after lyophilization). Values in parentheses are the standard deviation n = s 'Li ICP-MS Individual results 15.32( 0.64) 15.03 (0.34) 15.09(0.39) 14.73 (0.59) 15.04 f0.39 19.25 f0.55 ( ETAAS)54 14.8 f 2.8 ( ICP-MS)53 Mean f 95% confidence limits Certified or literature values ICP-MS individual results Mean f 95% confidence limits Certified or literature values ICP-MS individual results '14Cd 2.05( 0.47) 2.93( 0.24) 1.65(0.38) 2.22( 0.24) 2.21 f 0.85 138Ba 9.77( 0.45) 13.3 ( 1 .O) 12.06( 0.5 1 ) 9.12(0.30) Mean k 95% confidence limits Certified or literature values 11.1 f3.1 'OB 217.7( 6.8) 241 (10) 21 8.5 ( 3.6) 230.0( 7.7) 227 f 18 222 6 ( ICP-MS)53 '18Sn 9.97(0.51) 9.88( 0.93) 11.3( 1.1) 8.84( 0.30) 10.81 (0.67) 8.93(0.47) 9.71(0.46) 9.92 10.83 10 f 2.6 (NAA)' 7.6 & 1.4 7.96 f 0.97; 8.38 f0.48 (NAA)57 '"Hg 6.77( 0.40) 6.87(0.54) 6.64( 0.50) 5.66(0.75) ( ICP-MS)54 6.49 & 0.89 6.6 f0.445 9 8 ~ ~ 6.89( 0.69) 8.06( 0.73) 7.63(0.81) 7.46( 0.68) 7.51 f0.77 7.5 k 0.845 "*Sn 9.65(0.46) 10.38( 0.61) 10.90( 0.90) 8.69(0.91) 10.83 (0.77) 8.67(0.45) 9.30( 0.44) 9.77 f 0.88 'O'Hg 6.43( 0.80) 5.67(0.41) 5.44( 0.7 1 ) 7.7( 1.2) 6.3 -t 1.6 "'Cd "'Cd "'Cd 2.67( 0.43) 1.96(0.63) [ 3.24(0.88)]* 1.73(0.72) 2.56( 0.32) 1.92( 0.5 1) 1.64( 0.6 1 ) 2.55( 0.39) 1.57( 0.53) 2.45(0.95) 1.83( 0.74) 1.87( 0.68) 2.12 -t 0.8 1 2.23 f 0.61 1.79 f 0.47 2.0( 1.7-2.5)45 '"Sb 123Sb 133cs 0.89( 0.37) 1.07 (0.46) 10.88( 0.83) 0.61 (0.40) 0.83 (0.33) 10.23 (0.3 1) 1.04(0.26) 0.70( 0.36) 9.76( 0.44) 0.93( 0.25) 0.44( 0.30) 9.9 1 (0.27) 0.87 & 0.29 0.76 f 0.42 10.20 f 0.79 0.25f0.15 10.0 f 2.345 (NAA)' ZOgBi '08Pb 42.62(0.66) 0.708( 0.096) 40.9( 1.2) 0.61 3( 0.064) 48.1(2.5) 0.773( 0.048) 44.6( 2.1) 0.683 (0.039) 49.66( 0.77) 52.2( 2.9) 46.3 & 4.6 0.69 & 0.1 1 * Outlier6'28 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL.9 Table 7 Results (mean & 95% confidence limits) for Human Serum SRM 909 Element ICP-MS Certified value Lithium (pg m1-l) Cadmium (ng ml-') 1 1.45( 1 1.1 8-1 1.80) 'Li 11.66 k 0.28 'I2Cd 1.124 & 0.090 'I4Cd 1.19 f 0.1 3 '"Pb 18.98f0.52 1.24( 1.15- 1.34) Lead (ng ml- ') 20.0( 17.9-22.5) Table 8 Ultra-trace element concentrations (ng ml-') in normal human serum Element Li B Mo Sn Sb c s Ba Pb Bi Hg Concentration 0.60 0.20( 0.25-0.87) 13.9 6.9(4.1-25.8) 0.61 f 0.17( 0.30-0.86) 1.02 f 0.26(0.66-1.46 j 0.137 & 0.028( 0.095-0.165 j 0.70f 0.12(0.50-0.96) 1.04 & 0.65(0.23-2.35) < 0.20-0.71 < 0.10-0.71 < 0.007-0.067 No.of persons 12 12 12 12 6 12 16 12 12 19 Literature values 0.2-44 8.3-48.1 0.28-1.17 0.4-350 0.01-3.1 0.45-2.06 < 30-1900 0.05-4.8 0.02-14.5 0.1-6.0 and 33%) it does not limit the RSD on the measurements (between 1.6 and 4.1%) as the blank signal originates mainly from impurities present in the reagents and is therefore fairly stable. For normal human serum a range between 4.1 and 25.8 ng ml-' with a mean B concentration of 13.9 ng ml-' was found which is in good agreement with literature values (8.3-48.1 ng ml-I). Besides ICP-MS the only technique that can determine accurately the low levels of B in human serum is NA-MS.55 The determination of B in human serum and other biological materials has been discussed in more detail in a previous p~blication.~' Molybdenum Determinations of Mo were made at m/z 98 Cg8Mo (24.1%)].It must be emphasized that neither 95Mo (15.9%) nor 97Mo (9.5%) can be used as both nuclides suffer from spectral overlap respectively from 79Br160 + and 40Ar39K160+ 9 a nd ''Brl60+ and 40Ar41K'60+.56 A mean Mo concentration of 7.51 ng g-' was obtained (corresponding to 0.68 ng ml-' for liquid serum) which is in excellent agreement with the certified value. For the certification only NAA was used. The RSD on the measurements is about lo% mainly owing to the important contribution of the blank signal to the total signal in human serum (between 7 and 20%). The RSD is however comparable with that obtained by NAA. For normal human serum a range between 0.30 and 0.86 ng ml-' with a mean Mo concentration of 0.61 ngml-' was found which is in agreement with the value of 0.6 ng ml-' reported by Versieck and Cornelis.'2 Cadmium Cadmium has eight stable nuclides lo6Cd (1.25%) '"Cd (12.2%) 'I4Cd (28.7%) and '16Cd (7.5%).Some of these are interfered with isobarically '12Cd is interfered with by '12Sn (0.90/) "'Cd (12.5%) "'Cd (12.8%) '12Cd (24.1%) '13Cd (0.97%) '13Cd by '131n (4.3%) '14Cd by '14Sn (0.65%) and '16Cd by '%n (14.5%). Since In is used as an internal standard the use of 'I3Cd is excluded. In addition as the Sn concen- tration in human serum is between 0.40 and 0.64 ng ml-1,57 the contribution of Sn to the Cd signals at m/z 112 114 and 116 in human serum is not negligible. Therefore '16Cd cannot be used. Determinations were made at m/z 110 111 112 and 114 (Table 6). In general it can be stated that for all four nuclides good agreement with the certified value is obtained.The RSD on the measurements is relatively high (between 8.2 and 38.8%) mainly due to the important contribution of the blank signal to the signal for Cd in serum (between 30 and 60%). For '12Cd and '14Cd the isobaric overlap from Sn was corrected for by calculating the contribution of the Sn nuclides at m/z 112 and 114 respectively using the signal intensity for 12'Sn and the natural isotopic abundances for the Sn isotopes involved. Additional proof of the accurate correction of the isobaric overlap with the Sn nuclides is given by the analysis of Human Serum SRM 909 (Table 7). It should be noted that the concentration of Cd in this reference serum is about a factor of 7 higher than that found for the 'second-generation' biological reference material Freeze-Dried Human Serum (0.18 ng ml-' for liquid serum).The latter is in agreement with the concentration expected in normal human serum.12 Versieck and Vanballenberghe4 reported a range between < 0.105 and 0.192 ng m - ' of Cd. In order to perform accurate determinations of Cd it is necessary to get maximum sensitivity (at least a count rate of 3 x lo5-5 x lo5 for 100 ng ml-I of In). This was however not always attainable and varied from day to day. For that reason determinations of Cd in human serum could not always be carried out and therefore no study on the concentration of Cd in normal human serum was performed. Several studies are underway in order to overcome this problem. Flow injection whereby undiluted serum can be measured and electrothermal volatilization with extremely low absolute detection limits are the most promising techniques.In addition the use of a modified high-performance interface was evaluated With this interface a sensitivity of up to 2 x lo6 (for 100 ng ml-' of In) can be reached. Tin Determinations were made at m/z 118 C118Sn (24.2%)] and 120 ["'Sn (32.6%)] 9.925033 ng g-' for "'Sn and 9.77 +OX8 ng g-' (95% confidence limits) for 12'Sn (approxi- mately 0.9ngml-' for liquid serum). These results are in reasonable agreement with the values obtained by ICP-MS and NAA. The RSD of the measurements varies from 3.4 to 9.7% whereas the contribution of the blank to the total signal in human serum is between 10 and 25%.For normal human serum a range between 0.66 and 1.46 ng ml-' with a mean Sn concentration of 1.02 ng ml-' was found. Literature values for the concentration of Sn in human serum are rather scarce and are high in comparison with the results obtained here. They range from <40 to 350 ng ml-'. Recently Versieck and Vanballenberghe57 reported a mean Sn concentration in human serum of 0.505 ng ml-' with a standard deviation of 0.096 ng ml-' (range 0.400-0.636 ng ml-') which is comparable with the results obtained in the present study. Probably the concen- tration of Sn in human serum has been overestimated in the past similar to that of V Cr and M o . ~ ~ Antimony Determinations of Sb were made at m/z 121 [12'Sb (57.3%)] and 123 [123Sb (42.7%)] 0.87k0.29 ngg-' for "'Sb and 0.76k0.42 ng g-' (95% confidence limits) for 123Sb (approxi- mately 0.075 ngml-' for liquid serum).The RSD of theJOUKNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 29 measurements is high (between 25 and 70%) due to the important contribution of the blank to the total signal in human serum (up to 60%) and to the fact that the concentration found is between the detection limit (3s 0.03 ng m1-l) and the determination limit ( ~ O S ' ~ 0.1 ng ml-'). The results can only be compared with those obtained with NAA. Since both results have a great uncertainty nothing can be said about the agreement. Literature data on the concentration of Sb in normal human serum are scarce. Versieck and Cornelis" mentioned a range between 0.5 and 1 ng ml - of Sb.More recently Minoia ef a/.'' reported a range between 0.01 and 3.1 ng ml-' of Sb which is in agreement with the range found in this work (0.095-0.165 ngml-' of Sb). It can be concluded that the possibility of performing accurate determinations of Sb in human serum depends on the actual sensitivity. As already mentioned previously this can vary from day to day. Caesiun1 Since Cs is mono-isotopic determinations were made at nil- 133 (133Cs) resulting in a mean Cs concentration of 10.20 ng g - ' (corresponding to 0.93 ng ml-' for liquid serum) which is in excellent agreement with the certified value. For certification only NAA was used. The RSD of the measurements varies from 2.7 to 7.6%. whereas the contribution of the blank to the total signal in human serum is between 2.7 and 7.6%. For normal human serum a range between 0.50 and 0.96 ng m-' with a mean Cs concentration of 0.70ngml-' was found which is in agreement with the concentration expected for normal human serum (about 1 ngml-').'2 It should be noted that up to now only NAA could be used for the determination of Cs in human serum. B m ii4ni Determinations of Ba were made at m/z 138 [13*Ba (71.7%)] resulting in a mean Ba concentration of 11.1 ngg-' (corre- sponding to 1.01 ng ml- for liquid serum).The RSD of the measurements varies from 3.3 to 7.5% whereas the contribution of the blank to the total signal in human serum is between 8 and 20%. No certified or indicative values are available. For normal human serum a range between 0.23 and 2.35 ng ml-' with a mean Ba concentration of 1.04 ng ml-' was found.Literature values on the concentration of Ba in human serum mainly obtained by NAA are rather scarce and are high compared with the results obtained here. They range from <30 up to 1900ng ml-'. Some of these values are a factor of 1000 higher than the range found in this work. The concentration of Ba in human serum has probably been overestimated in the past. hlercwj Since the concentration of Hg in human serum is very low determinations could only be made at m/z 200 [200Hg (23.1 YO)] and 202 ["'Hg (29.8%)]. The results obtained with both nuclides do not differ significantly from each other and are in good agreement with the certified value. The RSD of the measurements is relatively high (between 5.9 and 15.6%) mainly due to the important contribution of the blank to the total signal in human serum (between 25 and 40%).The analysis of sera from 12 healthy subjects provided a result for only four cases 0.35 (2 x) 0.54 and 1.35 ng ml-I. For the other cases the signal measured did not differ signifi- cantly from the blank signal. Literature data range from 0.05 to 4.8 ng m1-l. It is clear that in some cases the concentration of Hg is between the detection limit (3s 0.20 ng ml-') and the determination limit (lOs 0.60 ng ml-I) so that determinations of Hg in normal human serum are not always possible. L e d Determinations of Pb were made at mlz 208 ['O*Pb (52.4%)] resulting in a mean Pb concentration of 46.3 ng g - ' (corre- sponding to 4.21 ng ml-' for liquid serum).The RSD on the measurements varies between 1.5 and 5.6% whereas the contri- bution of the blank to the total signal in human serum is between 9 and 15%. No certified or indicative values are available. In order to check the accuracy of the proposed method Human Serum SRM 909 with a certified Pb content was analysed (Table 7). The results obtained are in good agreement with the certified value. Analysis of sera from 12 healthy subjects resulted in a detection limit (0.10 ng ml-') for seven cases. In addition the highest concentration of Pb found was about 0.7 ng ml- ' which is a factor of 6 lower than that in the 'second-generation' biological reference material Freeze-Dried Human Serum. Literature values for the concentration of Pb in normal human serum mainly obtained by ETAAS and isotope dilution MS are rather scarce and contradictory.12 They range from 0.02 to 14.5 ng ml-'.Contamination with Pb from the air reagents etc. can introduce an important error when low concentrations of Pb have to be determined. Bism ti t h Since Bi is mono-isotopic determinations were made at rn/z 209 ('09Bi) resulting in a mean Bi concentration of 0.69 ng g - ' (corresponding to 0.063 ng ml-' for liquid serum). The RSD of the measurements varies from 5.7 to 13.6% whereas the contribution of the blank to the total signal in human serum is between 7.5 and 20%. No certified or indicative values are available. Analysis of sera from 19 healthy subjects results in a detection limit (0.007 ng m1-l) in ten cases.The highest concen- tration found was 0.067 ng ml-'. Literature data which are scarce range from 0.1 to 6 ng ml-'. Normal concentrations of Bi and levels after the intake of a therapeutic dose of colloidal bismuth subcitrate have been discussed in more detail in a previous p~blication.~' Conclusion It has been shown in this work that apart from Fe Co Cu Zn Br Rb Sr and I ICP-MS is able to determine the following ultra-trace elements in the 'second-generation' biological refer- ence material Freeze-Dried Human Serum Li B Mo Cd Sn Sb Cs Ba Hg Pb and Bi. For Cd Sb and to a lesser extent Bi some reservations must be made. The actual sensitivity is the limiting factor for these elements. Studies are underway to overcome this problem by using FI or electrothermal volatiliz- ation instead of pneumatic nebulization as a means of sample introduction.Extrapolation of these conclusions to normal human serum is not applicable to all the elements because the concentration for some of these e.g. Pb is significantly higher in the reference serum than in normal human serum. Moreover the concentration of these elements in normal human serum is not far above the detection level so that in some cases only a detection limit is obtained. If accurate and precise determinations are to be performed maximum sensitivity must be attained. This situation can be achieved by optimization of the nebulizer gas flow rate and the electrostatic lens settings using the signal of the internal standard. In addition for the determination of ultra-trace elements only a small mass range can be scanned so that the measuring time is used optimally for each element.Finally an internal standard with a mass close to that of the analyte must be used in order to correct for matrix effects. Furthermore,30 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 since the contribution of the blank is not negligible in the determination of ultra-trace elements in human serum only the most pure reagents can be employed. For Li Mo Sn Cs Ba and Bi the contribution of the blank was shown to be between 10 and 20% whereas for B Cd Sb Hg and Pb this could rise to more than 30%. The origin of the blank signal varies from element to element. Apart from the continuous background which is more or less constant over the whole mass range impurities present in the reagents or the argon gas (for B Sn Ba Hg and Pb) contribute to the observed blank signal. Accurate correction is possible with a blank solution containing 0.14 moll-' HN03 and the internal standard.It can be concluded that with the method developed here which includes a simple and short sample preparation followed by rapid determination with a high sample throughput a great number of elements that can be determined by ETAAS INAA or RNAA can also be determined by ICP-MS. The elements with a mass between 40 and 80u form an exception V Cr Mn Co Ni and As are spectrally interfered with and cannot therefore be determined accurately after a simple dilution of the serum samples6 "V suffers from interference from 35C1160+ 52Cr from 35C1160H+ and 40Ar'2C+,55Mn from 37C1180 + and 39K160+ 59C0 from 43Ca160+ and 36Ar23Na+ 60Ni from 23Na37C1+ and 75As from 40Ar35C1+. A more extensive sample preparation including separation of the matrix or the use of alternative sample introduction systems such as electrothermal volatilization is necessary.An example is the determination of As in human serum after the separation of C1 by an anion exchanger.61 As can be deduced from this work ICP-MS in combination with the optimized procedure yields for some elements such as Li Sn Ba Pb and Bi concentrations in normal human serum which until now have been more or less unknown. Therefore ICP-MS is a good alternative and/or a complementary technique to NAA and ETAAS. Grateful acknowledgement is made to L. Vanballenberghe for her contribution to the preparation of the serum samples.Thanks are also due to C. Vandecasteele for his interest in the work and to F. Vanhaecke for evaluating the manuscript. The PlasmaQuad was acquired by a grant from the Fund for Medical Scientific Research (FGWO). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 References Versieck J. Hoste J. Barbier F. Vanballenberghe L. De Rudder J. and Cornelis R. Clin. Chim. Acta 1987 87 135. Versieck J. Hoste J. Barbier F. Steyaert H. De Rudder J. and Michels H. Clin. Chem. 1978 24 303. Cornelis R. Versieck J. Mees L. Hoste J. and Barbier F. J. Radioanal. Chem. 1980 55 35. Versieck J. and Vanballenberghe L. in Trace Elements in Man and Animals eds. Mills C. F. Bremmer I. and Chesters J. K. 1985 vol. 5 Commonwealth Agricultural Bureaux Farnham Royal p.650. Xilei L. Van Renterghem D. Cornelis R. and Mees L. Anal. Chim. Acta 1988 211 231. Halls D. J. and Fell G. S . Anal. Chim. Acta 1981 129 205. Veillon C. Patterson K. Y. and Bryden N. A. Anal. Chim. Acta 1984 164 67. Lewis S. A. O'Haver T. C. and Harnly J. M. Anal. Chem. 1985 57 2. Andersen J. R. Gammelgaard B. and Reimert S. Analyst 1986 111 721. Slavin W. J . Anal. At. Spectrom. 1986 1 281. Ericson S. P. McHalsky M. L. and Jaselskis B. Talanta 1987 34 271. Versieck J. and Cornelis R. Trace Elements in Human Plasma or Serum CRC Press Boca Raton Florida 1988. McLaren J. W. Beauchemin D. and Berman S . S. Anal. Chem. 1987 59 610. Beauchemin D. McLaren J. W. Willie S. N. and Berman S . S. Anal. Chem. 1988 60 687. 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 Lyon T.D. B. Fell G. S. Hutton R. C. and Eaton A. N. J . Anal. At. Spectrom. 1988 3 265. Ridout P. S. Jones H. R. and Williams J. G. Analyst 1988 113 1383. Satzger R. D. Anal. Chem. 1988 60 2500. Vanhoe H. Vandecasteele C. Versieck J. and Dams R. Anal. Chem. 1989,61 1851. Friel J. K. Skinner C. S. Jackson S. E. and Longerich H. P. Analyst 1990 115 269. Schmit J.-P. Youla M. and Gelinas Y. Anal. Chim. Acta 1991 249 495. Wiederin D. R. Smyczek R. E. and Houk R. S. Anal. Chem. 1991 63 1626. Lyon T. D. B. Fell G. S. McKay K. and Scott R. D. J. Anal. At. Spectrom. 1991 6 559. Gelinas Y. Youla M. Beliveau R. Schmit J.-P. and Ferraris J. Anal. Chim. Acta 1992 269 115. Vanhoe H.J. Trace Elem. Electrolytes Health Dis. 1993 in the press. Ward N. I. Durrant S. F. and Gray A. L. J. Anal. At. Spectrom. 1992 7 1139. Abou-Shakra F. R. Havercroft J. M. and Ward N. I. Trace Elem. Med. 1989 6 142. Tothill P. Matheson L. M. Smyth J. F. and McKay K. J. Anal. At. Spectrom. 1990 5 619. Vaughan M.-A. Baines A. D. and Templeton D. M. Clin. Chem. 1991 37 210. Lasztity A. Viczian M. Wang X. and Barnes R. M. J . Anal. At. Spectrom. 1989 4 761. Nygren O. Vaughan G. T. Florence T. M. Morrison G. M. Warner I. M. and Dale L. S. Anal. Chem. 1990 62 1637. Viczian M. Lasztity A. and Barnes R. M. J. Anal. At. Spectrom. 1990 5 293. Wang X. Lasztity A. Viczian M. Israel Y. and Barnes R. M. J. Anal. At. Spectrom. 1989 4 727. Serfass R. E. Thompson J.J. and Houk R. S. Anal. Chim. Acta 1986 188 73. Smith F. G. Wiederin D. R. Houk R. S. Egan C. B. and Serfass R. E. Anal. Chim. Acta 1991 248 229. Mulligan K. J. Davidson T. M. and Caruso J. A. J. Anal. At. Spectrom. 1990 5 301. Baumann H. Fresenius' J. Anal. Chem. 1990 338 809. Vanhoe H. Van Allemeersch F. Versieck J. and Dams R. Analyst 1993 118 1015. Dean J. R. Ebdon L. and Massey R. J. Anal. At. Spectrom. 1987 2 369. Delves H. T. and Campbell M. J. J. Anal. At. Spectrom. 1988 3 343. Campbell M. J. and Delves H. T. J. Anal. At. Spectrom. 1989 4 235. Vandecasteele C. Vanhoe H. and Dams R. J. Anal. At. Spectrom. 1993 8 781. Park C. J. Smith D. C. and Vanloon J. C. Trace Elem. Med. 1990 7 103. Vandecasteele C. Vanhoe H. Vanballenberghe L. Wittoek A. Versieck J.and Dams R. Talanta 1990 37 819. Vandecasteele C. Vanhoe H. Dams R. and Versieck J. Anal. Lett. 1990 23 1827. Versieck J. Vanballenberghe L. De Kesel A. Hoste J. Wallaeys B. Vandenhaute J. Baeck N. Steyaert H. Byrne A. R. and Sunderman F. W. Anal. Chim. Acta. 1989 204 63. Vanhaecke F. Vandecasteele C. Vanhoe H. and Dams R. Mikrochim. Acta 1992 108 41. Schmit J. P. and Chtaib M. Can. J. Spectrosc. 1987 32 56. Vaughan M. A. and Horlick G. Spectrochim. Acta Part B 1990 45 1301. Tanner S. D. Spectrochim. Acta Part B 1992 47 809. Vanhoe H. Vandecasteele C. Versieck J. and Dams R. Anal. Chim. Acta 1991 244 259. Vanhoe H. Vandecasteele C. Versieck J. and Dams R. Anal. Chim. Acta 1993 281 401. Vanhoe H. Versieck J. Vanballenberghe L. and Dams R. Clin. Chim. Acta in the press. Abou-Shakra F. R. 1990 personal communication. Baumann H. 1989 personal communication. Iyengar G. V. Clarke W. B. and Downing R. G. Fresenius' J . Anal. Chem. 1990 338 562.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JANUARY 1994 VOL. 9 31 56 Vanhoe H. Goossens. J.. Moens L. and Dams R. J . A d . At. Spectrom.. submitted for publication. 57 Versieck J. and Vanballenberghe L. Anal. Chem. 1991 63 1143. 58 Versieck. J. and Cornelis R. Anal. Chim. Acta 1980 116. 217. 59 Currie L. A. A n d . Chem. 1968 40 586. 60 Minoia. C. Sabbioni E.. Apostoli P.! Pietra R. Pozzoli L. Gallorini. M.. Nicolaou G. Alessio L.. and Capodaglio E. SL'I'. Totd Eiwiron. 1990 95 89. Goossens J. Moens L. and Dams R. J . Anal. At. Specrrom.. 1993 8 921. 62 Miller J. C. and Miller J. N. Stutisticsfor Analytical Chemistry Wiley New York 1984 Ch. 3. pp. 59-62. Puper 3/028#8G Receiued MUJ? 20 1993 Accepted Augtcst 11 1993 61