Mikrochim. Acta 131, 145±151 (1999)
Contribution to the Development of Indirect Atomic Absorption Methods: Application of the Ion Pair 1,10-phenanthroline-mercury(II)-iodide to Iodide Determination in Water and Infant Formulae Samples Pilar Bermejo-Barrera , Rosa Ma. Anllo-SendõÂn, Manuel Aboal-Somoza, and Adela Bermejo-Barrera University of Santiago de Compostela, Department of Analytical Chemistry, Nutrition and Bromatology, Faculty of Chemistry, Avda. de las Ciencias, s/n, E-15706-Santiago de Compostela (La CorunÄa), Spain
Abstract. A method to determine iodide is developed, based on the formation of an ion pair between 1, 10phenanthroline, mercury(II) and iodide that can be selectively extracted into IBMK. When the IBMK layer is analyzed by electrothermal atomic absorption spectrometry (ETAAS) for mercury, iodide can be quanti®ed too. Parameters related to the preparation of the ion pair and to the measuring process are studied. Thus, 7.2±7.4 reveals to be the best pH interval to carry out the extraction, and 250 and 1000 C, respectively, the mineralization and atomization temperatures for mercury determination by ETAAS. The procedure is applied to drinking tap water and commercial infant formulae milk samples. The accuracy of the method has been investigated by means of the analytical recovery for water samples (mean analytical recovery obtained at different concentration levels 98.1%) and by using the certi®ed reference material BCR CRM No 151 Skim Milk Powder (Spiked) for the infant formulae (results within certi®cation interval). The repeatability of the measurements at different concentration levels gave a RSD lower than 10% for both types of samples and the repeatability of the whole procedure for infant formulae shows a RSD of 1.33%. In addition, the limits of detection and quanti®cation were 2.5 mg/L and 8.5 mg/L, respectively, for drinking water and 1.1 mg/g and 3.8 mg/g, respectively, for infant formulae. Key words: indirect ETAAS; iodide; ion pair; drinking tap water; infant formulae. To whom correspondence should be addressed
The dif®culties in determination of non-metal by conventional atomic absorption spectrometry (AAS) have been widely described in the literature, and these problems led to the development of indirect AAS procedures. Some of them are based on the previous preparation of a compound (complex or ion pair) that contains the non-metal and an element easily measurable by AAS. For iodine, the formation of a compound that contains iodide is the most usual [1±4], though sometimes other species such as iodate have been used [5]. AAS can be used with different atomization modes depending on the analyte to be studied. For indirect iodine methods that use mercury as the metal to be determined, the cold vapour has been the most common mode, but the ¯ame and the graphite furnace modes can also be used [1]. The interest in iodine determination in food samples is due to the fact that iodine is an essential trace element for humans, and a de®ciency as well as an excess in its intake can produce well-known pathologies. In water, iodine is present as iodide, but in milk there are different opinions: Some authors state that iodine is present naturally in milk entirely as iodide [6]; for others, most of iodine is dissolved in the whey, a small amount bound to proteins and about 2% included in the fat globules [7]. The of®cial AOAC method proposes iodine determination in milk as iodide [8]. The ,0 -diimine group (±N=C±C=N±) enables 1,10-phenanthroline (``phen'') to form stable ®vemembered-ring chelate ions with a variety of metallic cations, such as Cu(I and II), Co(II), Fe(II and III),
146
Hg(II), etc. Thus, for Hg(II), two chelate ions are possible with phen, [Hg(phen)2]2 and [Hg(phen)3]2 , with stability constants of 19.65 and 23.35, respectively (expressed as log k and calculated at 20 C and 0.1) [9]. [Fe(phen)3]2 , [Co(phen)3]2 , [Ni(phen)3]2 , and [Cu(phen)3]2 show stability constants of 21.0, 19.8, 24.3 and 21.0, respectively (expressed as log k and obtained at 25 C and 0.1) [9, 10]. Because these chelate ions are positively charged, they can form ion pairs with different anions, for example bromide, iodide or iodate. In the present work, the formation of an ion pair between the chelate ion phen-mercury and iodide, its extraction into IBMK and the measurement of mercury by ETAAS is proposed. The procedure is applied to iodide determination in drinking tap water and in infant formulae milk samples. Experimental Apparatus The measurements were performed by means of a Perkin-Elmer 1100B atomic absorption spectrometer, equipped with an HGA700 graphite furnace atomizer and an AS-70 autosampler. The measurements were carried out at 253.7 nm, using a mercury electrodeless discharge lamp operated at 4 W (lamp and power supply by Perkin-Elmer). The background corrector was a deuterium lamp, and 0.7 nm was the spectral bandwidth of the monochromator. Pyrolytic graphite coated graphite tubes with pyrolytic graphite (L'vov) platforms (also by Perkin-Elmer) were used throughout the work. Reagents Potassium iodide. ACS reagent grade. Sigma Chemical, St. Louis, MO, USA. Mercury (II) nitrate standard solution. Containing 1.000 0.002 g/L and prepared in 1 M HNO3 solution. Panreac-Montplet & Esteban, Barcelona, Spain. 1,10-phenanthroline monohydrate. P.a. Scharlau, FEROSA, Barcelona, Spain. To prepare solutions, solid reagent was dissolved in the smallest possible volume of methanol and then, ultrapure water was employed to made up the solutions to their ®nal volume. Palladium powder. 99.999%. Aldrich Chemical, Milwaukee, WI, USA. Nitric acid. Suprapur. BDH Chemicals, Poole, UK. Acetic acid. 96% Merck, Darmstadt, Germany. Sodium acetate trihydrate. Probus, Badalona, Spain. Sodium diethyldithiocarbamate. Analytical reagent grade. BDH. Ammonium dihydrogen orthophosphate. Analytical reagent grade. BDH. Diammonium hydrogen orthophosphate. ACS reagent grade. Sigma. Isobutyl methyl ketone (IBMK). Analytical reagent grade. BDH. Sodium carbonate. Merck. Argon N50. I.e., 99.9990% purity (sheating gas for the atomizer). SEO, Madrid, Spain. Methanol. Gradient grade. Merck.
P. Bermejo-Barrera et al. Ultrapure water, resistivity 18 M cm, obtained by means of a Milli-Q Water Puri®cation System. Millipore, Bedford, MA, USA. Due to the fact that IBMK dissolves plastic autosampler cups, the use of PTFE cups was necessary. In addition, to avoid contamination, all material (glass-, plastic- and PTFE-ware) was washed and kept for at least 48 h in 10% (v/v) nitric acid solution, and then rinsed several times with ultrapure water before use. Samples The samples analyzed were drinking tap water from several town supplies located in Galicia (NW Spain) and commercial infant formulae samples. Procedure for Preparation and Extraction of the Ion Association Complex First, aliquots of the following solutions ± in the order given ± are dispensed into a glass tube: Hg (II) solution, phen solution, KI solution and NH4H2PO4/(NH4)2HPO4 buffer solution (which provides the ®nal solution with a pH between 7.2 and 7.4). The volume of the resulting aqueous solution is made up to 5 mL with ultrapure water. Then, 2 mL of IBMK are added and the tube is gently shaken for 30 s. Afterwards, the organic upper layer containing the complex is removed by means of a Pasteur pipet, to be ®nally transferred to an autosampler cup and subjected to ETAAS. Pretreatment Procedure for Infant Formulae Samples In a porcelain cup, 0.2 g of infant formulae are weighed and 1 g of Na2CO3 and 1 mL of a 6 M NaOH solution, as well as 10 mL of MeOH are added. The cup is allowed to dry in a heater at about 100 C, to remove any moisture and alcohol (for avoiding sample ignition). After this drying step, the cup is placed inside a cold oven and the temperature is then rised slowly up to 500 C (the door of the oven is opened from time to time, to avoid violent sample combustion, what would easily imply losses of analyte). Once the sample is white, the calcination is considered to be completed. The cup is then cooled down in a desiccator to room temperature. Then, the cup is placed in a sand bath and the sample redissolved with 10 mL of hot ultrapure water, what is followed by a ®ltration of the solution obtained and subsequent made up of the ®ltrate to 25 mL with ultrapure water. Finally, a 5 mL-aliquot of this solution is neutralized to give a pH between 7.2 and 7.4 and the resulting neutralized solution is added to suitable amounts of mercury and phen to carry out the solvent extraction above described for water samples, the extracts being subjected to AAS.
Results and Discussion Study of the Graphite Furnace Temperature Programme The graphite furnace temperature programme obtained in a previous work [11] about indirect iodide determination by ETAAS, was taken as a start point. To carry out the study of the best conditions for each step (i.e. ramp and hold times, temperatures and purge gas ¯ow), an extract was prepared by following the
Contribution to the Development of Indirect Atomic Absorption Methods
described procedure, and for obtaining the most suitable temperatures and times, the usual way of working in graphite furnace AAS was followed. Moreover, to evaluate the effect of a matrix modi®er, Pd, was added to the sample extracted into IBMK and the programme of temperatures was again studied. The preparation of that matrix modi®er, implies the extraction of Pd diethyldithiocarbamate (stoichiometry: Pd(DDTC)2 [12]) into IBMK [11]. The results of this studies are described below. The drying step was divided into two, one at 100 and the other at 110 C, with ramp and hold times of 5 and 15 s, respectively (for both sub-steps). This is enough to remove all the solvent present, although it shows a boiling point of about 117 C. The best ashing temperature obtained was 250 C: at higher temperatures, the analyte losses are considerable. 10 and 15 s were the ramp and hold times selected for this step, respectively. The fact that the mineralization temperature considered now as optimal is slightly higher than that proposed before [11], 200 C, can be explained on the basis of the relative stability of the complexes formed. With phen, Hg(II) forms more stable chelates than with 2,20 -dipyridyl (``dipy''), as revealed by the values of the corresponding stability constant: 19.65 and 23.35 (for [Hg(phen)2]2 and [Hg(phen)3]2 , respectively) before the 16.7 and 19.5 (for [Hg(dipy)2]2 and [Hg(dipy)3]2 , respectively [9]), being all these data expressed as log k and obtained at 20 C and 0.1. Moreover, phen exhibits a melting point of 117±120 C, whereas dipy only remains solid up to 70±73 C. When studying the atomization step, a 1 s-ramp was selected, and this revealed to be a good choice to attain an adequate atomization of the analyte and integration of the absorbance peak. In addition, 8 s was the hold time obtained for this step, and the absorbance was integrated for 9 s (1 s-ramp plus 8 s-hold). Finally, 900 C was the atomization temperature considered as optimal within the temperature interval studied. However, 1000 C is a more practical choice, mainly because at 900 C after many runs of the graphite tube the atomization of the analyte seemed to be incomplete. This was shown by ``tailed'' AA peaks. This did not happen with new tubes or tubes having been used for only few runs. To avoid this dependence on the use of the tubes, 1000 C is the temperature proposed for this step since the described differences in atomization and in the peaks obtained did not occur at this temperature. Moreover, this value does not
147
Fig. 1. Mineralization and atomization curves with (curves A and B, respectively) and without (curves C and D, respectively) the use of matrix modi®er
induce analyte losses, because 900 and 1000 C showed the same signal, as shown in Fig. 1, explained below. Finally, a 1500 C-cleaning step was added to remove any residue from the tube before the next injection is performed. 1 and 3 s were the ramp and hold times selected for this step. The purge gas ¯ow, Ar, was also studied in each step, and the best results were achieved with maximum ¯ow (namely 300 mL/min) in every step except for the atomization which required no Ar ¯ow to carry out the measurements adequately. Background correction was used throughout. In Fig. 1, the ashing and atomization curves obtained with and without the use of modi®er are shown. Higher signals are obtained with Pd but, since the ashing and atomization temperature are the same and the signal recorded without Pd is high enough, it was considered that the use of Pd was not essential. So it was discarded in order to simplify the procedure. In view of the analytical recovery results described below, it can be concluded that the use of matrix modi®er is not essential. Table 1 summarizes the graphite furnace temperature programme proposed. pH In¯uence Among the parameters concerned with the formation of the complex, pH is of primary importance due to the acid-base properties of phen. Moreover, pH is also
148
P. Bermejo-Barrera et al.
Table 1. Graphite furnace temperature programme Step
Temperature ( C)
Dry 1 100 Dry 2 110 Mineralization 250 Atomization 1000 Clean
Ramp (s)
Hold (s)
Ar ¯ow (mL/min)
5 5 10 1
15 15 15/20
300 300 300 0 (read, delay: 0 s) 300
1500
8
1
3
of the NH4H2PO4/(NH4)2HPO4 buffer solution. For pH lower than 6.3 a great turbidity was observed in solutions. Therefore only two pH values lower than 6.3 were investigated. In addition, the Hg(II)-phen chelate, is formed at slightly basic pH values, and this led us not to study pH values much higher than 8. The results obtained when measurements were carried out following the conditions described, are shown in Fig. 2. The background absorption signals were also recorded for each point. It can be seen that between 7.2 and 8.1 the absorbance signal exhibits a quite constant value, but at pH above 7.4 the corresponding background absorption signals are too high for the deuterium correction system (they are higher than 0.5 units). Thus, 7.2±7.4 was the pH interval proposed for the best formation and extraction of the complex as well as adequate recording of the ETAAS signal.
The programme is exactly the same for water and for milk samples, except for the mineralization hold time: 15 s for water samples and 20 s for milk samples.
Stability of the Complex To evaluate the stability of the complex formed, a series of 8 extracts was prepared and measured the same day and also the next day (the extracts were conserved in a fridge overnight). Each extract was analyzed several times each day, and Table 2 shows the mean and the RSD values of the measurements for each extract and day, as well as the percentage of increase or decrease of the signal between both days. It can be seen that the differences observed between the means corresponding to the same extract are not important, since though some extracts show more than a 5% decrease in the signal, others show variations lower than 1%, being the mean variation of signal considering the eight extracts ÿ1.7%, what can be considered as acceptable. In addition, the RSD of the measurements of each extract does not vary signif-
Fig. 2. In¯uence of pH on the recorded absorbance (&) and background absorption (*)
relevant in extraction processes. In order to ®nd out the best pH interval, a series of extracts was prepared from the corresponding aqueous solutions in which pH varied from 6.0 to 8.4, by using adequate amounts
Table 2. Stability of the complex Extract
mean (n 3) RSD(%) Second day mean (n 3) RSD (%) Variation of signal with respect to the ®rst day (%)
First day
1
2
3
4
5
6
7
8
0.168 3.6 0.159 2.6 ÿ5.4
0.170 4.1 0.161 2.8 ÿ5.3
0.176 1.4 0.168 1.6 ÿ4.5
0.161 0.4 0.164 1.8 1.9
0.174 0.9 0.174 0.7 0.0
0.163 1.6 0.161 0.6 ÿ1.2
0.160 3.4 0.161 0.9 0.6
0.162 2.8 0.163 2.0 0.6
Mean absorbance of three runs for each extract. The RSD values are also related to those three values.
149
Contribution to the Development of Indirect Atomic Absorption Methods
icantly from the ®rst to the second day. Therefore, the extracts obtained can be supposed to exhibit at least, a two-day stability. Application to the Determination of Iodide in Drinking Water
the sensitivity of the method: characteristic mass (m0), limit of detection (LOD) and limit of quanti®cation (LOQ), de®ned as follows: m0 V C0:0044=
Asample ÿ Ablank LOD 3 SD=slope
LOQ 10 SD=slope
Standard calibration and standard addition graphs. To investigate the linearity of the method, two calibrations were performed: One of them using aqueous standards and the other applying the standard addition method (in both cases, with the corresponding extractions). In the latter, drinking tap water was employed as sample. Fig. 3 represents the results obtained. Both series of results were ®tted to straight lines by using the least squares method, giving the following lines:
(where V stands for volume of sample, C for concentration of sample, A for absorbance, SD for standard deviation and slope for the slope of the standard addition line). Performing the suitable measurements, the mean characteristic mass obtained within the linear range was 19.1 pg of iodide. The values obtained for LOD and LOQ were 2.5 mg/L of iodide, and 8.5 mg/L of iodide respectively, and referred to the water sample.
Aqueous standards : A 1:8 10ÿ3 C
Accuracy. The accuracy of the method was evaluated by studying the analytical recovery, since no certi®ed reference material was available. The analytical recoveries (%) achieved for 5, 10, 15 and 20 mg/L of iodide added to a water sample were 84.9, 102.4, 108.2 and 96.8, respectively. The mean analytical recovery is 98.1% (with an RSD of 10.1%), showing an acceptable value, given the complexity of the whole procedure.
Standard addition :
A 5:3 10ÿ3 4:2 10ÿ3 C
(where A is integrated absorbance and C is concentration of iodide in mg/L) The slopes of these lines are rather different, what was also supported when the t-test was used to compare both slopes statistically: a signi®cant difference between them does exist. Then, any quantitative measurement should be performed by using the standard addition method. The linearity of the method extends at least up to 20 mg/L of iodide. Precision. In order to evaluate the precision of the method, a study of repeatability was carried out at different levels of iodide concentration. Thus, starting from a drinking water sample, a series of extracts was prepared after adding to the aqueous solution adequate amounts of iodide solution to give 0, 5, 10, 15 and 20 mg/L of iodide. Five extracts were prepared for each of the ®ve levels of iodide concentration. At least, each extract was analyzed twice. The values obtained, expressed as meanRSD(%), for 0, 5, 10, 15 and 20 mg/L of iodide were 0.0634.3, 0.0815.4, 0.1065.7, 0.1294.1 and 0.1433.6, respectively. The RSD values achieved are acceptable (mean RSD is 4.61%). Sensitivity. A ®rst idea of the sensitivity of the method is given by the slope of the plot of signal vs concentration. In our case, this slope is that of the standard addition line: 4:2 10ÿ3 L/mg of iodide. Moreover, other three parameters were studied to investigate
Application. Finally, the method was applied to the determination of iodide in 6 drinking tap water samples proceeding from different cities in Galicia (NW Spain). The results obtained were between 60.0 and 154.5 mg/L of iodide. Regarding these iodide levels, they show an adequate agreement with those obtained previously by a different indirect procedure. However, these values can not be compared with maximum allowed levels, since there is not any legal normative in Spain referred to the subject. Application to the Determination of Iodide in Infant Formulae First, an attempt was carried out directly with the liquid milk sample to perform the formation and extraction of the ion pair. Unfortunately, when shaking it was completely impossible to distinguish the two usual phases due to the sample matrix, and a method of sample pretreatment was mandatory previous to the extraction. The main problem when removing organic matrix to determine volatile elements such as iodine, is the possibility of analyte losses during sample decomposition. Therefore, a
150
P. Bermejo-Barrera et al.
careful control of temperature is essential. The procedures proposed in the literature to mineralize samples usually involve an acid digestion or an alkaline calcination. The of®cial method by the AOAC [13] for iodine determination in mineral mixed feeds involves an alkaline calcination of the sample. To adapt this procedure to the infant formulae milk samples, a study of the NaOH and Na2CO3 amounts was performed. Study of the amounts of NaOH and Na2CO3. A series of calcinations was carried out varying the amounts of each base. Thus, 6 possible mixtures were tested: 0.5 g0.5 mL, 1 g1 mL, 2 g2 mL, 3 g3 mL, 4 g4 mL and 5 g5 mL (g Na2CO3mL 6 M NaOH). For each of the six conditions, four sample calcinations (sample size 0.2 g) were performed, as well as two reagent blanks. These calcinations and the corresponding extractions and measurements were carried out as describe above and each extract was measured at least twice. The result obtained, Table 3, show that there are no signi®cant differences among the various conditions employed. Moreover, the RSD values are quite acceptable. Although with the mixture 0.50.5 the highest absorbance values were attained, to ensure a complete calcination in every type of samples, the mixture 11 was selected. An important problem were the high blank values, and a study to reduce them was performed. Studies to reduce the blank values. Some experiments were carried out for reducing the blank values. The proposed calcination: 1 g Na2CO3 and 1 mL 6 M NaOH and the use of only NaOH, for a sample and a reagent blank were compared. The results obtained were similar, but the use of Na2CO3 was necessary especially when the neutralization to 7.2±7.4 was performed, due to the formation of a buffer solution that facilitates that neutralization considerably. In addition, a comparative study using various NaOH, supplied by different
manufacturers, was carried out and no signi®cant differences were obtained. Finally, some studies were carried out to check the cup materials for the high blank values. In the literature, the use of crucibles or cups for alkaline calcinations is widely described. Dolezal et al. [14] point out that the corrosion of the crucibles or cups is the cause of dif®culties in trace element analysis because some elements are adsorbed onto the crucible or cup surface, while others are released by the surface. The result of this kind of ion exchange processes (due to the extreme conditions the cups or crucibles are subjected to), is the cups or crucibles show their surfaces corroded when the calcination is ®nished, and this can be the reason for the high blank values. In order to overcome this disadvantage, some tests were performed by using and comparing the results obtained when nickel crucibles (the most recommended in literature) and porcelain cups were used. In each case, an experiment with a reagent blank and a calcination of a sample was carried out. The results obtained, starting from 0.2 g of sample and carrying out the corresponding calcinations and extractions as already described, were 0.183 when using nickel crucibles and when using porcelain cups for the calcination, but the signals recorded for the reagent blanks were 0.261 and 0.120, respectively. Therefore, the employment of porcelain cups was the choice selected. Standard calibration and standard addition graphs. Standard calibration and standard addition were compared, as for the water samples, and the equations of both lines are: Aqueous standards :
A 2:2 10ÿ3 2:2 10ÿ3 C
Standard addition :
A 5:3 10ÿ3 3:2 10ÿ3 C
(where, again, A is integrated absorbance and C is concentration of iodide in mg/L)
Table 3. Variation of absorbance with the amounts of NaOH and Na2CO3 Absorbance recorded using the x g y mL of Na2CO3 and 6 M NaOH, respectively
Mean RSD (%) Blank
0.50.5
11
22
33
44
55
0.184 2.5 0.158
0.154 2.4 0.138
0.150 9.3 0.121
0.129 4.6 0.116
0.126 9.8 0.089
0.117 0.5 0.104
n 4, the four means of the data recorded for each of the four extracts measured. n 2, the two means of the data obtained for each of the two reagent blanks.
151
Contribution to the Development of Indirect Atomic Absorption Methods
The t-test was again used to compare the slopes of both lines, leading to conclude that the quantitative measurements must be performed again by means of the standard addition method. In addition, the response is linear up to iodide concentrations of about 35 mg/L. Precision. To study the precision of the method, the fact was considered that the repeatability studies above described (which were performed with water samples) were applicable in this case, since those studies were carried out with the extracts. Therefore, the precision studies were now devoted to the whole procedure. Thus ®ve 0.2 g-aliquots of powder milk were subjected to the procedure developed and analyzed for iodide. The value obtained for RSD, 1.33%, reveals the method is precise enough. Sensitivity. Similarly to the case of water samples, the slope of the standard addition line, 3:2 10ÿ3 L/mg provides an idea of the sensitivity of the procedure developed. The m0, LOD and LOQ were also calculated, as for the water samples, giving a mean m0 within the linear range of 29.4 pg of iodide, bigger than that achieved for water, and this can be attributed to the higher matrix complexity of the infant formulae. The values achieve for LOD and LOQ were, respectively, 3.7 and 12.5 mg/L, which when referred to the infant formulae sample are 1.1 and 3.8 mg/g, again respectively. Accuracy. To study the accuracy a certi®ed reference material from BCR was used, BCR CRM No 151 Skim Milk Powder (Spiked), with a reported certi®ed iodine content of 5.3500.014 mg/g. The value achieved with the present method was 5.300.99 mg/g of iodide. The certi®ed value is within the iodide concentration interval obtained. Application. When, the method was applied to six commercial infant formulae samples, and the iodide levels obtained were lower than the LOD of the method, 1.1 mg/g, and only in one sample the concentration found was 1.2 mg/g of iodide.
Conclusions In this paper, the use of the ion pair formation as an approach to perform an indirect non-metal determination by ETAAS is shown. Iodine determination is possible by measuring the mercury content in the phen-Hg(II)-iodide ion pair extracted into IBMK. Although other ions such as Ni(II), Co(II), Tl(III) or Fe(II and III) could form similar chelate ions, the levels of these metals in the studied samples are not high enough to produce interferences, as revealed by the accuracy achieved in the method. This can be explained with the high excess of chelating agent that enables the formation of the Hg(II) ion pair. Moreover, the possible interferences in the mercury measurements are controlled by the use of the standard addition procedure. The proposed method allows iodide determination in different matrices, such as drinking tap water and infant formulae milk, achieving precise and accurate results. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
T. Nomura, I. Karasawa, Anal. Chim. Acta 1981, 126, 241. A. Kuldvere, Analyst 1982, 107, 1343. A. M. Wi¯adt, W. Lund, R. Bye, Talanta 1989, 36, 395. D. Chakraborty, A. K. Das, At. Spectrosc. 1988, 9, 189. D. Chakraborty, A. K. Das, Talanta 1989, 36, 669. E. Renner (Ed.), Micronutrients in Milk and Milk-Based Food Products, Elsevier, Barking (Essex), 1990, p. 28. P. Walstra , R. Jenness ``QuõÂmica y FõÂsica LactoloÂgica'', Acribia, Zaragoza, 1987, p. 126. AOAC, ``AOAC Of®cial Methods of Analysis 1995'', AOAC, Washington DC, 1995, p. 29. R. M. Smith, A. E. Martell, Critical Stability Constants, vol. 2. Amines, Plenum Press, New York, 1975, p. 252. R. M. Smith, A. E. Martell, Critical Stability Constants. Vol. 5. First Supplement, Plenum Press, New York, 1982, p. 254. P. Bermejo-Berrera, M. Aboal-Somoza, A. Moreda-PinÄeiro, A. Bermejo-Barrera, J. Anal. At. Spectrom. 1995, 10, 227. H. Bode, Z. Anal. Chem. 1955, 144, 165. AOAC Of®cial Methods of Analysis 1980, AOAC, Washington DC, 1980, p. 138. J. Dolezal, P. Povondra, Z. Sulcek, Decomposition Techniques in Inorganic Analysis, Iliffe Books, London, 1963, p. 90 and ss.
Received December 15, 1997. Revision September 21, 1998.